A Companion To The Philosophy of Mind

RIa ck lPcl1
( ~ o m pan ion s to
Philosophy
ACOMPANION
TO THE PHILOSOPHY
OF MIND
Edited by
SAMUEL GUTTENPLAN
IiJ
Blackwell
Companions to
Philosophy
A Companion to the
Philosophy of Mind
Edited by
SAMUEL GUTTENPLAN
II
BLACKWELL
Reference
Copyright Blackwell Publishers Ltd, 1994, 1995
Editorial organization copyright Samuel Guttenplan, 1994, 1995
The right of Samuel Guttenplan to be identified as author of this work has been
asserted in accordance with the Copyright, Designs and Patents Act 1988.
First published 1994
First published in paperback 1995
Reprinted 1996 (twice)
Blackwell Publishers Ltd
108 Cowley Road
Oxford OX4 IJF, UK
Blackwell Publishers Inc.
238 Main Street
Cambridge, Massachusetts 02142, USA
All rights reserved. Except for the quotation of short passages for the purposes
of criticism and review, no part of this publication may be reproduced, stored
in a retrieval system, or transmitted, in any form or by any means, electronic,
mechanical, photocopying, recording or otherwise, without the prior permission
of the publisher.
Except in the United States of America, this book is sold subject to the condition
that it shall not, by way of trade or otherwise, be lent, re-sold, hired out, or
otherwise circulated without the publisher's prior consent in any form of binding
or cover other than that in which it is published and without a similar condition
including this condition being imposed on the subsequent purchaser.
British Library Cataloguing in Publication Data
A CIP catalogue record for this book is available from the British Library
Library of Congress Cataloging in Publication Data
A Companion to the philosophy ofmind/edited by Samuel Guttenplan.
p. cm. - (Blackwell companions to philosophy)
Includes bibliographical references and index.
ISBN 0-631-17953-4 - ISBN 0-631-19996-9 (Pbk)
1. Philosophy of mind. I. Guttenplan, Samuel D. II. Series.
BD418.3.C62 1993 93-39595
128' .2--dc20 CIP
Typeset in 10.5 on 12.5pt Photina
by Acorn Bookwork, Salisbury, Wilts.
Printed and bound in Great Britain by Hartnolls Ltd, Bodmin, Cornwall
This book is printed on acid-free paper
Contents
List of Contributors
Preface
Part I
An Essay on Mind
Part II
A Companion to the Philosophy of Mind, A-Z
action (1)
action (2)
agency
anomalous monism
artificial intelligence
behaviourism
belief (1): metaphysics of
belief (2): epistemology of
Chomsky, Noam
cognitive psychology
computational models of mind
concepts
conceptual role semantics
connectionism
ix
xiii
1
111
117
121
122
122
132
140
146
153
167
176
185
193
200
v
CONTENTS
consciousness 210
content (1) 219
content (2) 225
Davidson, Donald 231
Dennett, Daniel C. 236
desire 244
developmental psychology 250
Dretske, Fred 259
dualism 265
eliminativism 270
emotion 270
epiphenomenalism 277
explanans/ explanandum 288
externalism/internalism 289
first-person authority 291
Fodor, Jerry A. 292
folk psychology (1) 300
folk psychology (2) 308
functionalism (1) 317
functionalism (2) 323
history: medieval and renaissance philosophy of mind 333
history: philosophy of mind in the seventeenth and
eighteenth centuries 338
holism 347
identity theories 348
imagery 355
imagination 361
innateness 366
intensional 374
intention 375
intentionality (1) 379
intentionality (2) 386
vi
introspection
language of thought (1)
language of thought (2)
Lewis, David: Reduction of Mind
Leibniz's Law
memory
mental representation
modularity
naturalism
natural kind
normative
ontology
pain
perception
perceptual content
phenomenal/phenomenological
physicalism (1)
physicalism (2): against physicalism
possible world
practical reasoning
property
proposition
propositional attitudes
psychoanalytic explanation
psychology and philosophy
Putnam, Hilary
qualia
Quine, Willard Van Orman
radical interpretation
rationality
reasons and causes
CONTENTS
395
401
408
412
431
433
441
441
449
449
450
452
452
459
463
471
471
459
484
485
486
486
488
493
500
507
514
520
526
526
531
vii
CONTENTS
reduction 535
representation 536
Ryle, Gilbert 541
Searle, John R. 544
the self 550
self-deception 558
sensation 560
simulation theory and theory theory 561
Stalnaker, Robert 561
subjectivity 568
supervenience 575
syntax/semantics 588
teleology 585
thought 585
thought and language 589
Turing, Alan 594
twin earth 596
type/token 596
the unconscious 598
weakness of will 608
the will 610
Wittgenstein, Ludwig 617
Index 623
viii
Contributors
Lynne Rudder Baker
University of Massachusetts,
Amherst
William Bechtel
University of Washington, St Louis
Ned Block
MIT
Michael E. Bratman
Stanford University
Malcolm Budd
University College London
Alex Byrne
MIT
Christopher Cherniak
University of Maryland
Noam Chomsky
MIT
Paul M. Churchland
University of California, San Diego
Andy Clark
University of Washington, St Louis
B. Jack Copeland
University of Canterbury,
New Zealand
Tim Crane
Donald Davidson
University of California, Berkeley
Lawrence H. Davis
University of Missouri - St Louis
Ronald de Sousa
University of Toronto
Daniel C. Dennett
Tufts University, Boston
Fred Dretske
Stanford University
Howard L. Fields
University of California,
San Francisco
Jerry A. Fodor
Rutgers and City University of
New York
Sebastian Gardner
Birkbeck College, University of
London
Jay L. Garfield
Hampshire College, Amherst
Alan Garnham
University of Sussex
Alvin 1. Goldman
University of Arizona
ix
CONTRIBUTORS
Samuel Guttenplan
London
John Haldane
University of St Andrews
D. W. Hamlyn
London
William D. Hart
University of Illinois at Chicago
Christopher Hookway
University of Birmingham
Jim Hopkins
King's College London
Terence E. Horgan
University of Memphis
Annette Karmiloff-Smith
MRC Cognitive Development Unit
and University College, London
Jaegwon Kim
Brown University
Kathleen Lennon
University of Hull
Ernie LePore
Rutgers University
David Lewis
Princeton University
William G. Lycan
University of North Carolina
Edwin McCann
University of Southern California
Colin McGinn
Rutgers University
x
Brian P. McLaughlin
Rutgers University
J. Christopher Maloney
University of Arizona
M. G. F. Martin
Kirstie Morrison
Wolfson College, Oxford
John Morton
MRC Cognitive Development Unit,
London
Norton Nelkin
University of New Orleans
Brian O'Shaughnessy
David Papineau
Christopher Peacocke
University of Oxford
John Perry
Stanford University
Donald D. Price
Medical College of Virginia
Hilary Putnam
Harvard University
Georges Rey
University of Maryland
David M. Rosenthal
City University of New York
J ames Russell
University of Cambridge
Stephen Schiffer
City University of New York
Robert Schwartz
University of Wisconsin -
Milwaukee
John R. Searle
University of California, Berkeley
Gabriel Segal
Sydney Shoemaker
Cornell University
Paul Smolensky
University of Colorado at Boulder
and Johns Hopkins
CONTRIBUTORS
Robert Stalnaker
MIT
Dennis W. Stampe
University of Wisconsin - Madison
Stephen Stich
Rutgers University
Michael Tye
Barbara Von Eckardt
University of Nebraska - Lincoln
xi
Preface
Like most volumes in the Blackwell Companions to Philosophy series, this one
contains alphabetically arranged entries covering its subject matter. However,
there are differences between this Companion and the others, which it might
be helpful to highlight.
1 An Essay on Mind, which forms Part I of this book, is intended to provide an
alternative, non-alphabetic, way of navigating through the entries. No doubt
the alphabetic arrangement will suit those who are familiar with philosophy of
mind, but this will not be the position of all readers. The idea is for the reader to
use the Essay to get his or her bearings in respect of some topic and then to
pursue it in more depth by reading the entry which is cross-referenced in the
Essay. SMALL CAPIT AL LETTERS are used to indicate cross-references, as they are
in individual entries throughout the Companion. The Essay is not a compendious
survey of the whole of philosophy of mind as those areas which are extensively
covered in the entries are touched on only lightly in it. Instead, it is a selective
narrative which attempts to adumbrate a picture of the mind, and some of the
philosophical problems it generates. It is my hope that it will tempt, rather than
merely introduce, the reader into the subject.
2 There is a tendency to identify many viewpoints in the philosophy of mind
by the authors with whom they are most closely associated. One often hears
views described, for example, as 'Davidsonian', where this name conjures up a
number of theses which together have been influential in the subject. It was
this fact which led me to commission a number of 'self-profiles': straightforward
accounts by certain well-known philosophers of their particular conception of
the mind (or at least some of the central features of that conception).
3 Reflecting sometimes the lack of agreement, and sometimes the mere differ-
ence of approach which exists in respect of even basic notions in the philoso-
phy of mind, I have in some cases commissioned two entries on the same topic,
and these are always marked by (1) or (2) after the entry heading. Also, by
commissioning longer entries than one finds in some of the other Companions,
I have encouraged authors to go beyond introductory material. As a result,
there is much contained in the volume which will be of interest to those who
already work in the relevant fields.
4 Within philosophy, the philosophy of mind is easily the most active sub-
discipline. It is virtually impossible to pick up a mainstream philosophy journal
without finding one or more articles on some topic in philosophy of mind, and
xiii
PREFACE
there is a constant change in what are considered to be the most burning
issues. To some extent this made the choice of entry headings difficult, and this
was compounded by the fact that philosophy of mind has become inextricably
linked to such related areas as computational modelling and cognitive psychol-
ogy. However, I did not want to make this a Companion to all of these areas -
to the whole of what is often called 'cognitive science'. So, whilst there are
carefully selected entries on fields adjacent to the philosophy of mind, the focus
of the book is definitely philosophical.
I cannot say that editing this Companion has been an easy task, but I have
learned a great deal from it. And here I refer to what I have learned about the
philosophy of mind and not to my now greater knowledge of the difficulties
of working with sixty or so contributors. (Though that too is undoubtedly
something that could be put to work in the philosophy of mind. See FOLK
PSYCHOLOGY.) What I hope now is that others too will learn from it.
In compiling the original list of entries and in matching them to prospective
authors. Ned Block gave me much helpful advice. Special thanks are due to Kirstie
Morrison for the work she did in providing both philosophical commentary and
editorial control over what turned out to be a typescript some twelve inches thick.
xiv
Samuel Guttenplan
Birkbeck College
London
PART I
AN ESSAY ON MIND
Preliminaries
Human beings definitely have minds. Other creatures on this planet or else-
where may have minds. Inanimate objects such as rocks do not have minds.
These claims will no doubt seem unexceptionable to all but the most perverse.
Yet in attempting to understand them fully, the original intuitions on which
they are based can get pushed and stretched to such an extent that, in the end
and without any perversity, we can come to have doubts. It is as if our every-
day and unexamined conception of the mind contains features that, when
examined, undermine the very conception itself.
This observation may strike some as an unnecessarily pessimistic way to
begin an introduction, but such a reaction ignores the central role that per-
plexity has always played in philosophy. In most disciplines, problems define at
most the outer boundaries - the frontiers - of investigation, not the subject
matter itself. For example, whilst the research programme of molecular biology
is determined by what it is about biological structure and chemistry that we do
not know, the subject itself - what someone would study in a textbook - is a
growing compendium of what we have already found out. With philosophy,
matters are, if anything, the reverse. Uninformed opinion sometimes mockingly
implies there isn't any thing like philosophical knowledge, that philosophy
makes no advances. This is not true. There has been over the centuries a con-
siderable accretion of insight and analysis that could count as philosophical
knowledge. However, in philosophy this accumulation serves at most as a
background. The core of any philosophical subject matter - what is distinctive
about it - is not what we know, but what continues to puzzle and perplex us.
Indeed, it would not be an exaggeration to say that when someone speaks
about 'The philosophy of X' for some specification of X, what is intended is a
budget of unresolved questions - questions whose very form is sometimes a
matter of intense debate. And even this way of putting it is not strong enough.
For sometimes one finds not merely questions, but the dizzying prospect of
paradox which threatens to overwhelm any attempt to think about some
subject matter in a systematic way.
Nowhere is this shown more clearly than in the philosophy of mind. In
trying to layout our supposed wisdom about the mind, and, further, in trying
to integrate that putative knowledge into our wider understanding of the world
of nature, we end up with the fascinating, stubborn and even paradoxical
problems that have come to define the subject matter of philosophy of mind.
And, as was noted above, the very stubbornness of these problems has some-
3
AN ESSA Y ON MIND
times tempted philosophers into proposing radical solutions - solutions that
challenge the original starting points of the investigation.
This Essay will outline the trajectory leading from our initial conception of
the mind to the problems that can be uncovered in respect of that conception,
and finally to the kinds of resolution that have been suggested for these prob-
lems. At numerous points, the outline can be filled in by consulting the entries
that are highlighted in the text by small capital letters. Since there are pro-
blems enough in the philosophy of mind itself, I don't want to add to their
number by making you wonder at each stage where we are headed. So, con-
sidering the mind to be something like a newly discovered territory, here is a
brief account of what you will find in each of the three stages of this Essay.
Stage 1 Surface Exploration
Surveying a new territory - finding out what kinds of resources there are - we
first have to make a sort of surface map and then try to bring some minimal
order to what is there recorded. Thus, our first task will be to chart the mind
and note the main features of its landscape.
Stage 2 Digging Deeper
Having a reasonably neat map of the terrain is only the first step, for surface
features often have a deceptive appearance. In this second stage we shall begin
to probe the surface landscape with various questions; this probing will reveal
faults and fissures not evident, as it were, to the naked eye. The questions we
shall ask are of the sort often labelled 'philosophical', but this does not mean
that they are specialist in any way. Indeed, the fascinating thing is precisely
that so many problems and puzzles begin to emerge when we ask certain
obvious questions about the lay of the land characterized at stage 1.
Stage 3 Bedrock
The mysteries of what lies just under the surface of our initial map might well
be thought work enough for philosophers of mind. But sometimes it is neces-
sary to dig even deeper in order to understand the contours of the upper layers.
In particular, it has seemed to many philosophers that we can never really be
certain of what the mind is like without understanding how the whole of the
structure sits on the relatively stable bedrock of the physical world. This is the
world of material stuff, of atoms and molecules weaving their patterns in
accordance with physical laws. We know that at the extreme limit of physical
complexity are such things as biological organisms and the physiological struc-
tures that make them up. We know also, or at least strongly suspect, that
certain of these physiological structures - brains and their attendant neuro-
physiological mechanisms - are deeply implicated in the very possibility of a
mental life. But how? How, if at all, does what was charted at stages 1 and 2
fit onto the bedrock of physical reality described by sciences such as physics,
chemistry and biology? This will be the third and final stage of our investiga-
tion.
4
PRELIMIN ARIES
The metaphor that runs through this introduction is intended as more than
a rhetorical flourish: it is extremely important that each of the stages be kept as
separate as possible, and it may be easier to do this if we think of the mind as
being subject to different layers of exploration. Stage 1 maps the surface, stage
2 explores just below the features of that surface, and stage 3 brings to bear a
kind of geological knowledge on what we have already uncovered. Less meta-
phorically, the difficulties we will come across at stage 2 - certain philosophical
problems of the mind - arise from the very conception of the mind with which
we began. The difficulties considered in stage 3 - problems about how the
mind is related to the world as described by science - have an origin outside
that conception, though they are clearly not independent of it. As you will see,
it is not always easy to keep these investigations separate - some ways of for-
mulating various specific questions in these two areas can make them sound
more or less the same. But it is worth trying, since not a little confusion can
result from mixing them up.
5
Stage 1: Mapping out the territory
1.1 First Steps
The starting point for our map of the mind is description. With the minimum
of theoretical (that is, philosophical) baggage, we need to describe those fea-
tures of the mind that figure in the landscape we are trying to map. And,
unlike any real charting of a territory, this task will not require field trips.
Without going anywhere, each of us is perfectly well-placed to do the job, since
we come equipped with (at least one of) the very things we aim to describe. Of
course, it may well be that the proximity of the mind can be, in the end, a
source of error. As you will come to appreciate, there can be two views here:
one stresses that the mind is special precisely because it is knowable from the
'inside', whilst the other view, insisting that real knowledge must be observer-
independent, demands that we study the mind from somewhere more object-
ively 'outside'. Exactly what the 'inside/outside' metaphor comes to will be
considered in later sections, but for the present we can proceed without worry-
ing too much about this.
Ideally, I should like to ask my readers to think about how they would
answer the following question: what things or phenomena count as mental, as
showing the presence of minds? These answers would then serve as the start-
ing point of our investigation. Though circumstances do not allow me to
gather this information directly, I can do the next best thing. For, over the
years, I have handed out a questionnaire to students before they have done
any philosophy of mind, asking them to list the sorts of things that they would
count as showing the presence of minds. Below is a lightly edited collation of
their answers.
ability to learn
awareness
ability to represent
consciousness
dreaming
experiencing happiness
having a point of view
imagining
loving
perceiving
6
acting intentionally
believing
choosing
deciding
emotions
feelings
having free will
intending to write an essay
melancholy
perceiving
agency
building a house
ability to value
desiring a holiday
experiencing a pain
getting the point of a joke
hearing a violin
introspecting
painting a picture
pleasure
reasoning
seeing a tree
theorizing
understanding language
will power
STAGE I: MAPPING OUT THE TERRITOR Y
reflecting on a problem
self-consciousness
the self
understanding symbols
anger
remembering
speaking
thinking
wanting
No doubt one could think of ways in which this list could be altered. First.
among other things. there might be felt to be a lot of unnecessary redundancy.
For example. seeing a tree seems to be at the same level of generality as hearing
a violin. and both would count as perceiving something. It is thus not clear why
we need to have all three in the list. In defence of my editing let me say this: in
response to my questionnaire. students tend to include items of radically differ-
ent degrees of generality. This may itself provide important clues and. therefore.
items should not be left out merely because of certain intuitions about what
goes with what - at least not at this point.
Secondly. it should be remembered that the list I have given is a collation of
the answers given by many different students. and you may not agree with a
number of the choices. Most importantly. you might feel that some item does
not belong on the list - is not genuinely of the mind. For example. it must be
said that a number of students argue that actions should be counted as at most
the outcome of what goes on in minds. and therefore as not deserving the same
status as such things as feelings. To this I can only say that further discussion
can show if this is a reasonable attitude. For there were many students con-
vinced that human action was just as important to the characterization of the
mind as other phenomena. and we must not begin our inquiry by closing off
the possibility that they are right.
Let us call the subject matter that is defined by the above list the 'mental
realm'. This somewhat grand-sounding title has a certain vagueness. but the
items on the list are such a heterogeneous bunch that any less vague term
would prejudice further discussion.
1.2 Order out of chaos
When discussing this list. it is possible to query various items and to see why
they were chosen for inclusion. This interchange is important because it leads
directly to the next task - putting the features into the order necessary for a
map. Merely having a list of landmarks in the mental realm is not enough. A
map must show the relationships among them.
The first thing to note about the list is that it contains broadly two sorts of
item: (i) things people (or other possessors of minds) can be said to do or
undergo which are naturally reported by verbs; and (ii) things that are.
roughly. the products or outcomes of such activity and which are described by
nouns. For example. thinking of a number between 1 and 10 is certainly some-
thing done. whilst the thought of a number between 1 and 10 could be considered
7
AN ESSA Y ON MIND
the product or upshot of some such activity. We use a verb to describe the
former and a noun phrase to describe the latter. (But don't think of 'product' in
its most literal sense. Certainly, I do not want to be taken as saying that a
thought is manufactured by thinking.)
Leaving 'products' on one side for the moment, it seems to me (and to the
students with whom this was discussed) that there are three importantly differ-
ent kinds of thing which minds get up to and which are represented in the list
in more or less generality. With several specific examples of each, these main
categories of the mental realm are given as follows:
Experiencing (having a pain, 'seeing' stars when you bump your head)
Attitudinizing (wanting a piece of chocolate cake, believing that the Earth is round)
Acting (signing a cheque, making a chair, reaching for a glass)
Each of these is an activity of mind, at least in the sense that the classificatory
word is in each case a verb, though that alone does not tell us much. More-
over, there is bound to be some puzzlement about the second of these items.
Experiencing and acting are themselves represented in the original list and I
have simply drafted them in to be the names of general categories in the
mental realm. But we do not ordinarily speak about 'attitudinizing' and this
term requires, and will be given, further comment. However, everyone knows
(sort of) what it is to want or believe something, so I shall let the examples
serve for the moment, returning later to the mysterious 'attitudinizing'.
Insofar as each of the above is an activity, each of them will have a char-
acteristic or associated 'product'. They are as follows:
Experiencing --------- > consciousness
Attitudinizing - - - - - - - - - - - - > attitudes
Acting -- -- -- -- -- -- - - - > actions
It might be thought odd that I have used the word 'consciousness' as the
partner of the activity of experiencing rather than 'experience'. In fact, nothing
much hangs on this, and my reason for having broken the symmetry is simply
that 'experience' can be either a noun or a verb, whereas what was wanted
was something more clearly a noun. Also, the point of the strange word 'atti-
tudinizing' might now be clearer. Speaking of such things as beliefs and wants
as attitudes is closer to ordinary usage. Nonetheless, to want something - to
adopt that attitude - is a kind of doing; it is something we report with a verb.
All I did was to make up the general verb which (interestingly) seems to be
lacking in our language.
8
STAGE 1: MAPPING OUT THE TERRITOR Y
As you will come to see, these three pairs are particularly important to
anyone trying to chart the mind's landscape. Like mountains, they constitute
the most prominent features. Yet before using them as fixed points in our map-
making activities, I should like to say something specific about each of them.
Considering that they all figure, at least initially, in most people's inventory of
the mental realm, they are surprisingly different from one another.
1.2.1 Experiencing and Consciousness
The laughter of the class, graduating from the first shrill bark of surprise
into a deliberately aimed hooting, seemed to crowd against him, to crush
the privacy that he so much desired, a privacy in which he could be
alone with his pain, gauging its strength, estimating its duration,
inspecting its anatomy. The pain extended a feeler into his head, and
unfolded its wet wings along the walls of his thorax, so that he felt, in
his sudden scarlet blindness, to be himself a large bird waking from
sleep. The blackboard, milky slate smeared with the traces of last night's
washing, clung to his consciousness like a membrane. The pain seemed
to be displacing with its own hairy segments, his heart and lungs; as its
grip swelled in his throat he felt he was holding his brain like a morsel
on a platter high out of hungry reach. (From The Centaur by John
Updike, pp. 3-4)
Perhaps the most persistent view that I have come across from students is that
our ability to experience and, thereby, to be conscious or aware of certain
things is a central activity of the mind. Indeed, some consider that the very
essence of the mind lies here. But what sort of things figure in this awareness?
Well, as the above quotation shows - graphically - there seems to be a special
kind of awareness of the state of our bodies and of our perceptual interactions
with the world. If you have been damaged or if certain bodily events are taking
place, then this will usually result in a consciousness of pain or pleasure, pres-
sure or fatigue, hunger or satiation, etc. Or, if you are seeing something there
is often a particular consciousness of what it is like to have such a perception;
the teacher described in the above passage sees the blackboard, and, in seeing
it, experiences it in a particular way. Additionally there is a kind of experience
that seems related to these but does not apparently depend on there being a
particular kind of damage or event in the body, or a perception. Think of the
moods and feelings that rise in us and accompany our other activities, often for
no obvious reason. A sense of well-being, a lurking anxiety that all is not going
well, these are just two of the many shades of experience that are like pains
and bodily pleasure, but which do not seem to have a particular location in the
way those do.
An important thing to notice about all of the above phenomena is that they
count as experience of what goes on 'inside', even when, as in the case of
9
AN ESSAY ON MIND
perception, there is something external to our consciousness. Walking down a
city street in the cool of March, you feel the wind in your face as it is funnelled
through the gaps in the taller buildings, you have the experience of greys and
browns of drab buildings and leafless trees, and you hear the hum of the traffic
punctuated here and there by louder sounds of impatient drivers using their
horns or trucks accelerating away from traffic lights. The wind, buildings, trees
and traffic are 'outside' of us, but we nonetheless count our experience of them
- what goes on when we perceive them - as 'inside'.
This whole show of experience - inside and outside, repeated in thousands of
varying ways as we move from place to place - is what counts for many as the
core of the mental realm. The view of some of my students tends to be: to have
a mind is nothing other than to have what is often described as a 'stream of
consciousness' - a kind of show that is going on most of the time. And the
metaphor of a show is the one that crops up most often when I ask for a
description of experience - a description of what it is like to be the possessor of
a stream of consciousness. 'It is as if you were in a cinema watching a film
from so close and with such involvement that you were only aware of what
was happening and not that it was happening on a film in an auditorium.'
Fine, I say to this recurrent sort of answer to my question, but it seems to
require us to understand what it is to be aware of a film in some particularly
close way, so it is not all that much help in telling someone what awareness
itself is. Moreover, this account seems to apply best only to our perceptual
experience, to the experience - itself inside - of what is happening outside. But
what about such things as pains and other wholly inner sensations? The
needed revision often runs as follows: 'Well, it's not exactly like the show in a
cinema, but it does seem to involve witnessing various things - observing
them, paying attention to them - even if sometimes from a very short meta-
phorical distance. When I have a pain, I direct my attention to it, just in the
way that I direct my attention to my present experience of, say, colours in my
visual field. This is sort of like a film or theatrical performance which I can
witness and with respect to which I can differentially direct my attention from
one character to another.' (See CONSCIOUSNESS.)
Does this sort of metaphorical description help? Perhaps it points you in the
direction of what I mean to speak about under the heading of 'experience',
but I doubt it is much more informative than that. Indeed, it raises more
questions than it answers: for example, who or what does the directing of
attention in this case? 'The self' comes the reply. But this reply also gets us
into very deep waters. Is the self separate from experiential activity and its
attendant consciousness or are they united rather as the dancer and the
dance? Here we are beginning to see some of the problems that lie just
beneath the surface of our conception of experience and consciousness, and
for the present we shall leave well enough alone. In any case, perhaps there is
not much more that one can do in directly characterizing experience than to
reach for metaphors such as are found in Updike's wonderfully lurid descrip-
tion of a pain.
10
ST AGE 1: MAPPING OUT THE TERRITOR Y
1.2.2 Attitudinizing and Attitudes
All this was lost on Alice, who was still looking intently along the road,
shading her eyes with one hand. 'I see somebody now!' she exclaimed at
last. 'But he's coming very slowly - and what curious attitudes he goes into!'
(For the Messenger kept skipping up and down, and wriggling like an eel, as
he came along, with his great hands spread out like fans on each side.)
'Not at all: said the King. 'He's an Anglo-Saxon Messenger - and those
are Anglo-Saxon Attitudes. He only does them when he's happy. (From
Through the Looking Glass by Lewis Carroll, p. 175)
In the subtle shift of perspective in this passage - the shift from attitude as
posture to attitude as a feature of a mind - Carroll has given us several impor-
tant hints about mental attitudes. We are invited to imagine the Anglo-Saxon
Messenger as taking up odd postures, setting his limbs in awkward or uncom-
fortable positions. However, in ways it is perhaps more tactful for me to leave
unsaid, the Anglo-Saxons have the reputation of having odd (even sometimes
uncomfortable) attitudes - beliefs and desires - in respect of a variety of sub-
jects.
The appeal of this passage is that it effortlessly manages to shift our attention
from a set of bizarre postures to a set of perhaps equally bizarre attitudes
towards life. In using the two senses of 'attitude' in the same context, Carroll
succeeds in getting us to pause over something that we don't usually bother
much about - the aptness of the word in its 'posture' sense for characterizing
such things as beliefs, desires and the like. A posture is something we man-
reuvre ourselves into and which is therefore observable in our behaviour. Simi-
larly, we usually tell what someone believes or desires by things done and said
- by behaviour; an attitude in this sense is a mental state which we often
'read' off from behaviour. Moreover, it is true of some attitudes, even in the
posturing sense, that they are directed or indicative of something. When
someone is said to adopt a menacing attitude towards another, what is in
question is not merely how the first person is standing, though some such
bodily position is being described. Rather what is special about a 'menacing
attitude' is that it is a posture that is directed towards someone or something.
And of course this is precisely what is typical of such things as beliefs and
desires. They are not merely states of mind we discern through behaviour, they
are states of mind that have a special kind of directedness. I don't just believe
or desire - I believe that something is the case, or I desire someone or some-
thing.
The two crucial defining features that any case of attitudinizing displays are:
(a) A kind of behaviour that is typically characteristic of the particular atti-
tude in question. (Imagine how you could tell the difference between
someone who wanted something, believed something, intended something,
etc.)
11
AN ESSA Y ON MIND
(b) A 'something' towards which the attitudinizing is directed, as when we
say that:
Harry believes that his telephone is out of order, or
Jane desires a new car, or
Bill intends to boil a kettle.
Note that the items towards which the attitude is directed can be quite various:
in the above three examples we have these three items:
that the telephone is out of order
a new car
to boil a kettle.
Focusing on the sentences we use to report attitudes and borrowing a term
from grammar, we shall call the 'something' towards which attitudes are dir-
ected the complement of the attitude. That is, in the sentences given in (b) there
are complement phrases which report the particular direction of the attitude.
Note that the first of these has a declarative sentence as a complement. This is
important because sentences like this are typically used to say, truly or falsely,
how things are. One way to put this is to say that declarative sentences express
PROPOSITIONS. Moreover, it seems possible (even if it might sound awkward in
given cases) to report virtually all attitudes using complement phrases that
contain whole sentences. We could have expressed the other examples in (b)
as:
Jane desires that she has a new car.
Bill intends that he will make the kettle boil.
Because complements of belief reports typically contain a complete declarative
sentence that expresses a proposition, and because the other attitudes can be
twisted into this shape, philosophers have settled on the idea that the products
of attitudinizing can all be called 'PROPOSITIONAL ATTITUDES'. SO, the Anglo-
Saxon messenger strikes odd postural and propositional attitudes.
Note, by the way, that one can also call the item to which an attitude is
directed its content. The notion of a 'complement' seems to many to be too
grammatical and too closely tied to the report of an attitude, whereas the word
'content' seems to capture something about the attitude itself. But for the
present it won't matter much whether you think of the attitudes as having
complements or contents.
It remains to be seen just how much trouble the propositional attitudes
create for our understanding of the mental realm, but it is difficult to deny that
they represent a large part of our everyday conception of that realm. And this
comes as a bit of a surprise to those who are convinced that experience is the
central feature of the mind. For, whatever else they are like, the propositional
12
ST AGE 1: MAPPING OUT THE TERRITORY
attitudes are not obviously items of experience. For example, suppose someone
were to ask you, out of the blue, whether the present government will be
returned to power at the next election. I have no doubt that your answer
would be readily forthcoming and would begin like this: I believe that . . . But
to get to this answer did you have to search the elusive stream of conscious-
ness we just discussed? Does that stream contain a sort of banner on which is
written 'the present government will not be returned to power at the next elec-
tion'? Hardly. Of course, I don't doubt that images of governments - a sort of
collage of images of politicians, government buildings, television coverage of
elections and perhaps even images of words - might be prompted by the ori-
ginal question. Yet these do not constitute the belief itself. In fact, those not so
wedded to the experiential picture of the mind as to rule out everything else
tend to report that consciousness plays very little role in our ability to know
and say what attitudes we have.
This last observation points the way down a number of difficult roads. If
consciousness figures less (and sometimes not at all) in our apprehension of our
beliefs, then how do we tell what we believe, want, intend, etc.? We certainly
don't do it in the way we tell these things about other people, Le. by looking at
what they do and say. Moreover, what relation is there between the 'self'
which made its appearance in our discussion of experience and the item that is
the subject of attitude reports? In what way is the 'I' of 'I am in pain' related to
the T of 'I believe that it will snow'? These sorts of question are typical of the
next stage of investigation. But our interest at present has only been in the
kind of thing that comes under the headings 'attitudinizing' and 'attitude', and
we have completed that task. The activity of attitudinizing results in our
having attitudes towards the ways things are or might be; each attitude has its
typical manifestation in behaviour; and all can be provided with propositional
contents that are reported by complement sentences.
1.2.3 Acting and Actions
The astonishing thing about action is that it is possible at all. For, if a man
is making a chair, you will find a physical causal explanation of the move-
ment of each piece of wood from its initial to its final setting; everything
that happens is in accordance with law; but you will look throughout this
world or universe forever in vain for an analogous physical explanation of
their coming together in the form they did, a form that mirrors human
need and the human body itself. (Try it.) (From 'Observation and the Will'
(Journal of Philosophy, 1963) by Brian O'Shaughnessy)
As I mentioned earlier, there is a strong tendency to overlook actions when
thinking about what to count as items in the mental realm. Those who find
themselves only reluctantly admitting attitudes into the fold, dig in their heels
at what they regard as too physical a thing to count as anything mental. Such
is the pull of the idea that the mental consists in the 'inner' - the show of
13
AN ESSA Y ON MIND
experience and consciousness - that actions can seem just too far removed
from this centre to count as anything more than the mind's wake as it moves
through the physical world. But this view is by no means universal, and one
would do well to listen to those who oppose it.
The point of the above passage is to illustrate just how difficult it is to fit
actions into the picture of the world encouraged by science. Thus, whilst each
movement of the arms and hands, hammer and nails might well be explicable
in terms of science, the fact that all these things come together as the making
of a chair can seem quite mysterious. Discussion of how the mental realm fits
in with the scientific picture of the world will only come in stage 3 of this
introduction. But the situation described in the passage can be used to illus-
trate something more pertinent to our present concerns.
Begin by supposing that everything is as described in the passage except that
the agent making the chair is invisible. To an unsuspecting witness the pieces
of wood seem to rise up and be nailed and glued in place, and the chair just
comes into being. This would of course be astonishing, but we can leave this
on one side for the moment. What I want to ask is this: would the witness
actually observe the action, the making of the chair? Clearly, by hypothesis, the
agent goes unseen, but if you are one of those who think of the action as
nothing but some sort of change in the physical world, you should be prepared
to say that the action is seen, even if not the actor. Yet that is surely not how
we would describe it. Why? Well, the very idea of an action - even of a purely
'physical' action - seems to require us to identify some sort of mental compon-
ent. As the passage notes, were there not human desires and needs, as well as
the further beliefs, desires and intentions to fulfil them, at work, then we would
not have the faintest idea of what was going on. When we do see the actor, we
see some or all of these attitudes in the transformation of the materials, and
unless we can see the mind in the process unfolding before us, we simply don't
count that process as an action; for all we know it might just be the accidental
product of some strange cosmic wind.
The idea that an action is in this way at least partly a mental phenomenon
is what one of my students had in mind with the comment: 'actions are the
mind's purposes in movement'. But those who insist that actions are not them-
selves mental, still have something to say. Here is a typical rejoinder:
What the example shows is that you couldn't imagine the pieces coming
together unless there was some mind orchestrating the movements. But the
action itself - the physical movement of the pieces - is not mental. What
happens is that you see these movements - the action - and then infer
that there are mental states directing them. In seeing the action, you don't
literally see the mind.
This rejoinder throws up many intricate problems and these must await further
discussion in stages 2 and 3. However, whatever we end up saying about an
action such as making a chair, it must be pOinted out that the class of things
14
called actions is much broader than we have so far allowed. Making a chair is
what is called a 'physical action' - an action in which some change is effected
in some physical object or event. Examples of this kind of action are what most
people think of when they are asked to imagine an action taking place, and it
is this kind of action that leads to the greatest disagreement in measuring the
boundaries of the mental realm. However, there is another kind of action,
which has been staring us in the face, the mental status of which must be
beyond doubt. I have in mind here the very activities of experiencing and atti-
tudinizing. Recall that I was careful to insist that the main categories of the
mental realm had both an activity and a product sense: experiencing and con-
sciousness, attitudinizing and attitudes, as well as acting and actions. But
surely, for example, to direct one's attention to some item in the stream of con-
sciousness - to experience it - is nothing short of an action, and a purely
mental one at that. Moreover, once you begin to think about it, there seems to
be a whole host of other things that we do which are 'in the mind' in this way.
Think of your favourite colour! Work out (but don't say) the sum of 15 and
22! When you accede to these requests. you are certainly doing something -
acting - only in neither case is there any ordinary change wrought in your
physical environment. These episodes of thought and inference would thus
seem to be the tip of a very large iceberg consisting of actions whose claims to
belong in the mental realm are unimpeachable. (See ACTION.)
As with experience and attitudinizing. each case of an action comes with a
subject (SELF). or. perhaps more appropriately in the case of action. an agent.
Indeed. just as for particular items of consciousness or attitudes. it is simply
impossible to have an action without an agent. The kind of impossibility here
seems to be conceptual: we cannot conceive of an unowned pain. a subjectless
belief. nor can we conceive of an action that lacks an agent. And now we have
another element to add to the problem raised earlier: what relations obtain
between the T of 'I am in pain', and 'I believe that my keys are in the cookie
jar' and the T of 'I pruned the ceanothus too late in the year'? Clearly. there is
an enormous pull in favour of saying that the items picked out by each
pronoun are one and the same self. Indeed, this tends to be such a universally
held view among my students that it takes them some time to see that there
might be a problem - that the differences between experiencing, attitudinizing
and acting might make it less than obvious why one and the same thing does
all three.
1.3 Estimating Distances
As was mentioned earlier, experience, attitude and action are the three fixed
points in the mental landscape. All that remains then before we are ready to
produce a sketch-map of the terrain is some way to locate all the other items
(from the first list) in respect of these landmarks. If this was more a real than a
metaphorical map-making exercise, what would be required would be some
15
AN ESSAY ON MIND
way to estimate the distances and directions of each of these items from one or
more of the fixed points. But, unlike the real case, there is more to it than mere
spatial distance and direction. For as we have seen, the three categories are
quite different from one another. Indeed they seem only to share this one
feature: they are all reckoned to belong to the mental realm. So, it is no
straightforward matter to decide how to locate the other items. For example,
take the case of emotions such as anger. Should we show them to be closer to
experience or to attitude? And what about acting? There is certainly a case for
saying that emotions are expressed in action. Clearly, emotions share some fea-
tures with each of experience, attitude and action. But which features? Before
we can do any map-making we have to discuss the grounds on which we
decide how near or how far to place an item with respect to our three fixed
points.
Just to keep the metaphor going, you can think of the features or respects
that distinguish the three main categories as like dimensions. Thus, there are a
number of ways in which, for example, experience differs from attitude. To
locate some particular item on the map what we have to do is to say it is more
like experience in such-and-such a respect and less like it in another. It is the
possibility of speaking this way that allows us to describe these respects as
dimensions. But what are these respects and where do they come from? The list
is as follows:
Observability,
Accessibility,
Expressibility,
Directionality,
Theoreticity.
As with everything discussed at this surface-mapping stage, they have their
basis in the untutored judgments most people would volunteer. In that sense
they come from the same source as the list with which we began. However,
the fact that I have had to invent my own names for these dimensions does not
mean that I am imposing my particular views on the shape of the final map.
For, though the labels are mine, the conclusions reached about each of the
dimensions is distilled from discussion with the students who supplied the data
for the original questionnaire. So, I expect that the brief discussions below will
both clarify the labels and strike familiar chords.
1.3.1 Observability
Confronted with a mind (someone else's), how easy is it to tell whether you are
in the presence of experiencing, attitudinizing or acting? This is not meant to
be a deep question. There is a long tradition in philosophy of considering how,
if at all, we can justify our faith in the mindedness of others. This is not what
we are up to here. Assume that others do have minds, that the extreme scep-
tical stance is inappropriate, and ask yourself this: how easy is it to tell just by
16
STAGE 1: MAPPING OUT THE TERRITORY
looking that some mind is experiencing something, maintaining an attitude
towards something, or acting? To many, at least part of the answer is straight-
forward. Philosophical argument might well shake our convictions in respect of
all three, but it is certainly easier to wreak sceptical havoc in respect of experi-
encing than in respect of acting. The usual thought is that we can conceal
what we experience, sometimes with no effort at all, but that what we do - our
actions - are there for the looking. However, even in respect of this apparently
obvious conclusion, one must be careful.
There are experiences that would be regarded as easily observable and
actions that are not. It is natural to think that the victim of a serious accident
can be seen to experience pain, whereas someone can do something completely
away from even the possibility of prying eyes - something like adding up two
numbers, as we say, 'in the head'. That is, there would seem to be cases where
experiences are out in the open, and also cases of actions that are 'inside'.
Moreover, the idea of observation that is in play here cries out for further elu-
cidation. Still, let us agree that, though there is much more to it, the proper
place for this is in our stage 2 investigations. For now we can say that 'in
general and for the most part' experience comes at the low end of the observa-
bility spectrum, while action lies at the other. A typical case of experiencing
something - having an ache in a limb - is usually counted as fully discernible
only to the subject of the experience, whereas a typical case of acting - signing
a will - is rated as something anyone in the right place can witness.
But what about attitudes? How easy is it to see that someone wants an ice-
cream or believes that it is about to rain? The temptation is to say: it all
depends. If the circumstances are right, for example if there is enough beha-
viour to go on, it would seem to be quite easy. The child irritably resisting his
parents' best efforts to distract him from the ice-cream vendor can be clearly
seen to want an ice-cream, whereas the academic comfortably engaged in
reading a book might well believe that it is about to rain without giving our
observational abilities any purchase at all. Still, if we abstract away from
special cases and, as in respect of experiencing and acting, think only in
general and for the most part, the attitudes seem to be somewhere in between
the two extremes in respect of observability. It is easier to see what people do
than what they believe, but it is also easier to see what they believe than what
they experience. As before, there is a lot to be said about exactly what we
think is going on when we are said to observe that someone wants or believes
something, but discussion of this will come later.
1.3.2 Accessibility
How easy is it for you to tell of yourself that you are experiencing, wanting or
doing something? That is, how accessible is your own portion of the mental
realm? Do we always know what we are doing, or what we believe and want?
No, but perhaps this is because we don't always attend to these things; the idea
would be that if we did attend, we would know. Yet couldn't there be cases in
which no amount of thinking about it would lead us to acknowledge particular
17
AN ESSA Y ON MIND
actions or beliefs and wants. as our own? Indeed, aren't such cases perfectly
familiar? Smith sets out to help Jones dig the garden; he believes that he is
doing this from the goodness of his heart, and that is what he would avow
after reflection. But, to those who know him, what he is doing seems more
appropriately described as competitively displaying his horticultural superiority
over Jones; the way in which he goes about 'helping' seems to give him away.
Ask Smith what he is doing, believing and wanting and you get one answer.
Ask his friends and you get another. Perhaps Smith could be brought to see
himself in the way others do, but that is not really relevant. All that I want
this example to remind us of is the perfectly ordinary fact that we don't always
have instant accessibility to what we believe, want or are engaged in.
Experiencing, however, seems to be in stark contrast to these. Not only do
we think that such things as pains and itches are highly accessible; we would
find it difficult to imagine cases in which there was any attenuation of access-
ibility. Could you be in pain, for example, and not notice that you were? And
here, by 'pain', I mean some fairly robust example of the kind, not a barely
perceptible sensation which comes and goes too fleetingly to count as one thing
or another. You could of course be stoical about it, not show others that you
were in pain; you could even push it to the background so that it didn't inter-
fere with your present activities. But could you have a pain and not notice it at
all? This is a difficult question, a question whose very status has been debated.
In particular, is it a question about how things are as a matter of fact in
respect of pains, or is it somehow a more conceptual question: is the very
concept of pain such that it is logically impossible to have an exemplar of it
without noticing? (See INTROSPECTION.)
As in the case of observability, nothing we are engaged in just now requires
us to deal with these worries. Whatever is to be said in the long run when we
start to dig deeper, here it is enough to note what seems the unvarnished truth
to most people (and, in particular, to the students who so forcefully expressed
this view): we have a much greater degree of access to items of experience
than we do to attitudes and actions.
How do attitudes and actions compare in respect of accessibility? There is a
tendency to think that we know more about what we believe and want than
about what we do. The reason most often given for this is that acting requires
some cooperation on the part of the world: we have greater accessibility to
what we intend to do (an attitude) than to what we are actually doing or
achieving because we are only doing or achieving something if certain worldly
events are actually taking place, and we may be in error about whether they
are. Dreams illustrate the point nicely.
If, in a dream, you are about to sign a cheque then you seem to have the
intentions, desires and beliefs appropriate to that commonplace action. But if
you were actually signing a cheque, not only would there have to be this atti-
tudinal background, your hand would have to hold the pen and move in some
appropriate way. And it is precisely the latter that is missing in a dream. When
you dream that yet another bill is overdue and, in a state of generalized
18
anxiety, reach for your chequebook and write out a cheque hastily and
without due care and attention to the balance remaining in your fragile
account, you have a keen awareness of the attitudinal background - it seems
wholly accessible to you. But, as you often come to realize on waking, one
thing that didn't happen was that you signed a cheque.
Dreams are the extreme case here but there are less dramatic cases of actions
being inaccessible in ways that the attitudes are not. So, summing up, we
usually rank experiences at one extreme - immediate and full accessibility -
whereas attitudes come somewhat further down the line with actions bringing
up the rear.
1.3.3 Expressibility
It would seem equally easy to tell someone that you have a pain in your arm,
that you believe right will triumph over wrong and that you are cooking your
dinner. But many feel that this way of putting things misses an important
feature of these categories. In particular, there is a prevalent idea that, though
we can tell someone that we have a pain in the arm, we cannot express or
communicate the experience itself. As one puts it colloquially, 'what it is like'
to have a particular pain seems something that escapes even the most imagin-
ative use of language. As we have found with the other dimensions, intuitions
like this one raise more questions than they answer. For example, what exactly
would it be like to express an experience if we could? What would constitute
success in this apparently difficult task? If we don't know even that much, then
perhaps our conviction that experiences cannot be expressed is less interesting
than it seems. Still, we must not stop just yet at such deeper questions; there is
a consensus that experiences are very low on the expressibility scale, and that
is good enough for the present.
But what about attitudes and actions? Actions seem to be straightforwardly
expressible: insofar as you know what you are doing, you just put it into words
- you describe your action in some appropriate way chosen among all the ones
available to you. In appropriate circumstances, you just say: 'I'm signing a
cheque', 'paying the gas bill' or 'practising my signature on this already ruined
cheque'. To be sure, there are cases where it is not quite that easy. I can
imagine myself engaged in some intricate physical manamvre which is neces-
sary to the well-being of my bicycle, but which I cannot properly describe - it
is just too complex, even though the aim of the action itself is simple. Of
course, I could always just say: 'I am adjusting the brakes', or 'fixing my
bicycle' and this might do. Telling someone what I am doing does not always
require detailed description. In the end, then, there doesn't seem to be much of
a problem about expression here.
With belief, want, and other attitudes, the problem comes down to getting
hold of some appropriate sentence to use in the complement place in the atti-
tude report. In most cases, this is straightforward. To be sure, there are times
when you are not quite sure whether you believe something to be the case, or
merely hope that it is. And there are also bound to be times when you, say,
19
AN ESSA Y ON MIND
expect something to happen, but would be hard put to find the exact sentence
that captures the content of your expectation. (I am assuming here that expec-
tation is a specialized form of belief - belief about some future course of events.)
In sum, there are problems for both actions and attitudes in respect of
expressiblity - problems that make them about equal in this dimension. But
they are nowhere near as severe as the problems encountered in respect of
one's experiences.
1.3.4 Directionality
An attitude is a mental item which can show itself in activities and behaviour.
Of course, this is not invariably so; one can easily conceive of beliefs, desires
and the like that happen never to leak out into the realm of action. Still, it is
not unreasonable to think of the attitudes as having particular and typical
kinds of manifestation in activity. A desire to buy a new coat, for example, will
'look' very different to observers from a belief that coats keep you warm in
winter.
However, what is particularly characteristic of the attitudes is that they are
attitudes about something - they are reported in sentences which contain com-
plement clauses, or, using the other idiom, they have contents. Yet another way
of putting this is to say that attitudes are never merely expressed in behaviour,
they are also, and essentially, directed to, or at, something. For example,
compare desire with, say, vanity. Both have a claim to be the kind of thing
appropriate to the mind, but there is an important difference between them. A
vain person, like a desirous one is disposed to act in various ways, but to
understand the desire fully, we must know what it is a desire for. There is no
counterpart to this directedness in the case of vanity.
On the face of it, directionality is virtually absent in those items that most
naturally group themselves around the category of experience. Taking pain as
the first example, imagine that you have overdone some exercise and that you
are now suffering for it. You have various aches and pains and these seem to
be located in various parts of your body. They are located - and they have spe-
cific characteristics, each different from one another - but they don't seem to
be about anything; they lack directionality. Your aching thigh is not an ache
for anything - it is not reported in a sentence containing a complement clause,
and thus it does not have a content.
One must be careful here. The notion of content as just used is somewhat
specialized. It is that item to which an attitude is directed. The content of the
desire that you have a new coat is, roughly, the state of affairs of your having
a new coat. If you had it badly enough, one could describe your state - some-
what fancifully - as an ache for a new coat. In a more general sense of the
word 'content', of course it is true that a pain has a content. But this is not the
sense of the word in question.
I said that directionality is virtually absent in typical cases of experiencing
such as the pain case. Certainly, there is nothing that corresponds to the
robust use of complement clauses with which we report beliefs. But why only
20
'virtually'? Well, it seems to many as if there is a kind of directionality in the
pain case which it is easy to overlook. It is a lower grade of directionality than
we have in connection with the attitudes, and it may in the end be nothing
more than a phenomenon based on the attitudes, but it is worth remarking on.
Certainly, pain is not usually a neutral item of experience: it is something
unpleasant and which we seek to avoid. The directionality of an experience of
pain may be no more than: 'would that it would go', but it is at least possible
to see this as a primitive relation of the kind of directionality we have in full-
fledged attitude cases. Of course, one might take the view that pains just
happen to have (in us, and for the most part) a kind of connection to the atti-
tudes. On this view, it is not the pain that has any kind of directionality, it is
just that pains bring with them desires to get rid of the pain. The idea is that
the desire, not the pain, is directed.
Somewhat differently from the pain case, think of what it would be like to be
standing in front of a blue wall and looking directly at it. Your perception
would be directed: it would be described as a perception of a blue wall. But that
is not quite what is at issue. Try to forget about the fact that you perceive a
blue wall (which is surely directional, like an attitude), and think instead of the
conscious experience occasioned by the expanse of blue. This is something that
happens when you perceive the blue wall, but is distinct from the latter. It is
the experience found, as is said, by introspection in the stream of conscious-
ness, and it can be separated from what causes it (the wall), or what it is about
(the blueness of the wall). As the struggles of the last few sentences show, it is
not an easy matter to use words to point in the right direction, but most people
are, on reflection, only too eager to admit that there is such a thing as the
what-it-is-like-to-see-blue sensation in their stream of consciousness when they
direct their attention to a blue object.
Does the qualitative experience you have when you are perceiving a blue
wall constitute a case of pure, non-directed experiencing? It certainly doesn't
seem to have even the most primitive form of directionality. Unless the colour
is particularly shocking, your experience of blueness does not come with the
feeling: 'would that it would go away'. So perhaps the colour perception case is
a better example than pain of non-directedness. Or perhaps the pain case really
constitutes just as good an example, which only seems different because pain is
hooked up in us to genuine attitudes such as the desire to get rid of the pain?
Well, whether pain has a kind of primitive directionality, is not something we
need to settle here. For whatever we end up saying, it seems at first that direc-
tionality is typical of, and central to, attitudinizing, and is only of marginal
importance to experience.
What of the third category - acting? We have briefly discussed the question
of the degree to which actions are mental items. Our discussion was admittedly
inconclusive, but it was suggested that even a physical action could not be
thought of simply as a sequence of physical movements: the mind is either
actually present in the action or is intimately involved in it in some way. Thus,
signing a cheque certainly involves various hand movements, but these are (at
21
AN ESSA Y ON MIND
least) mind-directed. The movements have as their aim, for example, the
payment of the electricity bill.
As always, there is much more to be said here. But for now, it is enough to
note that, with respect to directionality, attitudes have it as a central feature,
actions include elements which are directional. and items of consciousness
have at most a minimal kind of directionality. (See INTENTIONALITY.)
1.3.5 Theoreticity
Is it possible to see electrons? Not an easy question, nor one we have to settle
definitely here. But this much seems true: whether or not one can stretch the
notion of 'seeing' sufficiently to allow it to be said that we see electrons, any
seeing of them would be a wholly different kind of thing from our seeing of
tables and chairs. Though not a precise business, it does seem that some items
count as immediately or directly observable, whereas others are less directly
observable (if observable at all). What have been called 'middle-sized dry goods'
- taking tables and chairs to be representative - falls under the first heading,
whereas electrons fall firmly under the second.
Recognizing that electrons are at best indirectly observable, the next question
to ask is: do they really exist? Here again, brushing aside the deeper rumina-
tions of certain philosophers, the answer is surely 'yes, there really are elec-
trons'. But having admitted that electrons are only indirectly observable, what
grounds do we have for saying that they exist? Undoubtedly, many people
regard the best grounds for something's existence to be its direct observability,
but there are other grounds. For instance, one could say this: the notion of an
electron forms an essential part of a theory we have about the nature of matter
- a theory that is by now established in the scientific community. Even though
we may never be able (even in principle) to observe electrons directly, we are
generally happy (give or take a few philosophical qualms) to say that they
exist. They exist because they are integral to our well-established theoretical
understanding of the universe.
Against this background, here are some things we can say about the feature
of theoreticity: chairs and tables - things we regard as directly observable -
have a low degree of theoreticity. We don't believe in the existence of these
things on the basis of our theory of the universe - we just see them. On the
other hand, electrons have a high degree of theoreticity: their very existence is
bound up with our theoretical understanding of nature.
What about the items in the mental realm? There is generally a consensus
for the view that, whatever else we say about other items, experiences have a
very low degree of theoreticity. We do not regard a pain, a visual appearance,
an experience of a sound, the changing coloured image which comes before
our closed eyes just after we have seen a bright light, as things whose existence
depends in any way on a theory we may have about how things work. Items
of experience seem to be immediately apprehended. Indeed, there is a tendency,
which has been encouraged, though not invented, by some philosophers, to
consider items of experience as more directly observable than the middle-sized
22
dry goods that surround us. Introspection can seem a more direct and reliable
guide to what exists than modes of 'extrospection' such as seeing. touching and
hearing.
Allowing strength of opinion on this to be our guide. and leaving on one side
any investigation of the basis of that opinion. we shall count experiencing as at
the lowest end of the scale of theoreticity. But what about the other two cat-
egories: attitudes and acting? Here matters get more complicated. In discussing
the accessibility of the attitudes. it was noted that. whilst we sometimes either
make mistakes as to the direction of our attitudes. or. on occasion, just fail to
register attitudes that others can more accurately gauge from our behaviour.
we often have fairly immediate access to what we desire. intend. believe, etc.
But one thing we also noticed was that. even when the access we have is fairly
immediate. it doesn't appear to be like the access we have to such things as
pain. For example. if asked whether next Sunday was the 15th. you would
surely do some kind of ruminating before answering. However. compare this
rumination with what you would go in for if I asked whether you could feel
the pressure exerted by the chair you are now sitting on. Your answer in the
second case seems something like a case of looking and discovering; that is
why the expression 'introspection' seems so apt. But this kind of introspection
seems the wrong sort of method for discovering whether you believe next
Sunday to be the 15th.
In cases of the attitudes and experiences of others. the contrast seems even
more pronounced. You find out what someone's attitudes are by being sensitive
to behaviour. Of course. you may be told point-blank what someone believes.
but even this may not settle the issue. Perhaps they are not facing up to
things. or are trying to see things in a better light. In cases more complicated
than the one about Sunday the 15th. perhaps they are mistaken about what
they believe. However. in the case of experience. it would seem that the verdict
of the subject is both necessary to an accurate judgment. and final.
How can one explain this difference? One way is this: an experience is some-
thing that is directly observable - though only by the person whose experience
it is - whilst an attitude is something not directly observable by either the
subject or his friends. On this view. attitudes are items we attribute to ourselves
and each other as part of trying to make sense of - to explain - behaviour. One
way of putting this would be to say that attitudes are part of our theory of
human nature. Clearly. a consequence of this view would be that attitudes are
more theoretical in nature than experiences. Of course. this is not to say that
they are just like electrons. After all. the explanatory theories of physics would
seem to be quite different from the 'theories' with which we explain human
activities. But the discussion of electrons was only meant to illustrate the
notion of theoreticity.
Accepting then that attitudes come out as more theoretical than conscious
experiences. what about actions? Do we directly observe actions. or do they
have a somewhat more theoretical and less directly observable nature? Here
the old wounds open up again. Those students who regarded actions as
23
AN ESSA Y ON MIND
nothing much more than physical movements would see them as directly
observable. Those who considered them to be 'purposes in movement' would
demur, since a purpose is an attitude. And, of course, one must not forget
actions that are generally classified as mental - actions such as thinking of a
number between 1 and 10. Without even trying to sort all this out here, I shall
take the easy way out by placing actions somewhere in between attitudes and
experiences on the theoreticity scale. (See DENNETT.)
1.4 The Map of the Mind
With the help of the five dimensions it is now possible to construct a working
map of the mental terrain. See figure 1.
Attitude
A
0
knowledge of language
/0 O b e ~ f
thinking 0 desire
/
imagining 0
o perceiving
OhoPin?'\
o pleasure 0
o fearing
consciousness 0
o awareness
o pain
Experience
Figure 1.
anger
Agency
o intending
" o willing
inferring 0 "
I 0 choosing
deciding 0 b.reaching
Action
The main categories are represented as three points equidistant from one
another. Think of them as mountains (seen from above) whose summits mark
boundaries derived by extrapolation from the five dimensions. Thus, for
example, Experience is that peak at the summit of which one would put any
feature of the mind that was wholly accessible, not observable, not expressible,
not directional and not theoretical. Of course, no actual feature of the mind has
this stark profile. Pain tends to be cited as the paradigm case of an experience
but there are ways in which even it falls short of being what might be called a
'pure' experience. First, whilst pain is thought of as highly accessible to its suf-
ferer, there is arguably room in our idea of the mind for pain that is not noticed
at a given time. Secondly, we do think of pain as sometimes observable - think
of the accident victim - even if it is in many cases difficult to discern from the
24
third-person perspective. Thirdly, it is not easy to express (describe) pain, but it
is not impossible to go some way towards it: one need only think here (again)
about Updike's description of the teacher in pain. Fourthly, as was noted, there
is a kind of directionality that seems to accompany a painful experience - a sort
of 'would that it would go away' content. Finally, it is possible to imagine cases
in which pain was appealed to more on theoretical than observational grounds.
Thus, a doctor might explain why you have a certain tiredness in your back by
citing the fact that you have had a pain in your leg which caused you to walk
and sit awkwardly. When you protest that you felt no such pain, the doctor
might well say that the pain never expressed itself - that it remained, as is often
said in medical circumstances, 'sub-clinical'. Given these considerations, pain
must be placed short of the summit of Experience: its dimensional profile shows
it to be some little way towards both Attitude and Action.
All the other features of the mental realm have been assigned places on the
same basis: due account has been taken of their relative distances from the
three summits. Of course, I don't want to insist that I have got these locations
precisely right: figure 1 is only intended to be a sketch map - something we
can use to orient ourselves. None the less, before we embark on the next stage
of investigation, here are a few notes explaining some of the reasons for
various placements.
(1) To keep the map uncluttered, I have left out some of the items that figured
in the full list of items belonging to the mental realm, but it should be obvious
where they would go. Thus, hope, fear, anger are placed more or less centrally
and they mark out a region within which one could put any other EMOTIONS.
This central location seems right because emotions look towards each of the
peaks without being markedly closer to anyone of them. Certainly, one can be,
for example, angry that such-and-such is the case - anger is certainly something
like an attitude with a content. Yet anger is often spoken of as a feeling, as
something accessible in the stream of consciousness. And finally anger not only
causes us to do various things, it is itself said to be expressed in action. Of
course, differences will emerge as soon as one moves from anger to one or
other of the emotions, so you should think of the central location labelled
'emotions' as a region within which more accurate placements can be made.
Perhaps a 'calmer' emotion like regret will be closer to the attitudes than anger
and further away from the other two fixed points, whilst love might be closer
to experience and further away from attitude.
Note also that feelings are placed slightly closer to experience and further
away from action than emotions. In some contexts, 'emotion' and 'feeling' are
used interchangeably, but in others, feeling owes more to experience. Its location
near the edge of the region is meant to cater for both of these possibilities.
(2) It may not be obvious why pleasure comes just within the emotion region,
whilst PAIN is firmly outside and closer to Experience. After all, isn't it virtually
a cliche to speak of 'pleasure and pain' as a contrasting pair? Yes, there is this
25
AN ESSA Y ON MIND
cliche, but there are also certain more pressing considerations which favour the
placements in figure 1. In particular, the most common reference to pleasure is
within what are clearly attitude-type contexts. Thus, one speaks of 'being
pleased that' or 'taking pleasure in'. Pleasure here comes closer to an emotion
than to any sort of bodily sensation: when, for example, it is appropriate to say
that you are pleased to have gone to the dentist, it is unlikely that such pleas-
ure would show itself in any particular part of your body, though of course
any pain that the dentist caused certainly does show up in this way. This is
not to deny that 'pleasure' can describe bodily sensations, nor that 'pain' can
be used to describe a specially intense kind of sorrowful emotion. But the more
typical uses of both justifies my having placed them as in figure 1.
(3) Why is consciousness not shown in exactly the same place as experience?
Admittedly, I have at times used the two expressions interchangeably, but
there is a reason - so far unremarked - that is responsible for this placement.
One can use the word 'conscious' and its related forms of speech in two ways:
either as a synonym for 'experience' or as qualifying such things as belief, deci-
sion and action. In this second sense, one says such things as: 'he consciously
decided to .. .', or, 'she consciously believed that .. .', or, 'he conSciously
inferred that .. .'. Here the contrast is with cases in which decisions, beliefs
and actions are somehow not directly available to the subject. Thus a con-
scious decision is one that has been reflected upon, taken after due deliberation,
and to which the subject has the kind of access required for reporting the deci-
sion to others. It is not necessary for a decision to be conscious that it be
experienced in the way that a pain is. In any case, it is not all that clear what
such an experience would be like.
Sometimes this distinction is described as that between 'access' consciousness
and 'PHENOMENAL' consciousness, and, without any particular commitment to
this way of putting it, I wanted figure 1 to reflect something of the dual nature
of the word. So, as would be expected of a notion that can figure in attitudes
and actions, 'consciousness' is shown as some small distance towards each of
them. (See CONSCIOUSNESS.)
(4) Reaching - stretching out one's hand and arm - is about as central a case
of bodily action as one could have, and its location on the map reflects this. On
the other hand, inferring - as in 'noticing that the shutters were closed, he
inferred that they were not home' - is a clear example of an act that does not
involve the body. For this reason it is placed further from action and closer to
experience.
Intending, willing, choosing and deciding are intimately connected with
actions of all sorts, and, according to some accounts, they are themselves forms
of mental act. I have included these, and the more typical cases of action,
within a region labelled 'agency' because, though I haven't discussed it at any
length, these are the notions that together give us our idea of an agent - what
we might call a 'self in action'.
26
(5) Speaking of which. notice that I have not been able to find a place for the
self. But then again. it would have seemed odd if I had. For whatever it is. the
self is not something you would expect to find along with the other items in
figure 1. It is not as if you could say: let's put the self just here next to belief.
or consciousness. Yet. as I suggested earlier. the self seems to accompany each
and every item in the mental landscape. So. perhaps the best I can do here is
to suggest that a full-scale map of mental features in the style of figure 1 - one
in which everything is shown in great detail - would count. not as a mental
map in general. but as of some one person in particular. The SELF thus would
belong to the legend of each map - the bit that tells one whose specific terri-
tory the map depicts.
(6) Finally. as has been noted. we are not always the best judge of our atti-
tudes. nor of our actions and decisions; our states of mind can be hidden from
us. Sometimes this happens because the states of mind in question are as a
matter of fact inaccessible to us and sometimes because we have in some sense
made them so. Examples of the first sort usually involve a sort of knowledge
that we have and use. but do not. and largely cannot. remark upon. For
example. when we hear the sounds of our language. we are able to interpret
them - indeed it is impossible not to - because of the vast number of things we
know about the sounds. grammar and meanings of that language. Yet many
skilled listeners (most of us in fact) are unable to describe these crucial bits of
knowledge. They guide us though they remain in some way tacit. To mark this
kind of circumstance. I put knowledge of language near the attitudes but on
the other side from experience. (The whole idea of tacit knowledge. especially
in connection with language. has been heatedly discussed by linguists and
philosophers. (See CHOMSKY).)
The second way we can lose track of our own states of mind revolves around
the notion of the UNCONSCIOUS. as this notion is used in psychoanalytic theory.
The idea here is that some of our attitudes. decisions and actions are under-
taken for reasons that we somehow manage to conceal from ourselves. Why
we do this - and how - are questions that form the subject matter of PSYCHO-
ANALYTIC EXPLANATIONS. Unconsciousness is a bit difficult to draw on the map.
In one sense it is everywhere. since virtually any mental phenomena might be
unconscious in the psychoanalytic sense. But if we think of it as a receptacle of
such states - as the unconscious - then it would be better to include it in the
legend along with the self. In any case. given the controversy which surrounds
this notion. any placement should be thought of as provisional.
For more on specific items see BELIEF; DESIRE; IMAGINATION; INTENTION;
MEMORY; PERCEPTION; THE WILL.
27
Stage 2: Digging Deeper
Natural historians of the seventeenth century were dismayed that the Earth
had so many mountains and hills. They regarded these features as irregular,
and hence as disfigurements of the landscape, some going so far as to think of
mountains as God's particular way of telling us of our sinfulness. Leaving aside
any aesthetic or theological speculation, it must be admitted that contours of
the mental realm are far from regular - its landscape is just as lumpy, meta-
phorically speaking, as that of the Earth. The idea that the things of the mind
are sufficiently uniform for us to separate them off from the rest of reality by
some simple criterion is just not on. This is bound to be a disappointment to
those who may have thought that the mind was a unified realm whose core is
revealed in the so-called 'stream of consciousness'. Whether fortunately or not,
this expectation cannot withstand the sheer force of the number and character
of the items that queue up for inclusion in the mental order. If we take this
variety seriously - and that is precisely what I have tried to do - then we must
be satisfied with a more piecemeal approach to our deeper investigations in this
stage.
That being said, however, I do not want to rush around the mental land-
scape taking soundings at every point. There are many interesting questions
that could be asked about each and everyone of the items in figure 1. In most
cases our dicussion of them was cursory, and even basic things such as their
location on the map could be profitably discussed further. But even though the
mental order is lumpier than some might have hoped, I should like here to
stick to the big picture as much as possible.
Throughout our investigations, there have been three constant landmarks
towering above the landscape: Experience, Attitude and Action. These cat-
egories are the basis for the organization of the mental realm, and they are in
large part responsible for its lumpiness. So, what I propose is that we confine
our attention to these categories; that we set up our seismic apparatus so as to
see what lies under them. Of course, in doing so we will have to make constant
reference to those mental items that are typical of the categories - items such
as pains, beliefs and specific types of action. But in every case our inquiry will
be general. For example, we shall be interested in belief to the extent that it is
representative of the category of attitudinizing; the more specific characteristics
of belief - what distinguishes it from other attitudes - will not figure promin-
ently. This way of proceeding should not lead us into error in regard to the
attitudes, as long as we bear in mind that it is only a first step. The hope is
28
STAGE 2: DIGGING DEEPER
that if we understand things that are typical of all the attitudes, we will be in a
better position to understand their idiosyncrasies.
In every case, the procedure will be quite simple: certain questions - which
are in no sense technical or specialist - will be asked about the assumptions
that underpin each of the categories. What I shall try to show is just how diffi-
cult it is to answer them - how the most straightforward things we think
about the mind just don't help. Or, worse still, how what we think actually
makes answers more, rather than less, difficult. In short, my aim will be to
reveal just how much trouble we can get into just by investigating the features
of the mind's landscape.
2.1 Attitudes and Attitudinizing
2.1.1 Attitude Problems
I begin in the middle with the second of the categories - attitudinizing. This is
not because of the well-known literary advice about where to begin a story.
Rather, it is because attitudinizing looks both ways: many think that it has
roots in experience - both of the bodily and perceptual variety - and it cer-
tainly seems to look towards actions in shaping and anticipating them. Besides,
the attitudes have so far raised very few awkward questions, and it is now time
to be awkward.
What is most striking about attitudinizing is the feature we have called
directionality, and this will be the focus of our discussion. The idea that beliefs
are directed, that they are reported by sentences that themselves contain sen-
tence complements, that beliefs therefore have contents, these things are so
familiar to us that it is easy to overlook just how special they are. When we
say that Anne believes she left her coat in the hallway cupboard, we are, one
would suppose, saying something true about Anne. Looked at from this per-
spective, there is a lot in common between our reporting that Anne has a belief
and our reporting - truly - that she has a cold. In each case, there would seem
to be state that Anne is in, albeit it is a mental state in the case of belief and a
physical state in the case of the cold. However, from this perspective, we miss
the crucial feature that differentiates the one state from the other.
A physical state of a person can be quite complex - it can take many words
to describe it accurately. But no matter how complex it is, it never has the
feature of directedness that goes with attitudinizing. Anne has a cold - certain
unpleasant biological and chemical things are going on in her - but this state
is not directed to anything, it is not about anything. Her physical state is
somehow complete in itself and can be described without including reference to
any other state of affairs. In contrast, Anne's belief is directed to another state
of the world - her coat's being in the hallway cupboard. It is this latter state
that forms the content of her attitude, and there is just no way of describing
Anne's belief without referring to this content. (See PROPOSITIONAL A TTI-
TUDES.)
29
AN ESSAY ON MIND
In the literature the directedness characteristic of the attitudes, and the sen-
tences we use to report them, is often called 'INTENTIONALITY'. However, in
some of its forms ('intentional') this word can be misleading, so I prefer to use
here the less common but more descriptive word 'directionality'. (Actually,
there is a close connection between the attitude verb 'intend' and the idea of
intentionality. But, even so, it can be confusing to speak of some belief being
intentional when we usually reserve such a description for actions.) Also, 'the
problems of intentionality' tends to be a label for difficulties that go well beyond
any we shall discuss at stage 2, so it is as well to use the more restrictive name
for the time being. (I shall later say something about the problems of intention-
ality in the wider sense.)
Whatever we call it, this remarkable feature of the attitudes is so familiar it
is all too easy to be blase about it, and to take for granted our command of the
apparatus with which we report them. Anne's coat is there hanging in the
hallway cupboard. This is as far from being a mental fact as anything can be;
it is merely a state of the world at a given time. But Anne is in, or has, a
special mental state - she has a belief - and the very description of this second
state requires us to make reference to the first. What allows us to connect up
these two claims in this way? Even if we are usually unreflective about it, what
guides our ways of reporting beliefs and the other attitudes?
The first thought that someone might have here is tantalizingly simple.
There is the coat hanging in the cupboard, and Anne is believingly related to
this fact. That is, just as she might be owningly related to the house on 23 Elm
Street, so is she believingly related to the coat's being in the cupboard. In the
one case, she bears a complex physical. social and legal relation to a house, in
the other a perhaps no less complex mental relation to something that is going
on in the world. (See REPRESENTATION.)
Unfortunately, this simple account is beset with problems. The most obvious
fault is that it signally fails to make room for false belief. Anne may truly
believe her coat is in the cupboard but it may well not be - her belief may be
false. In this perfectly ordinary case, it cannot be the relation between Anne
and the coat's being in the cupboard that grounds our attribution of the belief
to Anne, for there is no coat there and, hence, no such relation.
This is even more obvious in the case of desires. Suppose that you have a
long train journey ahead of you and you absolutely must have a newspaper to
make it pass more quickly. You see a shop and desire very much to acquire
from it a copy of your usual daily. In the somewhat awkward idiom discussed
earlier, we can say:
You desire that you come to own one of the copies of the Guardian in the
shop at the station.
But, sadly, there has been a run on Guardians, and there simply aren't any in
that shop. So, you truly have the desire, but the desire can hardly be a relation
between you and a newspaper in the shop which is (as you think) waiting
there. There simply isn't any newspaper there answering to your desire.
30
In the face of these difficulties, it might be tempting to try to keep the spirit
of the simple proposal by retreating a little. (Indeed, retreating is one of the
main strategies for dealing with the problems of directionality.) We are pre-
vented from seeing Anne as believingly related to the coat's being in the cup-
board because it just isn't there. But why not still insist that she is believingly
related to the coat (wherever it is), the cupboard and the relationship described
by the words 'being in'. That is, why not invent a complex entity made up of
these three items, which certainly exist in the world even if they do not exist
together in the way envisaged by Anne? Thus, we can say what her belief state
is directed at, without thereby guaranteeing that it is true. (Rather as we can
say what an archer's target is without committing ourselves to whether the
arrows hit the mark.)
The entity made up of the coat, the cupboard and the relationship is some-
what odd - it is a sort of abstract thing made up from concrete objects and
relations - but that in itself is no special problem.
1
After all, one wouldn't
expect mental attitudes to be completely straightforward. But oddness is only
the beginning of the difficulties with this proposal. Consider for example this
belief attribution:
Henrietta believes that Macbeth was not misled by a greedy wife.
1 This entity is sometimes called a 'PROPOSITION', but - confusingly - it is not the only
sort of item that claims this title. Some think that we can get away with simply saying
that a proposition is whatever it is that serves as the content of a propositional attitude.
The minimalism of this view has a lot to recommend it, but there are inevitably going
to be questions about what kind of thing can serve that role. And the idea that a pro-
position is a sort of abstract confection with real objects sort of baked into it is one way
of dealing with these questions. Though it does raise many more questions that are too
technical to discuss here.
Another way in which to be less than minimalist about propositions is based on the
idea of a possible world, and works as follows. Imagine all the different ways in which
our world could be different - however slightly. Then call each of these ways a 'possible
world'. Finally, say that a proposition is the group or set of possible worlds in which the
relevant content sentence is true. Thus, there are many possible worlds in which Anne
has a coat and it is in the cupboard, and many in which either the coat doesn't exist or
it is not in the cupboard. The sentence: 'Anne's coat is in the cupboard' is of course
true only in the first set of possible worlds, and we can define the proposition that the
coat is there as that set. Anne can then be said to be related to this set of possible
worlds so defined, though her belief will only be true if our own actual world happens
to be in that set.
The possible-worlds treatments of propositions also raises many questions that won't
be discussed here. However, I do not mean to suggest an aversion to this treatment just
because the text employs the other. It is just that the confectionary sense of a proposi-
tion - sometimes called 'structured propositions' - lends itself to my particular account
of attitude problems.
31
AN ESSA Y ON MIND
Are we saying that Henrietta is believingly related to Macbeth, his wife and an
appropriate 'misleadingness' relationship? It certainly has seemed to many
people as if one of the fascinating things about attitudes is that we can have
them toward fictional characters and other non-existents (remember the desire
for the newspaper). But doesn't this put a great deal of strain on the idea that
belief consists in a relation to a complex of existing items?
Actually, whilst the answer to my last question is clearly 'yes', I don't think
that worries about fictional entities constitute the main stumbling block to
accepting the idea that attitudes are relations to things. Fiction creates havoc
wherever it goes - whether in our accounts of language or in our accounts of
attitude ascriptions. And it creates havoc for virtually every account in these
areas. So, whilst it is worth mentioning, I shall put it on one side. There is a
larger obstacle standing in the way of our present proposal regarding the world-
directness of the attitudes, and we must deal with it before we can consider
such relatively minor matters as so-called 'fictional' objects. (See STALNAKER.)
2.1.2 Deeper into Trouble in London and Paris
So far we have been trying to hang on to the idea that beliefs and other atti-
tudes are directed at the things they most certainly seem to be about - the
items referred to in the complement sentences of attitude ascriptions. However,
we have seen that Anne cannot simply be related believingly to her coat's
being in the cupboard because it might not be there. So we have retreated to
the idea that she is believingly related to a made up, complex entity consisting
of the coat, the cupboard and the feature of one thing's being contained in
another. This seems to make room for falsity, since these items might not actu-
ally be in the relationship to each other that Anne thinks they are. But falsity
(and fiction) are not the only problems. For the relation between the believer
and the believed is much more elusive than we have so far noticed. For there is
a kind of slack between believers and what beliefs are about which can make
one despair of ever coming to understand the directionality of the attitudes.
The following two examples will illustrate what I mean.
2.1.2.1 London Monique, on her first visit to London, takes a lightning tour
from the top of a double-decker bus. Among the things pointed out to her is
the British Museum. As a result of what she is told, and what she can see for
herself, she comes to believe that the British Museum has two lions guarding
its entrance.
On a later visit to London - this time in order to attend a literary party in a
publishing house in Bloomsbury - she notices, as she arrives, that there is a
large, sombre building with imposing columns opposite the offices she is to
visit. She wonders whether this could be the British Museum (she knows it is
in Bloomsbury), but, looking about in vain for the lions (there are none at the
front entrance), she dismisses this possibility. But of course what she is looking
at is simply the British Museum from another angle.
If we construe the belief formed on the sightseeing bus according to the pro-
32
posal described earlier. we would say that Monique is believingly related to the
'triplex' entity consisting of:
the British Museum,
the two lions,
the relationship of guarding.
On the second visit, one would seem forced to say that she is believingly related
to:
the British Museum (she is looking at it),
the two lions (she clearly remembers these),
the absence of the relationship of guarding (she can see that there are
none flanking the entrance).
The first belief was attributed using the form of words:
Monique believes that there are two lions guarding the entrance of the
British Museum,
and the second:
Monique believes that the building over there does not have two lions
guarding the entrance.
However, in the example 'the building over there' in the second belief attribu-
tion is in fact none other than the British Museum. and this is what leads to
trouble. What trouble? After all, the first part of the story describes in a coher-
ent way the kind of mistake that we can all make, so it may seem that we
have not uncovered any particularly shattering gap in our everyday under-
standing of the attitudes. But this is to forget how we got to the present posi-
tion. Our question was: what counts as the content of a belief? The eventual
answer was: it consists in an amalgam of those items to which the belief is
directed: just describe that amalgam and you have the content of a belief attri-
bution. What the above example apparently shows is that this cannot be right.
In Monique's case, the key items that form the content of her two beliefs are
the same - it is the British Museum and the lions that figure in both. There is
as one might say just 'one reality' toward which Monique's beliefs are directed.
But Monique has two beliefs about this reality which are simply not compa-
tible. And it would be unfair to Monique to say that she believed both that the
British Museum was guarded by lions and that it wasn't. Monique. we can
assume, is much too rational for such an attribution to make any sense.
Anyway, she would have to be quite mad to believe such an obvious contra-
diction. So, does she believe that the British Museum is guarded by lions? And
if we say 'yes' (or 'no') to this, where does that leave our account of the direc-
tionality of the attitudes?
33
AN ESSAY ON MIND
This first example seems to show that one can adopt two different (even
incompatible) attitudes toward one and the same set of things, and this intro-
duces a puzzling element of slack into the relationship between our attitudes
and what they are about. It is as if our beliefs can slide about whilst still being
apparently directed at the same things. (See INTENTIONALITY.) Before trying to
alleviate this situation, let us consider the second example.
2.1.2.2 Paris Richard hasn't been to Paris for years. When he was last there,
he stumbled on a small brasserie near his hotel which he used to go to every
day. In fact, it was in this brasserie that he first met the woman he would now
describe as the love of his life. He was sitting at a table near the window, and
all the other tables were fully occupied when she came in ...
Now, all these years later, Richard is in Paris on business and he decides to
see whether the brasserie is still there. Heading to the quartier of his former
hotel. he has not reckoned on one thing: the tendency, in our times, for suc-
cessful businesses to try repeating their success by forming chains of identical
establishments. So, when he stumbles onto the second of the brasseries, he
takes it for the one that was the object of his search. Indeed, so close is the
resemblance between the two, and Richard's memory is very precise, that he
finds the tables and chairs, the metal-covered comptoir, the curtains, the menus
- everything seems to be just as it was before, give or take a new coat of
paint.
On and before entering the brasserie, Richard had many beliefs and other
attitudes which together contributed to what we would describe as his feeling
of nostalgia. But for the purposes of the example, let us focus on some specific
belief. Richard had believed that the chairs in that local brasserie were very
comfortable. This belief has not changed in the intervening years, and, as he
enters, one of the things he thinks is this: I hope they haven't changed the
chairs. He is thus relieved to see the same (sort of) chairs set neatly around the
tables. He now could correctly be described as believing that the chairs in the
brasserie he used to go to every day have not changed, they are still comfor-
table. In short, he continues to believe what he had believed before.
But is what he believes true? This is not an easy question to answer. For the
sake of definiteness, let us suppose that the brasserie he actually used to go to -
the one he mistakenly thinks he is now in - has changed its chairs, and that
they are no longer what Richard would describe as 'comfortable' if he were to
come across them. In this case, there is some pressure for thinking that his
present belief is false: the chairs in the brasserie to which he used to go are not
comfortable. After all, Richard would be quite insistent that, whatever else was
true, he had not changed his belief about the chairs, and we know - what
Richard has yet to find out - that the chairs have changed. Yet, sitting there in
the new brasserie entertaining the thought: these chairs are comfortable, one
may be tempted to think that his belief is true. But it is certainly bizarre to
describe one and the same belief as both true and false.
On the account of the directionality of beliefs within which we are working,
34
Richard's long-standing belief about the chairs was directed at an amalgam of
these things:
the chairs in the original brasserie,
the property of being comfortable.
On the apparently reasonable assumption that Richard's belief hasn't changed,
this belief continues to be directed at these things, only it is now false.
However. it is difficult to ignore the fact that he is sitting there in a brasserie
looking at some chairs and judging them to be comfortable. In this case, his
belief seems to be directed at:
the chairs in the new brasserie,
the property of being comfortable,
and it is thus true. So, either he is thinking the same thing and, hence, think-
ing something false, or he has changed his mind (without realizing it!) and he
is thinking something different and true. Neither seems very palatable, but the
second alternative seems particularly bizarre.
2.1.2.3 The Moral The example involving Monique showed that there is one
kind of looseness in the directedness of our beliefs: we seem to be able to have
quite different beliefs - even incompatible ones - about one and the same reality. The
case of Richard shows that this looseness can occur the other way around: we
apparently can have one and the same belief about two different realities. Put to-
gether, these two examples put intolerable pressure on the idea that Monique
or Richard can simply be described as being believingly related to features of
the world. There are standards that must be met by anything we could regard
as a genuine relationship, and these examples apparently show that the atti-
tudes do not come up to scratch. (See BELIEF; THOUGHTS.)
Consider again the ownership relationship between Anne and her house. As
was said. this is a complicated legal and social relationship - there may even
be problems in telling whether it holds definitively. But this much is true: if
Anne really does own a house, then she bears this relationship to it no matter
how that house is described. And she bears this relationship to just that house
and not to others merely because they resemble it.
Admittedly, the cases of Monique and Richard are those involving confusion
and mistake, and you might think that they are therefore of less importance
than I have made them out to be. But mistakes and confusion are just the
things on which we must focus if we want to understand the attitudes. An
attitude is a state of mind directed towards something or other, and it is there-
fore liable to go wrong. Indeed, where but in states of mind would you expect
to find confusions and mistakes? But it is precisely the possibility of these con-
fusions and mistakes that seems to undermine an account of the content of the
attitudes that makes them relations to extra-mental reality.
35
AN ESSAY ON MIND
At this point there are two ways to jump: either agree with the conclusion of
the last paragraph when asked what we should make of the directionality of
the attitudes. Or look for ways to get around the kinds of examples that created
the problem in the first place. As one might guess the philosophical community
has tended to go for the second option.
2.1.3 The Way Forward
Actually, though counting here is not a precise art, one can find at least four
ways forward: three of them could be understood as ways to improve on the
simple proposal of directionality that has led us into trouble, and the fourth as
rather more thoroughgoing revision of that proposal. But one must be careful.
As you will discern for yourself, these ways are not completely independent of
one another. Each consists in a suggestion of how we might begin to cope with
the puzzles about the directionality of belief described above, but there is
nothing to prevent - and everything to encourage - taking to heart more than
one suggestion at a time. For obviously if some combination of these strategies
can convincingly describe the directionality of the atittudes whilst defusing the
puzzling cases, then there is good reason to plump for it. I have chosen to
describe them seriatum, since I want to highlight the essentials and not
because I think that they are exclusive alternatives.
2.1.3.1 Language Nothing could be more obvious than that attributing atti-
tudes to someone involves careful choosing of one's words. To take an extreme
example: a five-year-old looking westwards on a late March afternoon may
well believe that the sun is setting. But it would be bizarre to say of him that
he believes that the medium-size, fusion-powered star now visible on the
western horizon is passing out of line of sight of the inhabitants of the British
Isles. One feels he just doesn't believe that. And the reason? Clearly, it is partly
a matter of language. The 'sophisticated' way of describing the setting sun
uses words that stand for concepts unavailable to a five-year-old. Had we said
simply: 'he believes that the sun is setting', this would have passed as reason-
able.
Taking a lead from intuitions such as this, there has been an enormous effort
to say just what about the choice of words (and/or concepts) governs our atti-
tude attributions. The hope is that if we can get this right, then we can deal
with the problems raised by examples such as those involving Monique and
Richard. If there is something about the use of the phrases 'the British
Museum' or 'the building over there with columns' that makes them function
differently in sentences attributing beliefs, then perhaps we can avoid having to
say that Anne is believingly related in different ways to the same reality.
Maybe this will allow us to regard the apparent slack between Anne's state of
mind and reality as due to the language we use to describe the beliefs and not
to the beliefs themselves. (See THOUGHT AND LANGUAGE.)
The details of linguistic attempts to deal with the propositional attitudes are
too complex to be described fully in this introduction. Indeed, there is a sense
36
in which this approach takes the problem of the attitudes outside the scope of
the philosophy of mind itself, and not merely beyond the reach of this survey.
For the question of how it is that words connect their users with reality - and
within what limits - falls squarely within the philosophy of language. More-
over, even a cursory look at the literature here will show just how complicated
the story can get. It is enough to make one wonder how we ever do manage to
say what people believe. However, for the sake of definiteness we ought to have
at least a rough description of how things might go.
In their ordinary employment, phrases like 'the British Museum' and 'the
brasserie I used to visit' refer to items in the world. That is, at least part of their
linguistic function is, broadly, to bring reference to such items into our con-
versational exchanges. Yet, in sentences reporting what people believe, we get
into trouble if we take these words simply to have that sort of function. Given
this, one way to get out of trouble would be to say that, when these sorts of
phrases (and indeed, words generally) occur in the context of propositional
attitude reports, they change their function somewhat. Perhaps they cease to
refer directly to items in the world and refer instead to how these items are
thought of by the believer. Here is the proposal put more concretely.
Monique believes that the British Museum is guarded by lions. Think of the
words 'Monique believes that' as having a strange effect on whatever it is that
follows them. Thus, the underlined space in 'Monique believes that _' is a
linguistic context in which words do not function as they would outside such a
context. To keep matters simple, let us just consider the phrase 'the British
Musuem'. This is a name of that famous building, but in the above-mentioned
context, this name does not simply refer to that building. Instead, it refers to
Monique's way of thinking about that building - we could say that it refers to
the way in which that building is presented to her. Thus, we should not think
of the building itself as taking its place in the content of her belief. Instead, it is
the 'mode of presentation' which is stuck into the propositional confection.
Calling it 'the British Museum mode' (for want of a better way to describe it)
here is the content of Monique's belief:
the British Museum mode of thinking of the British Museum,
the two lions,
the relationship of guarding.
Now this will differ significantly from her second belief - the one we reported
by the sentence, 'Monique believes the building over there is not guarded by
lions' because the first element in the confection for this one will be:
the building-over-there mode of thinking of the British Museum.
And we can say all of this even though the building her beliefs are about - the
one presented by these two different modes - is one and the same.
Of course there are problems with this account, many of them. Just to
37
AN ESSA Y ON MIND
mention one troublesome area, there is the obvious difficulty we might have in
spelling out what a mode of presentation is, and how it can happen that words
sometimes refer to buildings and sometimes to modes of presenting them. Still,
it is tempting to think that Monique's belief relationship to something or other
is like ownership, i.e it is not the slack kind of relationship that it might appear
to be. And it would therefore be nice if we could deal with the appearance of
slack by attributing it to some more complicated way in which language func-
tions in propositional attitude reports.
How does this kind of move fare in relation to Richard's encounter with the
unsuspected change in brasserie? It must first be admitted that the kind of
linguistic move described here was worked out with Monique-type cases in
mind. But perhaps it could be adapted. For example, one might say that since
his belief involves a relationship to the mode of presentation of the chairs
rather than to the chairs themselves, his nostalgic belief is about the former
brasserie even though he is no longer in that brasserie, and whatever truth
there was in it stays the same. However, the troubles with this suggestion
come thick and fast, and they point to certain deeper problems even in respect
of Monique-type cases. For what they all stem from is this: we want beliefs to
be about what is around us and to be made true or false by that reality.
Indeed, this is surely a large part of what belief-talk is for. Yet, if we slip too
deeply into the mode-of-presentation way of describing beliefs, we run the risk
of making them completely unresponsive to reality. For example, by allowing
Richard's belief to be tied to the original brasserie and to keep its truth value
we risk cutting him off from his new surroundings. Admittedly, one nice
feature about the present proposal is that it can explain why Richard would
say such things as, 'Nothing much has changed in the old place.' This is
because, so far as the content of his beliefs is concerned, nothing much has
changed. And as I have emphasized all along, the attitudes are our main tool
for explaining what people do and say. Still, we must be careful to preserve
both our intuitions about the attitudes: they explain what we do and they do
this by showing how the world seems to us to be. In other contexts, losing
the world may be a good thing to do, but it makes little sense in the present
one.
2.1.3.2 Styles of Believing An obvious thought to have, given the problems
there are in coping with the two aspects of belief, is that there may well be two
kinds of belief (and, as needed, two kinds of the other attitudes as well). One
kind has Monique believingly related to the relevant items in the world. With
respect to this kind, and appearances notwithstanding, Monique's beliefs just
are about the British Museum; she just does believe of that famous landmark
that it is both guarded by, and not guarded by, lions. In the trade, this kind of
belief is called 'de re' thereby indicating the connection between the believer
and the relevant 'things' (Latin: 'res') in the world.
The other kind, called 'de dicto', is not only sensitive to the choice of words
used in attributions, it goes so far as to have believers related not to the things
38
of the world, but to the linguistic items (Latin: 'dicta') that figure in the attribu-
tions. With respect to this style of belief, Monique does believe that the British
Museum is guarded by lions, but does not believe that the building over there
is guarded by lions even though the building over there is none other than the
British Museum.
It is sometimes difficult to get the hang of this, since it sounds so much like
the way of putting matters that got us into trouble in the first place. What one
has to do is to recognize that the de dicto style of belief has Monique believing a
linguistic item: the British Museum is guarded by lions. And it not implausible to
regard this linguistic item as quite different from the following one: the building
over there is not guarded by lions.
The basic idea behind this strategy (though this is very rarely made explicit)
is that by discerning two kinds of belief we can cope with the apparently puz-
zling cases in a divide-and-conquer way. The de re attribution shows Monique
to have a problem: she has beliefs that make the British Museum a strange
building indeed - one that both has and does not have lions guarding its
entrance.
By itself this shouldn't be all that surprising. After all, the story I told is
essentially one in which Monique is confused. So why shouldn't her confusion
take this form? Still, as noted earlier, if this were all we could say about
Monique's beliefs, things would not be very satisfactory. She is confused all
right, but not as totally mad as this attribution makes her sound. And this is
where the availability of the de dicto style comes in handy. It gives us the
chance to add - almost as a qualification - that Monique is not as crazy as all
that since she de dicto believes that the British Museum entrance is guarded by
lions and that the building over there is not so guarded.
2
These latter claims
might well make us less squeamish about the de re attribution because the two
styles, taken together, seem to give us a reasonable handle on the nature and
source of Monique's confusion.
2 Philosophers who go in for the de re/de dicta distinction tend to try to convince us that
there is an everyday linguistic device that does the work for us. Thus, the de re style is
said to be captured by:
Monique believes of the British Museum that its entrance is guarded by lions,
and the de dicta by:
Monique believes that the British Museum entrance is guarded by lions.
However, most of those I have interrogated (before they were trained by philosophers)
tend to use the 'believes of' and 'believes that' constructions interchangeably. Of course,
the distinction made by philosophers is no less (or more) clear in virtue of the presence
or absence of an ordinary language usage consistent with the distinction. It just would
have been nice. for those who believe in it, if it were enshrined in ordinary language.
39
AN ESSA Y ON MIND
As will be obvious, the detailed working out of the de dicto style overlaps
with the linguistic attempt to defuse the puzzles. For merely citing the de dicto
style is not enough. We need to have an account of the dicta - an account that
makes it plausible that Monique is related in various ways to the relevant bits
of language and the world - and this is central to the linguistic approach.
Leaving aside the Byzantine details that have grown up around attempts to
deal with the attitudes using linguistic resources, there are a number of more
straightforward problems with this second, divide-and-conquer approach. Most
of them begin with the obvious and simple question: what is the relationship
between de re and de dicto belief? Note that this is not just an idle question - one
that we would like to have answered but which can wait. Our original intuition
about the directionality of the attitudes was that it was some kind of relation-
ship between the attitude-taker and items in the world. We have been trying to
see (in outline) how to spell this out whilst remaining faithful to our intuitions
about how we actually use attitude-attributing sentences in particular cases. On
the present approach, it would seem that all the world-directedness is handled
by the de re style; the de dicto form helps us cope with those things we might say
about believers when they are confused or ill-informed, though it does so at the
cost of not being a relationship between believers and items in the world. But so
far from helping, this bifurcation of tasks risks losing everything. For unless
there is an intelligible relationship between the de re and the de dicto, each will
end up a failure at giving an account of directionality - one because it ignores
the attitude-taker and the other because it loses the world. We will have two
wrong approaches instead of one.
Additionally, there is a real question about whether, on the present proposal,
what we have are two different attitudes or merely two styles of attribution. If
there is only one kind of belief but it can be attributed in two different ways,
then our interest should be in the nature of this belief and not so much in the
styles of attribution. And if the proposal is that there are really two kinds of
belief (and two kinds, therefore, of each of the attitudes) where is the evidence
for this?
The history of attempts to cope with the de re/de dicto distinction would
require a book in itself. Indeed, the attempt to relate the two generated a thriv-
ing industry in philosophy - an industry whose output overlapped with the lin-
guistic approach to the attitudes. Nor has production completely ceased, even
though the main markets have moved elsewhere.
2.1.3.3 Styles of Content The de re/de dicto distinction (and the use of philo-
sophy of language to get clear about it) was historically connected with
Monique's kind of problem. But what about Richard's predicament? Does the de
re/de dicto distinction, problematic though it is, give us something to say in that
case?
The origin of Richard's kind of trouble is a certain thought experiment sug-
gested originally by Hilary Putnam, though it has gone through many vari-
ations since. In that experiment, you are asked to suppose that (a) there is a
40
glass of water in front of you and that (b) you believe there to be a glass of
water in front of you. Next you are invited to imagine that there is this other
place called 'Twin Earth'. The name is appropriate because everything on Twin
Earth is a molecule-by-molecule duplicate of things on our planet. In parti-
cular, there is a molecular duplicate of you, that duplicate is sitting in front of
a glass and has a belief about its contents. Crucially, though, there is this one
difference between Earth and Twin Earth: on Twin Earth the stuff in rivers,
lakes and in the glass is not H
2
0 but something chemically called 'XYZ'. For
reasons which are not really important to the present debate, we can take it
that this means that there is no water on Twin Earth, though there is some-
thing very much like it which we could call 'twater' (twin water). Now,
suppose that you are thirsty. Since you believe that there is water in the glass
in front of you, you would probably reach out for the glass and drink it. Simi-
larly, your Twin Earth counterpart would do the same. Why? Well, if you and
the twin were molecularly identical, one would expect that your brains and
nervous systems would support the same beliefs and other psychological states.
(There will be further discussion of brains and beliefs in stage 3.) But your twin
couldn't be described as having the belief that there is water in the glass in
front of him, because, as we have assumed, there is no water on Twin Earth.
There is only twater. Moreover, assuming that you don't know how to tell the
difference (by looking) between water and twater, what do you suppose you
would do if you (and your thirst) were miraculously transported to Twin Earth?
You would reach for the glass.
What these things are held to show is that you and your twin's beliefs can
be, in some respect, the same, even though the world they seem to be about is
different. Or, in the case of transportation, your belief remains the same, but
what it is about has changed. (The case of transportation is closest to the case
of Richard in Paris.) One consequence that has been drawn from Twin Earth
cases, is that there may well be two kinds of content. One kind is that you and
your Twin Earth counterpart can share, even though there are differences in
your respective worlds. This is called 'narrow' content. The other is that kind of
content that shows your belief, which is after all about water, to be different
from your twin's. This content is called 'wide' (sometimes 'broad') content.
Applying these two kinds of content to Richard's case, we can say that when he
walked into the new brasserie, he had a set of beliefs with narrow contents
responsible for the things he did and said. In being narrow, these beliefs were
not sensitive to the fact that Richard was not in the original brasserie. However,
from another point of view, we could view his beliefs as having wide content -
as being directed to real-world items and, hence, as sensitive to any swapping of
such items. The narrow content of Richard's beliefs was not falsified by the sur-
prising duplication of brasseries, but the wide content was. Insofar as Richard
widely thought that the chairs in the old brasserie were still comfortable, his
belief was false. (See EXTERNALISM/INTERNALISM/TWIN EARTH.)
Clearly, this proposal echoes some of those previously considered, especially
the de re/de dicta distinction. But one should not assume that, for example,
41
AN ESSA Y ON MIND
narrow content and de dicto content are the same. The problems each were
designed to deal with are quite different: narrow content is intended to help
with Twin Earth kinds of duplication of reality; de dicto content is a way of
dealing with the possibility of multiple beliefs about a single reality. One could
look at it this way: narrow content comes from a kind of subtraction - take
away from belief contents whatever it is that makes us think they connect
directly to particular things. In contrast, de dicto content comes from a kind of
addition - add to belief contents those elements that make them take on the
way some particular person thinks of things. Of course, this subtracting and
adding might get you to the same place - to a content which was expressible
by dicta - but there is no guarantee of this.
Aside from the problems of understanding how the narrow IWide content dis-
tinction is related to the others, there is the even more vexed question of how
these styles of content are related to each other and to our ordinary ways of
speaking. For beliefs do seem to be about the world; hanging on to that idea in
the face of certain problems is what we have been trying to do. Of course, it
would be nice if we could simply say that narrow content, entertained in a
particular environment, fixed wide content. That is, that what goes on in the
relatively narrow confines of our minds serves to attach us to the wider world.
But duplication cases like Richard's show that this just won't work; if it did,
Richard wouldn't have been in his predicament in the first place. As with de rei
de dicto styles of believing, there seems to be a danger that having two styles of
content simply pushes our original problem under another part of the carpet.
(See LANGUAGE OF THOUGHT; THOUGHTS.)
2.1.3.4 Troilism The approaches considered so far have closely followed our
talk about the attitudes in regarding them as essentially relations between atti-
tude takers and some second item. We say:
Anne believes that her coat is in the hallway cupboard,
and this has set philosophers on the trail of that second item - the thing to
which Anne is attitudinally related. Mesmerized by the language of attitude
reports, they have assumed that whatever else they are, attitudes show them-
selves to be two-place relations. But some things work better in threes. Perhaps
we could regard Anne's belief as really a three-place relation: one which holds
between Anne, the set of items consisting of the coat, the cupboard and the
relationship of being in, and a third element such as a sentence. In the
problematic case of Monique in London, the proposal would come out roughly
as follows:
Monique is believingly related in the lions-guard-the-British-Museum sort of
way to the British Museum, the lions and the guarding relation.
But this is a different relationship from:
42
being believingly related in the building-over-there-is-not-guarded-by lions
sort of way to the British Museum, the lions and the guarding relation.
This is a difficult proposal to come to grips with on first hearing, but the
central idea is simple enough. By adding some third element to the belief rela-
tion, we give ourselves extra room to manreuvre - something we can use, in
the appropriate circumstances, to get ourselves out of trouble. For example, we
can use the above three-place belief relation to say that Monique's two beliefs
really relate her to the same building, lions, etc., but that each of them does
this in a slightly different way.
Of course, there is an obvious problem here: where does this third item come
from? After all, the surface structure of attitude sentences reveals them to have
just one thing in the 'that p' place, whereas the present proposal requires two.
One serves to describe the reality to which Monique is fixed, and a second spe-
cifies the way in which she is fixed to it. To be sure, sometimes it may be pos-
sible to accomplish both of these jobs with a single sentence: just pick a content
sentence that both shows what the world must be like for the belief to be true
and displays the way in which the believer actually does her thinking about
that world. But there are many times when we attribute beliefs without being
all that careful. Indeed, there are times when we haven't a clue what sentence
would faithfully reproduce the subject's point of view.
The idea that propositional attitude attributions make subtle use of two dif-
ferent content sentences has obvious connections with the proposals discussed
earlier. One might even think that it was an amalgam of modes of presentation
and styles of believing. Unfortunately, it is not that simple. There are currently
a large number of different, incompatible and intricate suggestions for how to
engage in this particular menage a trois, and any attempt to summarize them
would be unfair.
2.1.3.5 Stepping Outside I said earlier that retreat was the strategy most often
used in trying to deal with the attitude problem. First, we retreated from
regarding belief as a relation between a believer and some actual state of
affairs, then we considered retreating from the idea that there was only one
kind of belief (or style of belief attribution) and, most recently, we have seen
that we may even have to retreat from the appearance that belief has of being
a two-place relation. But I have saved the most dramatic retreat for last.
In one way or another all of the proposals so far considered begin with the
idea that an attitude is a genuine state of a person, albeit a mental and rela-
tional one. There is something about Monique that is her mental state of
believing and that state is somehow related to the British Museum, or her
representation of it, or a sentence, or ... whatever. But perhaps this is com-
pletely the wrong way to go about it. Maybe we are taking our talk about
belief and the other attitudes too literally.
Look at it this way. We began with what is apparently a description of a
person:
43
AN ESSA Y ON MIND
S believes that p,
(where S is a person and p is some belief complement). Then we searched
around for something true of S - some state internal to the workings of S's
mind - which made the above attribution correct, did so in a way which met
various desiderata thrown up by the problem cases, and, most importantly, told
us something sensible about the relation between S and the topic of the belief.
(One must not forget this last thing since what started the discussion in the
first place was the need to understand the directionality of the attitudes.) But
perhaps we should be thinking of what goes on in attitude sentences in a com-
pletely different way.
Consider this sentence:
Henry weighed 80 kilograms.
In it, one seems to be claiming that there is some kind of relationship between
Henry and a number of kilograms. After all, the sentence seems to have the
same grammatical form as:
Henry sat in the armchair.
Yet there is clearly something misleading about this surface appearance. For
you wouldn't expect to find the 80 kilograms that Henry weighed in the way
you would expect to be able to find the armchair he sat on. The difference in
the two cases is easily explained. Henry is a certain size. That is just one of his
properties; it is not a relation between him and some other kind of object. But
when we come to describe that property, we have a special way of doing it: we
use a verb ('weighs') which relates Henry to a numerical scale, although the
scale is used merely as an index of Henry's weight and not as something that
can interact with him like an armchair.
The suggestion in the case of the attitudes is patterned on the case of weight:
attitudes are seen as non-relational features or properties of individual human
minds. However, when we come to describe these attitudes, we have devised a
special scale for the job - a special way of indexing them. We use the sentences
of our everyday language. Thus, going back to our first example: at a parti-
cular time, Anne is in a certain mental state. Call that state S. Now S is going
to be very useful for our understanding of what Anne is likely to do and say, so
it would be nice to have some revealing way of characterizing it. What we
therefmre do is to find some sentence (,The coat is in the hallway cupboard') in
our language which, for example, is the likely kind of thing we might say if we
were in a state like S. This sentence is then used in a verbal construction
('Anne believes that .. .') so as to provide us with a more useful description
of S:
Anne believes that the coat is in the hallway cupboard.
44
ST AGE 2: DIGGING DEEPER
Of course the sentence we choose has a meaning - it is about the way things
are or might be - and it is because of its meaning that this sentence is a rea-
sonable one to use in 'measuring' Anne's mental state. But in using it for
indexing or measuring her state, we are not committed to regarding that state
as intrinsically relational. There is no mysterious kind of thing to which Anne
is believingly related just as there is no mysterious kind of thing called '80
kilograms' to which Henry is related merely because he is of a certain bulk.
Exactly how to choose our sentences to measure people's attitudes is some-
thing open to lots of different kinds of interpretation (see FOLK PSYCHOLOGY;
BELIEF (2)). Above I suggested that we might choose sentences that reflect
what we, the attributer, might actually say when in the same kind of mental
state as the believer. But there are other possibilities. For example, the best
sentence might simply be the one that the believer would assent to, if asked,
though this tends to make it difficult for non-language users to so much as
have attitudes.
As with the other proposed 'solutions' to the problem of directionality, this
one has its drawbacks. Aside from the difficulty of giving an acceptable account
of how to go about measuring attitudes with sentences, one is left with worries
about these 'states' of believing (and desiring, hoping, etc.). Are they something
over and above the patterns of action they are responsible for? Or do they just
come into being when we measure them with sentences? Dealing with these
sorts of worries would take us far beyond the task we have set ourselves in this
stage.
2.1.4 Conclusion
We started with what seemed a straightforward question: how does the direc-
tionality of the attitudes work? In considering answers to it we have now gone
inconclusively through a number of proposals: attitudes are relations to the
world; to propositions; to modes of presentation; to things (res) and/or dicta, to
wholly self-contained (narrow) contents and/or to wide contents; and, finally,
we have even speculated that they might not be relations at all. Moreover,
inconclusiveness here just reflects the state of play in the philosophical com-
munity: there just isn't a best theory of the directionality of the attitudes. (See
BELIEF .)
And things get worse. The aim of stage 2 is to probe our conception of the
mind - to ask questions about it and to see what assumptions lie below the
surface. What we are not doing - yet - is to see how the mind fits into our
wider understanding of the world, into the kind of understanding most often
identified with scientific inquiry. But when philosophers talk about the 'prob-
lems of INTENTIONALITY', they tend to run together two things: what exactly is
the directionality of the attitudes and how can anything like that exist in the
world as described by science (see CONTENT). Of course, this is not necessarily a
bad thing to do. After all, you've got to know what you are talking about
before you see how it fits into your picture of the world's workings. Still, the
running together of these questions can lead to confusion, and that is why I
45
AN ESSA Y ON MIND
have chosen to keep them in separate stages. What I have described here in
stage 2 is how one of our most prominent ways of speaking about the mind -
the attributing of attitudes - has a feature that is downright perplexing. When
we dig just beneath beliefs, desires, etc. there is a labyrinth, and I have tried to
supply a thread to guide you, though it is not enough of a guide to show you
the way out. (See also CONCEPTS.)
2.2 Experience
2.2.1 An Experience
Before it was the usual practice to give injections, my dentist had a method
intended to help alleviate the suffering. He would arrange that my right arm
rested comfortably on the arm of the chair saying: 'If the pain is really bad,
raise your arm and I'll stop drilling.' A fine promise. But, on more than one
occasion, having suffered, as I thought bravely and for a long time, but
needing a rest from the mounting agony, I would raise my arm only to be told:
'What I am doing to you just doesn't hurt. You are not in pain.'
In retrospect, even I can find the dentist's claim amusing because the
thought behind it - if there was one - is so much at variance with both our
ordinary conception of mental items such as pain and with the dentist's own
affirmation of this conception. As noted in the first part of the Essay, bodily
sensations are generally regarded both as highly accessible and as only poorly
observable, where these terms are understood in the somewhat special ways
outlined earlier. Thus, the pain I experienced in the dentist's chair was highly
accessible (I was thoroughly and intimately aware of it) and only tenuously
observable (the dentist and any other third parties cannot directly tell that I am
in pain). In devising a system to signal him, the dentist apparently recognized
these features of the situation. Yet, his later refusal to count my hand move-
ment as a genuine indicator of pain contradicted that recognition. Perhaps
because of his then advanced age and long experience with suffering, he was
somehow more expert than most at telling whether someone was having pain.
But however much of an expert he took himself to be, it does strike us as
bizarre to regard his expertise as more reliable than his patient's, because we
are so committed to the idea that our bodily sensations are accessible to us and
not directly observable by others.
The above story brings out something else about the nature of experience.
The dentist asked me to 'tell' him when I was in too much pain by raising my
arm. Clearly, since I was unable to speak in those circumstances, some such
relatively crude signalling device was necessary. But suppose that I could have
spoken. Would this have allowed me to communicate my pain to him in any
more adequate way? I think we would be tempted to answer 'no'. As was
noted earlier, experiences such as pain tend to be very low on the expressibility
scale: we can say that we have them and we can say a few things about them,
but they have features that - as we commonly say - cannot be put into words.
46
In raising my arm, I was using a pre-arranged, uncomplicated signal to convey
the idea that I was in pain. But I was not thereby making the pain itself - the
content of my experience - available to the dentist. Nor would such con-
veyance seem easier, or even in principle possible, if I had been able to use the
resources afforded me by the public language I shared with the dentist.
PAIN is often taken to be a good example of the category of experience. But
when one realizes that this category includes the whole gamut of bodily sensa-
tions as well as the kind of thing that takes place when we perceive the outside
world, or introspect our current mood or trains of thought, it may be un-
reasonable to use the example of pain in this way. For certainly there are dif-
ferences between pain experiences and the others. Yet, pain has in common
with these other experiences these features: high accessibility (to their subject),
poor observability (by others) and poor expressibility (by the subject to others).
And, since it is the conjunction of these three general features that makes so
much trouble for our understanding of experience, we need not be worried
here about specific differences within that category.
2.2.2 Knowledge
There are many different ways in which such trouble can be made to appear,
but perhaps the most direct begins by a brief reflection on what it is to know
something. Here is a truth - almost as a truism - about knowledge:
When you know something, there is something that you know.
This claim owes its obviousness to our understanding of what makes some-
thing a propositional attitude. For, though knowledge may be special in all
sorts of ways, it is after all one among the attitudes, and as such it has
content. A specification of a knowledge claim has the standard form 'x knows
that p' where x is some knower and p is the knowledge content. So, of course
when you know something, there is something - a content - that you know.
A second claim about knowledge would appear to follow directly on from
this:
When you know something, you can say what you know.
Calling this the 'expressibility principle', you should be able to see why it might
be thought to be a consequence of the claim about content. If the standard
form of knowledge attribution is: 'x knows that p', then any such truth comes
equipped with what is needed for expressing that knowledge. I know that
Tower Bridge is in London, so I have available a form of words (,Tower Bridge
is in London') with which I can express my knowledge to anyone who cares to
listen.
There are many reasons one might have for resisting the idea that know-
ledge is essentially expressible, and some of them will emerge in the further dis-
cussion of experience. But there is one reason that is not particularly relevant
47
AN ESSA Y ON MIND
to our concerns, and it can be dealt with by a small qualification to the expres-
sibility principle. There can be things that we know, but only in an implicit
way. For example, I might be said to know all sorts of things about the
grammar of English, though there would be no point in asking me what I
know, since I might lack the requisite grammatical concepts to put my know-
ledge into words (see CHOMSKY). Usefully, such knowledge is called 'tacit' pre-
cisely because it is knowledge that we are unable to spell out. However, though
it is clear that tacit knowledge, if there is any, makes trouble for the idea that
we can always say what we know, this will not be relevant to the present
argument. For it certainly seems plausible that when we know something in a
completely explicit way - when we are non-tacitly aware of the content of our
knowledge - then we ought to be able to say what we know.
2.2.3 The Problem
Against these background claims about knowledge, the path leading to trouble
is uncomfortably straight. When someone has a pain, it certainly seems to be
something known to the person whose experience it is. Equally, when I look
out at an expanse of forest in the summer, the experienced quality of the shades
of green, as well as that of the shapes of the trees, seems to be something in
my CONSCIOUSNESS I not only know, but that I know in a way no one else does
or can (see PERCEPTION and PERCEPTUAL CONTENT). Yet, the contents of my
knowledge in these cases is not something that I can convey to someone else. I
can of course say that I am in pain or that I am experiencing a complex range
of shades of green. But neither of these would manage to convey the content of
what I experience to someone else; neither would they make it apparent what
it is like to have these experiences. So, it seems that I can know all sorts of
things in respect of my conscious or experiential states that I cannot properly
express. And this undermines the perfectly ordinary idea that if I know some-
thing, then I can say what it is that I know.
Initial reactions to this vary. I can even imagine someone thinking that there
isn't much of a problem here - just a case of knowing something we are
unable to find words to describe. Here we might be encouraged to think of
times when something was so surprising, or passed by so quickly, that we
struggled to find adequate descriptive language. And basically the thought here
is that I was just wrong in suggesting that we can always say what we know.
Perhaps, in spite of first impressions, the principle of expressibility is just not
true of knowledge. Unfortunately, the tension between the expressibility of
knowledge and the ineffability of experience will not go away so easily. Yet, if
you are struggling to keep the tension in focus, perhaps the following example
will help.
2.2.3.1 Simone's Story Simone was born colour-blind. In her case, this meant
that she could not distinguish any colour at all: her colour-discriminative abil-
ities were no better than yours would be if you watched a black and white film
and had to answer questions about the colours of the actors' clothes. However,
48
Simone did not accept her 'disability'. Instead, she set about trying to cope with
the world of colour by other means. By laborious study, she came to under-
stand the physical and psychophysical basis of ordinary colour perception. In
time, and using devices to measure the wave-lengths of reflected and trans-
mitted light, as well as allowing for the peculiarities of the way the human
brain processes such input, she was able accurately to tell the colour - indeed
even the shade of colour - of any object. Had you not been aware of the
complex measuring devices, not to mention the theoretical reasoning that went
into her judgments, you would not even suspect that her colour perception was
inferior to yours.
That is the example. But dwell for a minute on the idea that, in spite of all
her research and her uncanny accuracy in describing the colours of objects,
there is still something defective about Simone's colour judgment. What makes
it so natural for us to continue to think of Simone as colour-blind? I don't
think it is difficult to offer some kind of answer here. She gets the colours of
objects right, but she lacks something enjoyed by a non-colour-blind person -
her perceptual experience is markedly different. Speculating a bit, we might
imagine that her perceptual experience is something like that we would have if
we were watching a black and white film. But whether or not this is exactly
right, it seems clear that her inner world is markedly different from ours. And
one way to summarize this difference is to say that we know something about,
e.g. seeing the cloudless sky, that she doesn't.
Embellishing the story a little can make this point even more sharply.
Suppose that Simone undergoes an operation which makes her experience
exactly like ours. In this event, it is natural to imagine her saying something
like this: 'I always knew that the sky was blue, but I never knew that blue was
like that.' And this last word - the demonstrative 'that' - points directly at
what we are trying to capture. The story of Simone seems to lead us inexorably
to the idea of there being something it is like to perceive colours (and to have
pains), and to the further idea that we know about this inner realm. We are led
to these conclusions because - to repeat - it seems natural to think that
Simone comes to know something about her experience of seeing after her
operation that she didn't know before. (Remember that what Simone is said to
know after her operation is not a fact about the sky, but one about her experi-
ence of it.)
The trouble is, though, that whatever it is that she knows remains strangely
inexpressible. For, as we noted, there just doesn't seem to be a vocabulary suit-
able for giving any precise content to this knowledge. The best that I could
imagine Simone as saying about her newly acquired knowledge is: 'I now
know that blue is like that.' But this doesn't really convey the whole of what
we usually expect from someone who is said to know something. Thus, con-
sider how we manage with knowledge about our surroundings. I am just now
sitting in my study, so I certainly know what it is like. And, given this, you
would expect me to be able to tell you - to say things like: 'There is a desk in
the middle, a pile of books on the floor', etc. But what would your reaction be
49
AN ESSA Y ON MIND
if, instead of these things, 1 said: 'Well, my study is like that'? 1 suppose that
you would think that 1 either didn't know what 1 was talking about or that 1
was concealing certain things from you.
2.2.3.2 Introducing 'Qualia' It should be obvious that the problem we are
having with our knowledge of experience is unlike cases in which we witness
something so strange, or so terrifying, that we lack words to describe it. For it
is possible to keep your attention fixed on what you are experiencing long
enough to know (as we say) perfectly well what it is like. The trouble is that
we just don't have words to express this knowledge. Of course we say all sorts
of complicated metaphorical things about our states of mind: 'I felt as if 1 had
been run over slowly by a 73 bus' (hangover), or 'I feel as if someone has tied
a knot in my stomach and is twisting it' (anxiety). But our problem is not met
by these sorts of description. What we lack are words for the simplest of experi-
ences, such as the way it is with us when we see a perfectly ordinary shade of
colour.
At this point you might be tempted by the thought that this lack is really
quite superficial. After all, since we are mostly interested in telling each other
what is going on in the world, it is perhaps not surprising that our language
has developed without the necessary resources to describe our inner life.
However, what is to stop us inventing some terms to fill this gap? Nothing - or
so say those philosophers who have taken at least the first step in that direc-
tion.
Think of how things are when you look up at a cloudless sky. Asked, you
will say that it is blue. But of course the colour word is used here to describe a
feature or property of the sky, and no one would think otherwise. It is simply
bizarre to say that your experience is blue. But then how can you even begin
to describe how it is with you when you are seeing the blue sky? Here's how.
Concentrate on the nature of your sky-directed experience, taking special care
to keep fixed on what is happening to you, whilst ignoring as best you can
how things are with the sky. There certainly seems to be something going on
in your consciousness - something that has various properties. If you doubt
this, just imaginatively compare how different these things would be if you
were looking at the same sky, but that its colour began to change, having been
made to glow red by the setting sun. No one could doubt your ability to distin-
guish the experiences of the two differently coloured expanses. And what else
could explain this except that the experiences have different properties.
Now we of course don't have names for these properties - either the specific
ones in the blue-sky case, or even their general category. But - at least in
respect of the general category - why not just coin some suitable word? Why
not follow the lead of some philosophers by suggesting we call the properties of
experience (in general) 'QUALIA' (singular: 'quale')? The name seems apt since
what we are dealing with are the qualitative aspects of experiences - the 'how it
is to us' when we have them. Armed with this term, we can say this much
about the blue-sky experience: it has qualia which distinguish it from, for
50
example, the red-sky experience. We could also go on to invent terms for each
of the specific qualia that mark our experience - we could say that the relevant
quale of my experience when I look at the blue sky is 'bluish', though this
latter term should not be confused with 'blue'. But what matters in making
sense of such words as 'bluish' is that they pertain to qualia, so all the
weight really hinges on the general term. For example, we can now (appar-
ently) say what is defective about Simone's colour judgments before the opera-
tion: it is that she makes them without reference to the qualia of her
experiences.
Does the coining of the term 'qualia' solve the problem with which we
began? That is, can we now reconcile our knowledge of our experience with
the demand that knowledge be expressible? This certainly would be a quick fix
to the original problem - almost certainly too quick. Any new word only
counts as a genuine extension to our language if we are fairly sure that we can
understand what it is supposed to signify. And it is not clear that we have said
enough to imbue 'qualia' with a meaning suitable to get over the expressibility
problem. For the suspicion is that this new word doesn't do any more for us
than the demonstrative 'that' as it occurred in Simone's original judgment: 'So
that is what blue is like.' Indeed, I suspect that many of the proponents of this
new word would be perfectly satisfied with this outcome. They never wanted to
use the term to give a definitive answer to the expressibility problem. Far from
it. What they wanted was some way to talk about that very problem. The
word 'qualia' was invented as a way of answering those who take our lack of
expressive means as indicating that there may not be anything it is like to have
an experience. However, rather than spending any more time on qualia here,
we should now review more complete responses to the expressibility problem.
Considerations relevant to our understanding (or lack of it) of the word 'qualia'
will be more intelligible against this background.
(Those familiar with some philosophical literature may think that 'qualia' is
just another name for what used to be called 'sense data'. I don't think this
would be quite right, but spelling this out is not appropriate here (see PERCEP-
TUAL CONTENT). Suffice it to say that those philosophers who have found it con-
genial to speak of sense-data have used them in projects different from our
present descriptive one.)
2.2.4 Grasping the Nettle
I can imagine someone thinking that we have overlooked the best strategy for
dealing with our present trouble in respect of knowledge of experience: what
we should do is just brazen it out. If the accessibility and inexpressibility of our
experiences do not fit with the idea that knowledge ought to be expressible,
then why not just say - as far as experience is concerned - that the express-
ibility principle just doesn't apply? Why not just say that there is for each of us
an inner realm to which we have a directness of access that is denied to
anyone else and, further, that it is a mistake to regard the knowledge we have
of this realm as expressible in the ordinary way? I can tell you that my study
51
AN ESSAY ON MIND
has a desk and books, but that is because the study and its contents are pub-
licly observable - they are part of an 'outer' realm. However, when it comes to
experience, the best I can do is to make a sort of gesture: I can tell you (as I
did to the dentist) that I am in pain. I can tell you that my experience is of that
special kind that I have when I see the blue sky or green leaves. And I can
invent words like 'qualia' to make it seem easier to refer to these special fea-
tures of my experience. What I cannot do is to be as fully informative about
the inner realm as I am about the outer, but that is just how it is.
In essence, this suggested way with our problem comes under the heading
'grasping the nettle'. One admits that there are features of experience and
knowledge that are in tension with one another, but then, rather than allow-
ing this to bother us, we simply count the tension as itself a sign of the special
nature of experience. The trouble is that grasping nettles - even philosophical
ones - can be extremely painful. But before we see just how painful it is in the
present case, I should like to point out something about the history of the view
just canvassed.
As I have presented it, the view is a reaction to the apparent difficulties that
we get into when we add our conception of knowledge to our conception of
experience. However, the standard view of the history of these matters does not
treat it as in any way a reaction to problems. Rather, in the works of Descartes
(1596-1650), it is usually seen as the robust and confident statement of a
conception of the mind that has been worked on - and over - during the past
four hundred years. Indeed, it could be argued that Descartes himself only
articulated what anyone would say about these matters; many regard the Car-
tesian view as the common-sense view. Moreover, the underlying tensions to
be found in the conception have only been the centre of concerted philosoph-
ical attention in the past fifty years or so. Still, having admitted that my pre-
sentation is a-historical, I hope you will come to appreciate that it is none the
worse for that. (See HISTOR y.)
Grasping the nettle in the way described is tantamount to accepting that
there is a sharp divide between our knowledge of the world - the external
world of tables and chairs - and the inner world of our experiences which is
commonly said to be known by INTROSPECTION. Of course, the spatial metaphor
inherent in the words 'inner' and 'outer' is not mandatory. Less metaphori-
cally, we could describe the Cartesian view as requiring us to recognize the
divide as between first- and third-person knowledge - between what I know of
myself, and what I can know of someone else. But whatever words we use,
there is a high - and many think unacceptable - price to pay for maintaining
this divide. (See SUBJECTIVITY.)
First of all, the existence of a sharp distinction between first- and third-
person knowledge makes it difficult to understand how we can ever know that
our friends and neighbours so much as have mental lives. For the Cartesian,
each of the surveys of our feelings, perceptions, sensations and moods takes
place in the privacy of our own minds and, though the 'qualia' we thereby
come across can be as rich as anything, the attempts to convey these things -
52
to make them present in some way from the third-person perspective - never
amounts to anything; all that we have as evidence of each others' mental lives
are words issuing from our lips, and other bits of behaviour. But these words
and behavioural manifestations could issue from creatures completely lacking
in any inner life, so a complete scepticism about other minds is made possible
by the Cartesian picture. And if we think that any view encouraging such
scepticism is unacceptable, then this will be reason enough to look elsewhere to
deal with the tension between knowledge and experience.
Of course, there are those who will not react to this kind of scepticism with
horror: grasping the nettle a little harder (so to speak), they will insist that
scepticism about other minds is perfectly reasonable - at least as a possibility.
After all, no one has found a completely satisfying way of showing that philo-
sophical scepticism in general is incoherent. The image of philosophers as
people struggling to prove the reality of the kitchen table at which they are
sitting is an enduring one. And philosophers know that these struggles can be
interesting and helpful to our thinking, though they do not for a minute have
any serious worries about resting their elbows on the table. So, why not admit
that we might be unable to prove conclusively that the sceptic about other
minds is wrong, whilst unconcernedly living and acting as if we had no doubt
at all about the mental lives of other human creatures? What is to stop us
saying that we can never know that those around us have minds, but that we
can be pretty sure that they do - for practical purposes?
Leaving aside the issue of whether it is ultimately satisfying to deal with the
sceptic by conceding even that much, there is a deeper problem which is easy
to miss if you concentrate too hard on scepticism. In a nutshell, it is a question
of the very coherence of the first-/third-person divide, and it is best approached
by investigating the words that figure in our speaking about the mind. Take a
specific example: Harry has a pain in his elbow, and he tells you this. Given the
Cartesian picture, Harry has not thereby managed to convey fully what it is
that he knows; full expressibility is denied to this sort of first-person knowledge.
Yet Harry does use the word 'pain', so it is perfectly reasonable to ask what
this word means both in his mouth and in his audience's ears. Here is one
view: Harry uses the word 'pain' as the label of the experience he is presently
undergoing. After all, he judges this experience to be sufficiently similar to
other experiences he has previously labelled 'pains'. So, the word 'pain' has its
meaning fixed by the inner landscape that Harry (and no one else) is in a posi-
tion to survey.
This won't do. For how would we ever be in a position to understand what
Harry said? If the meaning of Harry's remark depends on his own con-
ceptualization of a realm completely unobservable by us, then he might as well
have said something like this (with apologies to Lewis Carroll): 'My elbow feels
uffish.' And actually the situation is even worse: for the conceptual location of
pain depends upon the availability of general terms like 'feeling' of which pain
is a sub-species. But the word 'feeling' too must get its meaning from Harry's
inner survey, and it is thus no more available to third parties than 'pain'. So,
53
AN ESSA Y ON MIND
Harry's remark might have been: 'My elbow whiffles uffish' and this certainly
leaves us in the dark about what is up with him. (Note that this line of reason-
ing would apply especially forcefully to the word 'qualia'. Indeed, the preceding
sort of reasoning is what grounds some people's conviction that coining terms
like 'qualia' does nothing to help us with the original problem.)
What can seem the natural move here is to admit that Harry's word 'pain'
labels something unavailable to us, but to note that his use of the word goes
with behaviour that is observable and is just like the behaviour we go in for
when we have such experiences. Included here is behaviour such as uttering
the sounds 'My elbow hurts.' Reasoning then by analogy with our own case,
we say that Harry's behaviour shows him to mean pain by 'pain', since when
we have pain, we behave in roughly the same way. But this won't do either.
What we are asked to imagine is that the word 'pain' applies to me because of
what I experience, but that it applies to others because they behave in roughly
the ways I do when I feel myself to be in pain. Unfortunately, this appeal to
analogy does not take seriously enough the confinement to our own case. For,
given what has been described, it may be more sensible to say that the word
'pain' is systematically ambiguous than it is to say that it has the same
meaning when I apply it to myself and when others apply it to themselves.
This is because I never do manage to experience someone else's alleged pain -
my analogical reasoning never gets a single confirmation. Even worse, it is not
clear that I could now coherently say what constitutes a confirming instance.
For, if I say that it is a case of someone else's having just what I have when I
use the word 'pain', this never happens - one thinks indeed it couldn't happen.
Yet, how could I even begin to say how someone's supposed inner life must
differ from mine whilst remaining faithful to the idea that the word 'pain' is
used in just the way I would use it? Given this problem, wouldn't it just be a
lot more plausible to say that 'pain' means one thing for me and something
else when used by others? The trouble is, this makes it wholly mysterious how
we ever do manage to communicate with one another.
I have put this point in terms of words and communication, and there are
those who might feel that such problems with language are relatively super-
ficial. However, whilst the difficulty is easiest to explain as one about the use of
words like 'pain', it is far more serious than that. For, given the Cartesian view,
we seem unable even to think coherently about experience. If the concepts in
terms of which we organize our thoughts about ourselves apply to a range of
experiences that are only available from a first-person point of view, then there
will be an unbridgeable divide between them and those concepts we use to
think about others.
The trouble with scepticism seemed to be that it made knowledge of other
minds impossible, though we imagined that we could at least formulate the
hypothesis that others enjoy experience like ours. Now it should be apparent
that there is a problem with the hypothesis even before we get to scepticism. If
our conception of the mindedness of other people is determined by their pat-
terns of action and speech, then of course we will often know that they have
54
minds. But the hypothesis that is the subject of the sceptic's interest is not
about publicly available behaviour. Yet how can we characterize the appro-
priate hypothesis? Is it that others can look inside themselves and see that
things are like this (where we concentrate very hard on our stream of con-
sciousness)? That can't be quite right. How can they find things to be 'like this'
when what that is like is forever hidden from them? (Remember that the word
'this' refers to one's own conscious experience even though the aim is to give
content to the thoughts of others about their experience.) How can I set out to
imagine someone else enjoying my stream of consciousness?
To this you may be tempted to reply that I have overstated the case for the
incoherence of the hypothesis: what is wanted is not that I imagine someone
else enjoying my stream of consciousness, but just something like it. The
trouble is. though. that on the Cartesian picture. the only experiences I ever
come across are my own. and the concepts I use in framing my thoughts about
that experience are forever bound to it. It is not easy to see what content I
could give to the idea of someone enjoying what is not my experience but only
very much like it. when the only idea of experience in play is my own. It is
easy to miss this. since there is a tendency to think of cases where it makes
perfect sense to make the first-/third-person shift. For example. though I have a
particular kind of watch. there is no problem in imagining someone else having
a watch of the same kind. But this is because the watch in question is in the
public domain to begin with; anyone with the relevant discriminative abilities
can frame a concept of that object and then use it to think about an object that
would be like that one except in respect of ownership. But the Cartesian view
makes it difficult to see how this sort of move is possible in respect of any given
experience since ownership in this case is simply not detachable.
2.2.5 The No 'Know' View
Our discussion of the Cartesian attempt to grasp the nettle seems to lead to an
impasse. But whether or not the case is proven. the time has surely come to
see if there are less painful ways out of the original difficulty. Recall that all
our trouble came from the fact that knowledge seems to be essentially express-
ible. whereas experience seems to be both known and inexpressible. I portrayed
the Cartesian as someone prepared to say that we just have to live with the
fact that knowledge of experience is an exception; someone prepared to say
that there can be an inner realm of things which we can survey and thereby
come to know about. though we cannot properly describe what we know.
However. since there do seem to be reasons for thinking that we cannot really
live with this picture. we ought to find something else to say about experience
and knowledge. And a promising thing to do - almost as a reaction against the
Cartesian picture - is to question the very idea of an inner realm known in this
way.
The simplest expression of this questioning attitude consists in denying that
the word 'know' is appropriate in the first-person case. Typically. when we
know something. or are said to know it. we have had to do some looking
55
AN ESSA Y ON MIND
around and checking. For example, I would now claim to know that there is a
bird sitting on the branch I can see from my window. Of course, I could be
wrong, and, if it was important, there are steps I could take to check this by,
among other things, consulting with others. Putative knowledge is essentially
fallible and it is also intersubjectively checkable. That is why the principle of
expressibility seems so apt. However, none of these features seem to apply to
what we have been calling 'knowledge' of our own experiences: it is not fallible
in the way that my perceptually based knowledge is, and it is certainly not
intersubjectively checkable. So, maybe the way forward is to deny that we
have knowledge of an inner landscape of the kind suggested by the Cartesian. I
shall call this the 'no-knowledge' view.
Here one must be very careful. It is all too easy to think that the suggestion
just made is in effect a denial of the very existence of experience. To take our
earlier example: when Harry says that his elbow hurts, the no-knowledge view
will insist that Harry is not thereby attempting to express something he knows.
And one might be tempted to think this was possible only on the assumption
that Harry simply had no inner life - that there was nothing there to be the
subject matter of any sort of knowledge. This temptation may even be further
reinforced by remarks made by proponents of the view when they attempt to
spell out just what it is that Harry is getting at. In particular, one suggestion
has it that we should count Harry's words only as a piece of behaviour to be
put alongside other things he might do in respect of his elbow, such as flinch-
ing when it is touched. When he says: 'my elbow hurts', he is in effect just
saying 'Ouch!', though in a more complicated way. The one thing he is not
doing is describing how things are with him, and this can lead one to think that
the no-knowledge view is actually denying that there is any way things are
with him. (See WITTGENSTEIN.)
A view that insists that mental concepts are in the end concepts that apply
only to people's doings and sayings is usually labelled 'BEHAVIOURISM', and this
view is beset with difficulities of its own. However, more important now is the
simple fact that behaviourism is not really much of a help with our problem.
We want to be able to say something about both experience and knowledge
which makes it reasonable to regard experience as highly accessible without
thereby threatening the expressibility of knowledge. A behaviourist interpreta-
tion of the no-knowledge view does not do this, since it just denies or ignores
the thought that Harry has experiences that are highly accessible only to him.
Of course, if the original problem proves completely intractable, someone might
urge us to do the radical behaviourist thing, facing whatever further problems
this leads to. However, for now it would be better to resist behaviourism and
see if there is a way to understand the no-knowledge view which does help
with the original difficulty. The no-knowledge view is after all the no-knowledge
view, and, however tempting, this should not lead us too quickly to see it as
the no-experience view.
As several of my earlier examples indicated, knowledge typically involves the
application of our concepts to the objects and properties we come across. Thus,
56
any knowledge I have of the bird sitting in the tree consists in my having con-
cepts of such things as birds, trees and sittings, and my putting them together
in appropriate ways in my thoughts. Note, by the way, that the objects and
properties mentioned here are publicly available - one might say that they
belong to no one and yet to everyone. However, on the Cartesian picture, the
objects and properties that make up the inner landscape (experiences and their
qualia) are quite different; they belong only to the person whose landscape it is.
And this is what upsets the no-knowledge theorist. Such a theorist has nothing
against experience or against the fact that experience is in some sense private.
What he objects to is seeing the realm of experience as just like the realm of
objects and properties that make up our public, intersubjectively available
world.
Spelled out a bit more, the thought might be expressed this way: of course
we all have mental lives and conscious experiences. And of course we inhabit a
world consisting of objects and properties which we can come to know about
and describe. But the conditions for our having concepts of, and describing, the
external world are simply not met by experience. The Cartesian mistake con-
sists in thinking that the 'inner world' is, for the purposes of knowledge and
description, just like the external one. This mistake is signalled by the will-
ingness to use the word 'knowledge' to describe our relationship to our experi-
ences, but the mistake itself is not merely one about when to use this word. It
is based on a deep misconception of the nature of experience. It assumes that
our experiences are arrayed in something like a landscape even though none of
the features of real landscapes apply.
The no-knowledge view is not an easy one to keep in focus. It has a ten-
dency either to shift before your eyes into a kind of behaviourism or to dis-
appear altogether. Returning to our example of Harry's pain, we certainly have
a right to be told a convincing story about what Harry is up to when he says:
'My elbow hurts.' As I mentioned, if this is treated as nothing more than a
complicated way for Harry to say 'Ouch!', then we are faced with a kind of
behaviourism. Surely (many think) these sounds are not fully characterizable
as a kind of ornate flinching. They say something about Harry's present experi-
ence - a kind of experience that explains why Harry made his remark and
that, if Harry did flinch, would explain that as well. The no-knowledge theorist
insists that we must not treat the phrase 'say something about Harry's experi-
ence' as just a paraphrase of 'tell us what Harry knows about himself'. For it is
insisted that to do this invites the profoundly mistaken Cartesian view that
Harry is the only possible witness to his 'private' pain, when the very idea of a
witness only makes proper sense in respect of things that could be witnessed by
others. Yet how else can we take Harry's remark? And, conSidering again the
case of Simone after her operation, what could be more natural than to see her
as saying: 'Now I know what it is like to see blue'?
Unless the no-knowledge theory can respond to these demands, it seems to
be no more than a warning about how to use certain words. And the pressure
on the no-knowledge theorist to let us use the word 'know' seems to many
57
AN ESSA Y ON MIND
overwhelming. Yet there does seem to be something right about resisting the
Cartesian picture and all the difficulties into which it tends to lead us. Given all
of this, perhaps the way forward is to think of some way in which we can con-
tinue to allow that there is a special kind of knowledge which each of us has in
relation to our experiences, but which is at the same time a kind of knowledge
that steers clear of any commitment to the Cartesian picture. This is where the
'know-how' theory might come to the rescue.
2.2.6 The Know-how Theory
When we think of knowledge as essentially expressible, we surely have in mind
knowledge having propositional content and characterized by the standard
form: 'knows that p'. But there are other sorts. Thus, among his other accom-
plishments, Harry plays the piano rather well and it is natural to describe his
talent by saying: he knows how to play the piano. But this attribution of
knowledge does not require anyone (and this includes Harry) to be able to
express what is known in some propositional form. The content of Harry's
knowledge is not something like a proposition that is even a candidate for
truth. Instead, it is an ability or capacity. (See LEWIS.)
Can the distinction between knowing that something is the case, and
knowing how to do something help in our present predicament? Here is a sug-
gestion. Suppose we agree that knowledge that is essentially expressible, and
that Harry (in pain) and Simone have knowledge of their experiences. But we
insist that, in the latter cases and appearances aside, there are no propositions
that Harry and Simone know. Instead they each have a special kind of know-
ledge how. Consider Simone: before her operation, we felt that her vision was in
some sense defective even though her judgments about the colours of things
was spot on. After the operation, her judgments were no different, but it is
natural to suppose that something had changed - that she had acquired some
kind of perceptual knowledge. Well, why not say that what she had acquired
was a kind of discriminatory knowledge - a knowledge of how to discriminate
colours in a new way? The old way consisted in her appealing to measuring
instruments and theories; the new way just involves looking.
Next consider Harry: he says that his elbow hurts, and it is natural to
suppose that he is thereby expressing something he knows about himself.
However, instead of telling the Cartesian story about Harry's inner landscape,
why not just say that Harry's remark shows him to have a kind of dis-
criminatory ability which is lacking in others? He can tell just by paying atten-
tion that his body is damaged in some way. Harry's specific ability is lacking in
us because we have to examine him to see what the damage is - we cannot
tell just by paying attention - though of course we each have an analogous
discriminatory ability in respect of ourselves. And here the appeal to an
analogy between Harry and ourselves is less fraught than it was in connection
with the inner landscape view. For what is analogous is a perfectly accessible
item: an ability to say - without instruments or detailed investigation - that
one is damaged in a certain way.
58
The appeal of this know-how view is considerable. It allows us to say that
the difference between the first- and third-person cases is one of knowledge: for
Harry and Simone do know something. Yet we can say this without fear of
contradicting the principle of expressibility of knowledge, since that principle
only applies to propositional knowledge - knowledge that. Neither Harry nor
Simone have that kind of knowledge about their states of mind. Additionally,
this view gives us something to say about the accessibility and observability of
our self-knowledge. For Harry and Simone not only know something - they
know something we don't. That is, their abilities to tell what colour something
is by looking, or to tell that they have suffered some kind of damage just by a
sort of paying attention, are abilities special to them. To be sure, we can also
tell what colour something is and whether their bodies are damaged, but not in
the way they do. And it is the particular way - the particular ability - that is
at issue here. So, we can say of this ability that it is highly accessible to those
who possess it, but only low on the observability scale. Finally, as far as
expressibility is concerned, it should come as no surprise to learn that it is
virtually impossible to express what goes on when we exercise this kind of
ability. No surprise, that is, just in the way that it is no surprise that Harry
cannot describe his knowledge of piano-playing in any terms but these: 'I know
how to do it.' Indeed, one could just about consider Simone's excited claim
'I never knew that blue was like that' as more or less equivalent to: 'I never
knew how to tell by looking that something was blue. But now I do know how
to do it.'
It is easy to get carried away by the know-how view, but all is not plain
sailing. The main problem is that our abilities don't stand on their own: it is
usual to ask for some kind of explanation of them. Thus, consider two abilities
that Harry has: he can play the piano and he can hold forth about the kings
and queens of England. The first of these is of course very complex and took
years to develop, but it isn't unreasonable to say that it is partly a form of
what is sometimes called 'hand-eye coordination'. Harry's eyes take in certain
notes on the page and his hands, fingers and arms react to them as a result of
the training he has received. Of course, if he is any good at the piano, he will
also have the more conscious ability to control these reactions in ways that
conform to his musical intentions. But for the moment let us think only of the
more basic technical ability that consists in getting from the written to the
played notes. How would we explain this? Clearly the answer to this would be
enormously complicated (and is probably only known in outline), but whatever
the details, we can with great plausibility say this much: our explanation will
not have to appeal to anything that Harry explicitly knows or believes. And by
'explicit' I mean things to which Harry has conscious access. Thus, Harry
doesn't consult some repository of consciously available information such as: if
the black note is on the bottom line, then one must use the middle finger of the
right hand. He will certainly know this sort of thing. but consciousness of this
knowledge is not what guides his hands. If it did. it is unlikely that he would
ever manage to move his fingers fast enough. (He might be guided by such
59
AN ESSA Y ON MIND
things tacitly, but, whatever this means, we can leave it on one side for the
moment for reasons that will become clear.)
Contrast piano playing with Harry's ability to tell you about the kings and
queens of England. The obvious way to explain this ability is by appealing to a
whole number of things Harry knows or believes, and to which he has con-
scious access. It is because he knows that Queen Victoria came to the throne in
1837 that he has the ability to say so. In a case like this, speaking about an
ability is really only an indirect way of speaking about what someone knows in
a fully propositional way. It is the knowledge of what is true that explains
someone's ability to come out with it.
Given these two sorts of case, the obvious next question is which of them is
the best model for cases of experiential know-how? That is, what explains the
abilities Simone has with respect to the experience of seeing blue and Harry
has with respect to his pain? Is it like the piano-playing case? Do they just
come out with judgments about colours and pains as if they were producing
these in unconscious response to something discerned - a sort of 'mind-mouth'
coordinative ability? This seems absurd. What one would most naturally say is
that the ability to say how things are (either with colour experiences or with
one's body) is based on a prior awareness of the relevant experiences. It is
because it is like that to see blue - and because Simone came to know this after
her operation - that she has the ability to say of an object, without any use of
instruments, that it is blue. And it is because Harry has an awareness of the
pain that is undoubtedly connected with the damage to his body that he is able
to make the discriminative judgments he does. In short, the more plausible
model of our know-how in respect of experience is the one based on some kind
of prior - and possibly propositional - knowledge. But, if this is the case, then
we are back where we started: the know-how interpretation only hid from
view the know-that problem.
Here I should insert a final note about the possibility of an explanation based
on tacit knowledge, for it might seem as if this would be the best way to deal
with our problem. This is because if tacit knowledge lies behind our abilities,
then we would have a kind of knowledge that supported the know-how inter-
pretation but avoided the expressibility problem (we already agreed that
expressibility didn't apply to tacit knowledge). This move might seem especially
appealing in view of the fact that many theorists of language (and piano
playing) appeal to tacit knowledge to explain the relevant abilities. However,
though all this is fine, it leaves out something crucial: we do want to preserve
the special kind of access that we have to our experiences. But it is difficult to
see how this can be preserved if we think that our experience of colour and
bodily conditions is nothing other than a discriminative ability grounded in
tacit knowledge. Just imagine being asked how it was that you could tell that
the sky had changed colour towards sunset. This would be like asking Harry
how he knew to begin the second bar with his right thumb. If, as we imagined,
this was just a trained reaction on his part, he would simply say: 'don't know -
just happened.' Though we could tell a story about his tacitly knowing a rule
60
that connected the note and the hand movement, this would not be knowledge
that Harry was aware of, and consulted, whilst he was playing. And could you
really be satisfied saying that you had no idea why you said the two sky scenes
differed - it just seemed right to say that they did?
2.2.7 Overall
That we know what we feel, that we know this in a way no one else can, and
that we cannot fully convey what it is like to have these feelings - these things
seem just obvious to many people. Unfortunately, when these are put together
with equally obvious sounding claims about knowledge, the result is less than
coherent. Moreover, as I have been labouring to show, it doesn't seem as if
tinkering with our conception of knowledge or experience gets us very far in
dealing with the threat of incoherence. So something has to give, something
we think about the experiential aspects of the mental realm must be wrong. Or,
at least, something we think must be re-interpreted in a fairly fundamental
way. Radical solutions seem to be called for, and the philosophical community
has not been slow in providing them. Here is a sample:
We don't really have experiences, we only think we do.
We don't really have experiences we only say we do.
We certainly do have experiences, and these are known only to us, but we
don't know anything else including, among other things, about the exist-
ence of other people. (This position is best put in the first person: I know
about my experiences and only about them, etc. It is called 'solipsism' and
it is certainly a radical solution to the expressibility problem.)
We do have experiences but these are known as well, or better, to others -
at least those who take an interest in the dispositions of our bodies. (On
this view, maybe the dentist had a point after all.)
Most of these are not really motivated by local conceptual problems about
experience that have been the subject matter of this part of stage 2. They arise
more directly from the attempt to trace the foundations of experience and other
mental items to the bedrock of the physical realm. This will be the subject
matter of stage 3, but first we must consider the third of the categories, that of
acting and actions. (See also DENNETT; IMAGERY; IMAGINATION.)
2.3 Acting and Actions
As was mentioned earlier, the mental status of actions is controversial: discus-
sion following my informal questionnaires always shows up sharp disagree-
ment about whether to count actions as mental phenomena. There are those
61
AN ESSAY ON MIND
who insist that they are no more than end-products of more genuinely mental
items, and others who think that the credentials of actions to be included in
the mental realm are just as impressive as those of the attitudes. The first
group does not deny the important role that the mind plays in acting, but the
point tends to be put this way: actions are the outputs of minds. As described
earlier, they are the wake our minds leave as they travel through life, and for
this reason they do not merit equal standing with the categories of experience
and attitude. The opposing camp denies this second-class status for acting. On
this view, when we see someone engaged in acting we are not merely seeing
some sort of output, we are seeing the mind itself.
Disagreement about the status of acting suggests that there may be some
fundamental difficulty in our thinking about actions, but it doesn't prove that
there is. After all, supporters of one or the other position may just be wrong,
and may change their minds when faced with a persuasive reason for doing so.
Yet the aim here - as it was in respect of attitudes and experience - is to
uncover any real difficulties infecting our understanding of actions. So, for the
time being, I shall ignore the disagreement over status, and start from the
beginning. What I intend to show is that there are several theses, each of
which would be accepted as obviously true of actions, yet that when these are
put together they yield unacceptable - even paradoxical - consequences. It will
then be easy to see just how far the problem about status is a real reflection of
the trouble with acting.
2.3.1 Four Theses about Action
The world contains lots of things - chairs, houses, people and the like. It also
contains things that happen. Indeed, it would be impossible to explain what our
world was like without saying both what things it contains and what happens
in it. Can we sharpen or deepen our understanding by getting more specific -
by replacing 'happening' with 'what takes place' or with 'event'? Perhaps, but
any such improvements will not be necessary here. For the first thesis about
action requires nothing more than the ordinary, if vague, word 'happen':
Thesis (I) Actions are things that happen.
Of course, there is a use of the word 'happen' that may seem less congenial to
this thesis. For sometimes people speak of a happening as almost accidental. It
might even be felt that this use contrasts with genuine acting, as in someone's
saying: 'there was no action, it just happened'. Yet I suggest that even a
moment's reflection will show that nothing in any of this undermines the truth
of thesis (I). For, when it is claimed that there is a contrast between actions
and happenings, the contrast is always between things done and things that
merely (or just) happen. In order to make the contrast some such restriction is
needed, and the felt need for this qualification actually supports thesis (I). The
supposed contrast between actions and happenings is really a way of saying:
look, actions are happenings alright, but they are not just happenings. And this
62
shows the next thing we must do. We must supplement thesis (I) with some
truth about actions that shows how it can be that the class of happenings is so
much broader than that of actions - how it can be that there are plenty of
happenings that are not actions. This supplement will take the form of a
second thesis:
Thesis (II) Actions are those happenings involving agents: they are hap-
penings that can be described as things done by agents.
This second thesis is a move in the right direction: it shows us how to draw
the line between mere happenings and actions in the broadest sense, but it
must be treated with caution. Here are two examples of happenings that count
as actions in virtue of thesis (II).
(i) The tide scours a riverbed and deposits the silt outside the mouth of the
river. In this case we say that the tide creates a harbour bar - this is
something it does. We also say that it acts on the riverbed.
(ii) Harry puts on his left shoe. This is something that he did - it was an
action on his part.
Now, clearly our interest is in actions of type (ii), but as stated thesis (II) seems
promiscuous enough to allow in both of the above types of action. Of course it
might be argued that tides only count as doing something - as agents - in a
metaphorical sense. There is a sort of harmless, if not poetic, animism in think-
ing of tides in this way. And if this is right, then perhaps thesis (II) could be
regarded after all as capturing just those cases that interest us: all we have to
do is to insist that any metaphorical uses of agency don't count. However,
whether we take this way out or not, we are going to have to say more about
the kind of agency that does interest us, so it is perhaps just as well not to
argue too much about thesis (II). If someone insists that type (i) cases are really
(literally) cases of action, so be it. But even such a diehard would have to
recognize that there is a world of difference between the kind of 'doing' brought
about by tides and the kind that involves what we might call 'personal' agents.
So, since there is this difference, and since our interest is wholly in personal
(mostly human) agents, we might as well come clean and say:
Thesis (III) Personal actions are things done by agents who possess minds.
In being explicit in this way, we can from now on economize by using the
simpler 'action' and 'agent' in place of having always to say 'personal action'
and 'personal agent'. But of course there is much more to be said about what
is involved in this kind of agency.
Theses (II) and (III) are unspecific about the relationship between agent and
action; all they say is that a happening is an action if it 'involves' an agent, or
63
AN ESSAY ON MIND
if it is a 'thing done' by an agent. But neither of these will do. The problem is
that there are many things that involve agents, or could be said to be done by
them, that we would not ordinarily count as actions. One example should
make the point clearly enough. Harry had a bad cold which left him with a
rasping cough. On any of the occasions when he gives voice to his affliction, it
makes perfect sense to ask: what did Harry just do? And for the answer to be:
he coughed. Yet, even though coughing is something that involves Harry - is
something that the agent Harry did - we would not count this among Harry's
actions. To see this, think of a case in which Harry coughs, but this time
because he wants to show his disbelief in what someone is saying. Both kinds
of coughing are things Harry does, but only the editorial cough seems up to
scratch as an action. And the difference between the cases is as plain as can
be: some things we do are done deliberately or intentionally, some are not.
Hence the next thesis is:
Thesis (IV) Things agents do count as actions if they are done intention-
ally.
There are lots of things to say about this thesis. On the one hand, it would be
nice if we could spell out somewhat what is required for something to be done
intentionally, and to show how the intentional is connected to the 'deliberate'
and the 'voluntary'. More on all this shortly. But there is a more pressing need
to say something to block what otherwise might be a devastating objection to
thesis (IV), namely that it is false.
Consider again hapless Harry. He is being shown the porcelain collection of
his neighbours and, when he reaches out in a suitably admiring way to grasp
a figurine, he unbalances another (known to its owner as the 'Hunter and
Dog') which smashes on the floor. No one thinks that Harry smashed the
Hunter and Dog intentionally. Not even those who know he has always
thought that porcelain figures were not much better than tacky souvenirs
would think that. Even so, smashing the porcelain is surely not like his cold-
induced cough; unlike the cough, the breaking was surely Harry's action. In
sum, we seem to have a case of an action that is unintentional, thus contra-
dicting thesis (IV).
One way to deal with this counterexample is to think about it a little more
carefully. Contrast it with a case in which Harry smashes the Hunter and Dog,
not by reaching for another piece of porcelain, but by a spasm that causes his
arm to thrash out in the Hunter and Dog direction. Like Harry's coughing this
kind of smashing is unintentional, but it seems quite different from the reach-
ing case, and the difference is fairly clear. In reaching, Harry didn't smash the
Hunter and Dog intentionally, but he did do something intentionally - he
reached for the other piece. This was what brought about the destruction. As
so often, what we intend and what we end up bringing about as a con-
sequence, are two different things. But it is usual to regard something as a
proper case of action as long as it can be described as intentional from some
64
point of view. With this in mind, we might be able to repair the damage to
thesis (IV) by changing it slightly:
Thesis IV (revised) Things agents do count as actions if they are done
intentionally, or if what they bring about results from something done
intentionally.
With thesis (IV) revised in this way, we are finally zeroing in on the essential
features of those happenings that would be regarded as actions. Indeed, all that
remains is to say what makes a doing intentional.
Note the heavy irony in my use of the phrase 'all that remains'. For one of
the most complicated topics in the philosophical discussion of action consists
precisely in saying what makes an act intentional. However, by sticking to the
most general considerations, it should be possible to say something un con-
troversial about all this. And given my ultimate aim of showing that our
ordinary thought about action tends towards certain incoherencies, it is crucial
that whatever is said be uncontroversially ordinary.
When Harry broke the Hunter and Dog, he reached for the neighbouring fig-
urine intentionally. Another way to say almost the same thing is to say that he
was minded to reach for the relevant object, and this quaint way of putting the
matter can tell us quite a bit. For in saying this, we are indicating that there
were features of Harry's mental condition directly implicated in the reaching -
features that made the reaching intentional. Which features? Well, in the
broadest sense, one could say that Harry believed certain things:
there was a reachable figurine in front of him,
it belonged to the collection of his friend,
his handling it in a respectful way would be appreciated,
etc.
and that he wanted certain things:
to put his hand on the reachable object,
to show interest in his friend's collection,
etc.
and finally that it was the combination of these beliefs and wants that con-
stituted something we would call Harry's reason for reaching in the way he did
(see REASONS AND CAUSES, RATIONALITY). Then all we have to add is that having
some such reason is what makes it true that Harry was minded to reach, that
his reaching was intentional. That much is surely uncontroversial, though
things get sticky when we try to be more specific about the connection
between having a reason and acting intentionally. Some are tempted to say
that the reason brings about a further mental condition - Harry's having an
INTENTION - and it is the presence of this that makes the reaching intentional.
65
AN ESSA Y ON MIND
Others say that there is no such further mental condition. and that having the
appropriately active reason is itself what constitutes the intentionalness of the
reaching. Moreover. there are long and contested stories told about the way
reasons 'bring about' or show themselves to 'be active'. (Some discussion of
this will figure in stage 3.) But since all we care about here is finding those
features that determine which kind of happening is an action. it is perhaps for-
givable if we pass over these disputes. For everyone accepts that if a happening
can be traced back to a reason. and if that reason suffices to make the happen-
ing intentional. then we have a case of action.
Just before we move on. I should tie up one loose end. Earlier I said that
being minded to do something would probably not be considered equivalent to
doing that thing intentionally. and pointing out why this is so will remove a
doubt about thesis (IV) that may have occurred to you. It has to do with the
question of our awareness of our reasons. There are of course times when we
ponder our next course of action. when we call to mind the beliefs and wants
that finally figure in our reason for doing this or that. (Think here of deciding
whether to make lunch now.) And there are other times when we act - and do
so for a reason - but when the reason is not one we have contemplated in
advance of acting. (Think here of suddenly squeezing the brakes on your
bicycle when a pedestrian carelessly strays in your path.) It makes perfect sense
to say that you were minded to make lunch. though it is doubtful that you
were minded to bring your bicycle to a sudden stop. Yet both are clearly cases
in which you acted intentionally. and both involve your having reasons. It is
just that in the bicycle case. the reason you had to squeeze the brakes was not
itself the subject of your deliberative awareness. Rather. it was something
plausibly attributable to you on the basis of attitudes and propensities to act
that were (one might say) silently at work as you were bicycling along. So.
when thesis (IV) speaks of the actions as intentional and when we connect
'intentional' with the having of reasons. we are not thereby committing our-
selves to any very fancy story about the explicitness or deliberative awareness
of reasons.
2.3.2 Identity and Individuation
As preparation for taking our next step. here are a few words - and a picture -
of where we have got to. Figure 2 is intended as a representation of the cumu-
lative wisdom of the four theses about ACTION. As the concentric circles show.
the successive restrictions we put on happenings have allowed us to narrow
our focus until we can be fairly sure that the innermost circle contains actions.
And being sure in this way that we have captured actions. we can now get to
work on saying more precisely what they are. Since this last thought may
come as something of a surprise to those who think we have been doing just
that. let me amplify it a bit.
The contribution of theses (I)-(IV) was in helping us to locate actions by
saying where to find them in our larger conception of the world. And what
we have said in thesis (I) is that the place to look is at a certain range of
66
Happenings
Doings (Happenings of agents)
0
\
Intentional dOings
(Happenings of agents with reasons)
Figure 2.
happenings, namely those involving the mind in roughly the ways outlined by
theses (II}-(IV). For example, we are now sure that Harry's reaching was an
action because it was a happening suitably involving something we called a
reason. However, though this helps us to recognize the reaching as an action,
it doesn't tell us what a specific reaching is, and how it differs from, say, a
grasping or pointing. One way to put this would be to say that the story so
far helps us to identify actions, but not to individuate them. Here an analogy
might help: you can identify something as a Beethoven symphony by saying
who composed it, but this does not tell you what a specific Beethoven symph-
ony is - what individuates it. It does not settle disputes about whether it is a
pattern of printed notes in a score authorized by Beethoven, or is a per-
formance according to that score, or is a pattern of imagined sound in Beet-
hoven's mind, or ... Similarly, in saying that actions are happenings located
by means of agents' reasons, we have gone quite far in identifying those
happenings that are actions without yet saying what the individual nature of
an action is.
Of course, in saying what you need to be able to identify something as an
action (or as a Beethoven symphony) you have to assume that your audience
has some inkling of what specific thing you are talking about. How else could
anyone understand the story about identification? So, perfectly reasonably, dis-
cussion of the identification of actions (or symphonies) has proceeded on the
assumption that you have some grasp of the individual nature of an action (or
symphony). The trouble is that in the case of actions (and here I think also in
the case of symphonies) the inklings many people have are not sufficient to
settle questions about individuation. For, as has been remarked, one gets
nothing but dispute when trying to decide, for example, whether a specific act
of reaching is a mental item or is merely the physical result of something
mental. And yet this dispute can take place against the background of complete
agreement in respect of theses (I)-(IV).
67
AN ESSAY ON MIND
2.3.3 Once upon a Time
Where does this leave us? Well, the obvious thing to do is to set to work un-
covering theses that bear directly on the issue of individuation of action. In this
way, we might end up seeing what the real ground of disagreement is, and
whether it can be resolved. Of course, this route has risks. For my main goal is
to show that there are problems arising from uncontroversial thoughts that
people have about action. If I were to begin now suggesting theses likely to be
either technical, disputable (or both), this would ruin any chance of reaching
that goal. Fortunately, we are not going to have to run that risk. As thesis (I)
has it, actions are happenings. And though this perfectly ordinary thought is
hardly rich enough to take us all the way to an individuating conception of an
action, it will serve my nefarious purpose suprisingly well. For what I suggest is
that a little further reflection on thesis (I) will show just what a mess we get
into when we think about actions.
Bypassing all the complicated things we could say about happenings and
actions, the following certainly seems beyond dispute:
Thesis (V) When something happens, there is a time at which it happens.
This thesis is not meant to be particularly demanding. Certainly, it isn't meant
to suggest that we can always be very precise in our saying when something
happens. If you ask someone 'when was the Battle of Hastings?', you are most
likely to be told '1066'. Though historians can no doubt do better than this,
even they would be hard pressed to say exactly on which days and at what
hours it began and ended. All that is fine as far as thesis (V) is concerned:
battles are just too complicated to allow timing by stopwatches. Moreover, the
thesis does not even demand that we should always be able to say even
approximately when a happening takes place. There are things that go on in
the world when we are simply not around.
Nonetheless, even bearing in mind these qualifications, thesis (V) does assert
something substantial. It reveals our commitment to the idea that happenings
have locations in the temporal order - that it just wouldn't make sense to think
that there could be a happening that didn't take place during some more or less
determinate stretch of time. Moreover, if the happening is one we actually
witness, then, within relevant limits of precision, we ought to be able to say when
it began and when it ended. Finally, since actions fall into the class of happenings,
it should always make sense to ask the question: 'when did someone inten-
tionally do such-and-such?' Moreover, since actions, unlike 'mere' happenings,
always involve an agent, this question should always be answerable. It just
doesn't make sense to think that someone could do something intentionally, but
that there would be no one around to count as a witness. Agents are witnesses.
Given all of this, what I propose is to consider some examples just to see how
we fare in placing human actions in the temporal order. Somewhat surpris-
ingly, it turns out that we do not fare very well. Moreover, the problem is
deeper than it might at first seem - or so I shall argue.
68
2.3.4 Moving. Breaking and Claudius
2.3.4.1 Moving Just to get us off on the right track. I shall begin with an
example that doesn't seem problematic. On the table in front of me is a bowl.
By bringing my hand into contact with it, and exerting some small, even pres-
sure, I have managed to displace it several inches to the left. What have I
done? I have moved the bowl to the left. When did I begin to do this? At that
very time when it began to make its way across the table. And when did I
finish this act? When the bowl came to rest. Nothing seems difficult here: the
happening that is my action can be located (within reasonable limits of preci-
sion) in the temporal order. My intentionally displacing the bowl would seem
to coincide in time with the observable displacement of the bowl.
2.3.4.2 Breaking The second case raises more questions. Arm outstretched, I
hold up a piece of chalk and then relax the grip of my fingers. The chalk falls
to the floor and breaks. What I did - quite intentionally - was to break the
chalk. When did this begin? Well, it is certainly tempting to think that I began
the episode of breaking when I released my hold on the chalk. But when did I
finish breaking it? At the moment when it began to slip through my fingers?
This doesn't seem right. since the chalk is at that stage unbroken. At the
moment when it hit the floor? This seems more like it, but there is a problem.
For having released my grip, there is a sense in which I didn't do anything at
all during the time it took to reach the floor; it is only a short time, but it is
not unreasonable to think that I was inactive during it. So, it may seem odd to
count that time as included in my action. Unlike the case in which I moved the
bowl, the intended effect of my action seems not to coincide temporally with
my being active. Still. since the time between my releasing my grip and the
chalk's breaking is very small, we might get away with saying that I broke the
chalk at 10 minutes past the hour, and leave it at that. After all, I did say
earlier that a high order of precision in our timing of actions is not required by
thesis (V).
Or we could take another tack - one less evasive. We could say that, strictly
speaking. the action was nothing more nor less than the releasing of my grip.
This happened at a particular moment in time. yet, as with many things we
do, it had further consequences, among which was the breaking of the chalk.
Knowing of this consequence, anyone who was a witness could redescribe the
action - the releasing - as a breaking, even though the action itself ended
before the breaking began. We go in for this kind of redescription all the time
with particular concrete objects. Thus. we can describe one and the same
object as 'the blue car in the drive' or as 'the car that stranded me on the
motorway'. Same object but two different descriptions, though the second of
these is only apt if certain facts about the car are known. (Before I set out for
the motorway, it would not have been possible for me to describe it in terms of
the stranding.) Similarly, the releasing of my grip - that very particular hap-
pening - can be described in many different ways, one of which becomes avail-
able only after the chalk reaches the floor and breaks. That particular releasing
69
AN ESSA Y ON MIND
was a breaking. This is something we come to know after it happens, and all
will be well so long as we don't allow ourselves to think about other releasings
which could have taken place, but didn't, These other releasings might have
misfired; one of them might have involved a carpet that cushioned the impact
and the chalk might not have broken. But these other non-destructive releas-
ings are not what took place.
Note that this way of dealing with the breaking was not apparently called
for in the bowl-moving episode because the movement of the bowl and the
action of moving the bowl were simultaneous. Of course, once we look at the
bowl-moving case in this new way, we are committing ourselves to a distinc-
tion between some effect (the bowl's moving) and some action (moving the
bowl), even though the effect and the action are simultaneous. Keep this in
mind as it will be important later on.
2.3.4.3 Claudius The following imagined case is meant to test the various
suggestions made in respect of the chalk-breaking. Changing the well-known
(though already fictional) story, consider the case of Claudius and the King.
Claudius wants the King out of the way, but he wants to avoid any chance of
being caught. So, he acquires a very slow-acting poison which he administers
by pouring it into the ear of the sleeping King. Having administered the poison,
he retires to his own castle and awaits the outcome. Sure enough, after a
period of some six months, the King succumbs to the effects of the poison and
dies. As he expected, Claudius appears to members of the court to be in the
clear: he has been miles away from the King for most of the six months. Yet
we who are in the know have no doubt in saying: Claudius killed the King.
Yes, but when? Well, Claudius moved his fingers and hand in such a way as to
empty the vial into the King's ear. Suppose this happened at 3 pm on the 5th
of May 1360. It then seems perfectly plausible to regard this as what we might
call the 'start time' of the deed. But a start time is not quite enough to get the
happening properly settled into the temporal order. For we need to know when
the killing ended, and many would be reluctant to say that Claudius killed the
King on the 5th of May. After all, the King didn't die until the following
November.
Someone might be tempted to say that the murder was a rather longish hap-
pening which began on the 5th of May and ended in November. The idea here
would be that happenings can have parts and that the pouring was the first
part and the dying the second. This would be like the move made earlier which
treated the chalk-breaking as beginning with the releasing and ending with the
breaking. However, though it just about sounded alright in that case, the
length of time involved in the murder makes this suggestion much less appeal-
ing. What seemed odd before was that someone should describe me as engaged
in breaking the chalk - being active in this way - when I was standing there
inactive, having already made my contribution to the activity by releasing my
grip. Whilst it is easy to think this was just picky in respect of the breaking, it
is a serious difficulty in the case of the murder. Do we really want to say that
70
what Claudius did happened from May to November? Suppose you see him
sitting peacefully in his castle in September reading a book: do you want to be
committed to saying that he is then both reading a book and murdering the
King? And what if, as is possible, he had died before the King? Posthumous
publication is one thing, but would we really countenance posthumous mur-
dering (other than supernaturally)?
What about the second suggestion? Why not treat Claudius' activity as
beginning and ending on the 5th of May when he poured the poison? The
death was of course a consequence of this pouring, but is not itself part of the
act. It is just that in November, when the King dies, we then have available a
new way of describing the original pouring: we can now say it was a killing,
indeed a murder.
As before this way of treating the matter makes more sense in connection
with the chalk-breaking than it does in respect of the murder. For, whereas it
may be just about acceptable to say that I broke the chalk when I released it, it
seems odd to say that Claudius murdered the King on the 5th of May. This is
because there seem to be so many things that must happen before the death
occurs, and so many ways it could go other than as Claudius had hoped. Of
course, we can stick to our guns here and say: 'Look, that very action - the
pouring - had the death as a consequence. If the death had not occurred, then
it would have been a different pouring of poison from the one that actually
took place.' As long as you keep thinking of what Claudius did as a wholly
specific thing which had the consequences it did have, then there is no harm
in saying that he murdered the King on the 5th of May. It is just that this
description of Claudius' act only becomes available in November.
It should be obvious that this kind of 'sticking to one's guns' brings with it a
conception of action that goes beyond what is contained in theses (I)-(IV). We
have now to think of each action as some kind of wholly particular thing
which has its causes and consequences tied to it in some unbreakable way. If,
even in thought, you change something about the consequences of an act, y ~ u
are then barred from speaking about the very same action. For example, in
connection with the present discussion, one might be tempted to appeal to a
case of 'Claudius Interruptus': Claudius pours the poison in just the way that
he did, but the King rolls over, or he later takes some substance which has the
effect of an antidote, or his doctor figures out how to cure him. However, on
the view we are considering, the very fact that the King does not die means
that Claudius' act on 5th May is not the same act as the one which began our
considerations - for that act resulted in the King's death.
Whether this conception of action is ultimately convincing enough to justify
the claim that Claudius murdered the King on the 5th of May is, I would
suggest, unclear. But remember that it or some other view is necessary to help
us answer the question: when did Claudius murder the King? For no matter
what deeper view we take of action, we are committed to the idea that actions
have temporal locations, and we are certainly having trouble locating this
dastardly deed. So, given that the view under consideration has less trouble-
71
AN ESSA Y ON MIND
some consequences than any other - for surely we do not want to say that the
act took Claudius six months to complete - let us continue our discussion
assuming that it is basically correct. It does commit us to saying some odd
things, but maybe this is just the price to pay for having something to say
about when the act took place.
2.3.4.4 Moving Reconsidered We have settled for what can be called the 'act/
effect' view. That is, it requires that we distinguish an agent's act from the
effects that follow such acts even when these effects are responsible for the descrip-
tions we give of the original actions. Thus, 'breaking the chalk' describes some-
thing that I did, but this is only appropriate because of the destructive effect of
my loosening my grip. Indeed, according to the act/effect view, one should
expect there to be lots of different descriptions of what will in fact be one and
the same action: loosening my grip, releasing the chalk, breaking the chalk,
making a mess on the floor, making a point to.a lecture audience, etc. And
since all of these descriptions point to the same bit of my activity (my loosening
of my fingers), we might as well distinguish this activity by calling it my basic
action. In the light of the distinctions between basic acts and descriptions, it is
worth reconsidering our first example, the case of moving the bowl.
When we discussed this example, it seemed to cause us no trouble at all.
And now we ought to be able to see why. The timing of my action in that case
coincided perfectly with the effect it brought about: my moving the bowl (my
basic action) and the bowl's moving (the effect of my act) took place simul-
taneously. So, of course we had no trouble in saying when I acted. Or so it can
seem. What we have to do now is to see how, if at all, the act/effect view
applies to basic actions. Don't be surprised about this. We have been assuming
that we know when these take place - that we know when I move my hand
whilst it contacts the rim of the bowl, that we know when I release the grip of
my fingers on the chalk, and finally that we know when Claudius does what is
appropriate to inverting the vial over the King's ear. But it is now important to
see what is going on 'inside' basic actions. And for this purpose, the bowl-
moving episode will serve as the example.
The first thing to do is to forget the bowl and its movement. We have agreed
that this is an effect of what I did, and our sole interest now is in what I did.
Well, the obvious thing to say here is that I moved my hand - that was my
basic action. However, let us look at this more closely. Focus on the claim that
I moved my hand. Can we see this - as per the act/effect view - as a complex
consisting of some action and some effect? Admittedly it is a very small
complex, but even so it does seem that there are two things discernible: one is
that my hand moved and the other is that I moved it. And the first of these -
the movement of my hand - could be seen as the effect of whatever it was that
I did. On the act/effect view, it is perfectly reasonable to describe my action in
terms of its effects, and in the present case we say 'I moved my hand' just
because whatever it was that I did produced the hand movement as an effect.
But what then did I do? And when? Neither of these questions is going to be
72
easy to answer. In answering the first question, we need to find something that
I did that can be described without mentioning its effect (the movement of my
hand). But there doesn't seem to be any candidate description for this. Perhaps
the best we can say is: I set myself to move my hand, and it was this that
brought about the movement. And when did I do this? Well, precise temporal
location isn't crucial, but, whenever it was, it certainly must have preceded the
movement of my hand. Are these answers satisfactory?
Setting myself to move my hand at a time before it moves must be some-
thing that happens, as it were, 'inside' me; it would seem to be some kind of
mental phenomenon (see THE WILL). We could even imagine a case where this
mental activity took place without its bodily effect. For example, suppose that
some demon psychologist gave you a drug which paralysed your hand and
arm, but did so in such a way as to leave you ignorant of this fact. She then
blindfolds you and asks you to move your hand as if you were pushing a bowl
across a table. What do you think you would do? Well, as far as anyone
watching you was concerned, you would do nothing. But how would it appear
to you? Well, you would certainly be able to set yourself to move your hand,
and, if the drug really did leave you ignorant of your state, it would appear to
you as if you were doing that very thing.
Taking apart basic actions in this way seems to have at least this advantage:
it allows us to locate the onset of an action in the temporal order. Setting
yourself to move your hand - what has been called 'willing' your hand to
move - begins in some small time interval before your hand actually moves.
But when does it end? Does it continue even after the hand begins to move? Or
are we to think of this act of will as somehow imparting its energy to the hand
even though it ceases to work when the hand starts to move? It certainly
doesn't feel that way in my own case: when I move my hand in the way
required to move a bowl, it seems to me as if I am guiding my hand - as if my
activity continues throughout. I can of course imagine what it would be like to
have my hand move after just one, so to speak, push, but that is not how it
seems in the normal case.
These questions are perhaps more difficult than the ones we asked of the
chalk-breaking and of Claudius' felony. Moreover, even if we could answer
them, we are still left with something very troubling. For, if we end up saying
that an action is something like setting oneself to do something, or exercise of
will, then, as was noted, this can take place without anyone seeing it. In the
normal case, it will have effects on the body, and we will see these, but the act
itself will escape the scrutiny of everyone but the agent. And do we really want
to treat an action as subject to all the problems of the first-/third-person divide
that were discussed in relation to experience?
2.3.5 Conclusion
We began with a few harmless theses about action and one thought about
happenings in general. However, several straightforward examples later, we
find ourselves in a bit of trouble. The theses told us that actions could be
73
AN ESSA Y ON MIND
identified as happenings with the appropriate mental pedigrees, and, further,
that as happenings, they ought to fit into the temporal order. But they don't fit
- or at least not without great discomfort. And even stranger, we have found
at least some reason to think that when we are genuinely active, the crucial
ingredient in that activity is rather like an item of experience - something that
raises the possibility that we will have some of the same problems with action
as we did with pains.
This last result will no doubt appeal to those who insist that actions are
mental phenomena. They might even take this as a victory over those who had
maintained that actions were at most the visible signs of the mind. But things
are not that simple. Recall that it was our need to assign a time to my break-
ing of the chalk that led to the doctrine of act/effect, and this doctrine seemed
the only sensible thing to say about the case of Claudius and the King. For in
that case the alternatives seemed to be: either say that the murder took six
months or say that we cannot properly fit the murder into the temporal order.
Admittedly, even the doctrine of act/effect has us saying that Claudius mur-
dered the King six months before he died, but at least we saw the beginning of
a more complex story about actions which might have made this palatable. Yet
when we brought this apparently most plausible option to bear on the simplest
of actions - the movings of our bodies which seem to be at the start of every-
thing else we do - things seemed to dissolve in our hands. Instead of some
well-behaved thing called a basic action, we ended up with an effect (a move-
ment of the body) and a mental item (an exercise of will) neither of which meet
all the criteria we expect in cases of action. The one is not really an action, just
an effect, and the other is active, but it is not observable from an outsider's
point of view except through its effects. In sum, it has turned out that when
we try to say when actions happen, they tend to disappear from view. And
even those who insist that actions are mental would not have been happy to
think they were that elusive.
There are of course things we can do. But all of them involve a great deal
more philosophy than one would have expected when we began. For example,
we could just insist that an action is a pair consisting of an exercise of will and
an effect on the body. Thus, we could simply refuse to count setting oneself to
move a hand as an action unless it was successful - unless the hand moved. In
this way, we could date the beginning of the action with the exercise of will
and its end-point with the visible effect on the body. But we would then have
to refuse to allow temporal assignment of this sort to any but basic actions, on
pain of having to say that Claudius took six months over killing the King. Or
we could bite the bullet and say that an action really doesn't belong in the
physical realm, and that too much was being read into the idea that actions
are happenings. Or, diametrically opposed to this, that much of the stuff about
the mental pedigree of actions is misguided, and that actions are just bodily
movements which are easily located in time.
These and many other options are all beyond my remit. For what they have
in common is that they force us in one way or another to revise some of the
74
things we ordinarily think true of actions, and my aim at stage 2 is restricted
to investigating the consequences of ordinary thought. But just as something
must be done about the attitudes and experience, it should be no less clear that
this most important feature of the mind - its acting in and on the world - is a
case for conceptual treatment.
75
Stage 3: Bedrock
3.1 Introduction
There is an enormous temptation to regard anything that happens as open to
explanation in scientific terms. As with all temptations we can either steel our-
selves to resist or give in. In the present case, I recommend giving in - though
in a considered way - and I am supported in this recommendation by most
writers in science, philosophy and related disciplines. Nonetheless, since so
much of what goes on in contemporary philosophy of mind depends on it, I
must say something about the background to this particular case of succumb-
ing to temptation.
The present scientific picture of everyday changes and happenings is very
impressive. Physical, chemical and biological phenomena which were once
deeply mysterious are now routinely discussed and explained in an array of
specialist journals and books. Indeed, the scientific enterprise has been so suc-
cessful that it is now a commonplace to say that anyone person cannot fully
grasp the details of all its sub-disciplines. But in outline the picture is clear
enough: the world seems to consist of such things as particles, atoms and
molecules governed by laws, and it is this orderly and increasingly complex
arrangement of energetic matter that gives shape to the world we experience.
One science - physics - studies the configurations of energetic matter at the
most basic level, and it is for this reason that the scientific world view is often
called 'physicalism'. Other sciences - from chemistry through biology - attempt
to unravel the laws that govern more complex configurations up to, and
including, the organisms that have populated this planet.
Of course, many mysteries remain - there are many phenomena that are not
explicable by the current physicalist picture of the world. And it this large
reservoir of ignorance that might well encourage someone to reject the claim of
this picture to any sort of comprehensiveness. This rejectionist line can seem
particularly attractive, accustomed as we now are to be suspicious of scientific
experts. Yet, in spite of its appeal, I would urge that taking the rejectionist line
is much less rational than acquiescing in the scientific world-view. To under-
stand why, some distinctions must be made between what actually comes with
the scientific picture and what is inessential to it.
First of all, there is the distinction between the physicalist view of the world
that goes with science, and the much more specific - and doubtful - view that
all of science itself is essentially reducible to that part of itself known as physics
76
STAGE 3: BEDROCK
(see REDUCTION). It is unfortunate that the commonly used word 'physicalism'
hints at the latter idea - the idea that everything is reductively explicable by
physics - but it isn't difficult to keep these things separate. Physics is the study
of the laws governing such things as elementary particles and atoms, forces
and fields. There may be every reason to think that the world is constructed
out of such things, but one can think this without also thinking that such
sciences as biology, zoology, chemistry, meteorology are themselves merely
branches of physics - that there will be laws of physics that tell us everything
about, for example, biological phenomena. All that the physicalist perspective
requires is that whatever happens will be explicable by some branch of science
or other; it does not require that all these specific scientific enterprises are
themselves reducible to physics. How we should deal with the 'branches' of
science will then be a further question, and not one on which we have now to
take any stand. (Some prefer to speak of 'naturalism' or 'materialism' in place
of physicalism here, but these labels can also be misleading. See PHYSICALISM;
LEWIS.) ,
Talking in this general way about what some science may come to explain
introduces a second distinction: that between the sciences we now have and
the idea of science itself. At present we have a general idea of what physics,
biology, chemistry and the other sciences say the world is like. But in adopting
the physicalist or scientific perspective, we are not committing ourselves to
believing that the present picture is correct. All we are claiming is that any
phenomenon that is a genuine happening in this world is in principle explic-
able by a science, albeit by a science that might be quite different from any we
now have at our disposal. Though it may be difficult to imagine, our present
scientific perspective may be deeply misconceived. But all that the scientific or
physicalist perspective requires is that the methods of science - not the laws or
theories we now have - can in principle provide an explanation for whatever
happens.
Finally, one must be careful not to misunderstand what it means to say that
any happening must be explicable by science. For it should not be taken - as it
so often is - to mean that what science has to say about a certain kind of hap-
pening is all that there is to say about it. The colours, shapes and images of a
painting can be the subject of our aesthetic judgment, without this challenging,
or being challenged, by the fact that science can explain everything about how
the surface molecules and incident light produce (in us) the painting's appear-
ance. Indeed, one may even find it possible to believe that there is a scientific
explanation of the aesthetic judgments themselves, without this undermining
their truth (or otherwise). In short, the universality of science should not be
mistaken for a kind of tyranny.
Keeping in mind these three points, the thought that all phenomena
(including the mental) must be in principle explicable by some science or other
shouldn't be all that controversial. Whilst nothing I could say would constitute
an air-tight argument in favour of it, I hope I have said enough to justify our
proceeding as if the scientific perspective were true. To do the opposite would
77
AN ESSA Y ON MIND
certainly be rash, as we would have to believe now that some phenomena - say
those of the mind - were not explicable by any present or future science. This
last is certainly a possibility - one that we may want to return to when the
going gets tough. But it is not sensible to start with it. (But see PHYSICALISM (2)
for a less optimistic assessment.)
Why should the going get tough in respect of the scientific understanding of
the mind? Answering this question will take a little time, but let us begin with
the brighter side of the picture. The mental realm consists of a quite hetero-
geneous collection of things which share at least this feature: they are, in the
broadest sense, things that happen. As you read this, as the light waves from
the page strike your eye, you undergo various conscious experiences and,
equally likely, you come to form beliefs and other attitudes, as well as going in
for actions such as turning pages. The question of how to understand all this
scientifically is thus no less pertinent here than it would be in respect of purely
chemical, atomic or biological changes.
Put in this way, there seems to be no especially pressing reason for investi-
gating the scientific basis of the mental. There are these phenomena - the
mental - and we have just as much reason to expect that we will understand
them in scientific terms as we have in regard to the weather or the behaviour
of proteins. Of course, if some special need arises - as when a major mind-
affecting drug is tested - then there is bound to be some more urgent interest
in the various mental 'side-effects' that follow administration of the drug. But
this kind of interest is much like the interest we all take in, say, seismology
when earthquakes threaten large population centres; where human concerns
figure prominently, the need to understand precisely why things happen - the
need for scientific understanding - moves up the list of priorities.
The idea that the mental realm sits on the bedrock of the physical world as
described by science, and that we shall gradually, and in our own good time,
come to understand more about the contours of this fit, is by and large the
common view in the scientific community (see also SEARLE). However, this relat-
ively sanguine attitude about the relationship between the mental and physical is
not typical of those in the philosophical community. Whereas the representative
textbooks on neurophysiology tend to suggest that we will one day come to know
more and more about the mind's relation to the brain, philosophical discussion of
this issue shows no such equanimity (see PAIN). Indeed, it is only a mild exag-
geration to say that what one finds instead is an urgent, even frenetic, search for
some way to reconcile the mental and the physical; there is even a hint in some
quarters of despair that the project will never get anywhere. Of course, even
though I use terms such as 'frenetic' and 'despairing', I do not for a moment
mean to suggest that the philosophical attitude is merely pathological - that,
having opted for the idea that the physical is the bedrock reality of things that
happen, philosophers have given in to a collective neurotic anxiety about the
present impoverished state of our knowledge in that regard. On the contrary, it
seems to me that the philosophical attitude is, if anything, more justified than
that of the neurophysiologists, and thus the most immediate task is to show why.
78
STAGE 3: BEDROCK
One possible reason for the philosophical attitude arises from our rumina-
tions in stage 2. The everyday notions of the mental are shot through with
problems. The simplest of questions and assumptions lead to difficulties which
can make us wonder whether we so much as have a coherent conception of
the mental realm. The loose behaviour of the attitudes, the elusiveness of
actions, the apparent incompatibilities of the first- and third-person perspect-
ives, these problems are only samples, and yet they might easily provoke
someone to worry about the possibility of ever having a proper science of the
mental. After all, if we do not have a firm grasp of the mental happenings
themselves how can we ever hope to understand their physical basis?
This last question seems calculated to invite the answer: 'we can't', but this
would be too hasty. Granted that success in scientific understanding is made
easier by our having a coherent grasp of the relevant phenomena, it does not
follow that the lack of such a grasp renders the search for a physical basis
impossible. For example, though the seventeenth-century account of what hap-
pened when things burned was largely incoherent, this did not prevent pro-
gress in the next century, and the eventual development of our now-accepted
ideas about combustion and oxidation. So, whilst it would be nice if our con-
ception of the mental was less problematic, this would not be a good enough
reason for the widespread malaise in the philosophical community.
3.2 Reasons and Causes
The real reason for philosophical anxieties about the relations between the
mental and the physical can be best brought out by backtracking a bit, i.e. by
considering an issue which I touched on briefly in stage 2. I have in mind here
the question of how acting is related to experiencing and attitudinizing. Con-
sider the following mundane example.
You are sitting on the sofa engrossed in reading. Whilst so doing you
begin to be aware of rumblings in your stomach - they are almost audible
- and a general feeling of emptiness. Could you be suffering from some
sort of indigestion? You look at your watch and realize that you have lost
track of time: it is now nearly 3 o'clock and you haven't had lunch. You
decide that you are more likely to be suffering from lack of something to
digest. Having been shopping the night before you know what is available,
and you mentally run through the items. Believing that a sandwich would
be the quickest thing to fix (though not necessarily the most tasty), you
recognize that your desire for food soon is much stronger than for food
quality. So, intent upon the idea of making a sandwich, you head to the
kitchen and open the door of the refrigerator ...
Clearly, as the dots show, the mental episodes reported in the example are
followed by lots of activity. But for specificity let us focus on the action of
79
AN ESSA Y ON MIND
reaching out and grasping the refrigerator door. This simple action is brought
about in some way or other by the bodily feelings. beliefs. desires and inten-
tions which you had in the minutes leading up to the action. It is clear
enough that items of our first two categories - experiencing and attitudinizing
- are in some way responsible for the eventual door-grasping behaviour. But
in what way? That is. in what way is that amalgam of attitudes and experi-
ences which could be called your 'mental condition' responsible for your sub-
sequent action?
Two somewhat different sounding answers have been proposed. On the one
hand. it would generally be held that your mental condition - or at least some
element of it - was the reason for your opening the door. That is. if asked why
you opened the door. you would make some more or less complicated reference
to your beliefs. desires and feelings. saying that they provided your reason for
so acting. On the other hand. it might be said (perhaps more naturally by an
observer than by you) that your mental condition was the cause of your acting.
Here the idea is that. just as we explain an earthquake by appeal to factors
that cause it. so can we explain your fridge-directed behaviour by saying that
your mental state caused it. And it is fair to say that these alternatives exhaust
the field: the philosophical community seems agreed that minds rationalize -
contain reasons for - actions. or they cause them.
Much effort has been expended in clarifying and defending these two possible
explanatory stories. Attempts to understand what constitutes a reason for
action began at least with Aristotle. and there is perhaps no more discussed
notion in philosophy than that of a cause. However. for the present line of
argument we do not need to spend a lot of time considering all of this back-
ground. For setting the stage to our problem. all we need to recognize is that
these two answers are not necessarily incompatible: someone could hold (and
many do) that a reason for an action can also be a cause of it.
Briefly. the story goes like this. A reason for an action is some set of attitudes
and feelings which together show the action in a favourable light to the agent.
Your mental condition in the above example made opening the refrigerator
door an intelligible first step in the sequence of actions which followed.
However. reasons tend to be a bit unspecific. Your reason for opening the door
by grasping the handle would have equally made intelligible a whole host of
slightly different movements. A reason doesn't explain precisely how our bodies
move in acting on it. And this is where the idea of a cause comes in. For our
original question was: what was the relationship between your mental con-
dition and the specific action of opening the door that took place? A not
implausible first reply is that your mental condition makes it reasonable or
rational (from your point of view) for you to go in for some sort of refrigerator
door-opening behaviour. But the full explanation of why you opened the door
when and in the way you did must make reference to the cause of that very
specific event. (See REASONS AND CAUSES.)
Bringing in causes helps. as they are the kinds of thing we appeal to when
we want to explain why some specific thing happened. For example. suppose
80
STAGE 3: BEDROCK
we are investigating the burning down of the house on 23 Elm Street. We
know that faulty wiring can result in houses burning down. This is a general
kind of truth which might well be in the background along with lots of other
general truths (such as that gasoline and matches, when mixed with insur-
ance policies, also bring about house burning). Knowing these sorts of thing
helps, but in order to explain this particular burning down we must locate
that actual state or event that preceded it and was, as one says, 'the' cause.
Supposing it was in fact a loose wire in the kitchen wall that short-circuited
the electrical supply, then the explanatory job is finished only if we can find
evidence of this loose wire. If we can, we will have established a causal chain
one could almost visualize: the loose wire in the wall touches another, there is
no fuse to break the circuit, gradually heat builds up in the inner wall, com-
bustion point is reached, the fire spreads in the wall, no one is home to notice
the charring, the timber frames ignite . . . the house on Elm Street burns
down.
Applying this to the refrigerator example, what we need is to find the cause
for the quite specific door-reaching/grasping behaviour that took place. Now
your mental condition functions to explain rationally why you went in for the
sort of behaviour you did - it is (or contains) your reason for having so acted.
But it doesn't by itself explain the particular door-grasping that occurred. So
why not, as it were, kill two birds with one stone and recognize that the reason
could itself also be the cause of the action? Philosophers have spent a lot of
time considering what is involved in one thing's being the cause of something
else, and at least the following three conditions are generally thought to be
necessary:
1. A cause is some state, event or datable/placeable happening.
2. It precedes its effect.
3. It is such that if it hadn't happened then neither would the effect. (Since in
a given case both the cause and effect have already happened, this last
condition is 'counterfactual' - it says what would happen if things had
been different. It thus places a very strong bond between cause and effect.
Note too that some would argue that there must be laws of nature ground-
ing these counterfactual claims. See ANOMALOUS MONISM; DAVIDSON.)
Unsurprisingly, these three conditions seem to be met by the very thing that is
your reason for opening the fridge door. Your reason is part of your mental
condition - a series of states and events which take place just before you open
the door. This shows that conditions (1) and (2) apply. Moreover, it doesn't
seem implausible to say that if you hadn't been in that condition you wouldn't
have opened the refrigerator door. And this is a way of saying that condition
(3) applies as well. So, the very thing that is a reason for an action also seems
to have the characteristics necessary to make it a cause of an action. At least
that will be the hypothesis which we shall adopt for our present expository
purpose, and which we can represent by figure 3 below.
81
AN ESSA Y ON MIND
Figure 3.
3.3 Brains
Experiencel Attitudes
Being hungry, wanting a sandwich,
believing that the fixings are in the
fridge
~
cause and rationalize
Focus on the stage in our example where your arm is by your side but you are
just about to reach for the handle on the fridge door. No one now knows
everything about how our nervous systems work, but we do have some rea-
sonable idea. Lots of electrochemical activity in the brain, channelled down
appropriate neural pathways, causes the very complicated and yet delicately
coordinated contracting and expanding of muscles in your arm. The precise
nature of this contraction and expansion is continuously controlled by the
brain's electrochemical activity, and this, in turn, is partly dependent on the
neural activity in the optical system. To cut a long story short: your muscles -
acting on the brain's electrochemical commands - extend your fingers to the
fridge door handle, and all of this is achieved with the help of visual and other
systems which monitor the position of your body and arm.
What is important is that what has just been imagined is a causal story,
indeed it is what we might call a 'purely physical' causal story. Tracing back-
wards from contraction of your muscles which force your fingers around the
handle to the neural excitation in your arm, and brain, and optic nerve, and so
on, there is nothing here that anyone would regard as an activity of the mind.
A textbook on the nervous system has no place in it for what would be a
miraculous intervention of the mind in the causal chain just described. The
physiologist doesn't even try to understand neural excitation and muscle con-
traction in any but chemical and physical terms. One would be stunned to read
a paper in a journal of brain sciences which said: '. . . and just after this parti-
cular nerve sends its spiking pulse, the subject's thought intervenes to carry
the message to the muscles in the arm.'
Have we any reason to believe that the neurophysiological story about your
arm movements can be satisfactorily completed? After all, at present we know
only a lot of general things about what goes on. I pointed out earlier that it
may be an article of faith that all the happenings in this world are at bottom
82
STAGE 3: BEDROCK
explicable in scientific terms. But I think that when it comes to that part of the
world which is the human body, such faith couldn't be better grounded. For,
unlike our picture of the sub-atomic basis of matter and energy, our general
understanding of the biochemical basis of human biology is virtually complete.
There are many details missing from the picture - and the details are crucial
for such things as medical intervention - but there really do not seem to be the
large scale mysteries in biology that there are in physics.
In respect of the hand poised for opening the fridge door, the picture shown
in figure 4 seems to just about sum up the way things are expected to be,
Figure 4.
Actions
Brains
The neurophysiological and
neurochemical goings on inside
your brain, nervous system and
musculo-skeletal structure just
before your arm and hand move
up from their side.
/
cause
/
The upward and then outward
movement of the arm and hand
resulting in a force being exerted
on the fridge door ...
scientifically speaking. There are two things about the way the boxes are label-
led in figure 4 which should be mentioned. Firstly, I use the label 'Brains' as a
shorthand for the whole - and as yet unknown - physical story of what goes
on. Clearly, where the mind is concerned, the brain itself will be particularly
important, but I don't want you to think that I have forgotten about the rest of
the nervous system, the muscles or the hormones. And there may well be other
important elements to the story about the causation of action.
Secondly, anyone who followed my detailed story about actions in stage 2
will be somewhat surprised to see that the physical movement of the hand and
arm is what gets the label 'action'. Given that discussion, it is far from obvious
that the movement is itself an action; indeed, it may well be wrong to think
this. Nonetheless, there is reason to think that an action such as reaching for
the fridge door handle does include (or, less assertively, involves) such a move-
ment, and that is the only thing inttmded by the label.
83
AN ESSAY ON MIND
3.4 The Eternal Triangle
In matters of the heart, triangular relationships are celebrated for their diffi-
culties. Two persons competing for the love of a third give the writers of soap
operas all they need to get going: only the details need to be filled in to gen-
erate endless different plots. What needs to be appreciated now is that putting
figures 3 and 4 together (figure 5 below) we get a triangular relationship
which, though hardly the matter of a soap opera, has at least at first glance
something of the same tensions - and leads to no fewer variations of plot.
Experiencel Attitudes
Being hungry, wanting a sandwich,
believing that the fixings are in the
fridge
~
cause and rationalize
Brains
The neurophysiological and
neurochemical goings on inside
your brain, nervous system and
musculo-skeletal structures just
before your arm and hand move
up from their side.
/
cause
The upward and then outward
movement of the arm and hand
resulting in a force being exerted
on the fridge door ...
Reaching for
the fridge door
Figure 5.
In figure 5, the left-hand arrow represents the claims of the mind to be the
cause of what we do, whilst the right-hand arrow represents the claims of our
neurophysiological make-up. Now human relationships can be surprisingly
accommodating, but causal ones are not. It may really be possible for there to
be something like harmony in a love-triangle, but no such possibility exists in
respect of the causal claims shown in figure 5. Here is why.
Recall that among the very few things said about causality was this: some
event or state - call it A - is the cause of another (call it B) when it is true that
were A not to have happened then B wouldn't have happened either. This require-
ment can be true even though both A and B have in fact happened - remem-
ber that it is 'counterfactual'. The trouble is that the situation in figure 5
makes it impossible for this feature of causality to apply to either the left- or
right-hand side of the diagram. For example, suppose that it is really true that
you have some reason for an action which in fact causes it. Is it true that had
84
STAGE 3: BEDROCK
you not had this reason, this action would not have taken place? No, because
so long as your neurophysiological state endowed you with the appropriate dis-
position, you would have acted, whether or not you had the supposed reason.
Equally, it cannot be said of a case in which your neurophysiological state
causes you to move in the active way that if you had not been in that state,
you wouldn't have so acted. This is because figure 5 allows it to be possible for
a reason to have brought about some action even in the absence of the neuro-
physiological state that in fact did cause it.
The problem here is known as that of overdetermination. It is arguably a
feature of causality that effects cannot have two completely different causes
competing for their attention. The counterfactual intimacy described above
means that if there seem to be two completely independent causes, then one of
them must give up its claim on the effect or, contrary to the original supposi-
tion, the causes are not fully independent. Here is a stark example to illustrate
what I mean.
A house burns down, and the investigation unit of the Fire Department
comes up with the view that it was caused by a short circuit in the kitchen
wall. However, a neighbour has another idea. He believes that it was caused by
a 'person or persons unknown' with a grudge to settle. Now, there seem to be
three possibilities here: (i) the Fire Department are right about the short circuit;
(ii) the neighbour is right about the grudge; or (iii) they are both right in that
the grudge was settled by rigging the wiring so as to produce a short circuit.
(They could of course all be wrong, but that is not worth considering here.) In
possibility (iii), the causes are not independent - they are part of the same story
- so this is not a relevant case of overdetermination. This is sometimes de-
scribed by saying the causes are overlapping. Which leaves the other two. Is it
reasonable to think that both (i) and (ii) obtain even when they do not
overlap? That is, could the Fire Department and the neighbour both be right
about the cause even though the grudge and the short circuit are both sepa-
rate and independent causes of the fire? This would be a case of over-
determination, but can you really conceive of it whilst keeping your grip on the
idea of a cause? I suspect not. How can the short circuit have been the cause if
some second, and wholly independent, chain of events also brought about the
fire? And how could the cause be the grudge if there was also a short circuit,
which had nothing to do with the grudge, and which would have burned
down the house anyway? Clearly, if we are careful about insisting on the non-
overlapping condition, then overdetermination messes up the claims of both
sides to be the rightful cause.
To be sure, there are cases where we might be tempted to think that we
could rationally speak of overdetermination - cases where independent causes
seem to be separately responsible for a single effect. For example, suppose that
someone is very ill and suffers both a massive heart attack and respiratory
failure - events that are followed by death. We may want to say that had the
heart attack not killed the patient, then the respiratory failure would have -
and vice versa - even though we would also insist that these do not overlap in
85
AN ESSA Y ON MIND
the way the short circuit and grudge did in (iii) above. Is this a counterexample
to the earlier claim about the incoherence of overdetermination? Not really. For
even though the two conditions are different, there is nothing to prevent us
seeing them as part of a larger condition which is itself the cause of death.
Indeed, this is just how such cases are described when they occur: a doctor in
such circumstances would cite the cause of death as 'a massive heart attack
and respiratory failure'. Moreover, any attempt to press this verdict by asking
which was really the cause of death would not unreasonably be resisted. For it
would be downright misleading to single out one or other part of the larger
circumstance as the real cause. What this shows is that, faced with a putative
case of 'overdetermination, our intuitive understanding of causality forces us to
see the competing events as parts of a single complex cause. In effect, the very
idea we have of certain events as parts of larger ones is one we reach for
naturally when overdetermination threatens. And this reaction shows just how
deeply entrenched in our idea of causality is the rejection of genuine over-
determination.
On the face of it, the situation in figure 5 is one where non-overlapping
causes compete to bring about an effect. Since overdetermination is simply not
reasonable, the situation in figure 5 is not a stable one - it is not a picture we
can live with. But notice that this unstable situation arose from what seemed
perfectly reasonable opinions about the causal prowess of both minds and
brains. And it is this situation, arising as it does from these considerations, that
grounds philosophical anxiety about the mental realm and its relationship to
the bedrock of scientific description. When a philosopher thinks about how
reasons cause actions and how our muscles are caused to move in response to
brain events, these seem to lead inexorably to the precipice of over-
determination. It will not help here to maintain the sanguine attitude that as
we learn more about the brain, we will come to understand how the mind is
related to it. For unless something is said right now about how to remove - at
least in principle - the instability of figure 5, we aren't going to come to
understand anything. Of course, there are many things we can say - and
which have been said - about how to cope. Further, one of the conveniences of
figure 5 (and this will make up for the initial embarrassment that it has caused
us) is that it can help us organize the survey of these possibilities.
[Note: I have shown that the problems of overdetermination arise on the
'output' side: mind and brain both having causal claims on action. However,
the whole triangle could have been based on the 'input' side: certain events
producing beliefs and other attitudes via perception and, apparently at the
same time, producing changes in the nervous system. This would have resulted
in an inverted figure in which mind and brain occupy the bottom left and right
vertices of the triangle respectively, with perceivable states of affairs at the apex
(see PERCEPTION and PERCEPTUAL CONTENT). Of course, to see what the problem is
we only need to look at one way in which overdetermination arises, and I
think it is easier to see this in terms of the 'outputs' of minds and brains rather
than in terms of their 'inputs'.]
86
STAGE 3: BEDROCK
3.5 Dualism
The triangular relationship is not stable, and we have to find acceptable ways
to remove this instability. Of course, one way to go about it would be simply to
eliminate or discount the claims of either Mind or Brain, though this rather
drastic move will be considered only after we have discussed the gentler
options. However, even before we get to them, I should like to say something
about a view that is at the other end of the spectrum from the elimination
strategy.
Descartes (1596-1650) held that mind and matter were distinct substances
(see HISTORY), and one might wonder whether, if this sort of DU ALISM can be
defended, it would make a contribution to the present problem. Unfortunately,
the answer here must be 'no'. For, whatever reasons one might have to share
Descartes' view, the problem arises, not merely from the alleged independent
existence of mind and matter, but from the fact that they both seem to make
causal claims in respect of actions. If you are inclined to be a dualist, then you
will be happy to drop the 'allegedly' in my last sentence. Mind and matter will
be taken to be in fact independent existences. But this still leaves us in the dark
about the rest of the picture. Indeed, insofar as we really do have distinct
mental and material substances, overdetermination is, if anything, even more
of a threat.
Descartes recognized that dualism alone was not a solution to the problem of
seeing how the mental is related to the physical. Indeed, he thought that the
mind brought about effects in the physical world; it acted on, and reacted to,
things that were undeniably material. One could almost say that Descartes was
the first to recognize the importance and difficulties of the triangular relation-
ship, though not precisely in the form described above. He wondered how the
mind could bring about effects in the physical world, since it was itself a sub-
stance not belonging to that world - his worries began with his dualism -
whereas I am considering whether dualism can count as a way of dealing with
the triangle. Still, both ways of approaching the matter lead to similar ques-
tions: are there mental causes of actions as well as material? If so, how does
one deal with the threat of overdetermination? And, in any case, how can
something mental cause something physical?
Descartes' answer to the third of these questions implied his answer to the
others. What he claimed was that there is a little, very 'subtle' organ in the
brain - the pineal gland - which is where mental states acquire the power to
causally affect physical ones. This interactive view of the mental and the phys-
ical, if successful, would avoid the problem of overdetermination, since the
mental would, in actual cases, provide links in a single causal chain leading to
action. In effect, the mental and the physical would be overlapping causes
rather like the unproblematic case of the grudge and short circuit in which the
person with the grudge rigged the wiring in the house so that it would short
and bring about the necessary heat in the wall to set off the fire.
Of course, as one is trained in philosophy to point out, in the case of mind
87
AN ESSA Y ON MIND
and matter, Descartes' suggestion just doesn't work. For either the pineal gland
is a mental or a material substance. If the former, then the problem of how the
mental has physical effects is simply pushed that bit further back - to the rela-
tion between the pineal gland and whatever is the first physical waystation on
the route to action. And if the pineal gland is material, as Descartes certainly
thought it was, then the problem of causal interaction between the mental and
physical doesn't go away - it simply takes place on the narrow stage of the
pineal gland. Mere dualism of the mental and the physical will not make the
problems go away, and Descartes' attempt to add a kind of interactionism to
his dualism is usually counted a failure. Of course, this doesn't mean that
dualism itself must be rejected. For all I have said, dualism might be true. But if
it is, then it is difficult to see how to resolve the triangle. Indeed, since it is
widely thought that such resolution cannot be achieved within the dualist
framework, there is an urgent need to find some better way.
It should be noted here, if only for completeness, that an even more thor-
oughgoing dualism than Descartes' could provide a way to resolve the trian-
gle. If one drops the idea that the mental and the physical interact, then one
could convert the triangle into two stable pairwise relations in which the
mental is partitioned from the material rather as in figure 6. However, this
'resolution' of the problem suffers from massive implausibility. For example,
suppose you were hungry and, having thought about it, reached for the
refrigerator handle. It would seem to you as if your thought processes even-
tuated in a movement of your arm, but, on the view being considered, this
would be denied. Instead, it would be maintained that your thought processes
eventuated in a mental action of some sort (note that I have carefully not
labelled the arrows and bottom boxes), and that this action was accompanied
by perceptions as of your arm moving. The whole of this train would be
purely mental and it would be shadowed by, though would not causally inter-
act with, a chain of purely physical events involving your brain, arm and the
refrigerator. Such a parallelism would certainly avoid the problem of over-
determination, but almost no one finds the idea acceptable. Aside from any-
thing else, it would require us to believe that there is a fortuitous
synchronization of mental and physical chains of events which only appear to
Mental Realm Physical Realm
Figure 6.
88
STAGE 3: BEDROCK
be the result of interaction. (See HISTOR Y for a brief account of a philosopher -
Leibniz - who did hold some such view.)
3.6 Coalescing
The problems would disappear at a stroke if some rationale could be found for
seeing the mental and the physical as somehow combined or cooperating
instead of competing. In particular, if we could coalesce two of the vertices of
the triangle, then all would be well, since of course we would no longer have a
triangle on our hands. Of course, though it is easy enough to suggest such a
rearrangement, the real work lies in making any kind of coalescing plausible.
And, as one might expect, there are a bewildering number of ways in which
this has been attempted. However, within this mass of detail, one can discern
two main approaches.
3.6.1 Mind into Action
BEHA VIOURISM counts as a manifestation of the coalescing strategy, insofar as it
attempts to absorb the mind into the realm of action. As a simple illustration,
consider the following behaviourist account of a belief. Begin by supposing it
true that I believe my lawn needs watering - that I am in that mental state.
Being in that state it is pretty likely that I would indulge in grass-watering
activity, at least in the right, non-rainy, not-too-busy circumstances. Now it is
this fact about what I am likely to do that behaviourists exploit. For they think
it reasonable to define my mental state - my belief that the lawn needs water-
ing - as a tendency to water the grass in certain circumstances. Continuing in this
way, managing to define each and every mental item as nothing more than
congeries of behavioural tendencies, they then insist that the mind need no
longer be represented as an independent element in the triangle. It will have
been, so to speak, superimposed onto action, and we will end up with figure 7:
there will no longer be a need for any separate thing labelled 'Mind'. (See also
DENNETT; LEWIS; QUINE; RYLE.)
Behaviourism that discerns the mind in action is at least concessive to the
idea of the mental, since the mental is treated as genuinely present in patterns
I Brain I
Action
Figure 7.
89
AN ESSA Y ON MIND
of action. But there is a more ruthless kind of behaviourism which makes no
such concession. Its adherents insist that there are brains and there is beha-
viour, but that mental states of any stripe are just surplus to requirements. As I
have divided things up, concessive behaviourism counts as a coalescing strat-
egy, since it is committed to seeing the mind as presented in complexes of
actions and tendencies to act. The other sort - which could be called 'elim-
inative behaviourism' - will be discussed along with other radically eliminative
strategies.
3.6.2 Mind into Matter
Behaviourism has fallen on hard times, in that few present writers actively
pursue lines of thought that they would characterize as behaviourist. The
example I used - counting my belief that the grass needs watering as a beha-
vioural tendency - just totters on the edge of plausibility. For it certainly seems
possible for someone to have such a belief without also having any specially
devoted cluster of active tendencies that could be used to define it. But things
really get difficult for the behaviourist with respect to experiences such as pain.
It is not easy to accept that my tennis elbow consists in my complaining,
tennis-avoiding and physiotherapist-visiting behaviour. What about the fact
that it hurts?
The second coalescing strategy sets out from a different starting point:
mental phenomena are superimposed, not on actions, but on the physical
phenomena I have called 'brain'. This is shown in a general way by figure 8,
but there are significant differences in the ways this superimposition is
achieved, so I shall have to divide my survey of this strategy into sub-sections.
3.6.2.1 Identity The IDENTITY THE OR Y takes the most direct route to the goal
of coalescing the mental and the physical: mental phenomena are said to be
physical phenomena - they are counted as one and the same. Stated in these
bald terms, this view is unlikely to satisfy anyone, or even to be wholly intelli-
gible. One needs to be told more specifically what is being identified with what,
and, even more pressingly, why any kind of identity between the mind and
material stuff is so much as reasonable. It is here that a certain philosophical
view of science comes to the rescue.
Brain
l Action J
Figure 8.
90
STAGE 3: BEDROCK
Many philosophers have been attracted to the identity account of mind
because of the way identity has figured in the evolution of scientific theories.
For example, consider how we have come to understand a phenomenon such
as lightning. Before the development of scientific theories of electricity, light-
ning must have seemed a very bizarre phenomenon indeed. But by the end of
the nineteenth century, it could have been truly be said that we had come to
know what lightning is, viz. the massive, sudden discharge of the collective
electrical charge generated by the movement of many slightly charged water
droplets or ice crystals that form the clouds. What had happened in the
passage from ignorance to understanding could be put this way: there was this
strange type of observed phenomenon (lightning) which was found to be one
and the same with a type of electrical phenomenon (static discharge) that we
had discovered in our search for a theory of the natural world. Note that one is
not here saying that lightning is caused by electrical discharge. That would be
to allow that they were in some sense different phenomena. Rather, the view is
that lightning just is electrical discharge.
This sort of idea offers a promising way to deal with the problem of the
mental and the physical: why not say that individual types of mental phenom-
ena (such as beliefs or pains) are really types of state in the brain (perhaps
certain patterns of neural firings)? On analogy with the lightning case, one
would be thinking of scientific knowledge as providing insight into the nature
of what are at first puzzling sorts of things. We are aware of such things
as beliefs and pains, but we are as ignorant of their true nature as we
once were of lightning. But if we can identify mental phenomena with the
physical, we will both cease to be ignorant of their real nature and, treating
mind and brain as one, we will overcome the difficulty posed by the triangle in
figure 5.
The above is the merest sketch of one sort of identity theory. In particular,
based as it is on the analogy with the scientific case, it features the identifica-
tion of types of mental state with types of physical state. If we take, say, the
pain resulting from a burn, as an example, then what the theory says is that
whenever someone has that kind of pain, there will be some particular type of
brain state or activity that is that pain.
Another way to look at it is to say that a type identity theory identifies one
property with another, and that it does so partly on the basis of special sorts of
laws of nature ('bridge' laws) connecting the two. Using the earlier example, it
could be said that the property of the heavens known as lightning is one and
the same as the property of cloud-generated electrical discharge because there is a
universal law connecting the two as they occur in one and the same spatio-
temporal location (,whenever and wherever you have the one, you have the
other'). Similarly, one might hold that a burning pain is a property of a mind,
and a configuration of neural activity is a property of a brain, though the dis-
covery of a bridge law connecting them would imply that burning pain is
nothing over and above some pattern of brain cell activity. (See CONCEPTS;
LEWIS; PROPER TY.)
91
AN ESSA Y ON MIND
3.6.2.2 Aspects and Identity If it were true that a mental property was
nothing over and above a physical one, then we could speak of the REDUCTION
of the mental to the physical. This is moreover just what one would expect
given the analogy between cases of scientific reduction and the type identity
account of the mental. For, given our current state of knowledge, what one
would say of lightning is that it has been reductively identified with electrical
discharges.
But, to put it mildly, not everyone is happy with such talk in the case of the
mind. There are a number of writers who think that mental phenomena are
not reducible to the physical. In particular, it is argued that properties such as
believing that p, or having a pain are just the wrong sort of thing to be identified
with physical properties. For instance, some writers have suggested that there
are a priori reasons why there could never be any bridge laws connecting mind
and brain. This, it is argued, is because mental properties form a closely func-
tioning network, not matched by any complex of properties in the physical
realm. It is therefore wrong to think one could so much as imagine a law that
singled out a mental property from this network and tied it to a physical one.
The idea is that such a singling out would be tantamount to giving up the idea
that the property was mental in the first place. (See DAVIDSON.)
Yet, in spite of their opposition to type identity, these writers are not in the
least tempted by dualism: they do not think of the mental as a different thing
from physical or material substance. What they want is some way to see the
mental as physical, but to resist the idea that the mental is reducible to the
physical on the model of, say, lightning and electrical discharge.
Given the thin line that is being trodden between dualism and reduction, it is
a delicate matter to so much as label this position. One could describe it as a
kind of dualism at the level of properties because mental properties are claimed
to be irreducible to physical ones. But there are deep issues in ONTOLOGY which
make talk of PROPERTIES suspect, at least to certain writers who are otherwise
adherents of the view being discussed. Or one could eschew talk of properties
and speak of a dualism of 'aspects': some phenomena have irreducibly both a
mental and a physical aspect. Yet there is not all that much to be gained by
using such a vague word: one suspects that an aspect is a property by another
name or, if not, then it is very difficult to say exactly what it is. Finally, and
least controversially, one could leave 'dualism' out of the characterization al-
together by describing the view as 'ANOMALOUS MONISM'. This label - coined by
DA VIDSON - has the virtue of emphasizing the coalescing credentials of the view
- it is after all 'monism' - whilst at the same time signalling a refusal to coun-
tenance reduction. Describing the mental as anomalous in respect of the physi-
cal just is a way of denying reducibility.
Now it is one thing to argue about labels, it is another to justify the view
itself. How can one be a monist - find that onto logically speaking there is only
physical stuff - and yet insist that the mental is not reducible to, or is anom-
alous in respect of, that stuff? It is here that a second kind of identity theory
comes into play, one made possible by the contrast between types and tokens.
92
STAGE 3: BEDROCK
To get hold of the TYPE/TOKEN distinction, think, for example, about the
book you are now reading. On one way of understanding the expression 'the
book' in my last sentence, it refers to that single work or type, the Companion
to the Philosophy of Mind, first published in 1994 and edited by me with con-
tributions from many different authors. But, given the right context, I could
have used that same expression to refer, not to the one thing that Blackwells
published under that title, but to the very object you are now holding. This
object, one of thousands which were printed, is said to be a 'token' of the
Companion (though in many ways it would do just as well to call it an
'instance' or an 'exemplar'). And just as one can think of types and tokens of
books, one can also make the type/token distinction in regard to other items,
e.g. mental (and physical) phenomena. The pain which, let us say, I now have
in my elbow, is of a certain type. It is that intermittently sharp kind of pain
one gets with so-called 'tennis elbow'. But I can also think of my pain as a
token mental occurrence. For example, if you asked me why I just winced, I
might say: 'I just felt a sharp jab in my elbow', and in this case, I would be
speaking about that particular token of pain that caused me to wince. Simi-
larly, there can be a certain type of brain state - say, a state consisting in the
firing of C-fibres - which will have various tokens or instances at particular
times.
As described earlier, the type theory sets out to identify types of pain with
types of mental state. Suppose, for example, that it attempts to identify tennis-
elbow pains with the rapid firing of certain neural C-fibres. As we saw earlier,
this kind of identity supports the reduction of mental properties to physical
ones: tennis-elbow pain is nothing over and above the firing of C-fibres. So if
we are determined to deny reduction, we must be prepared to deny type iden-
tity. But one can still cleave to the idea that the world is at bottom physical by
asserting token identity. Token identity theorists would expect there to be some
particular token of a physical state that was my pain (when I winced), but they
insist that this is the only kind of identity possible between the mental and the
physical. So we end up with only one kind of thing - the physical - whilst still
denying that mental states are reducible to physical states. Unfortunately, all is
not plain sailing.
What does it really mean to say that the world is at bottom physical? The
model provided by the case of lightning and electrical discharges shows one
way to unpack this claim: all phenomena are reducible to the physical - for
any real phenomenon we can always find out what type of physical thing it is.
However, this straightforward reductionist response is not available to token
identity theorists. Instead, they appeal to a notion of SUPERVENIENCE - a notion
that signals a sort of dependency relation. As an example of this dependency
consider the following case. Suppose that you judge a certain Corot painting of
a landscape to be autumnal. This feature of the painting clearly depends upon
the arrangement of pigments on the canvas, but one would not ordinarily say
that this feature was reducible to the disposition of pigments, that 'autumnal-
ness' in paintings just is that kind of arrangement. After all, another painting,
93
AN ESSA Y ON MIND
which was physically quite different, could also be autumnal. The commonly
used technical way of putting this is to say that autumnal-ness is 'multiply-
realizable' - that there are many ways in which this property could be realized
in different paintings. However, it is nonetheless true of the Corot painting that
its being autumnal supervenes on the physical structure of the canvas. Or we
could put it this way: if the Corot painting ceased to be autumnal - if this
feature of the painting changed - then we would expect that some physical
feature of the canvas would also have changed, perhaps because of ageing or
destructive 'restoration'. Thus, being autumnal in the case of the particular
Corot painting depends upon the physical arrangement of pigments on a specific
canvas. And we can say this without committing ourselves to the idea that
'autumnal-ness' is a type of thing reducible to that particular physical arrange-
ment.
Understood in this way, supervenience gives us a way of saying that mental
phenomena depend on, in the sense that they are realized by, their physical
embodiments without being reducible to them. Thus, the tennis-elbow pain I
am now experiencing is a token physical state in my brain. And insofar as this
sort of token identity holds for each mental phenomena, we can say that the
mental supervenes on the physical. But there is no reason to believe that
mental phenomena are themselves physical types of thing. Just as the property
of being autumnal could have been realized physically in lots of different ways,
so mental phenomena might well be realized by different types of physical
states in different organisms. This means that the mental depends upon the
physical without being reducible to it.
Substituting token for type identity and supervenience for reduction has an
obvious appeal. It is the closest we can get to preserving something of the tri-
angular relationship - keeping, that is, a certain distance between mind and
brain - whilst avoiding overdetermination. (See figure 9 for the best I can do to
represent anomalous monism. Mind lies outside brain - this indicates non-
reducibility - though the projective lines show Mind's dependence or super-
venience on the physical.) But there are deep problems with this view. In parti-
cular, it has been argued that the price to be paid for avoiding over-
Brain
Figure 9.
94
STAGE 3: BEDROCK
determination, whilst maintaining only token identity, is a certain idleness for
the mental. The details of this issue are complex and much discussed, but the
basics are apparent enough even from Figure 9. In that figure, the arrow of
causality connects physical phenomena to actions, whilst the mind, super-
veniently out on its own, has no such direct causal link. But many think that
we should only count something as genuinely real if it enters into causal rela-
tions. Using the fashionable word, the mind in figure 9 seems to be EPIPHE-
NOMENAL - sort of surplus to causal requirements - since what does the causal
pulling and pushing are brain states.
There are no shortage of replies possible to this worry. One might try to find
some special work for the mind by distinguishing between actions and move-
ments of the body; actions would have a special explanatory (even if not
causal) relation to the mind, though they too could be said to supervene on
appropriate movements of the body. Or one might try to show that super-
venience doesn't really render the mind idle. But these replies, and variations
on them, go beyond the brief of this Essay. (See' DA VIDSON; DENNETT; DRETSKE;
FODOR; LEWIS; STALNAKER.)
3.6.2.3 Content and Identity Before considering certain other objections to
identity theories, I must note something which, up to now, may have puzzled
you. Mostly I have written in a general way about identifying the mental and
the physical. though I have sometimes assumed that the identification would
be between the mental and a more specific kind of physical item, namely brain
states and processes. Indeed, my use of the label 'Brain' in figure 5 is explicitly
non-committal as between brain phenomena narrowly conceived and more
general physical happenings. It might at first seem that, if an identity theory is
right, then mental states are bound to be states of the brain together with
related parts of the central nervous system, so that it is just not necessary to be
non-committal in this way. After all, since the brain is surely responsible for
mentality, how could the physical embodiment of someone's mental state
extend beyond the brain and body of that individual? How could mental phe-
nomena either be or supervene on anything other than goings-on in the
central nervous system?
Using pain and experiences generally as examples of the mental tempts us to
think of the brain as the only possible physical foundation for mentality. But
one must not forget about the attitudes. When we attribute to Anne the belief
that her coat is in the hallway cupboard, we are characterizing Anne's state of
mind and we are doing so by reference to how things are, or might be, in the
world. In stage 2, we struggled with this feature of the attitudes without
coming to any definite conclusion. But concerned as we are now with the rela-
tion of the mental to the physical. this problem - the problem of directionality
- reasserts itself in a slightly different way.
Suppose for the minute that some identity theory is right, though it doesn't
matter for the present point whether it is of a type or token sort. Then when
Anne has the belief that her coat is in the hallway cupboard, there is some
95
AN ESSA Y ON MIND
physical state which is that belief. Suppose further that we accept the idea that
this state (or event or process) is to be found somewhere in Anne's central
nervous system. Call that item N. This means that we should accept that
Anne's belief is N. But how can N - some neurophysiological state internal to
Anne - be a belief about the coat's location, something wholly external to
Anne? It is far from obvious how to answer that question.
A crucial feature of the attitudes is that they are directed to some actual or
possible state of affairs. Understanding this is fraught with the difficulties
grouped together in stage 2 under the label 'the problem of directionality'. The
present worry extends beyond - though it includes - the problem of direction-
ality. For what we are now wondering is how a 'mere' physical state internal
to someone's nervous system can be about, or directed to, some feature of the
world external to that individual. Philosophers tend to speak about the
'problem of intentionality', using this label to cover both the conceptual diffi-
culties generated by the attitudes, and this issue of how to fit such direction-
ality onto the bedrock of the physical world. It is also described as the problem
of naturalizing intentionality.
One way in which an identity theory might try to cope is by denying the
very thing we have been supposing, that is, by denying that the relevant phys-
ical state is in fact internal to the individual. In Anne's case, this would mean
saying that her belief was one and the same with some physical state of the
world which may well include N but which also takes in the state of affairs
involving the coat, cupboard, etc. Of course, in doing this, one would be saying
that Anne's mental state is somehow spread out so as to include much more
than what is local to Anne. But this is not an easy consequence to accept,
since many find it difficult to understand how Anne's mental states could be
anything other than intrinsic to her, and thus located where and only where
she is.
A second tack would be to insist that there is a wholly internal state corres-
ponding to Anne's belief, and that the directionality of the belief is explicable by
relations between this state and the state of affairs involving the coat. Thus,
perhaps what makes some state of Anne's a belief about the whereabouts of
her coat is that there is a causal relation between that state and the coat. One
would here be trying to capture the directionality of the attitudes by citing
causal relations between internal (presumbably, brain) states and actual and
possible states of the world.
Naturalizing intentionality is a deeply perplexing problem. I have raised it in
connection with identity theories, but it is by no means confined to them. Any
coalescing strategy - any attempt to make out that the mental is superimposed
on the physical - has to provide some account of how an attitude can be both
a physicalistic ally respectable feature of an individual and yet have a content
that extends beyond that individual. I shall return to this issue again in con-
nection with the third attempt at coalescing, after first considering the sort of
general criticism that has been levelled against identity theories. (See CONTENT;
INTENTIONALITY; LANGUAGE OF THOUGHT.)
96
STAGE 3: BEDROCK
3.6.2.4 Against Identity Independently of any doubts about this or that form
of the identity account, there is a source of criticism that goes right to the core
notion in all of them, viz. identity. Any claim of identity carries with it a logical
commitment whose articulation is usually attributed to the seventeenth
century German philosopher G. W. von Leibniz. Though it is not uncon-
troversial in certain philosophical circles, LEIBNIZ'S LAW seems to state a per-
fectly obvious truth about any kind of identity claim: if an identity is true - if,
for example, one can truly claim of a and b, that a is identical to b - then any-
thing true of a will be true of b. Moreover, it is a principle we use all the time.
For example, suppose Smith, having had his bicycle stolen, is trying to find it in
the police lost property office. There are a number of bicycles there of the same
colour and make as his, and the police have given numbers to these recovered
bicycles. Suppose Smith conjectures that his bicycle is number 234, i.e. it is
one and the same as, or identical to, 234. If this conjecture is right, then,
whether Smith can prove it or not, absolutely everything true of 234 must be
true of Smith's bicycle. For instance, if it turned out that 234 was assembled in
France, whereas Smith's had been assembled in Belgium, then this would be
enough to falsify the identity claim that Smith made. For if 234 is Smith's
bicycle, then we are only really talking about a single thing. And how can it
possibly happen that one thing was both assembled in and not assembled in
France?
Leibniz's Law, as in the above everyday example, gives us a certain leverage
over claims to identity: it gives us a means of falsifying them. In the philosophy
of mind, this kind of leverage has been employed often. It has been held that
there are features of mental states not possessed by any brain states, and, of
course, if this were true, we would have to give up on any account that identi-
fied these states. Typical of the features in question are these: mental states do
not have spatial location in the way that brain states do; mental states such as
pains can be sharp or burning, brain states are neither; one can conceive of
mental states existing independently of brain states, but the same of course
cannot be said of the brain states themselves. However, whether fortunately or
not, none of these examples is un controversial, so identity theories are still
very much a live option. Applying Leibniz's Law considerations to mental phe-
nomena is just not as easy as applying them to stolen bicycles. (See IDENTITY
THEORIES, LEWIS.)
3.6.3 Mind onto Matter
3.6.3.1 The Functionalist Idea An identity theory answers the question 'What
are mental phenomena?' essentially by telling us how they are constituted.
Such a direct physicalist answer is one way to go about dealing with a 'What
is ... ?' question, but there are others. Moreover, thinking of things as phys-
ically constituted is sometimes a misleading way of trying to understanding
them. In order to illustrate these points, consider this strange ad which once
appeared in an Oxford newspaper:
97
AN ESSA Y ON MIND
Wanted: bicycle or similar.
What makes it strange is that a bicycle is an object with a certain basic shape
and construction, and one most naturally answers the question:
What is a bicycle?
by giving some account of this physical embodiment. So, when someone speaks
of wanting something 'similar' to a bicycle, it is difficult to understand this
except as expressing a want for a bicycle. The natural way to take the word
'similar' in this context makes it redundant. (Of course, in some other sense of
'similar', everything is similar to everything else. Thus, my broken steam iron
is similar to a bicycle: they are both manufactured objects. And yet one would
hardly have thought that the need expressed in the ad would be satisfied by
myoid iron.)
In contrast to this case, suppose that the ad had said:
Wanted: word processor or similar.
There are lots of ways in which word processing can be achieved: dedicated
machines, computers with word-processing programs, typewriters with mem-
ories, etc. Since these devices will look quite different from one another, the
most straightforward way to say what a word processor is will not involve a
description of a particular physical shape or construction. In answering the
question:
What is a word processor?
what would be given would be some account of what the device does, rather
than how it is physically embodied. For this reason, and unlike the bicycle
case, a demand for a 'word processor or similar' wouldn't be strange. Anyone
of the physically different word-processing devices might satisfy the need.
As described earlier, the type identity theory treats mental phenomena as
something like bicycles: understanding, for example, a pain or a belief involves
understanding how each of them is embodied in some physical realm such as
the central nervous system. (For the moment, leave token identity theories on
one side.) But perhaps a better way to think of pains and beliefs is as we think of
word processors: a pain or a belief is understood only when we know what each
of them does. This idea - that we understand specific mental phenomena by
describing what they do - is, in broad terms, at the heart of the view known as
FUNCTIONALISM. And, in one or other of its versions, it is currently the most
popular way to achieve coalescence between the mental and the physical.
Of course, it is not immediately obvious what it means to say that we under-
stand a belief when we know what it does - what its work or function is.
Certainly, it would be misleading to allow ourselves to be carried too far by the
98
STAGE 3: BEDROCK
word processor analogy. Word processors have functions in a rather rich sense:
they are designed to achieve certain purposes or goals. However, there are
other senses of 'function' which, at least at first, are more plausible in helping
us to understand the mental.
For example, consider my belief that Monday, 29 August 1994 (next
Monday) is a Bank Holiday. Without claiming that this belief is somehow
designed to achieve some goal or purpose, I can still describe it as having work
to do, as having a functional role. For example, given that I have this belief, I
am likely to go in for certain kinds of behaviour such as making sure I have
enough cash to get through the weekend, making plans to leave town, etc.
Also, there is a story to be told about the ways in which my seeing, hearing
and remembering various things together account for the fact that I have this
belief in the first place. If we describe how the belief affects my behaviour and
other mental states as 'output' and the sources of its formation as 'input', then
we can say that the functional role of the belief is some specific set of complex
inputs and outputs. Indeed, we can go further: in response to the question
'What is SG's belief that the 29th is a Bank Holiday?', a complete answer will
be: it is that feature of SG that comes about in virtue of such-and-such inputs
and results in such-and-such outputs. Of course, there is a lot of complexity
underlying this outline answer - filling in the detail of the 'such-and-suches'
would be an enormous task. But this complexity should not prevent one seeing
the core idea of functionalism: a mental phenomenon will be defined by its
inputs and outputs, its functional role.
It is important to note that, in spite of talk about inputs and outputs, a func-
tional account is not itself a behaviourist account. Behaviourists of the sort
considered earlier aim to define such mental items as beliefs as patterns of
behaviour, whether actual or potential. In contrast, a functionalist allows that
beliefs and relevant behaviour exist independently of any behaviour. Indeed,
given that mental states cause various patterns of behaviour, they cannot be
defined in terms of them. Moreover, among the inputs and outputs that con-
stitute the 'work' of a given belief - its functional role - there are bound to be
other mental states. Thus, my belief about the Bank Holiday may well cause
me to be mildly depressed, though this latter state of mind might never be
expressed in behaviour which could be used to define it. Perhaps I am just so
used to depression as to suffer a mild case of it without this in any way affect-
ing my behaviour.
One reason for the popularity of functionalist accounts is that they seem to
be untouched by various strong objections to type identity. Take pain as an
example. On the identity theory a given type of pain will be identified with a
given type of event in the central nervous system. Thus, burning pain might
come out as the rapid firing of C-fibres in one area of the brain. But this would
mean that a creature constructed differently from us - the proverbial Martian,
to take one example - just couldn't have such pain. For if a type of pain is a
type of neural event, then there can be no pain without neurons. Yet this
seems hard to swallow. Surely we would want to allow that a Martian, though
99
AN ESSA Y ON MIND
lacking neurons, could still experience pain. It is just that the physical basis for
this pain would be different in his world. And this is precisely what the func-
tional account allows. For the functionalist, pain of a certain sort is that phe-
nomenon that comes about as the result of certain inputs and produces certain
outputs. Any creature who possesses this pattern of outputs and inputs can be
said to have a burning pain, whether or not that creature has neurons, silicon
or something else in its physical make-up.
Burning pain, in being functionally characterized, is multiply-realizable, and
most functionalists insist that mental phenomena supervene on the physical.
These are of course features that functionalist accounts share with the token
identity theory, and it is thus unsurprising to find that those newly introduced
to these accounts tend to confuse them. The pictorial representation of token
identity (figure 9) could serve as well as a representation of functionalism.
However, what cannot be shown by such a simple sketch is the rationale for
regarding the mental as projecting from the physical, and it is in this respect
that certain token identity theories differ sharply from functionalist accounts.
DA VIDSON, and those prepared to count themselves as anomalous monists,
accept both multiple realizability and token identity, whilst insisting that
mental phenomena are not reducible to the physical. Indeed, these are the
defining features of their view. But notice that nothing in any of this depends
on seeing mental phenomena as functionally definable. In fact, one might have
reasons to resist the idea that mental phenomena are somehow understood in
terms of their location in a network of inputs and outputs without giving up
any of the theses which define anomalous monism.
In contrast, the functionalist is someone driven by a particular conception of
the mental - a conception that allows, and encourages, the token identity
thesis and multiple realizability as ways of guaranteeing the physical creden-
tials of the mind. However, it is perfectly conceivable that one could embrace
the core functionalist idea without going the token identity route. There are
two ways in which this could happen. On the one hand, one might be a func-
tional dualist: someone who thinks that mental items are to be understood in
terms of their characteristic work, without regarding the mind as in any way
material. Of course, as was seen, dualism does not qualify as a way of defusing
the overdetermination problem, but it is important in understanding function-
alism to see that it can even make peace with dualism. This can happen simply
because functionalism per se has no commitments to the material (or other)
constitution of mind. On the other hand, one could be a functional reduc-
tionist: someone who thinks that the mental is type identical to the physical
and, hence, to that extent, reducible. LEWIS among others holds this view,
though it should be noted that he attempts to avoid the earlier-mentioned
pitfall of type identity - multiple realizability - by confining identities to species
or even individual members of a species.
3.6.3.2 Functionalism and the Computational Model of the Mind It is one thing
to think that the core functionalist idea is attractive, it is another to describe it
100
STAGE 3: BEDROCK
in the sort of detail necessary to establish its plausibility. Recall the earlier
example: my belief that the 29th August 1994 is a Bank Holiday is function-
ally characterized when one can list all those states and events that cause me
to have it and all those effects that it leads to. Yet even for such a mundane
belief these inputs and outputs are going to be enormously complex - much
too complex to think it a practical project to list them. So, what grounds the
widespread conviction among philosophers of mind that functionalism of one
sort or another has any chance of success? Is it simply the beauty of the core
idea that blinds them to the fact that functionalism without the details is
nothing more than gesture towards a defensible view?
It is all too easy to think that these questions point unflatteringly at a certain
kind of philosophical gullibility. But underlying the large degree of 'hand-
waving' towards details that one finds in treatments of functionalism, there is
in fact a very solid foundation. For certain important results in the theory of
computation (due to TURING) have shown that enormous complexity is no real
obstacle to the project of functionalism. Indeed, what Turing showed was that
any complex pattern of interrelation between inputs and outputs could be
mimicked by very simple and orderly functional relations which have come to
be called 'Turing machines' (See ARTIFICIAL INTELLIGENCE; COMPUTATIONAL
MODELS). In essence, the work of Turing and others has made it possible to
think functionalism is plausible - so far as complexity is concerned - without
our having to work out a detailed functional account for any particular set of
mental phenomena.
Any specific Turing machine is essentially a very simple sort of computer
program - one which can transform certain inputs into some desired output.
(In spite of the label, a Turing machine is not so much a physical device as it is
a way of thinking about functional relationships based on analogy with an
(imagined) physical device.) Thus, there are Turing machines that take any
two numbers as input and produce their sum as output. One can think of these
as just like a program for adding that one might load into a computer. Of
course, the relationship between numbers and their sum is very straightfor-
ward. Yet, as was noted earlier, there are no patterned relationships between
inputs and outputs so complex as to escape some Turing machine.
Insofar then as the core idea of functionalism is accepted, we can relegate
the actual functional details to one or another way of employing the Turing
result. Here there are a bewildering number of possibilities, which are best left
to the more specific discussions of FUNCTIONALISM. Still, it might be helpful to at
least outline the range of possibilities:
(1) The most direct use of the Turing result consists in treating such things
as beliefs, desires, pains and other mental items as themselves states of a
Turing machine appropriate to the mind being functionally characterized.
Various features of the mental and Turing machine states have made this sort
of 'machine functionalism' the least-preferred option.
(2) At the other extreme is the view that treats the commonsense conception
of the mind as itself a causal/functional account which is underpinned by the
101
AN ESSA Y ON MIND
Turing result but which does not attempt to make any direct connections
between Turing machines and mental items. Because this view insists that the
commonsense account is already a functional characterization of the mental -
an account which gives us the essential features of each mental item - it is
sometimes called 'analytical functionalism'. (See LEW IS.)
(3) In between are a whole range of views that regard the commonsense
conception as tied more or less loosely to some computational level of descrip-
tion, though not necessarily a level as basic as that conceived by Turing.
Included here are views that employ the LANGUAGE OF THOUGHT hypothesis.
Because this range of views makes the connection between the computational
level and the commonsense conception a matter of empirical research, it is
often described as 'psychofunctionalism'. This label hints that the final picture
will be an empirical, though psychological, research result. (See also IN N A TE-
NESS; MODULARITY OF MIND.)
Though the above only sketches the possibilities, it brings out something
important about the connection between functionalism and detailed computa-
tional accounts: the fact that there is a certain slack between them. This slack
accounts for the arguments one comes across over the claim that the 'mind is
a computer program'. On the one hand, there are those who regard this claim
as no more than a metaphor, and on the other, there are those who think it
literally true. And it is not surprising that there should be this difference. For if
you regard mental items as computational states, then the mind will count for
you literally as a computer. But if you only think of the mental as loosely
grounded on the possibility of a computational account of our neural workings,
then the most you would allow is that the mind is like a computer program.
However, in either case, there is something appealing about the idea. For it
gives an interesting gloss to the functionalist idea that the mental is multiply-
realizable. After all, we are all familiar with the fact that a computer program
(the software) can run on all sorts of physically different machines (the hard-
ware). On this conception - which is common though not mandatory for func-
tionalists - the mind is like software and the brain is the hardware on which it
is 'run' (for human beings at least).
3.6.3.3 Functionalism and Content One of the problems discussed in connec-
tion with the identity theory was whether it could give a plausible account of
mental states with contents - the propositional attitudes. The particular diffi-
culty was this: if an attitude is identical to some wholly internal individual
brain state, then how can that attitude have a content directed to something
that is external to the thinker in space as well as time? And if the attitude is
said to be identical to a physical state of the world which includes the brain
state of the thinker but extends to those things that figure in the content, then
we end up with very implausible identities. After all, it seems very odd to say
that someone's thought is a physical state that includes items removed in space
and time from the thinker.
This problem - the problem of intentionality - is one that arises in connec-
102
STAGE 3: BEDROCK
tion with any attempt to coalesce the mental and the physical. This is
because, however such coalescence is achieved, one will end up by regarding
some physical or functional type as the embodiment of an attitude. But then
it will not be obvious how to accommodate the directedness of the attitude -
its content. Physical states seem to differ from the attitudes precisely because
they lack any sort of directedness, and functional states would seem to be no
better off. Nonetheless, the problem of intentionality tends to be discussed
mostly in the context of functionalist accounts, and this is probably because
of the feeling that some kind of functionalism offers the best hope of a solu-
tion.
Recall that a functionalist would define, for example, Anne's belief about the
whereabouts of her coat, by some complex of behavioural and mental inputs
and outputs. Insofar as these are wholly realized in states internal to Anne's
physical make-up, one might still try to explain the content of the belief by its
location in the whole network (see CONCEPTUAL ROLE SEMANTICS). Or, if one is
prepared to allow that items in the network have causal relationships to fea-
tures of the world outside Anne, then there is the glimmer of hope that inten-
tionality can be explicated in terms of that causal relation. Or if not a causal
relation of the most direct sort, then perhaps some more complicated causal
connection would do the trick. But many objections have been made to each of
these attempts. (See CONTENT; FODOR; SEARLE.)
3.6.3.4 Functionalism and Consciousness Perhaps the most widely discussed
problem for functionalist accounts of the mind is their apparent inability to
give an intuitively acceptable account of conscious or experiential states. And
though it is not necessarily the simplest way of putting it, a particularly
striking way to outline the problem is in terms of the so-called 'inverted spec-
trum' first discussed by Locke. Imagine that you and I are looking at an
expanse of blue. Both of us will say that what we see is blue and both of us
will be able to match what we see with other samples of blue. Summing up a
long story, it seems perfectly reasonable to assume that, insofar as we are
both equipped with physiologically normal visual systems. our behaviours in
respect of this expanse will be indistinguishable. We will be. or could be ima-
gined to be. functionally identical at least in respect of our colour judgments
and behaviours. Yet. as so many are willing to accept. none of this shows
that what we experience - what it is like 'internally' for each of us to see
blue - will match. For all I know. when you look at the blue expanse. what
you experience is what I would if I were to look at something yellow. The
matching of our behaviours only seems to show that we have. as it were.
adapted to the use of public language and behaviour in a uniform way. If. as
is imagined. our colour experiences are inverted with respect to one another.
there is no way that this would show up. though it is no less a difference
between us. If it is genuinely possible for there to be such spectrum inversion.
then this would seem to show that functionalism is in trouble. For it can
scarcely be true that our experiences of seeing blue consist in some complex
103
AN ESSA Y ON MIND
pattern of inputs and outputs since these match, though our experiences do
not.
There are many variations on this sort of example and many responses have
been made to them by functionalists. Nonetheless, there is a widespread feeling
that consciousness is just too elusive and subjective a phenomenon to be cap-
tured by functionalism. (See CONSCIOUSNESS; DENNETT; FUNCTIONALISM; INTRO-
SPECTION; LEWIS; QU ALIA; SEARLE; SUBJECTIVITY.)
3.7 Eliminating
As we saw at stage 2, the commonsense or 'folk psychological' conception of
the mind is shot through with problems. In itself this is perhaps not enough of
a reason to reject that conception, but when these difficulties are combined
with those arising from attempts to reconcile the mental and the physical
realms, one can understand why someone might be tempted by one or another
rejectionist line. Dispensing with the commonsense conception - eliminating
the 'Mind' box in figure 5 - would at a stroke save us from a host of philo-
sophical problems intrinsic to that conception and would remove the threat of
overdetermination. Eliminativism, as the view is described, has taken two
forms, and, though one is more worked out and current than the other, both
will be discussed.
3.7.1 Eliminative Behaviourism
The behaviourism discussed earlier aimed to superimpose the mind on action,
attempting thereby to make the relation of the mental to the physical un-
problematic. Crucially, this kind of behaviourism counts as an account of the
mind, rather than as any sort of attempt to dispense with it. Conceding, for
example, that there really are states of mind called 'beliefs', the aim would be
to define these mental states in terms of what the believer does or is disposed to
do. This concessive sort of behaviourism is largely the product of philosophical
labour, whereas the more ruthless eliminative behaviourism can be found in the
writings of certain psychologists during the period from about the early part of
this century until about the mid-1960s. These psychologists, unhappy with
what was they saw as the unscientific approach of psychology in the nine-
teenth century, insisted that only what was observable could be studied scien-
tifically. Because they implicitly accepted the idea that the mind - as described
in the philosophical and early psychological literature - was not observable,
they adopted, as a methodological principle, the view that only behaviour
measurable under experimental conditions was fit for psychological study. In
some writers, this methodological principle became more than that: it became
the view that there really was no such thing as the mind. (See PSYCHOLOGY AND
PHILOSOPHY.)
Post-behaviourist, so-called 'cognitive' psychologists, count themselves no
less scientific, but they have rejected methodological behaviourism in part
104
STAGE 3: BEDROCK
because they have rejected the too simplistic idea that one can only scien-
tifically study the directly observable (see COGNITIVE PSYCHOLOGY; DEVELOP-
MENTAL PSYCHOLOGY). In virtually every field of science, one deals both with
observable behaviour and with the unobservable - but no less real - under-
lying causes of this behaviour. This is as true of the chemical behaviour of
molecules as it is of the biological behaviour of genes, so why deny it in the
case of the human mind? In finding no reason to cleave to the methodology of
behaviourism, contemporary psychology offers no room for the more elimin-
ative scepticism about the mind which characterized the writings of earlier
psychologists such as B. F. Skinner.
3.7.2 Eliminative Materialism
The pull of rejectionism has never expressed itself in the philosophical literature
by any straightforward advocacy of eliminative behaviourism. To be sure, there
are deeply behaviourist leanings in many writers, but these are combined with
a recognition that behaviour alone is not a sufficient base from which to
mount the campaign. The aim being to eliminate the commonsense conception
of the mind from serious scientific consideration, there is every reason to appeal
both to behaviour and to those internal processes and events that cause, but
are not themselves identifiable with, behaviour. (See QUINE.)
Because this sort of rejectionism is based more broadly on accounts of our
physical constitution and functioning, it is generally called eliminative materi-
alism. And interestingly (though also confusingly) its most recent manifesta-
tions are grounded in the computational model of the mind - a model which,
as discussed earlier, also lies behind functionalism. Here in rough outline is
how this has come about.
The core idea of functionalism does not require the computational model,
though, as was pointed out, the availability of the computational account
makes it easier to accept as possible the defining of functional relationships of
great complexity. However, given this, it is possible to detach the idea that the
mind is like a computer program from its typical functionalist setting. In parti-
cular, one might be tempted to argue as follows:
1. The brain is responsible for those phenomena we call 'mental'.
2. But the brain itself is a complex input/output device whose workings can
be captured by some specific computational model.
3. There is no guarantee that the elements of this model will correspond to
the commonsense or folk psychological categories we employ.
4. Given the problems that the commonsense conception engenders. and the
fact that it has no real scientific credentials, there is reason to think that
there will be no correspondence between computational and everyday cat-
egories.
5. The most plausible computational model of the brain is a CONNECTIONIST
one. There is everything to suggest that attitude states and their contents -
105
AN ESSAY ON MIND
or anything like them - will simply not figure as elements in the connec-
tionist account.
Therefore, putting (4) and (5) together, we seem to have both philosoph-
ical and empirical reasons to think that our account of the brain - and
thus the 'mind' - will simply have no need of the commonsense concep-
tion. Or, even more strongly, (4) and (5) might be taken as showing that
all commonsense claims about beliefs, desires, intentions, and the rest are
just false.
Fully understanding this argument requires a great deal more than can be
provided here. Setting out the differences between connectionist and other
computational models as well as giving the detailed arguments supporting (3),
(4) and (5) would make this essay impossibly long. Added to this is the fact
that the argument combines elements from several different eliminative materi-
alist writers. (See FOLK PSYCHOLOGY, PSYCHOLOGY AND PHILOSOPHY.) The emphasis
on connectionism is by no means uncontroversial: there are connectionists
who are not eliminativists and there are non-connectionists who are. Addition-
ally, there are those whose eliminativism comes, not from any interest in
computational models, but more directly from physiological accounts of the
brain. (Many of those who work in neurophysiology hold a position which flirts
with eliminativism, though they call themselves 'reductionists'. The potential
for confusion here is enormous, but as this Essay aims to sort out philosophical
positions, terminology in the neurophysiological camp is not all that import-
ant.)
Evaluating eliminative materialism is not easy. Of course, in favour of this
extreme position, it must be allowed that it gives a neat and direct answer to
the problems posed by the triangle in figure 5. For if we needn't trouble our-
selves about a box labelled 'Mind', then there is no threat of overdetermination.
But, as has been pointed out in lots of different ways, the price paid for this
kind of resolution is high: it requires the almost complete abandonment of our
ordinary conception of ourselves. Some eliminative materialists point out that
such a wholesale overturning of everyday conceptions is not unprecedented.
After all, our ancestors are generally presumed to have thought of the natural
world as animated by all sorts of spirits, and at some point this conception
simply collapsed in the face of the growing body of scientific knowledge. Yet, in
spite of this consideration, many have found the price to be paid for elim-
inativism just too high. For example, it has been argued that there is some-
thing self-defeating about it: if there are no such things as beliefs, then we
cannot truly believe in eliminativism.
3.8 Conclusion
Eliminativism is the kind of view I had in mind in the opening paragraph of
this essay when I suggested that difficulties both within our ordinary concep-
106
STAGE 3: BEDROCK
tion of the mind and in respect of the mind's relation to the physical realm can
tempt us to take up extreme positions. And what could be more extreme than
recommending the wholesale abandonment of the scheme by which we have
up to now understood ourselves? Of course, there are many alternatives to this
extreme position, the main outlines of which have been described. That none of
them is wholly satisfactory would be something with which most philosophers
of mind would readily agree. Yet as will be apparent from reading this Com-
panion, philosophers have not been tempted by any sort of quietism - they have
not thought that, given all the problems, it would be better to have no opinions
at all.
107
PART II
A COMPANION TO THE PHILOSOPHY OF
MIND, A-Z
A
action (1) In contrast to what merely
happens to us, or to parts of us, actions
are what we do. My moving my finger is
an action, to be distinguished from the
mere motion of that finger. My snoring
likewise is not something I 'do' in the
intended sense, though in another, broader
sense, it is something I often 'do' while
asleep.
The contrast has both metaphysical and
moral import. With respect to my snoring,
[ am passive, and am not morally respons-
ible (unless, for example, I should have
taken steps earlier to prevent my snoring).
But in cases of genuine action, I am the
cause of what happens, and I may prop-
erly be held responsible (unless I have an
adequate excuse or justification). When 1
move my finger, [ am the cause of the
finger's motion. When 1 say 'Good
morning!' [ am the cause of the sound.
True, the immediate causes are muscle
contractions in the one case and lung, lip
and tongue motions in the other. But this
is compatible with me being the cause -
perhaps I cause these immediate causes, or
perhaps it just is the case that some events
can have both an agent and other events
as their cause.
All this is suggestive, but not really ad-
equate. We do not understand the intended
force of '[ am the cause' any more than we
understand the intended force of 'Snoring is
not something 1 do'. If 1 trip and fall in your
flower garden, 'I am the cause' of any
resulting damage; but neither the damage
nor my fall is my action.
We will consider four approaches to
explaining what actions, as contrasted with
'mere' doings, are. But it will be convenient
first to say something about how they are to
be individuated.
INDIVIDUATION OF, ACTIONS
If I say 'Good morning!' to you over the
telephone, 1 have acted. But how many
actions have 1 performed, and how are they
related to one another and associated
events? We may list several descriptions of
what 1 have done:
(1) move my tongue and lips in certain
ways, while exhaling;
(2) say 'Good morning';
(3) cause a certain sequence of modifica-
tions in the current flowing in our tele-
phones;
(4) say 'Good morning!' to you;
( 5) greet you.
This list - not exhaustive, by any means - is
of act types. 1 have performed an action of
each of these types. Moreover an asym-
metric relation holds: 1 greeted you by
saying 'Good morning!' to you, but not the
converse, and similarly for the others on the
list. But are these five distinct actions 1 per-
formed, one of each type, or are the five
descriptions all of a single action, which
was of these five (and more) types (see
TYPE/TOKEN)? Both positions, and a variety
of intermediate positions, have been de-
fended.
On the first view, a particular action (act
token) just is the exemplification by an agent
of an act type or property at a time
(Goldman, 1970, p. 10). A difference in any
of these three means a different action.
Assuming for example that 'saying "Good
morning!" to you' is an act property distinct
from the act property 'saying "Good
morning!" " my doing the one and my
doing the other were two different actions 1
performed on this occasion. We can call this
111
ACTION (1)
the 'maximizing' view. On the second view,
the 'minimizing view', particular actions are
never more than movements of (parts of)
the body; but these movements can be
described in terms of their consequences
and circumstances (Davidson, 1980, p.
58f). My saying 'Good morning!' just was
my moving tongue, lips, and so on, in ways
appropriate for producing a sound recogniz-
able by other English-speakers as an utter-
ance of 'Good morning'. This action had
modification of the telephone current as an
effect, and so it - the very same action -
may be described as my causing those
modifications. And because sound was pro-
duced at the other end, you were there to
hear it, and the conventions of English are
what they are, descriptions (4) and (5)
apply also. But there was only one action.
One type of 'intermediate' view regards
actions as events that can have 'smaller'
actions and other events, especially their
consequences, as components (cf. Ginet,
1990, pp. 49-50, for a similar view). My
saying 'Good morning' to you consisted of
the tongue and lip, etc., movements, and
also the resulting sound, current modifica-
tions, sound at the other end, and impact
on your auditory system. My greeting you
was the same action, differently described.
But my causing the current modifications
did not include the sound at the other end
nor the impact on you, and so was only a
component, a proper part, of my greeting
you.
Many variations of these three basic
views have been developed and defended in
the last two or three decades. Much of the
debate has featured odd-sounding implica-
tions of one or another view, countered by
attempts to down-play their significance.
The maximizing view implies that every
time we perform one action, we in fact
perform indefinitely many. For besides the
five so far mentioned, I made a sound, said
something in English, used an electronic
device, ... The minimizing view limits
actions in time and space to movements of
the body, even if the descriptions employed
entail the occurrence of more remote effects
of those movements. To use the standard
112
example, this means that I can shoot and
even kill you before the bullet reaches your
body. Intermediate views to the contrary
treat these consequences as literally part of
the action. But this means that my killing
someone is still in progress so long as the
victim clings to life - even if I myself die in
the interim! None of the views seems to
harmonize with all our intuitions, and it is
unclear what might be compelling reason to
favour one of the theoretical conceptions of
events and event-language these views
reflect (for general discussion, see Davis,
1979, ch. 2; Ginet, 1990, ch. 3). Our dis-
cussion, then, will be phrased neutrally as
far as possible; but occasionally the mini-
mizing view will be adopted for ease in
exposition.
WHA T ACTIONS ARE
We return to the problem of distinguishing
actions from other doings. Here are four
general claims:
(1) Actions are doings having mentalistic
explanations of a certain sort.
(2) Actions are doings that are intentional
under some description.
(3) Actions are doings that begin with a
certain kind of event.
(4) Actions are doings of which the doer
has a certain kind of awareness.
These claims are not mutually incompatible;
many theorists would assert all four. We
will discuss them separately, however, as if
each were offered as a complete account of
being an action. We will also limit our
attention to actions that involve bodily
movements, excluding purely mental
actions (e.g. mental arithmetic) and so-
called 'negative actions' (e.g. refraining
from nodding one's head).
(1) Actions as Doings having a Mentalistic
Explanation
Coughing is sometimes like snoring and
sometimes like saying 'Good morning!' -
that is, sometimes a mere doing and some-
times an action. It is the latter when it is
deliberate. And deliberate coughing can be
explained by invoking an intention to
cough, a desire to cough or some other 'pro-
attitude' toward coughing, a reason for
coughing or purpose in coughing, or some-
thing similarly mental. Especially if we
think of actions as 'outputs' of the mental
'machine' (see FUNCTIONALISM), the avail-
ability of such a mentalistic explanation
may well seem to be the crucial factor for
status as an action.
Advocates of (1) differ on just what sort
of mentalistic explain ability is requisite, and
how that explain ability is itself to be under-
stood. A popular view for example holds
that actions are things the agent does for a
reason and that doing something for a
reason is doing it because the agent believes
it will or may lead to something the agent
wants (or has some other 'pro-attitude'
toward). Many understand this 'because' as
causal: my desire to attract your attention,
say, together with my belief that coughing
will have this effect, actually causes my
coughing, and this is why my coughing is
an action (see REASONS AND CAUSES).
Others have offered non-causal interpret-
ations of this 'because'. (See Davis, 1979,
ch. 5, and Mele, forthcoming, for references
and general discussion.)
The causal interpretations face the
problem of deviant causal chains. Suppose, for
example, I want and intend to get down on
my knees to propose marriage. Contem-
plating my plan, I am so overcome with
emotion that I suddenly feel weak and sink
to my knees. Here, my sinking to my knees
was not an action even though it was
caused by my desire and intention to get
down on my knees. The problem is that my
sinking to my knees was not caused by the
intention in the right way - and advocates
of (1) have explored the possibility or
impossibility of spelling out what this 'right
way' is. One suggestion is that the intention
must 'proximately' cause the doing for the
latter to count as an action. In this
example, an emotional state intervened. Dif-
ficulties in determining when an action
begins (see below) may mean difficulties
also in determining whether a cause of a
putative action is really 'proximate'.
ACTION (1)
Another difficulty for (1) is the possibility
of genuine actions that have no mentalistic
explanation at all. For example, suppose at
dinner you ask me to pass the salt, and I do.
My passing the salt is a genuine action, but
may have been a direct response to your
request, with no role being played by any
desire to pass the salt or anything similar. I
have simply been trained to comply with
such requests in ordinary circumstances -
nor can this training be viewed, say, as
my having come to have a 'standing pro-
attitude' toward acts of compliance. It may
even be denied that I had any relevant
intention, though I did pass the salt inten-
tionally (see below, and INTENTION). True, I
had a reason for passing the salt: you asked
for it. But even my auditory perception of
your request is not the sort of reason propo-
nents of (1) have in mind. Similarly, a
purpose was served by my action of passing
you the salt; but it is doubtful that it was
my purpose.
Again, suppose while we are walking
outside together I happen to look up at the
sky. I may have had no reason whatsoever;
I just did it, spontaneously. Still, it was
something I did. I was the cause of my
head's motion. If there were any moral
issue, I could be held responsible, just as I
can be judged (favourably, I trust) for
passing the salt in the first example.
(2) Actions as Intentional Doings
The formulation 'Actions are doings that
are intentional under some description'
reflects the minimizing view of the indivi-
duation of actions. The idea is that for what
I did to count as an action, there must be a
description 'V-ing' of what I did, such that I
V'd intentionally. Recall the five descriptions
of my one action of greeting you by tele-
phone. I presumably did not know or care
about the modifications I was causing in the
current. Still, since (on the minimizing
view) my causing the modifications was the
same event as my greeting you, and I
greeted you intentionally, this event was
an action. Or suppose I did not know it
was you on the phone; I thought it was
my spouse. Still, I said 'Good morning!'
113
ACTION (1)
intentionally, and that suffices for this
event, however described, to be an action.
My snoring and involuntary coughing,
however, are not intentional under any
description, and so are not actions.
The adequacy of (2) depends on what it is
for an action to be intentional under some
description. If it requires being caused by an
intention, then (2) is just a version of (1)
and subject to the same counterexamples. If
it entails only that the agent have a certain
kind of knowledge of what he or she is
doing, then (2) is a version of (4), discussed
below. Another possibility is that being
intentional is a kind of goal-directedness in
the action itself. Thus we noted that if I pass
you the salt, I may have no purpose in
passing you the salt. But I had a purpose in
moving my hand as I did: to get the salt to a
position within your reach. We may specu-
late about feedback mechanisms involving
my eyes and hand which were 'set' to
'monitor' and 'guide' the motion of my
hand to ensure that that purpose was
accomplished. Perhaps this could be called
'intention in action', as opposed to an ante-
cedent intention, which might be a cause of
the whole action.
The adequacy of (2) given this last inter-
pretation again depends on whether there
are cases we would want to count as
genuine actions despite absence of the goal-
directedness described. To some extent this
may be an empirical matter. Consider again
the example of my spontaneously looking
up while we are walking. Was the motion
of my head goal-directed? Or did it result
from muscle contractions not 'monitored'
and 'guided' in any way? If the latter, my
looking up was not done intentionally
under any description - yet for all that, it
was a voluntary movement, and it seems
arbitary to exclude it from the class of
actions.
Digression: When Do Actions Begin? On one
important point the three theories of indi-
viduation presented earlier yield the same
result, or are at least compatible with it.
That is the question of when actions begin.
Examining this question will prepare us for
114
the question of what actions begin with, and
to the suggestion (3), that they begin with a
unique kind of event, from which their
status as action derives.
Recall the three theories. Whether my
killing someone ends when the victim dies
or earlier, it begins when the act by which I
do it - pulling the trigger, say - begins. I do
not pull the trigger and then kill the victim,
even if the two are entirely distinct actions,
as on the maximizing view. On the mini-
mizing view, there is only one action, so of
course there is only one beginning point.
And on intermediate views, my pulling the
trigger is the first part of my action of
killing; so again, there is just one beginning
point.
When is that beginning point? The ini-
tially surpriSing answer is: before the trigger
begins to move, even before the finger
pulling the trigger begins to move. This is
because I move that finger by contracting
certain muscles; and those muscles begin to
contract before the finger begins to move. I
do not contract the muscles and then move
my finger, so I must begin moving my
finger when I begin contracting the
muscles. Furthermore, the muscle contrac-
tions are caused by efferent neural impulses,
whose role in the action seems comparable
to that of the muscle contractions them-
selves. This suggests that the action does
not begin any later than those impulses. (Cf.
Hornsby, 1980, for similar argument. We
are, of course, generally ignorant of these
muscle contractions and neural impulses. It
is unclear what relevance this might have.
Note our similar ignorance of the tongue
and lip movements and current modifica-
tions in descriptions (1) and (3) of our first
example.)
In this connection we may note work by
Libet (1985), which ostensibly determined
moments by which agents judged that they
had begun certain intentional limb move-
ments. These moments preceded onset of
actual motion of the limb by some tens of
milliseconds, suggesting that English-
speakers implicitly understand their own
actions as beginning no later than certain
neural impulses which precede the muscle
contractions. (But this work and its inter-
pretation are very controversial. See e.g.
Dennett and Kinsbourne, 1992.)
(3) Actions as Doings Beginning with 'Act-
beginners'
To sayan action of moving a finger begins
'no later than' the efferent neural impulses
that cause the muscle contractions that
cause the finger's motion is still not to say
when the action begins. Nor is it to say
what it begins with. An 'act-beginner' in
such a case would be an internal event
which causes the neural impulses and is
itself something the agent 'does', in the
broad sense of 'doing'. Suggestion (3) is that
there are such act-beginners, and it is in
virtue of them that the doings they begin
are actions. Alleged examples include events
in which the agent:
(a) agent-causes some (presumably cere-
bral) event;
(b) wills to do something (e.g. move the
finger); or
(c) tries to do something (e.g. move the
finger).
Some versions of (3) treat act-beginners as
themselves always actions - doings in the
narrow sense - as well (e.g. Davis, 1979;
Ginet. 1990). According to others, an
action has been performed only if the act-
beginner has had suitable consequences
and/or has occurred in appropriate circum-
stances (e.g. Hornsby, 1980).
(a) 'Agent causation' is supposed to be a
causal relation that holds irreducibly
between an agent and certain events (see
Chisholm, 1966; Taylor, 1966). I move my
finger, and my doing so is an action,
because I certain of the neural
events that cause the muscle contractions
that cause my finger's motion. If 'e' refers
collectively to those neural events, then my
agent-causing e is my doing, and this doing
cannot be analysed, say, as the occurrence
of another, earlier, event c which causes e.
My action begins with my agent-causing e -
and the time of my agent-causing e is pre-
cisely the same as the time of e itself.
ACTION (1)
(b) and (c) Many older accounts treated
bodily actions such as moving a finger as
things done by 'acts of will' or 'volitions'.
Recently, a number of philosophers have
spoken of 'tryings' or 'attempts' instead. I
move my finger by trying to move it.
Success is so much to be expected that I
don't think of myself as 'trying' at all; I
'just' move the finger. Much is made, then,
of cases like one described by William
J ames: a blindfolded person is asked to move
his anaesthetized arm, which is held down.
The person is then 'astonished' to discover
that his arm has not moved (James, 1950,
II, p. 105). This is taken to show that the
person tried to move the arm, in what he
thought was a normal situation, implying
that we also 'try' whenever we perform
bodily actions. (See Hornsby, 1980, and
Davis, 1979, for additional argument and
references on the ubiquity of trying.)
Finally, several recent theorists have revived
the terms 'volition' and 'willing'. Ginet
(1990) thinks volitions are mental actions,
known to us by their 'actish phenomenal
quality', and in every case having some
'exertion of the body' as their object. For
Davis (1979), they are events of a kind
characterized (in the spirit of functionalism)
by their relations with their typical causes
and effects; and their objects may extend
beyond the body (e.g. a volition to tie one's
shoe). Both authors use the terms 'volition'
and 'willing' virtually interchangeably with
'trying' and 'attempt' (See THE WILL).
All three theories (a), (b) and (c), give us
the same picture of bodily action involving
act-beginners and the series of events they
cause, culminating in motion of limbs or
other parts of the body. Precisely where the
action itself is located varies with one's
theory of individuation. According to all,
the action begins when the act-beginner
begins. On intermediate views of act indi-
viduation, the action consists of the act-
beginner plus the motion of, say, the finger,
and all the events 'in between'. On the
maximizing view, the actions are abstract
entities distinct from these mereological
sums, but having the same temporal limits.
Most interesting is the minimizing view:
115
ACTION (1)
according to it, actions just are the agent-
caused events, tryings, or willings with
which they begin (so the term 'begin' is
misleading), Moving one's finger, for
example, is entirely something that takes
place inside one's skin - presumably inside
one's brain. The motion of the finger is just
a consequence - though the act-beginner is
describable as 'moving one's finger' only
because it has this consequence.
(Different views are possible regarding the
end-points of the act-beginners. We noted
earlier that some actions seem to involve
what could be called 'intention in action' -
perhaps feedback mechanisms for ensuring
that a goal. a certain end-state, is reached.
Such mechanisms could be considered part
of the act-beginner. This would mean that a
trying, for example, lasts until the relevant
body part's motion is completed. One says
'Good morning!' by trying to say it, and the
trying lasts until the last syllable is pro-
nounced.)
A main difficulty facing all these theories
is the possibility that the act-beginners
posited simply do not exist, or are not
present in all cases we want to call cases of
action. Many critics find the very idea of
agent-caused events obscure, and the
accounts of volitions and even tryings
unconvincing. Perhaps there is no moment
at which an action begins, even if there are
moments (even prior to limb motion) by
which it has begun. (Think of rivulets
joining to form a stream and then a river:
must there be a precise point at which the
river begins?) There may be an intention,
and then an action-in-progress, with
nothing determinate marking the transition
from the one to the other. If so, it may be
actually misleading to posit act-beginners.
Or there might be act-beginners, but not
for all actions. Consider what one does
while driving - especially when, as we say,
one's attention is elsewhere, perhaps on a
conversation one is having simultaneously.
A stop sign suddenly becomes visible, and
one steps on the brake pedal. The stepping
on the brake pedal here seems clearly to be
an action, even intentional under that
description; though like the earlier example
116
of passing the salt it is unclear what, if any,
mentalistic explanation it can be given. Our
new question is whether it began with an
act-beginner, especially a trying or volition.
How could we tell? Perhaps the physiology
and psychology of this sort of leg-movement
is so different from what happens in the self-
conscious and isolated movements used
most often as philosophers' examples, that
there is no point to saying it begins with the
'same' sort of event. a trying, for example.
Do we know enough to rule this possibility
out?
(4) Actions as the Objects of Awareness of
Agency
The last claim to be discussed is a sugges-
tion based on features of our knowledge of
our own action. A number of writers have
stressed that this knowledge has at least
two sources: perceptual (including kinaes-
thetic) and non-perceptual (Davis, 1979,
pp. 15f, 6lf; Ginet. 1990, p. 28). This is one
of the lessons of the case described by
William James: an agent deprived of percep-
tual information believes he has acted,
entirely on the basis of some non-perceptual
source. Normally, one knows by perception
that one's arm, or whatever, is moving; but
awareness that one is moving it, awareness
of agency, is non-perceptual. (Awareness of
why one is moving it. what one's reason or
intention is, if any, is also non-perceptual.
but presumably has a different etiology.)
The view just discussed posits a kind of
event, an act-beginner, as the source of this
non-perceptual awareness, but we ended
the discussion suggesting that there may
not be a single kind of event playing this
role in every case of action. It may turn out
that there is no kind of event playing this
role, or that there are a number of very dif-
ferent kinds of events playing it. But still, it
may be that in each case of action there is
non-perceptual awareness of acting. Or,
rather, we can regard the class of doings
that are actions as in effect defined by the
doer's readiness - based on non-perceptual
awareness - to claim the doing as his or her
own, to express sentiments tantamount to '[
am the cause'.
This suggestion may sound circular and
vacuous. But it is not quite as empty as
'Actions are whatever a person believes are
actions' would be. First, there is the require-
ment that the belief have a non-perceptual
source. This rules out the possibility, for
example, of someone's falling, or stumbling,
being classed as an action. One knows of
either only perceptually. Second, the
'persons' to which it is meant to apply are
rational agents (like us) who have learned
(like us) to act for reasons, etc., to be aware
of their actions, to use this awareness
purposively to modify their subsequent
conduct, and so on. This ensures that their
understanding of agency, their 'feel' for the
distinction between what they 'do' and
what merely happens, will arise from their
being agents, in line with the often-repeated
claim that agency can be understood, the
concept of action acquired, only 'from the
inside'.
See also ACTION (2); BELIEF; DESIRE; An
Essay on Mind section 2.3; INTENTION-
ALITY; PROPOSITIONAL ATTITUDES; THE
WILL.
BIBLIOGRAPHY
Chisholm, R.C. 1966. Freedom and action. In
Freedom and Determinism, ed. K. Lehrer. New
York: Random House, pp. 11-44.
Davidson, D. 1980. Essays on Actions and
Events. Oxford University Press.
Davis, L.H. 1979. Theory of Action. Englewood
Cliffs, N.J.: Prentice-Hall.
Dennett, D.C., and Kinsbourne, M. 1992. Time
and the observer: the where and when of
consciousness in the brain. Behavioral and
Brain Sciences, 15, 183-247.
Ginet, C. 1990. On Action. Cambridge Uni-
versity Press.
Goldman, A.I. 1970. A Theory of Human
Action. Englewood Cliffs, N.J.: Prentice-Hall.
Hornsby, J. 1980. Actions. London: Routledge
& Kegan Paul.
James, W. 1950. Principles of Psychology, 2
vols. Henry Holt, 1890; New York: Dover,
1950.
Libet, B. 1985. Unconscious cerebral initiative
and the role of conscious will in voluntary
ACTION (2)
action. Behavioral and Brain Sciences, 8,
529-66.
Mele, A.R. 1992. Springs of Action. Oxford Uni-
versity Press.
Mele, A.R. forthcoming. Recent work on inten-
tional action. American Philosophical Quar-
terly.
Taylor, R. 1966. Action and Purpose. Engle-
wood Cliffs, N.J.: Prentice-Hall.
LAWRENCE H. DA VIS
action (2) A central question in the philo-
sophy of action is what distinguishes a per-
son's action from something that merely
happens to him, such as sneezing or
snoring. A popular answer to this question
appeals to a causal theory of action (see
REASONS AND CAUSES). Genuine actions or
deeds, on this approach, are events with a
distinctive internal cause, such as a desire,
intention, or volition (see THE WILL). Sneez-
ing, snoring, and stumbling over a rug are
not actions because they are not caused by
suitable desires or intentions.
Two problems face this kind of theory.
First, many actions, especially automatic
ones, do not seem to be preceded by any plan
or intention to perform them. In playing a
complex passage on the piano, it does not
seem necessary for the performer to have a
distinct intention or plan for each note
played, at least if this means a conscious or
introspectively available plan. In fact deliber-
ate forethought may interfere with smooth
performance. Second, there is the problem of
'deviant' or 'wayward' causal chains. To
borrow an example from Davidson (1973), a
climber might want to rid himself of the
weight and danger of holding another man
on a rope, and he might know that by
loosening his hold on the rope he could rid
himself of the weight and danger. This desire
and belief might so unnerve him as to cause
him to loosen his hold, and yet it might be
that he never chooses to loosen his hold, nor
is this something he intentionally or volun-
tarily does. Apparently, not every causal
relation between seemingly appropriate
mental events and behavioural upshots
qualifies the latter as voluntary acts. The
117
ACTION (2)
right sorts of causal relations would have to
be specified if the causal theory is to succeed.
Despite these difficulties, the causal
approach remains promising. Models of
action in cognitive science may be able to
explain what is common to automatic and
deliberate action, and to identify the causal
routes that are distinctive of action as
opposed to mere happening. We shall con-
sider a model due to Norman and Shallice
(1986; see also Shallice, 1988, ch. 14),
which postulates two complementary pro-
cesses that operate in the selection and
control of action. The first process is
invoked to explain the ability of some action
sequences to run off automatically without
conscious control or the use of attentional
resources. The second process allows for
deliberate conscious control to initiate,
guide, or modulate the course of action.
The first process is utilized in selecting
and executing simple, well-learned, or ha-
bitual skills such as typing, or knotting a
necktie. It is assumed that the motor control
of well-learned movement is represented by
means of a motor 'schema' or family of
motor schemata (see Rumelhart and
Norman, 1982). A motor schema is an
organized unit of MENTAL REPRESENT A-
TION, which has as its output the control of
body movements. A set of such schemata
operates as a motor 'program', understood
as a flexible, interactive control structure
capable of calling upon sub-programs and
making local decisions as a result of current
conditions. The basic framework in this
approach is called an 'activation trigger
schema' (ATS) system. Each schema has, at
any time, an activation value that reflects
the total amount of excitation it has
received. The normal, resting value for a
schema is zero. It can increase when the
schema is activated, or decrease when the
schema is inhibited. Various factors can in-
fluence the momentary activation value of a
schema, including activation of a related
schema (especially a 'parent' or 'source'
schema), inhibition by rival schemata, and
satisfaction of 'trigger conditions'. which
specify environmental circumstances that
make the act in question feasible or appro-
ll8
priate. A schema is selected when its activa-
tion level exceeds a specified threshold. Once
selected, it continues to operate (unless
actively switched off) until it has satisfied its
goal or completed its operations. The opera-
tion of a schema often consists in calling
further ('child') schemata.
A concrete example can be given from
the domain of typing (Rumelhart and
Norman, 1982). A skilled typist will have a
schema for a keypress action associated
with each character on a standard key-
board. Each keypress schema specifies a
target position, where position would be
encoded in terms of finger and palm move-
ments within a keyboard-centred coordinate
system. A response sub-system feeds back
information to the keypress system about
the current location of the fingers. If a
certain schema is highly activated, and if
the current finger position is within some
criterion distance of that schema's target
position (thereby satisfying the trigger con-
ditions), the actual keystroke is launched.
Shortly after the launch, the keypress
schema deactivates itself, resulting in a
release of inhibition for all the succeeding
keypress schemata. The system operates
normally during the launch, and other key-
press schemata may have their triggering
conditions met and launch their own key-
strokes before the earlier ones have been
completed. People can type very fast; typing
champions reach close to 200 words per
minute (17 letters per second). This is pos-
sible, according to the Rumelhart-Norman
model, because people carry out many
actions at once. Responses are prepared and
executed in parallel, so although the inter-
val between two completed keystrokes may
be, say, 60 milliseconds, each response
takes much longer than that to complete.
Norman and Shallice call the foregoing a
'contention scheduling' mechanism. It
permits simultaneous selection and execu-
tion of cooperative acts and prevents simul-
taneous performance of conflicting ones.
This mechanism normally proceeds quite
automatically. Some tasks, however, require
deliberate attentional resources, viz., tasks
that fall within the following categories: (1)
they involve planning or decision making,
(2) they require troubleshooting, (3) they
are ill-learned or contain novel action
sequences, (4) they are judged to be danger-
ous or technically difficult, or (5) they
require overcoming a strong habitual
response or resisting temptation. Under any
of these circumstances, use of an additional
system is necessary: the 'supervisory atten-
tional system' (SAS). SAS is a source of con-
scious control of the selection of schemata,
which Norman and Shallice identify with
the 'will'. SAS does not operate autonom-
ously, however, but only via the contention
scheduling mechanism, by applying extra,
'attentional' activation or inhibition to bias
the selection of certain schemata. In other
words, SAS does not directly control action
selection, but only exerts itself indirectly
through its effect on activation values.
Thus, the basic mechanism of selection by
activation values occurs in all voluntary
behaviour, whether it is consciously willed
or purely automatic.
Evidence for two levels of action control
comes from clinical neuropsychology. It is
well known that lesions confined to the pre-
frontal structures leave the execution of
basic skills such as the use of objects, speak-
ing, and writing unaffected, but can impair
the performance of tasks involving novelty,
error correction, or planning. Such lesions
appear to involve damage to the SAS but not
to the contention scheduling mechanism.
Let us return now to the problem of dis-
tinguishing action from mere undergoing of
bodily movement. If the Norman-Shallice
model is correct, voluntary action obviously
cannot be identified with movement caused
by 'will', or by conscious plans or inten-
tions. Not only mayan action sequence be
run off without attention to each step, but
an action may even be initiated without
deliberate attention or awareness, as in
beginning to drink from a glass while ab-
sorbed in conversation. The critical causal
events distinctive of action, then, need not
be conscious or deliberate. Just a sequence
of schema activations and selections may
constitute the necessary and sufficient
causal elements to qualify a movement as a
ACTION (2)
voluntary action. Passive bodily happen-
ings, such as accidentally loosening one's
grip on the rope, would not be preceded by
this sort of sequence.
Against this proposal it may be argued
that an account of the ordinary concept of
voluntary action should appeal to factors
within the ken of ordinary people (cf.
McCann, 1974, pp. 462-3). But the causal
factors introduced here seem inaccessible to
the ordinary person; they are apparently
available only through scientific research.
How can these factors, then, be the ones the
ordinary person uses in distinguishing
between voluntary action and bodily hap-
pening?
The second premise of this argument is
questionable: the critical causal factors
posited here may well be accessible to the
ordinary person. Philosophers (e.g. Ans-
combe, 1957) speak of 'non-observational
knowledge' of what one is doing, where the
knowledge is not of a successfully executed
movement but of an attempt to act. This
may well consist in an awareness of the
selection of an action schema, or a
'command' to the motor system. Cognitive
theorists of motor action commonly postu-
late feedback processes by which informa-
tion about the ongoing execution of motor
programs is given to the controlling system
(e.g. Wright, 1990). The philosopher's talk
of 'non-observational knowledge' may be
translatable into the cognitive theorist's lan-
guage of 'feedback'. Is there empirical evid-
ence of feedback from events of schema
selection or action commands, as opposed to
feedback from actual muscle movements?
Yes there is. A study by McCloskey, Cole-
batch, Potter, and Burke (1983) found that
normal subjects attend to different signals
(feedback) when asked to judge the precise
time of an internal 'command' to contract
their hand muscles than when they are
asked to judge the time of actual movements.
Their subjective timing of movement com-
mands was also studied in the absence of
actual movements or local sensations from
movements. (This was achieved by requiring
subjects to attempt movements of the hand
during local anaesthetic or blockade of the
119
ACTION (2)
nerves to the forearm or hand). So there
seem to be distinct signals from movement
commands and from actual movements. The
former may provide the sort of 'feeling of
agency' characteristic of voluntary action, a
feeling which is absent in cases of mere
happening. A distinctive feeling would thus
be associated with the normal selection or
command of an action such as (voluntarily)
loosening one's hold on a rope, a feeling that
would be absent in a case of accidentally
losing one's grip on the rope.
It may still be queried whether it is legit-
imate to appeal to the neural events in
question in an account of the ordinary
concept of action. Is it not necessary that
the ordinary person be able to specify the
events in detail, which obviously is not pos-
sible here? No. The situation may be com-
pared to Putnam's (1975) treatment of
natural kind concepts. Just as a specification
of the detailed structure of water or gold
may be relegated to the special sciences, so
the special sciences, in this case psychology
and neuroscience, are needed to spell out
the detailed conditions distinctive of volun-
tary action. This fits an earlier suggestion of
the present author that a solution to the
problem of 'deviant causal chains' should be
assigned to the scientific study of mind
rather than to pure philosophy (Goldman,
1970, pp. 62-3. For doubts about this
strategy, see Bishop, 1989.)
The approach developed here has much in
common with certain conceptions of a voli-
tional theory of action, in which a volition is
not understood as a conscious occurrence
but as a postulated cognitive-neurophysio-
logical event. Our proposal is especially close
to that of Davis (1979), who gives a func-
tional characterization of a volition as 'an
event which is normally a cause of the
agent's belief that he is acting in a certain
way, and which normally causes such
doing-related events as make it true that he
is acting in that way' (Davis, 1979, p. 16; cf.
Ginet's account of volitions in Ginet, 1990).
Hornsby's (1980) notion of a trying may
also have kinship to the present conception,
and perhaps Brand's (1984) quasi-technical
notion of an intention as well.
120
It is fortunate that the present approach
does not place much weight on conscious
desires, intentions, or plans as the crucial
causes of voluntary action, because some
recent neurophysiological research points to
the possible conclusion that conscious
events have little causal role to play.
Research conducted by Libet (1985) mea-
sured both cortical activity that initiates
voluntary action and subjectively experi-
enced onset of an intention to initiate it
(Libet's 'W'). Libet found that the intention
was experienced about 350 milliseconds
after the cortical activity (though still 200
milliseconds before the motor performance),
which invites the conclusion that the con-
scious experience itself plays no causal role.
Libet's experiments are subject to rival
interpretations, however, so no such con-
clusion should be drawn prematurely.
In addition to the problem of distinguish-
ing action and happening, cognitive science
may help resolve other problems and
puzzles in the theory of action. Consider the
much-debated puzzle of WEAKNESS OF WILL.
So-called 'akratic' action is action wherein
the agent intentionally does something that
runs directly counter to his predominant
desire, intention, or judgment (Davidson,
1970). How is such action possible?
A solution to the puzzle, at least a partial
solution, may emerge from the same sort of
psychological theory examined earlier. As
we have seen, psychologists recognize that
actions can be triggered quite independently
of deliberate choice. Indeed, such actions
can sometimes proceed contrary to choice
or intention. One example is reading. When
we see a billboard on a highway, we cannot
help but read what it says, whether we
want to or not. The forms on the sign pro-
claim that they are letters and words, and
this is enough to initiate reading routines
(La Berge, 1975). A striking demonstration
of this phenomenon is the so-called 'Stroop
effect' (Stroop, 1935). Subjects are shown a
list of words printed in colour, and are
asked to name the colours of the print.
Quite diabolically, however, the words
themselves are colour words, and they
name colours distinct from those in which
they are printed. For example, 'green' is
printed in yellow, 'yellow' is printed in
black, and so forth. Since the subjects are
supposed to name the colours of the print,
they should not read the words at all,
which will only confuse them. But they
cannot help reading the words and there-
fore find the task extremely difficult. This is
an example of what philosophers call
akratic action, since the subjects deliber-
ately try not to read the words but do so
anyway. If the psychologists are right, the
phenomenon is explained by the presence of
an automatic selection mechanism that can
set actions in motion irrespective of, and
even in opposition to, the decisions of the
SAS, which embodies conscious, attentional
motivation. This explanation might work
for a wide range of so-called akratic actions.
See also ACTION (1); BELIEF; DESIRE; An
Essay on Mind section 2.3; COGNITIVE PSY-
CHOLOGY; CONSCIOUSNESS; INTENTION;
INTENTION ALITY; PROPOSITIONAL A TTI-
TUDES; THE WILL.
BIBLIOGRAPHY
Anscombe, G.E.M. 1957. Intention. Oxford:
Basil Blackwell.
Bishop,J. 1989. Natural Agency: An Essay on
the Causal Theory of Action. Cambridge Uni-
versity Press.
Brand, M. 1984. Intending and Acting: Toward a
Naturalized Action Theory. Cambridge, MA.:
MIT Press.
Davidson. D. 1970. How is weakness of the
will possible? In Moral Concepts, ed. J. Fein-
berg. Oxford University Press, pp. 93-113.
Davidson, D. 1973. Freedom to act. In Essays
on Freedom of Action, ed. T. Honderich.
Routledge & Kegan Paul. pp. l37-56.
Davis, L.H. 1979. Theory of Action. Englewood
Cliffs, N.J.: Prentice-Hall.
Ginet, C. 1990. On Action. Cambridge Uni-
versity Press.
Goldman, A.1. 1970. A Theory of Human
Action. Englewood Cliffs, N.J.: Prentice-Hall.
Hornsby, J. 1980. Actions. London: Routledge
& Kegan Paul.
La Berge, D. 1975. Acquisition of automatic
processing in perceptual and associative
AGENCY
learning. In Attention and Performance, vol.
5, ed. P. M. A. Rabbitt and S. Dormic.
London: Academic Press.
Libet, B. 1985. Unconscious cerebral initiative
and the role of conscious will in voluntary
action. Behavioral and Brain Sciences, 8,
529-539.
McCann, H. 1974. Volition and basic action.
Philosophical Review, 83, 451 -73.
McCloskey, 1'.1., Colebatch, J.G., Potter, E.K.,
and Burke, D. 1983. Judgements about
onset of rapid voluntary movements in
man. Journal of Neurophysiology, 49, 851-
863.
Norman, D.A., and Shallice, T. 1986. Atten-
tion to action; willed and automatic control
of behavior. In Consciousness and Self-
Regulation: Advances in Research and Theory,
vol. 4, ed. R. J. Davidson, G. E. Schwartz,
and D. Shapiro. New York: Plenum Press,
pp.1-18.
Putnam, H. 1975. The meaning of 'meaning'.
In Language, Mind and Knowledge, Minnesota
Studies in Philosophy of Science, vol. 7, ed.
K. Gunderson. Minneapolis: University of
Minnesota Press, pp. l31-93.
Rumelhart, B.E., and Norman, D.A. 1982.
Simulating a skilled typist: a study of skilled
cognitive-motor performance. Cognitive
Science, 6, 1-36.
Shallice, T. 1988. From Neuropsychology to
Mental Structure. Cambridge University
Press.
Stroop, J.R. 1935. Studies of interference in
serial verbal reactions. Journal of Experi-
mental Psychology, 18, 643-62.
Wright, C.E. 1990. Controlling sequential
motor activity. In Visual Cognition and
Action, ed. D. N. Osherson, S. M. Kosslyn,
and J. M. Hollerbach. Cambridge, MA.: MIT
Press, pp. 285-316.
ALVIN l. GOLDMAN
agency A central task in the philosophy of
action is that of spelling out the differences
between events in general and those events
that fall squarely into the category of
human action. An earthquake is certainly
something that happens, but it is not some-
thing done. Whereas when someone picks
up a hammer and drives a nail into a piece
of wood, this is both something that
121
ANOMALOUS MONISM
happens and something done or under-
taken. In this second case, one has a clear
case of agency - a case of an agent under-
taking to bring about some change in the
world. However, whilst it is easy enough to
locate the notion of agency with respect to
actions, it is not at all clear just what con-
stitutes it, whether it is a special feature of
the mind or is composed of such things as
intentions, desires and beliefs.
See ACTION; DESIRE; INTENTION; THE SELF;
THE WILL.
SAMUEL GUTTENPLAN
anomalous monism Monism is the view
that there is only one kind of substance
underlying all objects, changes and pro-
cesses. It is generally used in contrast to
DU ALISM, though one can also think of it as
denying what might be called 'pluralism' -
a view often associated with Aristotle which
claims that there are a number of sub-
stances (see HISTORY). Against the back-
ground of modern science, monism is
usually understood to be a form of materi-
alism or PHYSICALISM. That is, the funda-
mental properties of matter and energy as
described by physics are counted the only
properties there are.
The position in the philosophy of mind
known as anomalous monism has its histor-
ical origins in Kant, but is universally iden-
tified with Donald DAVIDSON'S views, and it
was he who coined the term. Davidson has
maintained that one can be a monist -
indeed, a physicalist - about the funda-
mental nature of things and events, whilst
also asserting that there can be no full
REDUCTION of the mental to the physical.
(This is sometimes expressed by saying that
there can be an ONTOLOGICAL, though not
a conceptual reduction.) To put it more
concretely, Davidson thinks that complete
knowledge of the brain and any related
neurophysiological systems that support the
mind's activities would not itself be know-
ledge of such things as belief. desire, experi-
ence and the rest of the mental. This is not
because he thinks that the mind is
somehow a separate kind of existence;
122
anomalous monism is after all monism.
Rather, it is because the nature of mental
phenomena rules out a priori that there will
be law-like regularities connecting mental
phenomena and physical events in the
brain, and, without such laws, there is no
real hope of explaining the mental via the
physical structures of the brain.
See also SUPERVENIENCE.
SAMUEL GUTTENPLAN
artificial intelligence Marvin Minsky,
founder of the MIT AI laboratory, has
defined Artificial Intelligence as 'the science
of making machines do things that would
require intelligence if done by men' (1968,
p. v). The field was given its name by John
McCarthy, who in 1956 organized the
conference that many AI researchers regard
as marking the birth of their subject: the
Dartmouth Summer Research Project on
Artificial Intelligence. The last two words
stuck.
One of the earliest lectures on computer
intelligence - possibly the earliest - was
given in 1947 by the British logician Alan
TURING, then working at the National
Physical Laboratory in London. The lecture
was entitled 'Intelligent Machinery, A Here-
tical Theory'. (Turing's notes for the lecture
are reproduced in Turing, 1959, pp. 128-
34.) 'My contention is that machines can be
constructed which will simulate the beha-
viour of the human mind very closely', said
Turing with remarkable far-sightedness. At
that time there were no more than two
electronic computers in existence, the Colos-
sus in Britain and the ENIAC in America,
both of them extremely primitive. (The
Manchester Mark I. the world's first stored-
program general-purpose electronic com-
puter, did not run its first program until
June 1948.) In 1948 Turing circulated a
startlingly original report on the prospects
for machine intelligence. In it he anticipated
many later developments, including CON-
NECTIONISM. In 1950 he published an
article entitled 'Computing Machinery and
Intelligence' in the philosophical journal
Mind (he was by this time Deputy Director
of the Computing Laboratory at Man-
chester). The article began 'I propose to
consider the question "Can machines
think?" '.
The first working AI program was a
checkers (or draughts) program that incor-
porated a learning mechanism. The pro-
gram rapidly picked up the skills of the
game and was soon able to beat its creator,
Arthur Samuel. This was the first heuristic
program to be fully realized on a computer.
(A heuristic program is one that follows
'rules of thumb', as opposed to following a
rule that is guaranteed to lead to the desired
result - in this case, that of winning the
game.) Samuel gave a demonstration of the
program on American TV in the early
1950s. Some years later, in 1956, came the
most fecund of the early attempts at AI, the
Logic Theorist, written by Allen Newell, Cliff
Shaw and Herbert Simon. The Logic Theor-
ist succeeded in proving 38 of the first 52
theorems presented in Whitehead and Rus-
sell's Principia Mathematica. In the case of
one theorem, the machine's proof was
shorter and neater than the one Whitehead
and Russell gave.
I shall not here consider Turing's general
question of whether a machine may prop-
erly be said to think or display intelligence
(for a discussion see chapter 3 of Copeland,
1993). Assuming an affirmative answer to
that general question, my concern here will
be with the issue of whether the specific
types of machines currently under investi-
gation in AI are of the right sort to display
general intelligence. With that topic in view
I examine two of the fundamental assump-
tions of AI research: the symbol system
hypothesis and the algorithmicity assump-
tion. The symbol system hypothesis is asso-
ciated more closely with the 'traditional'
approach to AI than with the connectionist
approach (which did not become widely
pursued until the mid-1980s). The algo-
rithmicity assumption is common to both
approaches to AI. First articulated and
explored within AI, this assumption has
now become a foundational one in much of
cognitive science and in much con-
temporary philosophy of mind.
AR TIFICIAL INTELLIGENCE
THE SYMBOL SYSTEM HYPOTHESIS
From the most basic PC to the most
advanced Cray, commercially available
computers share the same basic principles of
operation. All are symbol-processing en-
gines. Information is stored in the com-
puter's memory in the form of symbols,
programs are internally stored lists of sym-
bolically encoded instructions, and each step
of a computation consists of a simple opera-
tion on a symbol string. These symbol
strings that the computer stores and pro-
cesses are rather like sentences of a lan-
guage or code. Typically the code used is
binary in nature, which is to say that it
employs only two basic characters. Binary
code is easier to implement electronically
than a code like the one I am presently
writing in, which of course employs over 26
characters. The symbol system hypothesis is
the hypothesis that a symbol-processing
system - that is, a computer - can be set up
so as to exhibit general intelligence.
As I have said, there are at present two
major approaches to AI, that of the tradi-
tionalists and that of the connectionists.
Connectionists distance themselves from
symbol-processing and are exploring an
alternative approach to computation. Here is
the recipe traditionalists propose for building
a machine that is as intelligent as we are.
Use a language-like symbolic code to
represent real-world objects, events,
actions, relationships, etc.
Build up an adequate representation of
the world and its workings (including
human creations such as commerce)
inside a computer. This 'knowledge base'
will consist of vast, interconnected struc-
tures of symbols. It must include a repre-
sentation of the machine itself and of its
purposes and needs. Opinions differ as to
whether programmers will have to 'hand
craft' this gigantic structure or whether
the machine can be programmed to learn
much of it for itself.
Use suitable input devices to form sym-
bolic representations (in the same code)
123
ARTIFICIAL INTELLIGENCE
of the flux of environmental stimuli
impinging on the machine.
Arrange for complex sequences of the
computer's hard-wired symbol-processing
operations to be applied to the symbol
structures produced by the input devices
and to the symbol structures stored in
the knowledge base. Further symbol
structures result. Some of these are desig-
nated as output. The sequences of opera-
tions are selected by the computer by
means of an algorithm. (This term is
explained below.)
This output is a symbolic representation
of appropriate behavioural responses
(including verbal ones) to the input. A
suitable robot body can be used to
'translate' the symbols into real beha-
viour.
The symbol system hypothesis is simply
this: the recipe will work. According to the
hypothesis any general-purpose computer
with sufficient memory can, through
further internal organization, acquire
general intelligence. The nomenclature
derives from Newell and Simon (1976). The
symbol system hypothesis as here presented
is the sufficiency-part of their physical
symbol system hypothesis (the necessity-
part being the claim that symbol-manipula-
tion is necessary for general intelligent
action). For ease I drop 'physical'.
This hypothesis is a bold empirical con-
jecture. At present no one knows whether it
is true or false. Initially the impetus for
advancing the hypothesis was provided by
the success of programs like the Logic The-
orist and Samuel's checkers player. In
1958, on the basis of their experience with
the Logic Theorist, Simon and Newell
wrote: 'Intuition, insight, and learning are
no longer exclusive possessions of humans:
any large high-speed computer can be pro-
grammed to exhibit them also ... The sim-
plest way [we] can summarise the situation
is to say that there are now in the world
machines that can think, that learn, and
that create' (1958, pp. 6-8). Their enthu-
124
siasm led them to predict that by 1967 a
digital computer would be able to beat any
human at chess. A more recent prediction,
by Doug Lenat and Ed Feigenbaum, is for a
symbol-processing system with human-level
breadth and depth of knowledge by the
early years of next century (1991, p. 224).
Artificial intelligence is 'within our grasp',
declare Lenat and Feigenbaum (p. 188). Yet
the truth of the matter is that after nearly
half a century of grappling with some of the
hardest problems known to science, AI has
achieved only very modest results. The pro-
grams developed to date are like toys when
matched against the overall goal of a
machine that can operate at human levels
of intelligence in the unruly complexity of
the real world. At present the symbol
system hypothesis has much the same
standing as the hypothesis that there is
intelligent life on other planets: people have
strongly held opinions, both for and against,
but as yet there is no firm evidence either
way.
THE ALGORITHMICITY ASSUMPTION
First I will explain what an algorithm is. I'll
call the steps of a procedure moronic if no
insight, ingenuity or creativity is necessary
in order to carry them out. A procedure for
achieving some specified result is known as
an algorithm when (1) every step of the
procedure is moronic; (2) at the end of each
step it is moronically clear what is to be
done next (Le. no insight, etc. is needed to
tell); and (3) the procedure is guaranteed to
lead to the specified result in a finite
number of steps (assuming each step is
carried out correctly). To give a simple
example, if you've mixed up your keys and
can't tell by sight which one fits your front
door, the well-known expedient of trying
them in succession is an algorithm for
finding the right key: the procedure is sure
to work eventually (assuming you haven't
actually lost the key) and is certainly
moronic. All (successful) computer pro-
grams are, of course, algorithms.
A system (real or abstract) is said to be
algorithmically calculable if there is an algor-
ithm - known or unknown - for calculating
its behaviour. To put this more formally, the
system is algorithmically calculable just in
case there is an algorithmic procedure for
deriving correct descriptions of the system's
outputs from correct descriptions of the
inputs it receives (and moreover a proce-
dure that works for all possible inputs that
will produce output).
ALGORITHMIC CALCULABILITY AND
INTELLIGENCE
The types of machines used in AI are algor-
ithmically calculable. Tautologically, a com-
puter with a given set of programs is an
algorithmically calculable system (provided
it doesn't malfunction) because, of course,
these programs are themselves algorithms
for passing from input to output. Connec-
tionist networks are also algorithmically
calculable systems. In short, all current
work in AI is based on the assumption that
the way to build an artefact capable of
general intelligent action is to build one or
another type of algorithmically calculable
system. The possibility that general intelli-
gence may transcend algorithms is ignored.
As early as 1936 Turing mooted the pos-
sibility of our own cognitive processes being
algorithmically calculable. Subsequently
this has become an article of deepest faith
among AI researchers, and has even
acquired the status of something 'too
obvious to mention'. For example, take the
following famous argument due to Newell,
which is intended to demonstrate that com-
puter intelligence is achievable, at least in
theory: 'A universal [symbol] system always
contains the potential for being any other
system, if so instructed. Thus, a universal
system can become a generally intelligent
system' (Newell, 1980, p. 170). What
Newell means is that a universal symbol
system (Le. a general-purpose computer
with no practical bound on the size of its
memory) can be programmed to simulate
any other algorithmically calculable system
(this is known as the Church-Turing thesis
or sometimes - less accurately - simply as
Church's thesis). The assumption that a
system exhibiting general intelligence will
be algorithmically calculable is thought so
obvious that Newell does not even mention
it. Some of AI's most notable opponents also
subscribe to the assumption that our cog-
nitive processes are algorithmically calcul-
able, for example John Searle. 'Can the
operations of the brain be simulated on a
digital computer? ... [G]iven Church's
thesis that anything that can be given a
precise enough characterization as a set of
steps can be simulated on a digital compu-
ter, it follows trivially that the question has
an affirmative answer' (Searle, 1992, p.
200). In fact, this follows only given the
premise that the brain is algorithmically
calculable. The thought that it follows trivi-
ally from the Church-Turing thesis that our
cognitive processes are algorithmically cal-
culable is a modern fallacy; there is no
reason to think that Turing himself saw the
issue of the algorithmic calculability of cog-
nition as being anything other than an
empirical one.
THE ALGORITHMICITY ASSUMPTION
STATED
There are two parts to what I am calling
the algorithmicity assumption. AI as pre-
sently conceived is feasible only if both parts
of the assumption are true.
(1) There exists, or can exist, some algor-
ithmically calculable system exhibiting
general intelligence.
(2) There is a practicable algorithm (known
or unknown) for calculating the beha-
viour of the system.
By a practicable algorithm I mean one that
can be implemented in a real machine in
real time, as opposed to one that could be
implemented only in an ideal world of
unlimited resources - a world in which time
is of no concern, in which issues of reli-
ability and cumulative error can be ignored,
and in which there are no bounds whatever
on the capacity of the machine's memory
devices.
The idealized computers invented by
125
Turing and now known as Turing machines
are a rich source of examples of algorithms
that can be run in an ideal world but not in
the real world. A Turing machine is very
simple: it consists only of a read/write head
and a memory tape - a paper tape of unlim-
ited length (Turing, 1936). In theory a
Turing machine can be given an algorithm
that will enable it to simulate a state-of-the-
art Cray supercomputer. But it would take
untold centuries for a Turing machine to
simulate even a few hours of the Cray's
operations, and the simulation might well
demand more paper tape than our planet
has to offer. (Turing's own formulation of
the Church-Turing thesis was in terms of
these idealized machines: a Turing machine
can simulate any algorithmically calculable
system (Turing, 1936).) Obviously, an
algorithm for general intelligent action that
could not be implemented on some real
machine in the real world would be of no
use to AI.
The remainder of this discussion focuses
on the first part of the algorithmicity
assumption. The mammalian brain is the
best - indeed the only - example we have of
a system capable of supporting general
intelligent action. Is this system algor-
ithmically calculable? In point of fact there
is no reason - apart from a kind of wishful
thinking - to believe that it is so. This is
terra incognita, and for all we presently
know many aspects of brain function may
be non-computable. To pursue this point we
need some terminology from computability
theory.
A LITTLE LIGHT COMPUT ABILITY THEOR Y
Let S be a set of sentences. For the moment
it doesn't matter what these sentences are;
I'm going to use S as the basis for a number
of general definitions - general in the sense
that they are applicable no matter what
sentences S happens to contain. Call the
language in which the sentences of S are
written L. (So L might be English or the dif-
ferential calculus or the first order predicate
calculus, etc.) To say that S is decidable is to
say that there is an algorithm for settling
126
whether or not given sentences are
members of S. The algorithm must give the
right answer for each sentence of 1. So for S
to be decidable there must be an algorithm
that can be applied to each sentence of L
and that will deliver either the answer 'Yes,
this sentence is in S' or 'No, this sentence is
not in S' (and moreover the answers the
algorithm gives must always be correct). S
is said to be undecidable if there is no such
algorithm. If there is such an algorithm it is
called a decision procedure for S.
To give a straightforward example of a set
that is decidable, let the only members of S
be the sentences 'London is in England' and
'Paris is in France'. It is easy to compose a
decision procedure for this two-membered
set: given any English sentence, scan
through it character by character and
determine whether or not it is identical to
either the first member of S or the second
member of S. If it is identical to one of them
then output 'This sentence is in S'; other-
wise output 'This sentence is not in S'. This
is called a lookup decision procedure (or
lookup algorithm). The program contains a
complete list of the members of S and tests
sentences for membership of S simply by
looking through the list.
Any finite set of sentences has a lookup
decision procedure. In other words, every
finite set of sentences is decidable. Lookup
decision procedures become less practicable
as the set in question gets larger. Indeed, it
is easy to imagine lookup decision pro-
cedures that are completely impossible to
implement in practice. Take a set contain-
ing N English sentences, where N is the
number of atoms in the known universe.
Since N is finite this set has a lookup deci-
sion procedure, but obviously it isn't one
that any real computer could run. (Though
a Turing machine could run it, since Turing
machines have unbounded resources.) This
is why lookup decision procedures for very
large sets are of no relevance to AI.
Many sets have decision procedures that
don't involve lookup. One example is the set
of English sentences that are less than 100
characters long (to test whether a given
English sentence is a member of the set just
count the characters). However, some finite
sets have no decision procedure apart from
a lookup procedure. The sentences of this
book form such a set. There is no rule for
generating the sentences that an algorithm
can exploit. I will describe such sets as
'undecidable save by lookup'. If the theory
of cognition involves large sets that are
undecidable save by lookup this could be
very bad news for AI.
An infinite set is a set containing at least
as many members as there are whole
numbers. If an eternal being were to begin
counting the members of an infinite set,
one, two, three, four . . . then no matter
how large a number the being gets up to,
there will always be further members left to
count. Infinite sets are not necessarily exotic
things. An extremely humdrum infinite set
can be constructed from nothing more than
the sentence 'Fred is just an ordinary bloke'
and the rule: if X is a member of the set
then so is 'it is false that X'. Here are the
first few of the set's infinitely many
members.
Fred is just an ordinary bloke.
It is false that Fred is just an ordinary
bloke.
It is false that it is false that Fred is just
an ordinary bloke.
It is false that it is false that it is false
that Fred is just an ordinary bloke.
Formal logic furnishes examples of both
decidable and undecidable infinite sets. The
set of valid sentences of truth-functional
logic is infinite and decidable. Truth-
functional logic is the branch of logic that
focuses on the expressions 'if-then', 'and', 'it
is false that' and 'either-or'. A valid sentence
is a sentence that is true come what may,
true in all circumstances. So an example of
a valid sentence of truth-functional logic is
'either Fred is an ordinary bloke or it is false
that Fred is an ordinary bloke'. The next
step up from truth-functional logic is first-
order quantifier logic. The quantifiers are
'some' and 'every'. In quantifier logic (but
not truth-functional logic) it is possible to
express such propositions as 'every cat has
nine lives', 'some cats dribble', and 'some
infinite sets are undecidable'. Logic books
brim with examples of valid sentences of
first-order quantifier logic: 'if it is false that
everyone in the bar is drunk then someone
in the bar is not drunk', 'either everyone
loves Suzy or someone does not love Suzy',
and so forth. The set of valid sentences of
first-order quantifier logic is infinite and
undecidable. (This was first proved by
Church (1936).) Any attempt to write an
algorithm that will test arbitrary sentences
of first-order logic for validity is bound to be
unsuccessful. Any procedure that anyone
dreams up is bound to be unreliable, bound
in the long run to give wrong answers or
no answer at all at least as many times as it
gives right ones. There simply is no decision
procedure for the set of valid sentences of
first-order quantifier logic.
However, this set is not completely
intractable computationally. There is an
algorithm (in fact many) meeting the fol-
lowing conditions.
( 1 ) Whenever the algorithm is applied to a
valid sentence of first-order quantifier
logic it will (given enough time) deliver
the result 'Yes, this sentence is valid'.
(2) Whenever the algorithm is applied to a
sentence of first-order quantifier logic
that is not valid it will either deliver the
result 'No, this sentence is not valid' or
will deliver no answer at all (Le. will
carryon computing 'forever' - in prac-
tice until someone turns it off or until it
runs out of memory and crashes).
An algorithm that meets these two condi-
tions is called a semi decision procedure. If
there is a semi decision procedure for a set S
then S is said to be semidecidable. Every
answer that a semi decision procedure gives
is correct, and every time the algorithm is
applied to a sentence that is in the set it
gives an answer, provided you wait long
enough. When the algorithm is applied to a
sentence that is not in the set it may give an
127
answer (the right one) or it may never come
to the end of its calculations.
Can you see why a semi decision pro-
cedure is not a decision procedure? The
sticking point is the proviso that you wait
long enough. If you're testing a sentence
and no answer is forthcoming then no
matter how long you wait you can never be
sure whether the algorithm is just about to
pronounce, perhaps positively, or whether
the sentence is one of those invalid ones
for which the algorithm never delivers a
result.
Are there such things as nonsemidecidable
sets - sets, that is, for which there is not
even a semi decision procedure? Yes. Again
formal logic furnishes an example. The set
of valid sentences of second-order quantifier
logic is known to be nonsemidecidable.
First-order logic is all about saying that
things have properties and bear relation-
ships to other things: 'every thing has the
property of being massive', 'some things
have both the property of being a cat and
the property of dribbling', 'there is some
thing that stands in the relation of being-
larger-than to every other thing'. The lan-
guage of second-order logic allows us to
quantify over not just things themselves but
also over the properties that things have
and the relationships that things bear to
one another. In other words, in second-
order logic we can write sentences contain-
ing such expressions as 'every property that'
and 'there are some relationships that'. In
second-order logic (but not first-order logic)
it is possible to express such propositions as
'Jules and Jim have some properties in
common', 'every thing has some properties',
'Napoleon has every property necessary to
being a good general' and 'every constitu-
tional relationship that holds between the
US President and Senate also holds between
the British Prime Minister and the House of
Commons'. The set of valid sentences of
second-order logic is as computationally
intractable as they come. Any attempt to
write an algorithm that will say 'Yes, that's
a member' whenever it is applied to a
member of the set (and never when it is
applied to a non-member) is doomed. There
128
is no such algorithm: that is what it means
to say that a set is nonsemidecidable.
Turing proved the existence of non-
semidecidable sets in his 1936 paper
(although not by reference to second-order
logic). Penrose (1989) contains interesting
discussions of some of the issues that Tur-
ing's proof raises.
NONSEMIDECIDABILITY AND THE
ALGORITHMICITY ASSUMPTION
There is an intimate connection between
the concept of a system's being algor-
ithmically calculable and the concept of a
set's being semidecidable. Let the members
of S be descriptions of the input/state/
output behaviour of a system B. That is,
each member of S is of the form 'If the
system B is given input I while in internal
state X then it goes into internal state Y and
produces output 0'. S contains all the sen-
tences of this form that are true of B. If S is
nonsemidecidable B is not algorithmically
calculable; for if S is nonsemidecidable there
can be no algorithm capable of computing a
correct description of the ensuing output (if
any) from each description of input into B.
If there were no nonsemidecidable sets
then we could rest assured that every pos-
sible system is algorithmically calculable;
and in particular we could be assured that
any system capable of general intelligent
action is algorithmically calculable. But
thanks to Turing we know that there are
nonsemidecidable sets; so we must accept
that there is no a priori guarantee that
intelligent systems are algorithmically cal-
culable.
CAN A COMPUTER SIMULATE THE BRAIN?
In the particular case where the system B is
the human brain I will call the set of
descriptions of input/state/output behaviour
~ . The members of ~ are highly complicated
sentences. Each records the total output
(electrical and chemical) that the brain will
produce in response to a particular total
input received while the brain is in a par-
ticular (total) internal state.
Is ~ infinite? This is an empirical question
and the answer is not yet known. It is cer-
tainly the case that the corresponding set
for a Turing machine is infinite, since there
are infinitely many possible inputs (infinitely
many ways that the limitless tape can be
inscribed with symbols before the machine
is set in motion). Many believe there is a
good prima facie case for thinking that ~ too
is infinite. It is, of course, true that a brain
can process only a finite number of finite
inputs in its finite life. But it is not at all
obvious that there are only a finite number
of potential inputs from which the actual
inputs that a given brain encounters are
drawn. The potential inputs seem endless -
in the same way that there are an endless
number of English sentences. If the number
of potential inputs is infinite then ~ is in-
finite, for ~ contains all input/state/output
descriptions, potential as well as actual.
The same goes for states. Each brain can
enter only a finite number of states in its
life, yet the set of potential states from
which this finite number is drawn may be
infinite. For example, 'thinking that P' is an
internal state, and there certainly seems to
be no limit to the number of different things
that a person might think. (Analogously the
electrical potential of a cell membrane can
take only a finite number of values in the
course of the cell's finite life, yet there are
an indefinite number of possible values
lying between the maximum and minimum
values for that membrane, from which the
set of actual values is drawn.)
Nothing is known to indicate that ~ is
either decidable or semidecidable. And if ~ is
nonsemidecidable the brain is not algor-
ithmically calculable. Even if ~ should turn
out to be finite it may be undecidable save
by lookup. The existence, in theory, of a
lookup decision procedure could be of no
conceivable relevance to AI. given the size
of ~ (not to mention the sheer difficulty of
listing the members of ~ ) . Ambitions to
simulate the brain by computer depend for
viability on there being a practicable algor-
ithm that can generate members of ~ . One
may hope that even if the brain should turn
out not to be algorithmically calculable, it is
nevertheless 'recursively approximable' in
the sense of Rose and Ullian ( 19 6 3 ). A
system is recursively approximable if and
only if there is an algorithm that computes
descriptions (not necessarily in real time) of
the system's input/output behaviour in
such a way that after a suffiCiently large
number of descriptions have been generated
the proportion of the descriptions that are
incorrect never exceeds a prescribed figure.
However, this is at present no more than a
hope. There are uncountably many func-
tions that are not recursively approximable.
One sometimes sees it claimed that any con-
ceivable physical device or process can be
approximated by a Turing machine to any
required degree of fineness; but this is true
only for devices and processes that are
recursively approximable.
A MODEL FOR A NON COMPUTATIONAL
BRAIN
Readers who have studied the section on
CONNECTIONISM may be interested to see a
sketch of a type of artificial neural network
that is not algorithmically calculable. In
connectionist jargon, the activation function
for a neural network is a description of
what it takes to make the neurons in the
network fire. For example, in a typical con-
nectionist network a neuron fires if the
weighted sum of its inputs exceeds its
threshold. I will call a neural network
second-order if the activation function for the
network can be formulated only by means
of second-order logic. That is, the activation
function can be formulated only by means
of quantifying over properties of neurons
and/or relationships between neurons. An
example of such an activation function is: a
neuron n fires if all relationships of a parti-
cular sort holding between the neurons in
some cluster of neurons containing n also
hold between the neurons in some asso-
ciated cluster. It remains to be seen whether
'global' activation functions such as this
one are physically realizable in a way that is
consistent with what is already known
about neural tissue. (The function is 'global'
in the sense that the neuron needs to know
129
more than what is happening just on its
own doorstep in order to be able to tell
whether or not to fire - that is. the neuron
needs non-local, or global, information.
Certain types of second-order neural net-
works are not algorithmically calculable.
For all that is presently known. significant
portions of the brain may be like this.
CONCLUSION
AI researchers are fond of saying that they
are looking for Maxwell's laws of thought.
Last century James Clerk Maxwell reduced
electrodynamics to a few elegant equations.
The idea of doing the same for cognition is
certainly appealing. Minsky speaks of the
search for 'the three algorithms'. His meta-
phor is a child throwing three pebbles into a
pond. The basic wavepattern produced by
each pebble is very simple - concentric
rings moving outwards - but the inter-
action of these three patterns produces a
confusion of waves and ripples. The hope is
- though it's fading - that the complexity of
our cognitive life will similarly turn out to
be the product of a small number of elegant
algorithms.
Lack of success in the search for Max-
well's laws of thought is ushering in a new
perspective. The mind is coming to be seen
as a rag bag of large numbers of special-
purpose algorithms. a motley assortment of
ad hoc tools assembled by Mother Pro-
grammer. the greatest pragmatist in the
universe. The latest generation of AI
systems mirror this new image. For
example, Lenat's CYC program contains
around 30 different special-purpose infer-
ence mechanisms (Lenat and Guha. 1990).
AI is polarized around these two view-
points - a small number of very powerful
algorithms versus a large number of weaker,
messier ones. Yet as we have seen there is a
daunting third possibility. Algorithms may
not be much of the story of intelligence at all
- at best one of its minor characters,
perhaps.
Traditional program-writing AI is, of
course. irrevocably bound up with the
assumption that intelligence is algor-
130
ithmically calculable. and moreover calcul-
able by means of practicable algorithms.
Present forms of connectionism, too. are
wedded to this assumption, in that all con-
nectionist architectures currently under
investigation are algorithmically calculable
(indeed. investigation typically proceeds by
means of computer simulation). I am cer-
tainly not denying the claim that a system
capable of exhibiting general intelligence is
algorithmically calculable; but I would
stress that at the present moment there is
precious little evidence either for or against
this assumption. on which AI as we pres-
ently conceive it depends.
See also COGNITIVE PSYCHOLOGY; COMPUTA-
TIONAL MODELS.
BIBLIOGRAPHY
Church, A. 1936. A Note on the Entschei-
dungsproblem. The Journal of Symbolic Logic,
1. 40-1.
Copeland, B.J. 1993. ArtifiCial Intelligence: A
Philosophical Introduction. Oxford: Basil
Blackwell.
Lenat. D.B., Feigenbaum. E.A. 1991. On the
thresholds of knowledge. Artificial Intelli-
gence, 47, 185-250.
Lenat, D.B., Guha. R.V. 1990. Building Large
Knowledge-Based Systems: Representation and
Inference in the CYC Project. Reading. MA.:
Addison-Wesley.
Minsky, M.L., ed. 1968. Semantic Information
Processing. Cambridge. MA.: MIT Press.
Newell. A. 1980. Physical symbol systems.
Cognitive Science, 4. 135-83.
Newell, A .. Shaw, J.C., Simon. H.A. 1957.
Empirical explorations with the logic theory
machine: A case study in heuristics. In
Computers and Thought. ed. E. A. Feigen-
baum, J. Feldman. New York: McGraw-Hill,
1963, pp. 109-33.
Newell, A .. Simon, A. 1976. Computer science
as empirical inquiry: Symbols and search. In
Mind Design: Philosophy, Psychology, Artifi-
cial Intelligence. ed. J. Haugeland. Cambridge,
MA.: MIT Press. 1981. pp. 35-66.
Penrose. R. 1989. The Emperor's New Mind.
Oxford University Press.
Rose. G.F., Ullian, J.S. 1963. Approximation of
functions on the integers. Pacific Journal of
Mathematics, 13, 693-701.
Samuel, A.L. 1959. Some studies in machine
learning using the game of checkers. In
Computers and Thought, ed. E. A. Feigen-
baum, J. Feldman. New York: McGraw-Hill,
1963, pp. 71-105.
Searle, J. 1992. The Rediscovery of the Mind.
Cambridge, MA.: MIT Press.
Simon, H.A., Newell, A. 1958. Heuristic
problem solving: the next advance in opera-
tions research. Operations Research, 6, 1-10.
Turing, A.M. 1936. On computable numbers,
with an application to the Entscheidung-
sproblem. Proceedings of the London Math-
ematical Society, Series 2, 42 (1936-37),
230-265.
Turing, A.M. 1947. Lecture to the London
mathematical society on 20 February 1947.
In A. M. Turing's ACE Report of 1946 and
Other Papers, ed. B. E. Carpenter, R. W.
Doran. Cambridge, MA.: MIT Press, pp.
106-24.
Turing, A.M. 1948. Intelligent machinery.
National Physical Laboratory Report. Repro-
duced in Machine Intelligence 5, ed. B.
Meltzer, D. Michie. Edinburgh: Edinburgh
University Press, 1969, pp. 3-23.
Turing, A.M. 1950. Computing machinery
and intelligence. Mind 59, 433-60.
Turing, S. 1959. Alan M. Turing. Cambridge:
W. Heffer.
B. JACK COPELAND
131
B
behaviourism Introductory texts in the
philosophy of mind often begin with a dis-
cussion of behaviourism, presented as one
of the few theories of mind that have been
conclusively refuted. But matters are not
that simple: behaviourism, in one form or
another, is still alive and kicking.
'Behaviourism' covers a multitude of
positions. Yet there is a common underlying
thread. The behaviourist takes minds not to
be inner psychic mechanisms merely con-
tingently connected with their outer beha-
vioural effects, but to be (at least to a
significant extent) constituted by those
outer effects. The behaviourist's motivation
is often epistemological: on the picture of
the mind as essentially inner, how can its
outer effects provide us with the wide-
ranging knowledge of others' minds we
confidently take ourselves to possess?
As an imperfect but serviceable analogy,
consider a clock. A clock has visible moving
exterior parts - the hands. To the beha-
viourist about clocks, a clock is simply
something with such time-indicating exter-
ior parts. The inner workings of any clock
are entirely irrelevant to its status as a
clock, provided they produce (or at least
don't interfere with) the movement of the
hands. The anti-behaviourist, by contrast,
thinks of a clock as an inner mechanism
which, in favourable circumstances, can
cause some exterior parts to move in a way
which reliably indicates the time. But there
is no a priori reason, according to the anti-
behaviourist, why these favourable circum-
stances should be even possible. There may
be clocks to which, given the laws of
nature, hands cannot be attached. More-
over, such clocks need not be abnormal
ones: they could even be paradigm ex-
amples of clocks.
132
Of course there are intermediate posi-
tions. One might hold that typical clocks
must have, or be capable of having, time-
indicating hands, while acknowledging that
there could be atypical clocks of which this
is not so. These atypical clocks would count
as clocks in virtue of sharing inner mechan-
isms with typical clocks. Or one might,
while insisting on the importance of the
hands, impose some minimal constraints on
the innards of a clock. For example, one
might say that something could not be a
clock unless the big hand and the little
hand were controlled by the same mechan
ism; but beyond that, anything goes.
Suppose, to press the analogy still further,
that we never open up any clocks to
examine their inner parts. Clock anti-
behaviourism would then seem to give us
an epistemological problem: how do we
know that there are any clocks?
Behaviourism flourished in the first half of
the twentieth century. Philosophers from
that period with behaviourist leanings
include Carnap, Hempel. Russell, WITT
GENSTEIN, and R YLE. Arranging some con-
temporary philosophers on a spectrum from
the most behaviouristically inclined to the
least finds QUINE at the behaviourist end,
and SEARLE at the other. DAVIDSON,
DENNETT and DUMMETT are closer to Quine
than Searle, with FODOR, DRETSKE (and
many others) closer to Searle than Quine.
Armstrong and LEWIS are squarely in the
middle.
PHYSICAL AND AGENTIAL BEHAVIOUR
Let us say that any instance of physical
behaviour is a physical change to an agent's
body (perhaps in relation to his environ-
ment), such as the rising of the agent's arm.
Let us say that any instance of agential beha-
viour is something an agent does, such as
raising his arm, (For a related distinction,
see Armstrong, 1968, p, 84,) We can simi-
larly define physical behavioural dispositions
and agential behavioural dispositions.
The relationship between physical beha-
viour and agential behaviour is controver-
sial. On some views, all ACTIONS are identical
to physical changes in the agent's body.
(However, some kinds of physical behaviour,
such as reflexes, are uncontroversially not
kinds of agential behaviour.) On others, an
agent's action must involve some physical
change, but is not identical to it.
Both physical behaviour and agential
behaviour could be understood in the
widest sense. Anything a person can do -
even calculating in his head, for instance -
could be regarded as agential behaviour.
Likewise, any physical change in a person's
body - even the firing of a certain neuron,
for instance - could be regarded as physical
behaviour.
Of course, to claim that the mind is
'nothing over and above' such-and-such
kinds of behaviour, construed as either
physical or agential behaviour in the widest
sense, is not necessarily to be a behaviourist.
The theory that the mind is a series of voli-
tional acts - a view close to Berkeley's - and
the theory that the mind is a certain config-
uration of neural events, while both con-
troversial. are not forms of behaviourism.
So either the behaviourist needs a less
inclusive notion of behaviour or, at the very
least, if he does allow some inner processes
to count as behaviour, he must minimize
their importance.
Waving to someone, or the consequent
movement of one's arm, are more beha-
viouristically acceptable than calculating in
the head, or the accompanying firing of
neurons. If a philosophical theory of the
mind emphasizes wavings or arm move-
ments over silent cogitations and brain
events, then it is, to that extent, behaviour-
ist. Accordingly, 'physical behaviour' and
'agential behaviour' will henceforth be
understood in a behaviouristically restricted
way. Wavings or visible arm movements
BEHA VIOURISM
(together with, for example, blushing and
standing still, which do not involve bodily
movement) are included, inner goings-on
are excluded. This is vague, but none the
worse for that.
ELIMINATIVE. ANALYTIC. AND RYLEAN
BEHA VIOURISM
Eliminative behaviourism is a forerunner of
the contemporary doctrine of eliminative
materialism (see ELIMINATIVISM). Elim-
inativists about the mental repudiate all or
most of our commonsense psychological
ontology: beliefs, conscious states, sensa-
tions, and so on. One argument for elim-
inativism, in contemporary dress, is this.
First, 'POLK PSYCHOLOGY', taken to be, inter
alia, our tacit theory of the behaviour of
others, suffers various deficiencies: wide-
spread explanatory failings, for instance.
Second, there are much better theories of
behaviour which do not quantify over
mental states. Therefore, in accordance with
good scientific practice, folk psychology
should be replaced by one of these superior
theories. Eliminative behaviourism takes the
replacement theory to be couched in the
vocabulary of physical behaviour. (In fact,
there are a number of choices for the voca-
bulary of physical behaviour: the austere
terminology of kinematics is one, a certain
rich fragment of English another. But at the
least the vocabulary will not contain men-
tally loaded terms.)
Eliminative behaviourism is a dominant
theme in the writings of Watson (1930)
and Skinner (see the papers collected in
Skinner et al.. 1984): two central figures
in the development of the now unpopular
doctrine of psychological behaviourism.
Psychological behaviourism is primarily a
claim about the correct methodology of a
scientific psychology, and arose in the early
part of the twentieth century as a reaction
to the 'introspective' psychology of Wundt,
James and Titchener. According to the
introspective school. the subject matter of
psychology is CONSCIOUSNESS, and the
proper methodology for its study is
INTROSPECTION. Against this, Watson
133
BEHA VIOURISM
argued that a scientific psychology should
just concern itself with what is 'objective',
and 'observable', namely, according to him,
behaviour.
There was more to psychological beha-
viourism than this, of course. Watson and
Skinner both thought that the behaviour of
an organism could be explained by its
history of stimulation together with relat-
ively simple processes of behavioural modi-
fication. Skinner's introduction of operant
conditioning as one of these processes
marked an improvement over the crude
stimulus-response behaviourism of Watson.
Watson found it more acceptable than
Skinner to go inside the organism to find
stimuli and responses, but Watson gave
these inner events no special status.
Further, Watson took the brain to be just
one inner part among others of equal
importance: 'the behaviourist [places] no
more emphasis on the brain and the spinal
cord than upon the striped muscles of the
body, the plain muscles of the stomach,
[and] the glands' (1930, p. 49).
Both Watson and Skinner shared unclar-
ities on two connected issues. They were
unclear whether stimuli and responses can
be described in mentally loaded terms, or
whether only purely physical descriptions
are allowed. They also vacillated between
endorsing: (a) eliminativism about the
mental; (b) the claim that mental states
exist but are irrelevant to the scientific
study of human beings; or (c) the claim that
mental terminology can be translated into
vocabulary of physical behaviour. But they
both had a strong tendency towards elim-
inativism. Watson, for instance, took 'belief
in the existence of consciousness' to go
'back to the ancient days of superstition and
magic' (1930, p. 2). And Skinner has
expressed similar sentiments (e.g. 1971, ch.
1). (For recent commentary on Skinner, see
Skinner et aI., 1984; Modgil and Modgil
1987.)
QUINE is another eliminative behaviour-
ist, but for quite different reasons. His beha-
viourism appears to be motivated largely by
his verificationism. He gives two reasons for
eliminativism. The first is that belief and
134
desire talk resists regimentation in first-
order logic, which Quine takes to be the
litmus test for complete intelligibility. The
second is his argument for the thesis of
the indeterminacy of translation, which
purports to show that there is simply no
'fact of the matter' as to what someone's
language means (Quine, 1960, ch. 2).
Quine assumes a sufficiently intimate con-
nection between language and belief for it
to follow that there is also no 'fact of the
matter' as to what someone believes.
Most behaviouristically inclined philoso-
phers are not eliminativists. The most pow-
erful and straightforward kind of (non-
eliminative) behaviourism is:
analytic (or logical) behaviourism: state-
ments containing mental vocabulary
can be analysed into statements con-
taining just the vocabulary of physical
behaviour.
Skinner can be interpreted as a part-time
analytic behaviourist (e.g. Skinner 1971, p.
24). And Hempel, stating a view common
to many logical positivists, wrote:
All psychological statements which are
meaningful, that is to say, which are in
principle verifiable, are translatable
into statements which do not involve
psychological concepts, but only the
concepts of physics. (1949, p. 18,
italics omitted)
Hempel derived this strong thesis from two
premises. First, he held (but later aban-
doned) the verificationist theory of meaning,
namely that 'the meaning of a statement is
established by its conditions of verification'
(1949, p. 17, italics omitted). Second, he
held that a person's physical behaviour was
a large part of the evidence for ascribing to
him particular mental states. Putting the
two together, he concluded that statements
about mental states were equivalent to
statements (largely) about physical beha-
viour.
Largely, but not entirely. Hempel was not
a thoroughgoing behaviourist in the sense
of ignoring inner processes altogether.
According to Hempel, the verification condi-
tions (which amount to the meaning) of
'Paul has a toothache' include certain
changes in Paul's blood pressure, his digest-
ive processes and his central nervous
system. But as gross bodily movements play
a large role in the verification of psychologi-
cal sentences, and hence playa large role in
determining their meanings, Hempel's posi-
tion is to a significant extent behaviourist.
We now briefly turn to Ryle's influential
quasi-behaviourist polemic, The Concept of
Mind (1949) (see RYLE). Somewhat confus-
ingly, Ryle's position is often called analytic
(or logical) behaviourism, but it is quite dif-
ferent, in content and motivation, from the
positivist sort of behaviourism exemplified
by Hempel.
Hempel's behaviourism is part of his
PHYSICALISM. Ryle, in contrast, was no
physicalist. He regarded the very question of
whether the world is ultimately physical as
conceptually confused. Accordingly, Ryle
spoke of agential behavioural dispositions,
and showed little inclination to analyse this
away in terms of physical behavioural dis-
positions.
Further, it is arguable whether Ryle's
primary intention was to offer analyses of
statements apparently about inner mental
occurrences in terms of agential beha-
vioural dispositions. At any rate, he does
not supply very many. Ryle was chiefly
concerned to deflate the idea that there
must be complex inner mental processes
behind a person's public actions and to
show how this dissolved the problem of
other minds. In doing this, he often did not
analyse the inner away, but merely derided
it. 'Overt intelligent performances are not
clues to the workings of minds, they are
those workings. Boswell described Johnson's
mind when he described how he wrote,
talked, ate, fidgeted and fumed. His descrip-
tion was of course incomplete, since there
were notoriously some thoughts which
Johnson kept carefully to himself and there
must have been many dreams, daydreams
and silent babblings which only Johnson
could have recorded and only a James Joyce
would wish him to have recorded' (Ryle,
1949, pp. 58-9).
BEBA VIOURISM
Ryle was indeed, as he reportedly said,
'only one arm and one leg a behaviourist'.
PhYSicalists such as Place, Smart, Arm-
strong, and Lewis, were able to find much
to agree with in Ryle, despite his antipathy
toward physicalism. They took Ryle to have
in effect shown that physical behavioural
dispositions were of great importance in
understanding the nature of the mind.
Geach (1957, p. 8) raised an important
difficulty for any behaviourist analysis of the
propositional attitudes, whether in terms of
agential or physical behavioural dispositions.
Geach noted that what someone does, or is
disposed to do, depends not only on the fact
that he holds a particular belief, but also on
his desires (and, it should be added, on his
other beliefs as well). Therefore there can be
no question of a simple atomistic behavioural
analysis: one which matches each belief with
a different kind of behaviour (whether speci-
fied in the language of agential or physical
behaviOur). A given belief may issue in prac-
tically any sort of behaviour, depending on
the agent's other attitudes (see BELIEF;
PROPOSITIONAL ATTITUDES.)
CONTEMPORAR Y BEBA VIOURISM
Geach's observation that belief types do not
have characteristic behavioural expressions
is now regarded as a datum of folk psycho-
logical explanation, and consequently few
contemporary philosophers are tempted by
any sort of (atomistic) analytic behaviour-
ism. But the failure of this kind of beha-
viourist analysis does not imply the failure of
behaviourism, construed as a metaphysical
thesis about the nature of the mind. All it
does imply is that the most direct route to
behaviourism - namely a simple analysis
pairing beliefs and behavioural dispOSitions
- is a dead end. A convincing argument for
behaviourism must proceed down a less
obvious path.
Contemporary behaviourist views derive
from three sources. The first is the analytic
FUNCTIONALISM of Armstrong, Lewis and
others (see LEWIS). According to this view,
the meanings of mental terms are deter-
mined by their role in our commonsense
135
BEHA VIOURISM
theory of behaviour: folk psychology. This
tacitly known theory is taken to consist of
generalizations linking perceptual input,
(physical) behavioural output, and mental
states. Some of these generalizations will
link beliefs to desires, thus incorporating
Geach's insight that a belief does not have a
fund of behaviour to call its own. However,
if analytic functionalism is correct, then the
analyses of mental terms will have a sig-
nificant behavioural component.
Note that analytic functionalism, while
not underwriting an entailment from having
such-and-such physical behavioural disposi-
tions to having a mental life, nonetheless
does appear to solve the problem of other
minds. According to analytic functionalism,
one may reasonably conclude that a crea-
ture has a mind on the basis of its beha-
viour, just as one may reasonably conchide
that a car has an engine on the basis of its
motion.
The second source is opposition to the
functionalist idea that an organism needs a
certain kind of inner causal organization in
order to be a genuine believer.
These two incompatible lines of thought
have their roots in Ryle, although he would
only have expressed agreement with the
second. The third source of contemporary
behaviourism is one on which Ryle's influ-
ence was at best indirect, although he
would have found it perfectly congenial.
That source is Wittgenstein's (1958) attack
on the possibility of a 'private language',
which some take to show that meaning and
belief must be 'manifestable' in behaviour
(see WITTGENSTEIN).
Roughly corresponding to these three
sources are three behaviourist theses. The
first is:
Behaviour-as-necessary: necessarily, any-
thing that has no physical behavioural
dispositions of a certain kind and com-
plexity does not have a mental life.
This vague formulation should be taken to
express the behaviourist thought that some-
thing like a stone could not possibly have a
mental life, because it is outwardly inert (not
because there is nothing sufficiently compli-
136
cated going on in the stone). It is a thesis
which, incidentally, can with some safety be
ascribed to Wittgenstein (1958, I, 281-4).
Analytic functionalism can evade the
behaviour-as-necessary view to some extent,
but not completely. Lewis's brand of analytic
functionalism, for instance, has the con-
sequence that anything which has no be-
havioural dispositions above a certain level
of complexity either (a) has no mental life;
or, (b) is an atypical member of its kind
(Lewis, 1980). Therefore an extreme kind of
permanent paralysis cannot be the typical
condition of any population of thinking crea-
tures: there cannot be a race of thinking
stones.
A natural companion to the behaviour-
as-necessary view, although logically inde-
pendent of it, is:
Behaviour-as-sufficient: necessarily, any-
thing that has physical behavioural
dispositions of a certain kind and com-
plexity has a mental life.
According to functionalism, whether of the
analytic variety or not, minds must have
some specific type of inner causal structure.
Functionalism in general is therefore incon-
sistent with the behaviour-as-sufficient
view. However, DENNETT - on whom Ryle
had a direct influence - has been a longtime
opponent of the idea that to have a mind is
to have a specific type of inner causal struc-
ture. On the one hand, Dennett is unmoved
by thought experiments - often involving
Martian scientists controlling hollow an-
thropoid puppets - which purport to show
that having a mind involves some restric-
tion on inner causal order. (See p. 138
below.) On the other hand, Dennett is
impressed, as was Hume, with the remark-
able predictive power of everyday psychol-
ogy. Adopting 'the intentional stance'
(Dennett, 1971, 1981) - the predictive
strategy of ascribing intentional states such
as belief and desire to a system - is indis-
pensable in many cases, its utility not
diminished by any discoveries of bizarre
inner causal organization.
Now if we can be assured that the
assumption that a system has beliefs is not
going to be defeated by future evidence -
although specific belief attributions may
well be - and that assumption is useful,
why insist that there is some further ques-
tion to be raised concerning whether the
system really has beliefs?
Thus, according to Dennett, 'any object -
whatever its innards - that is reliably and
voluminously predictable from the inten-
tional stance is in the fullest sense of the
word a believer' (1988, p. 496).
The two behaviourist theses considered so
far are certainly controversial, but both are
weak forms of behaviourism. Even their
conjunction does not imply that any two
behavioural duplicates necessarily share the
same mental states. That last claim is our
third behaviourist thesis:
Supervenient behaviourism: psychological
facts supervene on physical beha-
vioural dispositions: necessarily, if x
and y differ with respect to types of
mental states, then they differ with
respect to types of behavioural disposi-
tions (see SUPERVENIENCE).
(Note that analytic behaviourism entails
supervenient behaviourism, but not con-
versely.)
Supervenient behaviourism may be
broadened to include the supervenience of
linguistic meaning on behavioural disposi-
tions, or narrowed to exclude, say, sensa-
tions. Supervenient behaviourism can
accommodate the view that content is not
entirely 'in the head' by taking the super-
venience base to comprise physical beha-
vioural dispositions together with facts
about the subject's environment. (The pos-
sibility of this latter refinement will be taken
for granted in what follows.)
Supervenient behavourism is quite com-
patible with eliminativism (according to
which no x and y ever differ mentally!).
And indeed Quine appears to need the
former to derive the latter via the indeter-
minacy thesis (see e.g. van Cleve, 1992).
However, some philosophers who repudiate
eliminativism are at least attracted by
supervenient behaviourism. More precisely,
these philosophers hold a doctrine about the
BEHAVIOURISM
mind which, together with assumptions
they would probably accept, entails super-
venient behaviourism. This doctrine may be
explained as follows.
In the writings of DA VIDSON, and to some
extent Dennett and Dummett, there is a
strand of thought which amounts to a kind
of third-person Cartesianism (see HISTORY).
Where Cartesianism holds that someone
has, under ideal conditions, complete and
infallible access to his own mental life (see
FIRST-PERSON AUTHORITY), the third-
person version says that someone has,
under ideal conditions, complete and infal-
lible access to the mental life of another. Or,
in its typical formulation, complete and
infallible access to the PROPOSITIONAL
ATTITUDES (and/or language) of another.
Picturesquely: under ideal conditions a sub-
ject's belief-box and desire-box become
transparent. This kind of claim is sometimes
called interpretivism (Johnston, 1991).
We can dramatize interpretivism by intro-
ducing the familiar device of the ideal inter-
preter: an idealization of a human being, in
ideal epistemic circumstances. The ideal
interpreter is capable, according to inter-
pretivism, of discovering exactly what a
subject believes, desires (and/or means). The
powers, and data of the interpreter are a
matter for dispute, but at least three points
should be noted.
First, the interpreter's powers are not to
be construed - on pain of triviality - to be
just whatever powers are necessary in order
to deliver the facts.
Second, the interpreter's official evidence
does not include the contents of the sub-
ject's mental states, or the meaning of his
utterances, for that is what the interpreter
is supposed to be finding out.
Third, interpretivism does not offer a
reductive analysis - at least of the attitudes;
for the reference to our best judgments (cap-
tured in the heuristic device of the ideal
interpreter) is supposed to be an inelimin-
able part of the story. Lewis (19 74) makes
use, in effect, of the ideal interpreter, but
intends this to be a staging-post en route to
a fully reductive account. That is not a
version of interpretivism.
l37
BEHA VIOURISM
Bearing in mind the three points above,
we could formulate interpretivism in the
particular case of belief as the thesis that all
biconditionals of the following form are a
priori:
x believes that p if and only if, if there
were an appropriately informed ideal
interpreter, he would be disposed to
attribute to x the belief that p.
Something like this seems to be clearly
expressed by Davidson when he writes:
What a fully informed Interpreter could
know about what a speaker means is all
there is to learn; the same goes for what
the speaker believes. (1986, p. 315)
Davidson is making a substantial claim, and
so 'fully informed' should not be understood
as 'fully informed about what a speaker
means and believes'. He here has in mind
his 'radical interpreter' (Davidson, 1973),
starting the process of interpretation with
no knowledge of the meaning of the sub-
ject's language or the content of his beliefs
(although Davidson allows the interpreter's
initial data to include intentional facts,
namely what (uninterpreted) sentences the
subject 'holds true').
Dennett comes to a similar conclusion:
[AlII there is to being a true believer is
being a system whose behaviour is reli-
ably predictable via the intentional
strategy, and hence all there is to really
and truly believing that p (for any pro-
position p) is being an intentional
system for which p occurs as a belief in
the best (most predictive) interpreta-
tion. (1981, p. 29)
Now an interpretation requires a (hypo-
thetical) interpreter, and so Dennett would
appear to be saying that some ideal inter-
preter could deliver all and only the facts
about what a subject believes. Moreover, as
the failure of reductions in the philosophy of
mind has been one of Dennett's major
themes, he very likely thinks that this talk
of interpretations and interpreters cannot be
analysed away. But then Dennett is an
interpretivist.
138
Interpretivism does not entail super-
venient behaviourism. For it may be that
the interpreter must take into account
certain non-behavioural facts (e.g. facts
about inner causal organization) when
determining what someone believes. It
seems likely, however, that Dennett's and
Davidson's ideal interpreters would not
ascribe different beliefs to two subjects
unless they differed in their physical beha-
vioural dispositions: their respective inter-
preters would judge any two behavioural
duplicates alike. And as the ideal inter-
preter's word is gospel, this entails that no
two subjects would differ in beliefs (or, we
may assume, desires) unless they differed in
physical behavioural dispositions. But this is
supervenient behaviourism with respect to
the propositional attitudes (and linguistic
meaning, at least in the case of Davidson).
Dennett appears to arrive at inter-
pretivism by an ambitious extension of his
route to the behaviour-as-sufficient view.
That is: why not take the ideal interpreter
at his word, if we have no clear conception
of how he can be mistaken? Davidson's
reasons for interpretivism are less straight-
forward. The main ones emerge in the fol-
lowing passage, where Davidson argues for
a behaviourist version of interpretivism
from a Wittgensteinian premise about the
'publicity' of meaning:
As Ludwig Wittgenstein, not to
mention Dewey, G.H. Mead, Quine, and
many others have insisted, language is
intrinsically social. This does not entail
that truth and meaning can be defined
in terms of observable behaviour, or
that it is 'nothing but' observable beha-
viour; but it does imply that meaning is
entirely determined by observable
behaviour, even readily observable
behaviour. That meanings are de-
cipherable is not a matter of luck; public
availability is a constitutive aspect of
language. (Davidson, 1990, p. 314)
Dummett has expressed similar views (e.g.
1973, pp. 217-18).
Davidson's 'readily observable behaviour'
is intended to be understood as 'behaviour
observable as physical behaviour' (or, perhaps
as: 'behaviour observable as the holding true
of sentences'). For suppose that the relevant
deliverances of observation include linguis-
tic behaviour such as saying that snow is
white. Then Davidson's claim that meaning
is determined by this sort of behaviour is,
while perhaps not entirely trivial, at least
extremely weak. Alternatively, suppose that
the deliverances of observation include
psychological behaviour such as intending to
induce a certain belief in one's audience, but
exclude any kind of behaviour described
in a way which presupposes linguistic
meaning. Then Davidson's claim would cer-
tainly not be trivial, but would violate his
well-known insistence that a speaker's pro-
positional attitudes and the meaning of his
language are settled together, neither one
having priority over the other (e.g. David-
son, 1973).
It seems, then, that lying behind David-
son's argument is a controversial epistemic
thesis: our warrant for saying that someone
speaks a certain language, or has certain
beliefs, is ultimately founded on behaviour,
observed as physical behaviour (or, perhaps:
observable as the holding true of sentences).
But even if that is found acceptable, David-
son's apparent move from 'our judgements
are founded on such-and-such data' to 'the
truth of our judgements is determined by
such-and-such data' is questionable. We
have learnt much about the world that is
not 'determined' by what we have learnt it
from.
BEHAVIOURISM REFUTED?
To refute the view that a certain level of
behavioural dispositions is necessary for a
mental life, we need convincing cases of
thinking stones, or utterly incurable para-
lytics, or disembodied minds. But these
alleged possibilities are to some merely that.
To refute the view that a certain level of
behavioural dispositions is sufficient for a
mental life, we need convincing cases of
rich behaviour with no accompanying
mental states. The typical example is of a
puppet controlled, via radio links, by other
BEHAVIOURISM
minds outside the puppet's hollow body
(Peacocke, 1983, p. 205; Lycan, 1987, p. 5;
see also Block, 1981). But one might
wonder - and Dennett does - whether the
dramatic devices are producing the anti-
behaviourist intuition all by themselves.
And how could the dramatic devices make
a difference to the actual facts of the case? If
the puppeteers were replaced by a machine
(not designed by anyone, yet storing a vast
number of input-output conditionals) which
was reduced in size and placed in the pup-
pet's head, do we still have a compelling
counterexample to the behaviour-as-
sufficient view? At least it is not so clear.
Such an example would work equally
well against (the anti-eliminativist version
of) the view that mental states supervene
on behavioural dispositions. But super-
venient behaviourism could be refuted by
something less ambitious. The 'X-worlders'
of Putnam (1965), who are in intense pain
but do not betray this in their verbal or
non-verbal behaviour, behaving just as
pain-free human beings, would be the right
sort of case. But even if Putnam has pro-
duced a counterexample for pain - which
Dennett for one would doubtless deny - an
'X-worlder' story to refute supervenient
behaviourism with respect to the attitudes
or linguistic meaning will be less intuitively
convincing. Behaviourist resistance is easier
here for the reason that having a belief, or
meaning a certain thing, lack distinctive
phenomenologies.
There is a more sophisticated line of
attack. As Quine has remarked, some have
taken his thesis of the indeterminacy of
translation as a reductio of his behaviourism
(1990, p. 37). For this to be convincing,
Quine's argument for the indeterminacy
thesis has to be persuasive on its own
terms, and that is a disputed matter.
If behaviourism is finally laid to rest to
the satisfaction of most philosophers, it will
probably not be by counterexample, or by a
reductio from Quine's indeterminacy thesis.
Rather, it will be because the behaviourist's
worries about other minds, and the public
availability of meaning, have been shown to
be groundless, or not to require behaviour-
139
BELIEF (1): METAPHYSICS OF
ism for their solution. But we can be sure
that this happy day will take some time to
arrive.
See also CONTENT; DUALISM; An Essay on
Mind sections 3.6. & 3.7. IDENTITY THEOR Y;
INTENTIONALITY; THOUGHT; THOUGHT AND
LANGUAGE.
BIBLIOGRAPHY
Armstrong. D.M. 1968. A Materialist Theory of
the Mind. London: Routledge and Kegan
Paul.
Block. N., ed. 1980. Readings in Philosophy of
Psychology, vol. 1. Cambridge, MA: Harvard
University Press.
Block, N. 1981. Psychologism and behaviour-
ism. Philosophical Review, 90, 5-43.
Davidson, D. 1973. Radical interpretation.
Dialectica, 27, 313-28. Reprinted in David-
son, Inquiries into Truth and Interpretation.
Oxford University Press, 1984.
Davidson, D. 1986. A coherence theory of
truth and knowledge. In Truth and Inter-
pretation: Perspectives on the Philosophy of
Donald Davidson, ed. E. LePore. Oxford: Basil
Blackwell.
Davidson, D. 1990. The structure and content
of truth. Journal of Philosophy, 87, 279-328.
Dennett, D.C. 1971. Intentional systems.
Journal of Philosophy, 68, 87-106. Reprinted
in Dennett, Brainstorms. Brighton: The Har-
vester Press.
Dennett, D.C. 1981. True believers. In SCientific
Explanation, ed. A. F. Heath. Oxford Uni-
versity Press. Reprinted in Dennett, The
Intentional Stance. Cambridge, MA: MIT
Press, 1987; page references are to the latter.
Dennett, D.C. 1988. Precis of The Intentional
Stance. Behavioural and Brain Sciences, 11,
495-504.
Dummett, M.A.E. 1973. The philosophical
basis of intuitionistic logic. In Logic Collo-
quium '73, ed. H. E. Rose and J. e. Shep-
herdson. Reprinted in Dummett, Truth and
Other Enigmas. London: Duckworth, 1978;
page references are to the latter.
Geach, P.T. 1957. Mental Acts: Their Content
and their Objects. London: Routledge &
Kegan Paul.
Hempel, e.G. 1949. The logical analysis of
psychology. In Readings in Philosophical Ana-
140
lysis, ed. H. Feigl and W. Sellars. New York:
Appleton-Century-Crofts. Reprinted with
revisions in Block 1980; page references are
to the latter.
Johnston, M. 1991. The missing explanation
argument and its impact on subjectivism. MS.
Lewis, D.K. 1974. Radical interpretation.
Synthese, 23, 331-44. Reprinted with post-
scripts in Lewis, 1983.
Lewis, D.K. 1980. Mad pain and martian pain.
In Block 1980. Reprinted with postscript in
Lewis. 1983.
Lewis. D.K. 1983. Philosophical Papers. vol. 1.
Lycan, W.G. 1987. Consciousness. Cambridge,
MA.: MIT Press.
Modgil, S., and Modgil, C., eds. 1987. B. F.
Skinner: Consensus and Controversy. London:
Falmer Press.
Peacocke, C. 1983. Sense and Content. Oxford
University Press.
Putnam, H. 1965. Brains and behavior. In
Analytical Philosophy, Vol 2, ed. R. J. Butler.
Reprinted in Block, 1980.
Quine, W.V. 1960. Word and Object. Cam-
bridge, MA.: MIT Press.
Quine, W.V. 1990. Pursuit of Truth. Cam-
bridge, MA.: Harvard University Press.
Ryle, G. 1949. The Concept of Mind. New York:
Barnes & Noble.
Skinner, B.F. 1971. Beyond Freedom and
Dignity. New York: Alfred A. Knopf.
Skinner, B.F., et al. 1984. Skinner: canonical
papers. Behavioural and Brain Sciences, 7,
473-701.
Van Cleve. J. 1992. Semantic supervenience
and referential indeterminacy. Journal of
Philosophy, 89, 344-61.
Watson, J.B. 1930. Behaviorism. University of
Chicago Press.
Wittgenstein, L. 1958. Philosophical Investiga-
tions (2nd edn). Oxford: Basil Blackwell.
Thanks to David Lewis, Michael Thau and the
editor for helpful comments.
ALEX BYRNE
belief (1): metaphysics of The two most
important metaphysical theories of belief are
what I shall call the 'propositional' and the
'sentential' theories. The German philoso-
pher, Gottlob Frege, founded the proposi-
tional approach. The sentential view is
developed most fully in the work of Jerry
Fodor. I begin with the propositional theory.
PROPOSITIONAL THEORIES
According to Frege and his followers, propo-
sitions are abstract, non-mental entities
which form the contents of beliefs (and
certain other psychological attitudes, for
example, desires). Belief consists in the mind
standing in a special relation of direct
apprehension to a proposition. So, if you
and I both believe that unemployment is
rising, say, we (or our minds) grasp (in the
way distinctive of belief) the same abstract
proposition. What we believe is the same,
since there is a single proposition to which
we are both here related.
Frege presented no doctrine about what
direct apprehension comes to beyond saying
that since its objects are both abstract and
non-mental, it is different from either sense-
perception or INTROSPECTION. He did,
however, make a number of remarks about
the character of abstract propositions. In
Frege's view, propositions are structured
entities, having as constituents the mean-
ings or senses of the terms composing the
sentences embedded in contexts of the form
'that p'. So, for example, according to Frege,
the reason why
(1) Jones believes that the inventor of
bifocals is a Philadelphian
may be true when
(2) Jones believes that the first U.S. Post-
master General is a Philadelphian
is false is that in the context of (1) and (2)
'the inventor of bifocals' and 'the first U.S.
Postmaster General' do not denote a single
person (even though in other contexts, of
course, they do, namely Benjamin Frank-
lin). Indeed they do not denote people at all.
Instead they stand for their normal mean-
ings or senses (which Frege termed 'indivi-
dual concepts'). So, the proposition
expressed by the 'that' clause in (1) has a
different component from the proposition
expressed by the 'that' clause in (2). The
propositions themselves must be distinct,
then, and, as a result, belief in one can
occur without belief in the other.
Critics of Frege's approach have some-
times argued that identifying propositions
with the meanings or senses of sentences
amounts to making propositions 'denizens of
darkness', since the notion of meaning is
itself radically unclear (see Quine, 1960).
Some philosophers sympathetic to the
general Fregean approach have responded
to this charge by claiming that propositions
are better identified with functions from pos-
sible worlds to truth values or equivalently
with sets of possible worlds in which the
value of the function is 'true' (Stalnaker,
1984). Propositions, conceived of in this
way, do at least have the virtue of reason-
ably clear identity conditions. But these
identity conditions themselves generate a
serious problem for the abstract propOSition
theory; for the possible-worlds elucidation of
propositions entails that. if the proposition
that p is necessarily equivalent to the propo-
sition that q. then the former proposition is
identical with the latter. Hence, on the
abstract proposition theory, if a person
believes that p, he or she automatically
believes that q. And this result seems un-
acceptable. Surely, for example, I can believe
that all brothers are male siblings without
believing that all ophthalmologists are eye
doctors. The situation with mathematical
beliefs is even more counterintuitive. It
seems undeniable that I can believe that
7 + 5 = 12 without possessing all true
mathematical beliefs. But the proposition
that 7 + 5 = 12 is necessarily equivalent to
every true mathematical proposition, so that
really there is only one such proposition.
Hence, in believing that 7 + 5 = 12, I
must also believe that 793 - 132 = 661,
that 18 x 19 = 342. and so on.
One standard reply to the problem math-
ematical beliefs present for the proposition
theory is to argue that such beliefs really
relate persons to contingent propositions
about the link between overt, public math-
ematical sentences and the one necessary
141
proposition. Thus, to believe that 7 + 5 =
12 is to believe the contingent proposition
that '7 + 5 = 12' expresses the necessary
proposition. Since this contingent proposi-
tion is not necessarily equivalent to the pro-
position that '793 - 132 = 661' expresses
the necessary proposition, one can believe
that 7 + 5 = 12 without thereby auto-
matically believing that 793 - 132 = 661.
This reply faces a further objection,
however. Suppose a Frenchman has a belief
that he expresses by uttering the sentence
'Sept plus cinq fait douze'. If I have a belief
that I express by uttering the sentence
'Seven plus five equals twelve', then, unless
the contexts of utterance are atypical. it
seems intuitively reasonable to say that we
have the same mathematical belief. But
according to the above reply, my belief must
be different from that of the Frenchman.
It is worth noting that the original
Fregean view that propositions are senses or
meanings evidently handles the above
problem of mathematical beliefs. But there
remains another very significant objection
(see Burge, 1978). This is simply that, con-
trary to what Frege supposed, synonyms are
not always intersubstitutable salva veritae in
belief contexts. Suppose, for example, that
when asked about Jill and John's where-
abouts, Paul says sincerely that they have
gone to Paris for a dozen weeks. Suppose
also that Paul is under the misapprehension
that a dozen is 20 items, not 12, so that he
expects Jill and John to return 20 weeks
later. Then it seems reasonable to say both
that
(3) Paul believes that Jill and John have
gone to Paris for a dozen weeks
is true and that
(4) Paul believes that Jill and John have
gone to Paris for 12 weeks
is false. However, (4) results from (3) by the
substitution of terms that are ordinarily
classified as having the same meaning.
One possible reply the Fregean might
make to this problem is to argue that
142
'12' and 'a dozen' do not really have
exactly the same meaning. The difficulty
now is that it is far from clear which
expressions do have the same meaning.
Moreover, for any two expressions that are
synonymous, there might be someone who
believes that they differ in meaning. So a
case parallel to that in (3) and (4) could still
be constructed.
SENTENTIAL THEORIES
Let us turn now to the sentential theory.
According to sententialists, the objects of
belief are sentences. Some sententialists
maintain that public sentences are the
objects of belief. Gilbert R YLE, for example,
seems to have held that to believe that p is
to be disposed to assent to some natural
language sentence that means that p. And
Donald DA VIDSON is usually read as accept-
ing a version of the public sentence
approach. The dominant version of the sen-
tential theory, however, is the view that the
objects of belief are private sentences. This
view goes hand in hand with the com-
putational conception of the mind (see
COMPUTATIONAL MODELS; LANGUAGE OF
THOUGHT).
Computers are symbol manipulators: they
transform symbols in accordance with fixed
syntactic rules and thereby process informa-
tion. If the mind is a computer then its
states must themselves be symbolic states in
whatever inner language the mind employs
(call it 'Mentalese'). So, belief must involve a
relation to a string of symbols in Mentalese.
This symbol string is a sentence of Ment-
alese which has, as its natural language
counterpart. whatever sentence is used to
specify the content of the belief in a public
context. So, on the dominant version of the
sentential theory, believing that nothing
succeeds like excess, say, is a matter of the
mind standing in a certain computational
relation (distinctive of belief) to a sentence
in Mentalese which means that nothing
succeeds like excess. This sentence is physi-
cally realized in the brain by some neural
state just as symbol strings in electronic
computers are physically realized by
charged states of grids or patterns of elec-
trical pulses.
The sentential theory sketched above
involves no explicit commitment to abstract
propositions. But propositions can still enter
in the analysis of what it is for a given sen-
tence of Mentalese to mean that such and
such is the case. Thus, it is a mistake to
suppose that a sentential approach to belief
automatically repudiates propositions.
There are four main considerations that
proponents of the dominant version of the
sentential theory usually adduce as moti-
vating their view. To begin with, it is noted
that the view that the mind is a computer
is one that has considerable empirical
support from COGNITIVE PSYCHOLOGY. The
sentential theory, then, is an empirically
plausible theory, one that supplies a
mechanism for the relation that proposi-
tionalists take to obtain between minds and
propositions. The mechanism is mediation
by inner sentences.
Secondly, the sentential theory offers a
straightforward explanation for the parallels
that obtain between the objects or contents
of speech acts and the objects or contents of
belief. For example, I may say what I
believe. Furthermore, the object of believing,
like the object of saying, can have semantic
properties. We may say, for example,
(5) What Jones believes is true
and
(6) What Jones believes entails what
Smith believes.
One plausible hypothesis, then, is that the
object of belief is the same sort of entity as
what is uttered in speech acts (or what is
written down).
The sentential theory also seems sup-
ported by the following argument. The
ability to think certain thoughts appears
intrinsically connected with the ability to
think certain others. For example, the
ability to think that John hits Mary goes
hand in hand with the ability to think that
Mary hits John, but not with the ability to
think that London is overcrowded. Why is
this? The ability to produce or understand
certain sentences is intrinsically connected
with the ability to produce or understand
certain others. For example, there are no
native speakers of English who know how
to say 'John hits Mary' but who do not
know how to say 'Mary hits John'. Simi-
larly, there are no native speakers who
understand the former sentence but not the
latter. These facts are easily explained if
sentences have a syntactic and semantic
structure. But if sentences are taken to be
atomic, these facts are a complete mystery.
What is true for sentences is true also for
thoughts. Thinking thoughts involves manip-
ulating MENTAL REPRESENTATIONS. If
mental representations with a propositional
content have a semantic and syntactic
structure like that of sentences, it is no acci-
dent that one who is able to think that John
hits Mary is thereby also able to think that
Mary hits John. Furthermore, it is no acci-
dent that one who can think these thoughts
need not thereby be able to think thoughts
having different components - for example,
the thought that London is overcrowded.
And what goes here for thought goes for
belief and the other PROPOSITIONAL ATTI-
TUDES.
Consider next the inference from
(7) Rufus believes that the round object
ahead is brown
and
to
(8) The round object ahead is the coin
Rupert dropped
(9) Rufus believes that the coin Rupert
dropped is brown.
This inference is strictly parallel to the infer-
ence from
(10) Rufus uttered the sentence 'The
round object ahead is brown'
and (8) to
143
(11) Rufus uttered the sentence 'The
coin Rupert dropped is brown'.
If the immediate objects of belief are sen-
tences, we should expect the former infer-
ence to be invalid just as the latter is.
Another motivating factor is the thought
that. since the pattern of causal interactions
among beliefs often mirrors various infer-
ential relations among the sentences that
are ordinarily used to specify the objects of
beliefs, it is natural to suppose that these
objects have logical form. For example, cor-
responding to the inference from
( 12) All dogs make good pets
and
(13) All of Jane's animals are dogs
to
(14) All of Jane's animals make good pets
we have the fact that, if John believes that
all dogs make good pets and he later comes
to believe that all of Jane's animals are dogs,
he will, in all likelihood, be caused to
believe that all of Jane's animals make good
pets. Generalizing, we can say that a belief
of the form
(15) All Fs are Gs
together with a belief of the form
(16) All Gs are H,s
typically causes a belief of the form
(17) All Fs are Hs.
This generalization concerns belief alone.
But there are also generalizations linking
belief and desire. For example, a desire of
the form
(18) Do A,
together with a belief of the form
144
(19) In order to do A, it is necessary to
do B,
typically generates a desire of the form
(20) Do B.
Now these generalizations categorize beliefs
and desires according to the logical form of
their objects. They therefore require that the
objects have logical forms. But the primary
possessors of logical form are sentences.
Hence, the (immediate) objects of beliefs and
desires are themselves sentences.
Advocates of the propositional theory
sometimes object to the sentential approach
on the grounds that it is chauvinistic. Maybe
our beliefs are represented in our heads in
the form of sentences in a special mental
language, but why should all beliefs neces-
sarily be so represented in all possible crea-
tures? For example, couldn't belief tokens
take the form of graphs, maps, pictures, or
indeed some other form dissimilar to any of
our public forms of representation?
This objection is based on a mis-
understanding. The sentential theory is not
normally presented as an analysis of the
essence of belief, of what is common to all
actual and possible believers in virtue of
which they have beliefs. So, it has nothing
to say about the beliefs of angels, say, or
other possible believers. Rather it is a theory
of how belief is actually realized in us.
One great virtue of the sentential theory
is that it provides a satisfying explanation of
how it is possible to believe that p without
believing that q even when 'p' and 'q' are
synonymous. Consider, for example, the
earlier case involving the sentences (3) and
(4). Believing that Jill and John have gone
to Paris for a dozen weeks requires Paul to
token an inner sentence which means that
they have gone to Paris for a dozen weeks.
On the assumption that this sentence is not
the same as the one that is tokened when it
is believed that that Jill and John have gone
to Paris for 12 weeks, there is no difficulty
in explaining how (3) can be true when (4)
is false. The beliefs are separable because
they involve the same relation to different
sentences with the same meaning. Just as
writing down the sentence 'That is a vixen',
for example, is quite distinct from writing
down the sentence 'That is a female fox',
notwithstanding their identity of meaning,
so too the belief expressed in (3) is quite dis-
tinct from the belief expressed in (4).
So far I have focused exclusively on what
are known as de dicto beliefs, that is, beliefs
that are standardly attributed to people
using predicates of the form, 'believes that
p'. There is another important class of
beliefs, however. These are standardly
attributed using predicates of the form,
o'believes of x that it is F'. Beliefs of this sort
are called 'de re beliefs'. Consider, for eample,
my believing of the building I am facing that
it is an imposing structure. This is a belief
with respect to a particular building,
however that building is described. Suppose,
for example, the building is St Paul's Cathe-
dral. Then, in believing of the building I am
facing that it is an imposing structure I am
thereby believing of St Paul's that it is an
imposing structure. So, for a belief to be de re
with respect to some object 0, there must
really be an object 0 which the belief is
about. By contrast, if I simply believe that
the building I am facing is imposing - this is
the de dicto case - I need not believe that St
Paul's is imposing. For I may not believe
that the building I am facing is St Paul's.
Moreover, it is not a condition of my having
the belief that the building I am facing is
imposing that there really be any building
before me. I might. for example, be under
the influence of some drug, which has
caused me to hallucinate a large building.
De re beliefs, then, are beliefs held with
respect to particular things or people,
however described, that they have such and
such properties. On the propositional
theory, such beliefs are often taken to
require that the given thing or person itself
enter into the proposition believed. So,
believing of Smith that he is dishonest is a
matter of standing in the belief relation to
the proposition that Smith is dishonest,
where this proposition is a complex entity
having the person, Smith, as one of its com-
ponents.
The sentential theory can account for de
re belief in a similar fashion. The assump-
tion now is that the inner sentence is a sin-
gular one (consisting in the simplest case of
a name concatenated with a predicate).
This sentence has, as its meaning, a propo-
sition which meets the above requirements
(assuming a propositional approach to sen-
tence meanings).
Of the two theories I have discussed, the
sentential view probably has the wider
support in philosophy today. However, as
my earlier comments should have made
clear, the two theories are not diametrically
opposed to one another. For the sentential
theory, unlike the propositional view, is not
(in its standard form) an analysis of belief.
Moreover, its advocates are not necessarily
against the introduction of abstract proposi-
tions.
There is one further feature worth com-
menting upon that is common to both the-
ories. This is their acceptance of the
relational character of belief. The primary
reason for taking belief to be relational is
that existential generalization applies to its
syntactic objects. For example,
(21) Jones believes that gorillas are more
intelligent than chimpanzees
entails
(22) There is something Jones believes.
Not all philosophers accept that existential
generalizations like this one should be taken
at face value as indicating a metaphysical
commitment to some entity which is the
believed object. However, unless some
strong arguments can be given which show
that this case is anomalous, it is surely rea-
sonable to classify it with other standard
cases of existential generalization, and
hence to grant that there really are objects
to which we are related in belief.
See also BELIEF (2); DENNETT; DRETSKE; An
Essay on Mind section 2.1; FODOR; INTEN-
TION ALITY ; RATIONALITY; THOUGHT;
THOUGHT AND LANGUAGE.
145
BELIEF (2): EPISTEMOLOGY OF
BIBLIOGRAPHY
Burge. T. 1978. Belief and synonymy. The
Journal of Philosophy. 75. 119-38.
Davidson. D. 1975. Thought and talk. In Mind
and Language. ed. S. Guttenplan. Oxford Uni-
versity Press.
Fodor. J. 1978. Propositional attitudes. The
Monist. 61. 501-23.
Fodor. J. 1978. Psychosemantics. Cambridge.
MA.: MIT Press.
Fodor. J. 1990. A Theory of Content and Other
Essays. Cambridge. MA.: MIT Press.
Frege. G. 1968. The thought: a logical
inquiry. In Essays on Frege. ed. E. D. Klemke.
Urbana. Ill.: University of Illinois Press.
Quine. W. 1960. Word and Object. Cambridge.
MA.: MIT Press.
Ryle. G. 1949. The Concept of Mind. London:
Hutchinson and Co.
Salmon. N. 1986. Frege's Puzzle. Cambridge.
MA.: MIT Press.
Sellars. W. 1968. Science and Metaphysics.
London. Routledge & Kegan Paul.
Stalnaker. R. ] 984. Inquiry. Cambridge. MA.:
MIT Press.
MICHAEL TYE
belief (2): epistemology of The general
heading of 'The Epistemology of Beliefs'
covers an array of closely interrelated ques-
tions. I shall discuss just three areas of
interest. First. there is the question of how
we actually go about ascribing beliefs in
daily life. This question has a first- and a
third-person angle: how do I ascribe beliefs
to others? How do I ascribe beliefs to
myself? What are the relations between the
two? Second. there is a question concerning
the extent to which our ordinary practices
of ascribing beliefs are fallible. the extent to
which they stand in need of ratification by
something more - a scientific or philosoph-
ical theory of belief. Third. there is a ques-
tion concerning the limits of our ability to
know about beliefs. We seem to be very
good at finding out about our own beliefs
and those of other humans. But by what
methods. and to what extent. can we know
about the beliefs of non-humans such as
animals. extra-terrestrials and robots? I
shall discuss these issues in turn.
146
COMMONSENSE PSYCHOLOGY
As a lead into the topic I shall begin by
elaborating a view that has been defended
recently by a number of philosophers and
psychologists (e.g. Churchland. 1981; Stich.
1983; Fodor. 1987. Wellman. 1990). The
view concerns. in the first instance at least.
the question of how we. as ordinary human
beings. in fact go about ascribing beliefs to
one another. The idea is that we do this on
the basis of our knowledge of a common-
sense theory of psychology (see FOLK PSy-
CHOLOGY). The theory is not held to consist
in a collection of grandmotherly sayings.
such as 'once bitten. twice shy'. Rather it
consists in a body of generalizations relating
psychological states to each other. to inputs
from the environment. and to actions. Here
is a sample from Churchland (1981. p. 71):
(1) (x)(p)(if x fears that p. then x desires
that not-p)
(2) (x)(p)(if x hopes that p and x discovers
that p. then x is pleased that p)
(3) (x)(p)(q) (if x believes that p and x
believes that if P. then q. then. barring
confusion. distraction etc. x believes
that q)
(4) (x)(p)(q) (if x desires that p and x
believes that if q then P. and x is able to
bring it about that q. then. barring con-
flicting desires or preferred strategies. x
brings it about that q)
All of these generalizations should be
understood as containing ceteris paribus
clauses. (1). for example. applies most of the
time. but not invariably. Adventurous types
often enjoy the adrenalin thrill produced by
fear. This leads them. on occasion. to desire
the very state of affairs that frightens them.
Analogously with (3). A subject who
believes that p and believes that if p. then q.
would typically infer that q. But certain
atypical circumstances may intervene: sub-
jects may become confused. or distracted. or
they may find the prospect of q so awful
that they dare not allow themselves to
believe it. The ceteris paribus nature of these
generalizations is not usually considered to
be problematic, since atypical circumstances
are, of course, atypical. and the general-
izations are applicable most of the time,
We apply this psychological theory to
make inferences about people's beliefs,
desires and so on, If, for example, we know
that Diana believes that if she is to be at the
airport at four, then she should get a taxi at
half past two, and she believes that she is to
be at the airport at four, then we will
predict, using (3), that Diana will infer that
she should get a taxi at half past two.
The Theory Theory, as it is called, is an
empirical theory addressing the question of
our actual knowledge of beliefs. Taken in its
purest form it addresses both first- and
third-personal knowledge: we know about
our own beliefs and those of others in the
same way, by application of commonsense
psychological theory in both cases.
However, it is not very plausible to hold
that we always - or indeed usually - know
our own beliefs by way of theoretical infer-
ence. I ask myself, do I believe that a certain
student will get a First in her BA final
exams? To answer this I do not, of course,
theorize about myself and my beliefs: I
reason about the student and her abilities. I
decide that yes, she probably will achieve
the desired result. And thereupon I know,
immediately and without further reflection,
that I believe that she will attain a First. A
natural emendation to the pure Theory
Theory would thus be to add to it the rider
that, sometimes or often, we attribute beliefs
to ourselves in this non-theoretical way.
Since it is an empirical theory concerning
one of our cognitive abilities, the Theory
Theory is open to psychological scrutiny.
Various issues arise: we need to know the
detailed contents of the hypothesized com-
monsense psychological theory, we need to
know whether it is known consciously or
unconsciously, and we need to know how it
is acquired. I shall discuss one aspect of the
last of these issues, since it has been the
subject of interesting work in DEVELOP-
MENTAL PSYCHOLOGY. Suppose, then, that
adult human beings are in the possession of
a developed commonsense psychological
theory, and can apply it with reasonable
skill to the task of inferring the psycho-
logical states of others. How did this mature
capacity evolve?
Research has revealed that three-year-old
children are reasonably good at inferring
the beliefs of others on the basis of actions,
and at predicting actions on the basis of
beliefs that others are known to possess.
However, there is one area in which three-
year-olds' psychological reasoning differs
markedly from that of adults. Tests of the
following sort, 'False Belief Tests', reveal
largely consistent results. Three-year-old
subjects are witness to the following scen-
ario. A child, Billy, sees his mother place
some biscuits in a biscuit tin. Billy then goes
out to play, and, unseen by him, his mother
removes the biscuits from the tin and places
them in a jar, which is then hidden in a
cupboard. When asked 'Where will Billy
look for the biscuits?' the majority of three-
year-olds answer that Billy will look in the
jar in the cupboard - where the biscuits
actually are, rather than where Billy saw
them being placed. On being asked 'Where
does Billy think the biscuits are?' they again
tend to answer 'In the cupboard', rather
than 'In the tin'. Three-year-olds thus
appear to have some difficulty attributing
false beliefs to others in cases in which it
would be natural for adults to do so.
However, it appears that three-year-olds are
not lacking the idea of false beliefs in
general. nor does it appear that they strug-
gle with attributing false beliefs in other
kinds of situation. For example, they have
little trouble distinguishing between dreams
and play, on the one hand, and true beliefs
or claims on the other. By the age of four
and a half years, most children pass the
False Belief Tests fairly consistently. There is
as yet no generally accepted theory of why
three-year-olds fare so badly with the false
belief tests, nor of what it reveals about
their conception of beliefs. (See Wellman,
1990 and Perner, 1991 for detailed discus-
sion.)
Recently some philosophers and psycho-
logists have put forward what they take to
be an alternative to the Theory Theory: the
'Simulation Theory' (e.g. Gordon, 1986;
147
Goldman, 1989; Heal, 1986). The basic
idea of the Simulation Theory is that we
understand the psychologies of others by
using our own psychological processes to
simulate those of others. The Simulation
Theory builds upon the point, noted
earlier, that we do not typically know
about our own beliefs on the basis of
theoretical inference. Rather, in the typical
case, we know what we believe immedi-
ately and non-inferentially. Moreover, if we
consider how we would act, or what we
would believe, in some non-actual or
future situation, it again appears that we
do not proceed on the basis of theoretical
inference. Rather, we imagine ourselves to
be in the non-actual situation, and then
indulge in a sort of 'pretend play' (Gordon,
1986): I imagine the sound of footsteps in
the basement, I ask myself, in effect, 'What
do I do now?' To answer the question I
need to discover what I would believe,
desire and so on, in such a situation. And
to answer this, rather than theorize, I
pretend to be (or play at being) in the
situation, and I let my psychological pro-
cesses run as normal: footsteps in the base-
ment: an intruder! How did he get in? It was
a hot day, and I left the window open: he
climbed in through the window. I now know
that I would believe that an intruder has
climbed through the window. But I did not
infer that I would believe this from general
psychological principles. Rather, I let my
belief-forming mechanisms run 'off line',
within the scope of the pretence, and
merely note the result. The Simulation
Theory holds that the same method is
applied to discovering the beliefs of others:
one projects oneself into the other's situa-
tion and, once again, lets one's own psy-
chological processes run off line, and notes
the results. Of course, such projection
usually requires imagining oneself not only
in the other's spatio-temporal situation, but
also in his psychological situation: that is
to say, one adjusts for already known psy-
chological differences.
The Simulation Theory thus addresses
both first- and third-personal knowledge of
beliefs, and offers an appealing account of
148
the relationship between the two. While the
Theory Theory must either hold that we
know our own beliefs by theoretical infer-
ence, or allow for a sharp asymmetry
between the two cases, the Simulation
Theory argues that we know about the
beliefs of others by a natural extension of
our capacity to know about our own.
The Simulation Theory appears to face a
challenge: How can one run the necessary
simulations except on the basis of a psycho-
logical theory? After all, if one is to simulate
some process, one generally needs to know
how it works. And this would appear to
require, precisely, a theory of the process.
But this challenge, so put, misses the crucial
point. As believers, we ourselves already
have belief-forming processes to hand. And
we can deploy these processes with the aim
of discovering what beliefs they would
produce in a given situation, without
knowing how they work. We let the pro-
cesses run, we then record the results. As
Goldman (1989) points out, one can run a
successful simulation without deploying a
theory, providing (a) that the processes
driving the simulation are the same as the
actual psychological processes of the simu-
lated agent and (b) the initial states of simu-
lator and simulated are the same or
relevantly similar. And that is surely fair
enough.
However, the challenge does not end
there. We need also to consider the vital
element of making appropriate adjustments
for differences between one's own psycho-
logical states and those of the other. My
friend has been offered a job at a certain
Philosophy Department. Will he take it? I
cannot merely pretend that I have been
offered the job, and run through the reason-
ing I would go through to see whether I
would accept it. My friend and I differ in
respect of certain potentially relevant desires
and beliefs. The job is better paid than his
present one. How relevant is this? I need to
know how much he values money. Can I
discover this by further simulation? Either I
can or I cannot. If I cannot, then, it seems, I
must do some research and make a theoret-
ical inference. If I can, then surely a similar
question will arise when I perform this
further simulation: I will come to some psy-
chological state of my own that I cannot
assume to be shared by my friend. It is
implausible that in every such case simula-
tion alone will provide the answer.
The Simulation Theorist should prob-
ably concede that simulations need to be
backed up by the independent means of
discovering the psychological states of
others. But they need not concede that
these independent means take the form of
a theory. Rather, they might suggest, we
can get by with some rules of thumb, or
straightforward inductive reasoning of a
general kind.
A second and related difficulty with the
Simulation Theory concerns our capacity to
attribute beliefs that are too alien to be
easily simulated: beliefs of small children, or
psychotics, or bizarre beliefs deeply sup-
pressed in the unconscious. The small child
refuses to sleep in the dark: he is afraid the
Wicked Witch will steal him away. No
matter how many adjustments we make, it
may be hard for mature adults to get their
own psychological processes, even in
pretend play, to mimic the production of
such a belief. For the Theory Theory alien
beliefs are not particularly problematic: so
long as they fit into the basic general-
izations of the theory, they will be inferrable
from the evidence. Thus the Theory Theory
can account better for our ability to dis-
cover more bizarre and alien beliefs than
can the Simulation Theory.
The Theory Theory and the Simulation
Theory are not the only proposals about
knowledge of beliefs. A third view has its
origins in the later philosophy of Ludwig
WITTGENSTEIN. On this view both the
Theory and Simulation Theories attribute
too much psychologizing to our common-
sense psychology. Knowledge of other
minds is, according to this alternative
picture, more observational in nature.
Beliefs, desires, feelings are made manifest
to us in the speech and other actions of
those with whom we share a language and
a way of life. When someone says 'It's going
to rain' and takes his umbrella from his
bag, it is immediately clear to us that he
believes it is going to rain. In order to know
this we neither theorize nor simulate: we
just perceive. Of course this is not straight-
forward visual perception of the sort that
we use to see the umbrella. But it is like
visual perception in that it provides immedi-
ate and non-inferential awareness of its
objects. We might call this the 'Observa-
tional Theory'.
The Observational Theory does not seem
to accord very well with the fact that we
frequently do have to indulge in a fair
amount of psychologizing to find out what
others believe. It is clear that any given
action might be the upshot of any number
of different psychological attitudes. This
applies even in the simplest cases, like
saying 'It's going to rain' and taking an
umbrella from a bag. One might do this
because one believes it's going to rain. But
one might do it for any number of other
reasons. For example: because one's friend
is suspended from a dark balloon near a
beehive, with the intention of stealing
honey. The idea is to make the bees believe
that it is going to rain, and therefore believe
that the balloon is a dark cloud, and there-
fore pay no attention to it, and so fail to
notice one's dangling friend. Given this sort
of possibility, the observer would surely be
rash immediately to judge that the agent
believes that it is going to rain. Rather, they
would need to determine - perhaps by
theory, perhaps by simulation - which of
the various clusters of mental states that
might have led to the action, actually did
so. This would involve bringing in further
knowledge of the agent, the background cir-
cumstances and so on. It is hard to see how
the sort of complex mental processes
involved in this sort of psychological reflec-
tion could be assimilated to any kind of
observation.
THE FALLIBILITY OF COMMONSENSE
PSYCHOLOGY
Thus far I have considered three theories
concerned with how we actually attribute
beliefs to ourselves and to others. I now turn
149
to the second area of enquiry mentioned
above, the extent to which our lay practices
of attributing beliefs are fallible and the
extent to which they stand in need of cor-
rection or justification by other methods.
One extreme view holds that common-
sense psychology is not merely fallible, but
completely wrong. Thus Churchland (1981)
holds that commonsense psychology is a
terrible theory, riddled with explanatory
gaps and failures. And Stich (1983) also
finds commonsense psychology to be ser-
iously flawed. Both conclude that the theory
is a false one, and that therefore its basic
explanatory apparatus of beliefs, desires
and so on should be rejected from our
ONTOLOGY. They conclude, then, that all
belief ascriptions are false: there really are
no such things as beliefs (see ELIM-
INATIVISM).
By contrast, FODOR (1987) argues that
commonsense psychology is so successful
that we would be well advised to take its
explanatory apparatus most seriously. He
points out that commonsense psychology is
a far better predictor of people's actions
than any other theory currently available.
Fodor defends his claim by pointing to very
simple, everyday cases. Suppose, for
example, that I telephone Fodor at his office
in New York and ask him if he would like to
give a paper at King's College, London, on
the 20th of June. Fodor accepts the invita-
tion. I am immediately in a position to
predict where Fodor will be on the 20th of
June. This prediction is based on common-
sense psychology: I hold that Fodor under-
stood what I was asking him, that he was
speaking sincerely, that he believed he
would indeed come to London and so on. Of
course the prediction is not infallible. Fodor
might indeed not be in London on the rele-
vant day. Perhaps he forgot that he had
already agreed to be somewhere else. Or
perhaps his flight will be cancelled. But the
important point is that the prediction is very
likely to be correct.
Predictions of the simple, everyday kind
just illustrated are indeed very common,
and do come out right most of the time. It
can hardly be denied that the success of
150
commonsense psychology provides good
prima facie evidence that the theory is
largely on the right track and that human
beings really do have propositional attitudes
of the kinds we commonly ascribe.
Ultimately one's opinion of the extent to
which commonsense psychology is fallible
will depend upon one's general ontological
conception of belief. To illustrate this inter-
play between epistemology and ontology, I
shall consider just one philosophical theory
of belief, FUNCTIONALISM.
According to functionalism, psychological
states are defined by their causal relations
to each other, to inputs from the environ-
ment, and to behaviour. A psychological
theory - for example, commonsense psy-
chology, as illustrated by (1)-(4) above -
specifies in general terms the relevant
causal relations into which psychological
states enter. Now, one can abstract from
such a theory a specification of a pattern of
causal interrelations among a network of
states. For example (3) above tells us that if
a subject believes that p, and believes that if
p, then q, then (barring confusion etc.) he
will believe that q. This specifies a pattern
among three states, which we can describe
abstractly as Sl and S2, which (barring S3
etc.) cause S4. One can imagine taking the
whole psychological theory and abstracting
a much larger pattern that exhausts the
types of causal interaction among all psy-
chological states: all explicit psychological
terms are removed from the theory, and
replaced by variable Ss. What remain speci-
fied are just the environmental inputs, the
behavioural outputs and the causal inter-
relations among the intervening states.
Functionalism holds that any network of
states that fits this large causal pattern will
automatically be a network of the psycho-
logical states described by the original
theory. Thus being a given belief or desire
just is being a state with a particular kind of
causal role in relation to other, analogously
defined states, inputs from the environment
and behavioural outputs.
There are a number of different versions
of functionalism that disagree over such
matters as the kind of theory that defines
the causal network (commonsense psychol-
ogy. cognitive science. or something else)
and the appropriate specifications of the
inputs and outputs (e.g. are the inputs distal
or proximal? Are the outputs bodily motions
or impacts on the environment?). (For
various versions of functionalism and crit-
ical discussion. see Block. 1980.)
If functionalism is true. then common-
sense psychology is in principle open to
validation or refutation by further enquiry.
For functionalism imposes constraints on
the causal structure of the aetiology of
behaviour. A being might then act just as
if it is the subject of a certain ensemble of
propositional attitudes. without actually
being so. For it may be that its behaviour is
caused by some network of interrelated
states that forms a quite different pattern
from the one that would be required for it to
have the desires. beliefs etc. that it out-
wardly appears to have. (See Block. 1981.
for an example.) In such a case the being
might have no psychological states at all. or
quite different ones from those that would
be ascribed. on the behavioural evidence. by
commonsense psychology. It follows that to
determine conclusively what beliefs
someone has. one would need to investigate
the causal structure of the aetiology of the
behaviour. something that would pre-
sumably involve study of the brain.
THE LIMITS OF OUR KNOWLEDGE
I turn now to the third question I promised
to address. and to a general epistemological
problem about beliefs. one that seriously
affects the project of giving a philosophical
theory of belief. The problem concerns the
beliefs (if any) of beings that are not
human: gorillas. pigeons. sophisticated
robots. silicon-based extra-terrestrials. etc.
Which of these have beliefs? And what do
they believe? Each of us has his or her own
view of the matter. But what is striking is
how little agreement there is and how little
we know about how to resolve the ques-
tions. The problem seems to me to go very
deep. for the following reason. Current psy-
chological theories do not give us much
help. A fair amount is known about the
behaviour of. for example. gorillas. But
there is very little consensus as to the
correct psychological explanation of the
behaviour. The matter thus remains within
the province of philosophy. But it seems
impossible to arrive at a properly justified
philosophical theory of belief without first
having at least some idea about which non-
human beings have beliefs.
Functionalists. for example. maintain
that an appropriately programmed robot
whose control centre is made of silicon
chips would indeed have beliefs. Some
phYSicalists argue against functionalism
precisely on the grounds that such things
would not have beliefs (see e.g. Searle.
1980). How is the issue to be resolved?
This sort of impasse is typical: a given
philosophical theory entails that a given
type of being would (or wouldn't) have
beliefs. Objectors then object that this con-
sequence is unacceptable. Proponents of the
theory bite the bullet and accept the con-
sequence. Objectors deny the consequence
and reject the theory.
The difficulty is deep because neither the
theories of belief nor the intuitions about
which beings have which beliefs have any
independent justification. As matters cur-
rently stand. the theories get evaluated
largely by how well they do at delimiting
the range of possible believers. Since differ-
ent theories make different predictions
about the range of believers. we need some
independent means of finding out about this
range. But we have nothing to go on except
our intuitions. And these are erratic and
highly fallible. After all. why should we
expect to have trustworthy intuitions about
the beliefs of possums. extra-terrestrials or
robots? We therefore need some independ-
ent way of evaluating the intuitions. But
that. in turn. seems to require a theory of
belief. We are therefore in an aporia: we
can't trust our intuitions unless we have a
theory by which to test them. but we have
no way of testing a theory except by seeing
how well it accords with our intuitions.
Psychology and philosophy evidently have a
long way to go.
151
See also ACTION; BELIEF (1); DAVIDSON;
DENNETT; An Essay on Mind section 1.2.2:
TUDES; PSYCHOLOGY AND PHILOSOPHY;
SEARLE.
BIBLIOGRAPHY
Block, N., ed. 1980. Readings in Philosophy of
Psychology, vol. 1. Cambridge, MA.: Harvard
University Press.
Block, N. 1981 Psychologism and behaviour-
ism. The Philosophical Review, 90, 5-43.
Churchland, P. 1981. Eliminative materialism
and propositional attitudes. The Journal of
Fodor, J. 1987. Psychosemantics. Cambridge,
MA.: MIT Press.
152
Goldman, A. 1989. Interpretation psycholo-
gized. Mind and Language, 4, 161-85.
Gordon, R. 1986. Folk psychology as simula-
tion. Mind and Language, 1, 158-171.
Heal, J. 1986. Replication and functionalism.
In Language, Mind and Logic, ed. J. Butter-
field. Cambridge University Press.
Pemer, J. 1991. Understanding the Representa-
tional Mind. Cambridge, MA.: MIT Press.
Searle, J. 1980. Minds, brains and pro-
grammes. Behavioral and Brain Sciences, 3,
417-24.
Stich, S. 1983. From Folk Psychology to Cog-
nitive Science: The Case Against Belief. Cam-
Wellman, H. 1990. The Child's Theory of Mind.
GABRIEL SEGAL
c
Cartesian see DUALISM; FIRST-PERSON
AUTHORITY; HISTOR Y.
Chomsky, Noam What do I think is true
of the mind, and what is the best way of
studying it? These are the questions on
which I have been asked to comment - with
the understanding that the scope is so broad
that only some general directions can be
outlined, with no effort to motivate or to
justify. More extensive discussion can be
found in material cited at the end. In the
interest of clarity, I will indicate some con-
tinuities, and some areas of controversy and
disagreement, but without attempting any
explanation or resolution.
Study of the mind is an inquiry into
certain aspects of the natural world, includ-
ing what have traditionally been called
mental events, processes, and states. A
'naturalistic approach' would seek to invest-
igate these aspects of the world as we do
any others, attempting to construct intelli-
gible explanatory theories and to move
toward eventual integration with the core
natural sciences. This 'methodological nat-
uralism' is not to be confused with 'meta-
physical naturalism' or other varieties (See
NATURALISM). It should be uncontentious,
though its reach remains to be determined.
Plainly, such an approach does not
exclude other ways of trying to comprehend
the world. Someone committed to it can
consistently believe (I do) that we learn
much more of human interest about how
people think and feel and act by reading
novels or studying history than from all of
naturalistic psychology, and perhaps always
will; Similarly, the arts may offer appreci-
ation of the heavens to which astrophysics
cannot aspire. We are speaking here of the-
oretical understanding, a particular mode of
comprehension. In this domain, any depar-
ture from a naturalistic approach carries a
burden of justification. Perhaps one can be
given, but I know of none.
Keeping to the search for theoretical
understanding, are there alternatives to a
naturalistic approach, perhaps with broader
reach or deeper insight? In the case of lan-
guage, there certainly are approaches that
transcend the bounds of naturalistic inquiry
outlined below; say, in sociolinguistics. But
it is less clear that they are alternatives. The
question is whether they presuppose
(usually tacitly) an approach to language in
these terms. I have argued elsewhere that
they do, however vigorously the fact may
be denied. Are there genuine alternatives?
One naturally keeps an open mind. I will
return to the question later, in connection
with work in contemporary philosophy of
language and mind that takes a different
course.
There are, of course, important questions
as to how naturalistic inquiry should
proceed. These questions are most appro-
priately raised with regard to the advanced
sciences, in which depth of understanding
and range of success may provide guides to
inquiry and analysis: physics, not psycho-
logy. In a naturalistic approach to humans,
we may safely put such concerns aside,
unless some reason is offered to show their
unique relevance here. Again, I know of
none.
Like other complex systems, the human
brain can be profitably viewed as an array of
interacting subcomponents, which can be
studied at various levels: atoms, cells, cell
assemblies, neural networks, computational-
representational (C-R) systems, etc. We
cannot know in advance which (if any) of
these approaches will provide insight and
153
CHOMSKY, NO AM
understanding. In several domains, includ-
ing language, the C-R approaches currently
have the strongest claim to scientific status.
at least on naturalistic grounds.
We might ask whether a study of the
brain in such terms is improper or con-
troversial. If not. we then ask whether the
theories developed (say of language) are
true. Does the brain in fact have the archi-
tecture. subsystems. states. properties.
spelled out in such terms in some particular
theory? As for the first query. it is hardly
controversial to suppose that the brain. like
other complex systems. has subsystems with
states and properties. The properties attrib-
uted in C-R theories are by and large well
understood. No general conceptual issues
seem to arise. only questions of truth. the
second query. which we may put aside here.
A related question is whether achieve-
ments. or promise. at other levels of inquiry
eliminates the basis for recourse to C-R the-
ories. In recent years. a great deal of interest
has been aroused by neural net and con-
nectionist models. and there has been much
discussion of the implications of the possi-
bility that. once developed. they might
provide a preferable alternative (see CON-
NECTIONISM). These discussions appear to
be naturalistic in temper. but that may be
questioned. Suppose that someone were to
propose that unstructured systems with
unknown properties might some day make
it possible to account for development of
organisms without appeal to the complex
constructions of the embryologist in terms
of concentration of chemicals. the cell's
internal program. production of proteins.
and so on. Thoughts about the implications
of this possibility would be unlikely to
impress the biologist. even if such systems
were argued to have some of the properties
of cells. and to have successes in unrelated
domains.
It is common to try to relieve uneasiness
about C-R approaches by invoking com-
puter models to show that we have robust.
hard-headed instances of the kind (see COM-
PUTATIONAL MODELS): psychology then
studies software problems. That is a dubious
move. Artefacts pose all kinds of questions
154
that do not arise in the case of natural
objects. Whether some object is a key or a
table or a computer depends on designer's
intent. standard use. mode of interpretation.
and so on. The same considerations arise
when we ask whether the device is mal-
functioning. folloWing a rule. etc. There is
no natural kind or normal case. The hard-
ware-software distinction is a matter of
interpretation. not simply of physical struc-
ture. though with further assumptions
about intention. design. and use we could
sharpen it. Such questions do not arise in
the study of organic molecules. nematodes.
the language faculty. or other natural
objects. viewed (to the extent we can
achieve this standpoint) as what they are.
not in a highly intricate and shifting space
of human interests and concerns. The belief
that there was a problem to resolve. beyond
the normal ones. reflects an unwarranted
departure from naturalism; the solution
offered carries us from a manageable frying
pan to a fire that is out of control.
We naturally want to solve the 'unifica-
tion problem'. that is. to relate studies of the
brain undertaken at various levels. much as
nineteenth-century science looked forward
to the integration of chemistry and physics
that was finally achieved in the new
quantum theory. Sometimes unification will
be reductive. as when much of biology was
incorporated within known biochemistry;
sometimes it may require radical modifica-
tion of the more 'fundamental' discipline. as
when physics was 'expanded'. enabling it to
account for properties that had been dis-
covered and explained. at another level. by
chemists (see REDUCTION). We cannot know
in advance what course unification will
take. if it succeeds at all.
Take a specific current example. C-R
studies of language give strong reasons to
believe that linguistic expressions fall into
many categories of 'well-formedness': non-
deviant (though perhaps gibberish). and in
violation of various conditions on rule
systems that have been discovered. Recent
studies of electrical activity of the brain
(event-related potentials. ERPs) have suc-
ceeded in finding distinctive responses to
several of these categories of expressions.
These studies relate two levels of inquiry:
electrical activity of the brain and C-R
systems. In this case, the C-R theories have
much stronger empirical support, and are
far superior in explanatory power. The
current significance of the ERP studies lies
primarily in the correlations with the much
richer and better-grounded C-R theories.
Within the latter, the various categories
have a place, and, accordingly, a wide
range of empirical support. some of it indir-
ect, by way of general principles confirmed
elsewhere; in isolation from C-R theories,
the ERP observations are curiosities, lacking
a theoretical matrix. The situation could
change, in which case the C-R theories
would be confirmed, modified, or abandoned
on the basis of studies of electrical activity.
In naturalistic inquiry, the chips fall where
they may, in terms of explanatory success.
The aspects of mind that have particu-
larly concerned me are those involving lan-
guage: its nature, development, and use. By
'language' I mean 'human language'. There
is no more a study of 'language' for organ-
isms generally than there is a study of
'locomotion', ranging from amoeba to eagle
to science-fiction spaceship; or 'commun-
ication', ranging from cellular interaction to
Shakespeare's sonnets to 'intelligent' extra-
terrestials. The human brain appears to
have a subsystem dedicated to language.
Most of our current understanding of this
'language faculty' derives from inquiry at
the level of C-R systems, based on observa-
tion and study of perception, interpretation,
and action. We may call this part of the
study of mind, using the term only to refer
to a domain of inquiry, with no metaphys-
ical connotations, just as the study of chem-
ical interactions in the nineteenth century,
which could not be grounded in known
'physical mechanisms', entailed no meta-
physical distinction between physics and
chemistry.
Other components of the mind/brain,
about which little is known, provide what
we might call 'commonsense under-
standing' of the world and our place in it
(,folk physics', 'POLK PSYCHOLOGY', etc.).
CHOMSKY, NOAM
Some components, perhaps different ones,
make it possible for humans to conduct nat-
uralistic inquiry, and sometimes to achieve
remarkable insight: we may call them 'the
science-forming faculty', to dignify ignor-
ance with a title.
We hardly expect the constructions of the
science-forming faculty to conform to
common-sense understanding. The natural
sciences have no place for such common-
sense notions as liquid, energy, or wave, or
the terms in which we describe a missile
rising to the heavens and falling to the
ground. No matter how much physics we
know, we cannot help seeing the moon illu-
sion or the setting of the sun, or sharing
Newton's unshakeable belief in the
'mechanical philosophy' that he had refuted
(with no little perplexity). It does not seem
reasonable to expect folk psychology to fare
differently from folk physics: say, to expect
that belief and desire, or language, or rule, as
the terms are commonly used and under-
stood, will survive the transition to rational
inquiry, if it can be made in these domains.
Nor should we expect what may be dis-
covered about language, perception, or
thought to conform to folk psychology. That
aside, in invoking folk psychology we have
to be careful to observe the practice of
serious ethnoscience, distinguishing paro-
chial and culture-bound notions from the
elements of 'folk theories' that are a
common human endowment, grounded in
our nature. Not an easy problem, and one
that is, I think, too lightly dismissed.
Any organism has certain ways of per-
ceiving and interpreting the world, a certain
'Umwelt' or 'cognitive space', determined in
large part by its specific nature and general
properties of biological systems. In Hume's
terms, part of human knowledge derives
'from the original hand of nature' as 'a
species of instinct'. Cognitive systems are
grounded in biological endowment, shaped
in limited ways by interactions with the
environment (experience). In this regard,
they are like other components of the body,
which also may be profitably studied at
various levels of abstraction from mechan-
isms.
155
CHOMSKY, NOAM
Given an organism with its special cog-
nitive systems, we can identify a category of
'problem situations' in which it might find
itself: an array of circumstances that it per-
ceives and interprets in a certain way by
virtue of its nature and prior history,
including (for humans) questions that are
posed and background belief and under-
standing that are brought to bear on them.
Sometimes the problem situation is con-
trived on the basis of theory-driven con-
siderations and addressed with a degree of
self-awareness - the activity that we call
sCience'. Some problem situations fall
within the animal's cognitive capacities,
others not. Let us call these 'problems' and
'mysteries', respectively. The concepts are
relative to an organism: what is a mystery
for a rat might be only a problem for us,
and conversely. For a rat, a 'prime number
maze' (turn right at every prime choice
point), or even far simpler ones, is a per-
manent mystery; the rat does not have the
cognitive resources to deal with it, though a
human might. A radial maze, in contrast,
poses a problem that a rat might solve quite
well. The distinctions need not be absolute,
but they can hardly fail to be real.
If humans are part of the natural world,
not angels, the same is true of them: there
are problems that we might hope to solve,
and mysteries that will be forever beyond
our cognitive reach. We might think of the
natural sciences as a kind of chance con-
vergence between aspects of the natural
world and properties of the human mind/
brain, which has allowed some rays of light
to penetrate the general obscurity; chance
convergence in that natural processes and
principles have not 'designed' us to deal
with problems we face and can sometimes
formulate. Since Charles Sanders Peirce,
there have been proposals about evolu-
tionary factors that allegedly guarantee that
we can find the truth about the world, and
there are much earlier beliefs about our
unique access to the nature of our own
minds and their products. But such specula-
tions seem unpersuasive.
We might consider for a moment a tradi-
tional approach to problems of mind, one
156
that may reflect 'commonsense under-
standing' to a considerable degree: metaphys-
ical DUALISM. Take the Cartesian version. It
offered a sketchy account of the physical
world, baSically in terms of a kind of
'contact mechanics'. Certain aspects of the
world, it was then argued, do not fall under
these principles. No artefact, for example,
could exhibit the normal properties of lan-
guage use: unbounded; not determined by
external stimuli or internal state; not
random but coherent and appropriate to
situations, though not caused by them;
evoking thoughts that the hearer might
have expressed the same way - a collection
of properties that we may call 'the creative
aspect of language use'. Accordingly, some
new principle must be invoked; for the Car-
tesians, a second substance whose essence
is thought. We then have the problem of
determining its nature, and we face the uni-
fication problem that arises throughout the
natural sciences: showing how mind and
body interact, in the traditional formulation.
The approach is baSically naturalistic, and
the reasoning is unaffected when we move
from the complex artefacts that fascinated
the seventeenth-century imagination to
those that excite many of the same ques-
tions and speculations today.
As is well known, this programme col-
lapsed within a generation, when Newton
demonstrated that the theory of the material
world was fatally inadequate, unable to
account for the most elementary properties
of motion. Newton had nothing to say about
the ghost in the machine. He exorcized the
machine, not the ghost; the Cartesian
theory of mind, such as it was, remained
unaffected. Newton found that bodies had
unexpected ghostly properties; their 'occult
quality' of action at a distance transcends
the common notion of body or material
object. Like many other leading scientists of
the day, Newton found these results disturb-
ing, agreeing with the Cartesians that the
idea of action at a distance through a
vacuum is 'so great an Absurdity that I
believe no Man who has in philosophical
matters a competent Faculty of thinking,
can ever fall into it', a reaction that is
understandable, possibly even rooted in folk
psychology. He concluded that we must
accept that universal gravity exists, even if
we cannot explain it in terms of the self-
evident 'mechanical philosophy'. As many
commentators have observed, this intellec-
tual move 'set forth a new view of science',
in which the goal is 'not to seek ultimate
explanations', but to find the best theoretical
account we can of the phenomena of experi-
ence and experiment (Cohen, 1987). Con-
formity to commonsense understanding is
henceforth put aside, as a criterion for
rational enquiry. (See Newton, 1693; Kuhn,
1959; Cohen, 1987.)
These moves also deprive us of any deter-
minate notion of body or matter. The world
is what it is, with whatever strange proper-
ties may be discovered, including those pre-
viously called 'mental'. Such notions as
'PHYSICALISM' or 'eliminative materialism'
(see ELIMIN A TIVISM) lose any clear sense.
Metaphysical dualism becomes unstateable,
as does metaphysical naturalism, under-
stood as the view that the study of mind
must be 'continuous' with the physical. The
mind/body distinction cannot be formulated
in anything like the Cartesian manner; or
any other, as far as I can see, except as a
terminological device to distinguish various
aspects of the natural world. The domain of
the 'physical' is what we come more or less
to understand, and hope to assimilate to the
core natural sciences in some way, perhaps
modifying them as inquiry proceeds. Ideas
that yield understanding and insight are
judged legitimate, part of the presumed
truth about the world; our criteria of ration-
ality and intelligibility may also change and
develop, as understanding grows. If humans
have 'ghostly properties' apart from those
common to all of matter, that's a fact about
the world, which we must try to compre-
hend in naturalistic terms, there being no
other.
The natural conclusion from Newton's
demolition of the commonsense theory of
body is that human thought and action are
properties of organized matter, like 'powers
of attraction and repulsion', electrical
charge, and so on. That conclusion was
CHOMSKY, NOAM
drawn most forcefully by De La Mettrie, a
generation later by Joseph Priestley, though
neither attempted to deal with the proper-
ties of mind identified by the Cartesians.
(For De La Mettrie see Chomsky, 1966, ch.
I, and sources cited. Particularly Rosenfield,
1941. See also Wellman, 1992. For Priestley
see Yolton, 1983.)
We now face a series of questions. What
exactly are these properties of things in the
world? How do they arise in the individual
and the species? How are they put to use in
action and interpretation? How can organ-
ized matter have these properties (the new
version of the unification problem)?
On the last problem, progress has been
slight. Matter and mind are not two cat-
egories of things, but they may pose entirely
different kinds of quandaries for human
intelligence, a fact that is interesting and
important, if true, but in no way surprising
to the naturalistic temper, which takes for
granted that humans will face problems and
mysteries, as determined by their special
nature. It may be that central domains of
the 'mental' are cognitively inaccessible to
us, perhaps the creative aspect of language
use, which lies as far beyond our under-
standing as it did to the Cartesians. If so, we
shall have to learn about humans, as best
we can, in other ways.
In some areas, there has been consider-
able progress. In the case of language, it has
been possible, in the past generation, to for-
mulate and study a number of traditional
questions that had eluded serious inquiry,
and more recently, to recast them sig-
nificantly, leading to much new under-
standing of at least some central features of
the mind and its functioning.
To say that someone has (speaks,
knows, ... ) a language is to say that the
person's language faculty has attained a
certain state. The language, so construed, is
a state of the language faculty. Adapting
some traditional terms, we may call a theory
of this state a 'grammar of the language',
and a theory of the initial state, a 'universal
grammar'. As in the case of other aspects of
growth and development, the state attained
and the course of development are internally
157
CHOMSKY. NOAM
directed in crucial respects; external condi-
tions are far too impoverished to have more
than a marginal impact on the highly
articulated and intricate structures that
arise as the language faculty develops to a
'steady state', apparently before puberty,
afterwards undergoing only peripheral
change in the normal case. To the extent
that we can determine the properties of the
state attained, we can ask questions about
language use: how are expressions inter-
preted and produced? We can ask how prop-
erties of the initial state and external events
interact to determine the course of language
growth (what is called 'language learning',
though 'language growth' might be less
misleading). In all of these areas, there has
been substantial progress - the normal use
of language for expression of thought aside.
a not inconsiderable gap.
The state attained consists of a cognitive
system and performance systems. The cog-
nitive system stores information that is
accessed by the performance systems, which
use it for articulation, interpretation,
expression of thought, asking questions,
referring, and so on. The cognitive system
accounts for our infinite knowledge, for
example, our knowledge about sound and
meaning and their relations over an
unbounded range. There is by now a large
mass of reliable data about these matters
from a great variety of typologically differ-
ent languages, and non-trivial theories
that go some distance in explaining the
evidence.
It is natural to restrict the term 'lan-
guage' to the state of the cognitive system of
the language faculty. We say, then, that
Smith has (knows, speaks, ... ) the lan-
guage L if the cognitive component of
Smith's language faculty is in state 1. So
regarded, we may think of the language as
a way to speak and understand, a tradi-
tional conception. It is commonly assumed
that the performance systems are fixed and
invariant. The reasons are, basically, ignor-
ance: that is the simplest assumption, and
we have no evidence that it is false. The
cognitive systems, however, do vary.
English is not Japanese; the cognitive
158
systems have achieved different states -
though not very different ones, it appears. A
Martian scientist, observing humans from
the standpoint that we adopt towards
organisms other than ourselves, might not
find the differences very impressive, com-
pared with the overwhelming commonality.
The attained state of the cognitive system
is a generative procedure that determines
an infinite class of linguistic expressions,
each a certain collection of phonetic, struc-
tural, and semantic properties. Particular
'signals' are manifestations of linguistic
expressions (spoken, written, signed. what-
ever); speech acts are manifestations of lin-
guistic expressions in a broader sense.
Note that the concept of language so
developed is internalist, individualist, and
INTENSIONAL, in the technical senses. It has
to do with the internal state of the brain of
a particular individual. Accordingly, Peter
and Mary might in principle have different
generative procedures that determine the
same class of linguistic expressions, though
the highly restrictive character of the initial
state may rule out this possibility. If the
case existed, the grammar of Peter's lan-
guage would be different than the grammar
of Mary's, though they could not be dis-
tinguished empirically solely on the basis of
information concerning properties of expres-
sions of Peter's and Mary's languages. Note
that a naturalistic approach would never
even consider the far more restrictive condi-
tions of the familiar 'radical translation'
paradigm, on which. a few words below.
Similarly, Peter and Mary might, in prin-
ciple, have visual systems that provide the
same mapping of stimulus to percept, but in
different ways, in which case the theoretical
accounts of their visual systems would
differ, though evidence for this would have
to derive from sources beyond study of sti-
mulus - percept pairing.
We may choose to ignore differences
between states that are similar enough for
the interests and purposes at hand, whether
studying language, vision, or drosophila;
from that we cannot conclude that there
are 'shared languages' (or shared visual
systems or drosophila-types). From the fact
that two people speak alike, we do not
deduce the existence of a common shared
language, any more than we deduce the
existence of a common shared shape from
the fact that they look alike.
A linguistic expression has two 'interface'
aspects, each of which provides instructions
to the performance systems. One interface
includes information for the signalling
apparatus: the vocal musculature and aud-
itory system, among others. The second
interface provides information relevant to
the faculties of the mind/brain involved in
thought and action. Not surprisingly, the
first interface is much better understood.
The second is harder to study, because the
related systems are much more obscure, but
there is, nevertheless, a great deal of relev-
ant evidence and many interesting results
and insights.
We may regard each interface as a sym-
bolic system consisting of 'values', 'phonetic
values' in one case, 'semantic values' in the
other. Note that these are technical notions,
to be divorced of connotations drawn from
other domains of inquiry. The phonetic
values are syntactic objects, elements of
'MENTAL REPRESENTATION' postulated in
the (strictly naturalistic) study of states and
properties of the language faculty of the
brain (see SYNTAX/SEMANTICS). They are
not vibrating air particles. Similarly, the
semantic values are syntactic objects, not
part of some extra-mental world. Internalist
semantics seems much like the study of
other parts of the biological world. We
might well term it a form of syntax, that is,
the study of elements of symbolic systems of
the mind/brain. The same terminology
remains appropriate if the theoretical ap-
paratus is elaborated to include mental
models, POSSIBLE WORLDS, discourse repre-
sentations, and other systems of postulated
entities that are still to be related in some
manner to things in the world, or taken to
be in the world - no simple matter, and
perhaps even a misconceived project.
It is by virtue of the way the cognitive
system is embedded in performance systems
that the formal properties of expressions are
interpreted as rhyme. entailment, and so
CHOMSKY. NOAM
on, not - say - as instructions for locomo-
tion. We are studying a real object, the lan-
guage faculty of the brain, which has
assumed a particular state that provides
instructions to performance systems that
play a role in articulation, interpretation,
expression of beliefs and desires, referring,
telling stories, and so on. For such reasons,
the topic is human language.
Pursuing the inquiry into language in
these terms, we can account for many
curious and complex properties of sound
and meaning, and relations among expres-
sions, in a wide variety of languages. Much
of this appears to derive from our inner
nature, determined by the initial state of our
language faculty, hence unlearned and uni-
versal for human languages.
The inquiry seeks to attain both
'descriptive adequacy' and 'explanatory
adequacy.' A proposed grammar of L is
descriptively adequate insofar as it is
correct, that is, is a true theory of 1. A pro-
posed version of universal grammar, call it
UG, achieves descriptive adequacy insofar as
it is a true theory of the initial state, of lan-
guage in general. Insofar as UG is descrip-
tively adequate, it provides an interesting
kind of explanation for the facts described
by the grammar of L. It provides the means
for deriving the properties of the linguistic
expressions of L from the properties of the
fixed initial state of the language faculty,
under the 'boundary conditions' set by
experience.
We may put the point differently by
thinking of UG as an 'acquisition device'.
The initial state of the language faculty
changes in early childhood, as a result of
internally directed maturation and external
inputs, until it assumes a (relatively) stable
state: the language 1. We may, then, think
of the initial state as a procedure (algor-
ithm, mapping), which takes as its input
an array of data and yields as output the
language L; the 'output' is of course in-
ternal, a state of the language faculty. With
this perspective, the problem under investi-
gation is sometimes called 'the logical
problem of language acquisition'. A richer
theory will be based on assumptions as to
159
CHOMSKY. NO AM
the right answer to the logical problem.
seeking to fill in the details as to just how
the processes take place. perhaps rejecting
or modifying the assumptions. This is
empirical inquiry; particular subparts claim
no epistemological priority.
Many of the questions posed have a tradi-
tional flavour, though modern behavioural
science and structuralist approaches largely
avoided or denigrated them. It was evident
to Wilhelm von Humboldt in the early nine-
teenth century that language crucially
involves 'infinite use of finite means'. Otto
Jespersen recognized that the central
concern of the linguist must be free cre-
ation, the ability of each person to construct
and understand 'free expressions', typically
new, each a sound with a meaning. More
deeply, the task is to discover how the
structures that underlie this ability 'come
into existence in the mind of a speaker'
who, 'without any grammatical instruction,
from innumerable sentences heard and
understood ... will abstract some notion of
their structure which is definite enough to
guide him in framing sentences of his own.'
Though important and basically correct,
these ideas had little impact, unlike the far
narrower and more restricted Saussurean
conceptions of the same period, which were
enormously influential in many areas. The
ideas could not receive clear expression
until advances in the formal sciences pro-
vided the concept of generative (recursive)
procedure. The modern study of these ques-
tions might be regarded as a confluence of
traditional ideas that had been dismissed as
senseless or unworkable, with new formal
insights that made it possible to pursue
them seriously. (See von Humboldt, 1836;
see Chomsky, 1966 for discussion in the
context relevant here. See also Jespersen,
1924.)
Traditional grammars do not describe the
facts of language; rather, they provide hints
to the reader who already has, somehow,
the requisite 'notion of structure' and
general conceptual resources, and can use
the hints to determine the expressions of the
language and what they mean. The same is
true of dictionaries. Even the most elaborate
160
do not go beyond hints as to what words
mean, with examples to stimulate the con-
ceptual resources of the mind. Traditional
grammars and dictionaries, in short, pre-
suppose 'the intelligence of the reader'; they
tacitly assume that the basic resources are
already in place.
For grammars, the enormity of the gap
was not appreciated until serious attempts
were made to formulate the generative pro-
cedures that determine sound and meaning.
It quickly became evident that even short
and simple expressions in well-studied lan-
guages have intricate properties that had
passed completely unnoticed. The ensuing
work sought to attain descriptive adequacy
by filling these huge gaps - better, chasms -
and to undertake the far more interesting
project of explanatory adequacy for UG.
There is a tension between these two
goals. To achieve descriptive adequacy, it is
necessary to construct detailed and intricate
rule systems to account for the phenomena.
The rules for question-formation in English,
for example, involve intricacies undreamt of
in traditional accounts, and appear to be
specific to the interrogative construction in
English. The rules constructed to describe
such newly discovered facts were language-
particular and construction-particular,
much like the hints called 'rules' in tradi-
tional grammar. But the empirical condi-
tions of explanatory adequacy (language
acquisition) require that the structure of
language be largely predetermined. There-
fore the rules must be largely universal and
general in character.
To resolve the tension it is necessary to
show that the apparent complexity is epi-
phenomenal (see EPIPHENOMENALISM), the
result of interaction of fixed and probably
quite abstract principles, which can vary
slightly in the ways they apply, yielding
apparent complexity and large phenomenal
differences among languages that are basic-
ally cast to the same mould. The variations
will be determined by experience; the basic
principles are invariant, derived 'from the
original hand of nature'. Efforts to pursue
this path finally converged, about 1980, in
a conception of language that departed
radically from the 2 SOO-year tradition of
study of language. This 'principles-and-
parameters' model proposed principles of a
very general nature, along with certain
options of variation (parameters), perhaps
two-valued. The principles are language-
independent and also construction-
independent; in fact, it appears that tradi-
tional grammatical constructions (inter-
rogative, passive, nominal phrase, etc.) are
taxonomic artefacts, rather like 'terrestrial
mammal' or 'household pet'. These cat-
egories, with their special and often intri-
cate properties, result from the interaction
of fixed general principles, with parameters
set one or another way. Language acquisi-
tion is the process of determining the values
of parameters. There are no 'rules of
grammar' in the traditional sense: rather,
language-invariant principles and values for
parameters of variation, all indifferent to
traditional grammatical constructions.
We may think of a language, then, more
or less as a network that is not completely
wired up at birth. It is associated with a
switch-box, with a finite number of switches
that can be set on or off. When they are set,
the network functions; different settings
may yield quite different phenomenal
outputs. To the extent that the picture is
spelled out, we can 'deduce' Hungarian or
Yoruba by setting the switches one or
another way. Elements of the picture seem
reasonably clear, though a great deal is
unknown, and clarification of principles
regularly opens the doors to the discovery of
new empirical phenomena, posing new
challenges. Though much less is under-
stood, something similar must also be true
of the lexicon, with the links it provides to
the space of humanly accessible concepts
and signals.
Work of the past few years suggests
further and, if successful, possibly quite far-
reaching revisions of the general picture of
language. The language L consists of a
lexicon and a computational procedure that
uses lexical materials to construct linguistic
expressions with their sound and meaning.
It may be that the computational procedure
is fixed, identical for all languages; variation
CHOMSKY. NOAM
is restricted to the lexicon. Furthermore,
parameters seem to be of two types: either
they hold of all lexical items, or of such
formal elements as inflections; classes of
substantive elements do not seem to be
subject to parametric variation of this sort.
Languages also of course differ in the
sound-meaning linkage in the lexicon, and
in the parts of the computation that are
'close to' phonetic form, hence detectable in
the data. Differences of this kind are rela-
tively superficial, sufficiently so, one would
hope to show, to account for the observed
process of precise acquisition on the basis of
little usable evidence. The Martian observer
might conclude that, superficial differences
aside, there really is only one human lan-
guage, basically 'built in', much as we
assume to be the case, even without evi-
dence or understanding, for other systems
of the body.
It may be, furthermore, that many of the
principles that seem fairly well established
are themselves epiphenomenal, their con-
sequences reducing to more general and
abstract properties of the C-R system, prop-
erties that have a kind of 'least effort'
flavour. This 'minimalist' program also
seeks to reduce the descriptive technology to
the level of virtual conceptual necessity,
sharply restricting the devices available for
description, which means that the complex
phenomena of widely varied languages
must be explained in terms of abstract prin-
ciples of economy of derivation and rep-
resentation. A linguistic expression of L,
then, would be a formal object that satisfies
the universal interface conditions in the
optimal way, given the parameter values for
1. Such a programme faces an extremely
heavy empirical burden. If these directions
prove correct, they should yield much
deeper insight into the computational pro-
cesses that underlie our linguistic abilities.
There are other empirical conditions to be
met. It is well known that language is
'badly adapted to use'; of the class of 'free
expressions' determined by the 'notion of
structure' in our minds, only scattered frag-
ments are readily usable. Even short and
simple expressions often cannot be handled
161
CHOMSKY. NOAM
readily by our performance systems. Fur-
thermore, usability cross-cuts deviance;
some deviant expressions are perfectly
comprehensible, while non-deviant ones
often overload processing capacity. The
un usability of language does not interfere
with communication: speaker and hearer
have similar languages and (perhaps iden-
tical) performance systems, so what one can
produce, the other can interpret, over a
large range. But a really far-reaching theory
of language will want to provide an account
of the 'usability' of various parts of the lan-
guage and the categories of deviance, while
explaining a broad range of properties of
sound and meaning.
At this point we reach substantive ques-
tions that would carry us too far afield. Let
me finish with some words about the limits
of a naturalistic programme of the kind
sketched, supposing it to be essentially 'on
course'.
First, would it serve as a model for the
study of other aspects of mind? Probably
not, possible suggestiveness aside. As far as
we know, there are no 'mechanisms of
general intelligence', procedures of any gen-
erality that apply to various cognitive
domains. Inquiry into particular skills, abil-
ities, aspects of knowledge and belief, and so
on, has regularly found that the sub-
components of the mind function quite dif-
ferently. That is hardly a great surprise. It is
pretty much what we find in the study of
other complex systems: the visual cortex,
the kidney, the circulatory system, and
others. Each of these 'organs of the body'
has its properties. They fall together. pre-
sumably, at the level of cellular biology, but
no 'organ theory' deals with the properties
of organs in general. The various faculties
and cognitive systems of the mind can be
thought of as organs of the body (in this
case, of the brain) in much the same sense.
There is little reason to suppose that an
'organ theory' exists for such mental
organs, and to our (limited) knowledge, it
does not.
Over a broad spectrum in psychology,
philosophy, speculative neurophysiology,
artificial intelligence, and cognitive science,
162
the opposite has long been assumed. But the
belief in uniform mechanisms that apply in
different cognitive domains seems ground-
less. If that turns out to be the case, there
will be no serious field of 'cognitive science',
dealing with the general properties of cog-
nitive systems; and the study of language,
however successful, will neither provide a
useful model for other parts of the study of
mind, nor draw from them significantly.
Note that, if true, this implies nothing
about how language interacts with other
mental faculties and systems; surely the
interactions are dense and close, but that is
another matter entirely.
In the philosophical literature, cognitive
science is often construed as the study of
how behaviour is caused by a complex of
beliefs, desires, and so on (see REASONS AND
CAUSES). The approach to the study of mind
just outlined has nothing to say about these
topics. I'm not sure that is a defect, since
cognitive science in this sense does not
really exist, and may not even be a reason-
able goal. No principles are known, or even
imagined, that go beyond low-level descript-
ive observations of limited credibility and
scope. There are, furthermore, no strong
grounds for believing that behaviour is
caused, at least in any sense of 'cause' that
we understand. I know of no good reason to
suppose that the Cartesians were wrong in
their basic descriptive observations about
behaviour, in particular, about the creative
aspect of language use; that is, behaviour
that is appropriate and coherent but
uncaused, the normal case, they argued. It
could well be that we are approaching
mysteries-for-humans, at this point.
It seems a reasonable guess that there is a
crucial divide between causation of beha-
viour by situations and internal states, on
the one hand, and appropriateness of beha-
vior to situations, on the other; to adopt
Cartesian rhetoric, between machines
which are compelled to act in specific ways
(irrelevant random elements apart), and
humans who are only incited and inclined to
do so. In the post-Newtonian era, the divide
is not metaphysical, but mental-nonmental
(in the naturalistic sense) - a fact about the
special properties of one component of the
world, the human mind, and its cognitive
capacities. The divide might turn out to be
unbridgeable for a human intelligence. We
should not, I think, simply dismiss Des-
cartes' speculation that we may not 'have
intelligence enough' to comprehend the
creative aspect of language use and other
kinds of free choice and action, though 'we
are so conscious of the liberty and indiffer-
ence which exists in us that there is nothing
that we comprehend more clearly and per-
fectly', and 'it would be absurd to doubt
that of which we inwardly experience and
perceive as existing within ourselves' just
because it lies beyond our comprehension.
The 'internalist' approach to language
also fails to provide an account of such
notions as 'language of a community' or
'community norms' in the sense pre-
supposed by virtually all work in philosophy
of language and philosophical semantics. It
gives us no notion of a common public lan-
guage, which perhaps even exists
'independently of any particular speakers',
who have a 'partial, and partially erro-
neous, grasp of the language' (Dummett,
1986). It has no place for a Platonistic
notion of language, outside of the mind!
brain and common to various speakers, to
which each speaker stands in some cogni-
tive relation, for which a place is often
sought within theories of knowledge of a
dubious character.
Again, I do not think these are defects of
a naturalistic approach; rather of the
notions that it does not expect to capture.
We have beliefs about health, national
rights, the plight of the average man, and
so on. A person may know the construction
business, or the secret of happiness. But we
are not tempted to suppose that there are
corresponding Platonic objects to which the
person stands in some cognitive relation.
The same is true of beliefs about language
or knowledge of language (a locution of
English that should not be taken too ser-
iously, though having a language surely
entails possession of rich and varied know-
ledge).
To ask whether Peter and Mary speak the
CHOMSKY. NOAM
same language is like asking whether
Boston is near New York or whether John is
almost home, except that the dimensionality
of context-dependence provided by interest
and circumstance is far more diverse and
complex. In ordinary human life, we find all
sorts of shifting communities and expecta-
tions, varying widely with individuals and
groups, and no 'right answer' as to how
they should be selected. People also enter
into various and shifting authority and
deference relations. The world is not divided
into areas within which people are 'near
one another', nor do such areas exist as
idealizations. And human society is not
divided into communities with languages
and their norms. The problem is not one of
vagueness; rather, of hopeless underspeci-
fication. As for Swedish-versus-Danish,
norms and conventions, misuse of lan-
guage, and other similar notions, they are
fine for ordinary usage (as is 'near New
York'), but they should not, I think, be
expected to enter into attempts to reach
theoretical understanding in anything like
the ways that have been widely assumed. If
so, a good deal of work has to be seriously
reconsidered, it appears.
For similar reasons, such notions as
'competence in English' - which could hold,
say, of Martians - also have no place in this
approach. The question whether a Martian
knows English, or whether bees have a lan-
guage, is not meaningful within a natur-
alistic framework. The same is true of the
question whether a robot is walking or
reaching for a block on the table; or
whether a computer (or its program) is
doing long division, or playing chess, or
understanding Chinese, or translating from
English to German. These are questions of
decision, not fact: how seriously do we
choose to take a certain metaphor? Such
questions are not settled empirically, any
more than we can determine empirically
whether an airplane can fly (say, if it can
fool someone into thinking it is an eagle), or
whether a submarine really sets sail but
doesn't swim, or whether a camera sees
what it films, or whether a high jumper is
really flying, like a bird. In English, an
163
CHOMSKY. NOAM
airplane and an eagle fly, a high jumper
doesn't. In Japanese, all do. In Hebrew, only
an eagle does. No issues of fact arise.
In his classic 1950 paper that set off
much of the inquiry and debate about
whether machines can think, TURING
recognized that the question, and others like
it, may be 'too meaningless to deserve dis-
cussion', though in half a century, he
speculated, conditions might have changed
enough for us to choose to use such locu-
tions, just as some languages use the meta-
phor of flying for airplanes and high
jumpers. Turing seems to have agreed with
WITTGENSTEIN as to the pointlessness of the
debates that have raged in the years that
followed, until today. In any event, the
issues debated cannot be sensibly posed in
the naturalistic framework sketched; again
not a defect, but a merit, in my opinion. (See
Turing, 1950.)
Another limitation of a naturalistic
approach is that it has nothing to say about
the problems raised within W. V. Quine's
influential 'radical translation' paradigm,
which underlies a good deal of the work of
the past generation in philosophy of lan-
guage and mind (Quine, 1960). The para-
digm stipulates a certain epistemic situation
for the person trying to communicate, the
child acquiring a language, and the linguist
studying it. Far-reaching conclusions are
then reached about humans, their actions,
thought, capacities, and so on. The force of
the conclusions depends on the validity of
the paradigm, that is, the initial stipula-
tions. A number of questions arise. First,
there are crucial differences among the
cases grouped together. The epistemic situa-
tion of the infant equipped with the re-
sources of the initial state of the faculties
of mind cannot sensibly be compared with
that of a scientist seeking to determine what
these resources are; neither can be com-
pared with the situation of a person in a
communication situation whose mind has
'grown' the language L, or that of the lin-
guist seeking to discover L and understand
how it is used. More seriously, no stipula-
tions or conditions comparable to those of
the paradigm would be tolerated for a
164
moment in the study of the growth of an
embryo to a chicken, or the development of
the visual system, or the onset of puberty;
or, in fact, in the study of any aspect of the
natural world, apart from humans 'above
the neck', metaphorically speaking. If so,
then the failure of the naturalistic approach
sketched earlier to deal with these questions
is, again, exactly what we should want.
It is worth noting how radical is this
departure from normal science. It seems fair
to regard it as a new form of dualism, but
one without the virtues of the earlier
version. Cartesian dualism was a reasonable
theoretical construction, which collapsed
because of faulty assumptions about the
material world. The new 'methodological/
epistemological dualism', in contrast, insists
that the study of human thought and
action be pursued in a manner that is com-
pletely unacceptable in the sciences. No
justification has been given for these arbi-
trary demands, to my knowledge, though
their historical antecedents are fairly clear;
indeed the question is scarcely addressed.
Not surprisingly, on adopting the frame-
work we reach forms of sceptical doubt that
were dismissed for good reason centuries
ago.
It is sometimes argued that in the theory
of meaning (a term used broadly enough to
include much of the study of language) it is
necessary to distinguish 'psychological
hypotheses' from 'philosophical explana-
tions'. Being limited to the former (or more
broadly, to 'scientific hypotheses'), the nat-
uralistic approach is therefore inadequate, it
is argued. Suppose that naturalistic inquiry
were to discover a C-R theory that spells out
precisely what happens when sound waves
hit the ear and are interpreted. Suppose this
analysis is related to what cells of the body
are doing, thus solving the unification
problem. So far, we have a 'psychological
hypothesis', but not a 'philosophical explan-
ation', because the account does not tell us
'the form in which [the body of knowledge]
is delivered' (Dummett, 1986). For the
sciences, the account tells everything that
can be asked about the form in which the
body of knowledge is delivered; but for the
theory of meaning (and, presumably, lan-
guage and thought generally), some addi-
tional kind of explanation is required.
Suppose further that there is a Martian
creature, exactly like us, except that it can
become aware of the internal computations
and can truly answer questions about them.
The Martian understands the expressions
(1) and (2) exactly as we do, with the same
options for referential dependence of the
pronoun he on the antecedent the young
man - an option in (2) but not in (1):
(1) he thinks the young man is a
genius-
(2) his mother thinks the young man is
a genius.
By assumption, the Martian determines
these options exactly as we do, but when
asked, the Martian can state (correctly) the
rules that guide its decisions, those of the
C-R system postulated for humans. For the
Martian, we would now understand the
form in which the knowledge is delivered,
and could properly attribute knowledge of
the principles that determine referential
dependence to the Martian. Actually, we
would know nothing more of any relevance
about the Martian than about the human;
they differ only in that the Martian has
awareness where the human has none. But
we would have crossed the bridge to 'philo-
sophical explanation'. As Quine, John
Searle, and others put it, we would be
allowed to say that the Martian is following
rules or is guided by them, whereas the
human, who is doing exactly the same
thing, cannot be described in these terms.
To avoid immediate counterintuitive con-
sequences, Searle insists further on a prop-
erty of 'access in principle' that remains
obscure. (See Quine, 1972, and commentary
by myself and Ned Block in Searle, 1990.)
Similarly, if experiment, analysis, and
available theory convince us that Mary and
Peter do long division exactly the same
way, using the algorithm A, then a natur-
alistic approach will attribute A to both. If,
furthermore, Mary is conscious of using A,
but Peter not, the naturalistic approach will
CHOMSKY, NOAM
state these facts. That is where the story
will end - except, of course, to try to deter-
mine what is involved in conscious access
and why Mary has it but not Peter: interest-
ing questions, but irrelevant here. But on
the non-naturalistic assumptions invoked in
much of the literature, we must refrain
from attributing A to Peter, unless (in some
versions) 'access in principle' holds for
Peter, in which case we may attribute A to
him as well.
In the study of other parts of the natural
world, we agree to be satisfied with post-
Newtonian 'best theory' arguments; there is
no privileged category of evidence that pro-
vides criteria for theoretical constructions.
In the study of humans above the neck,
however, naturalistic theory does not
suffice: we must seek 'philosophical ex-
planations', require that theoretical posits
be specified in terms of categories of evi-
dence selected by the philosopher (as in the
radical translation paradigm), and rely cru-
cially upon unformulated notions such as
'access in principle' that have no place in
naturalistic inquiry.
However one evaluates these ideas, they
clearly involve demands beyond naturalism,
hence a form of methodological/epistemo-
logical dualism. In the absence of further
justification, it seems to me fair to conclude,
here too, that inability to provide 'philoso-
phical explanations' or a concept of
'rule-following' that relies on access to con-
sciousness (perhaps 'in principle') is a merit
of a naturalistic approach, not a defect.
A standard paradigm in the study of lan-
guage, given its classic form by Frege, holds
that there is a 'store of thoughts' that is a
common human possession and a common
public language in which these thoughts are
expressed (Frege, 1956). Furthermore, this
language is based on a fundamental relation
between words and things - reference or
denotation - along with some mode of fixing
reference (sense, meaning). The notion of a
common public language has never been
explained, and seems untenable. It is also
far from clear why one should assume the
existence of a common store of thoughts;
the very existence of thoughts had been
165
CHOMSKY. NOAM
plausibly questioned, as a misreading of
surface grammar, a century earlier.
The third component of the paradigm, a
relation of reference, could in principle
be accommodated within a naturalistic
approach: a relation holding between
certain terms of Mary's language and
things in the world, or things as Mary con-
ceives them to be, or natural kinds (under-
stood as kinds of nature). But the existence
of such relations is a matter of fact, not
doctrine; and in fact it does not seem that
terms of human language enter into such
relations, as noted by P. F. Strawson years
ago when he warned of 'the myth of the
logically proper name' (Strawson, 1952), to
which we may add related myths concern-
ing indexicals and pronouns. True, people
use expressions to refer to things and talk
about them, but when we study these
actions we typically find that the words and
expressions provide rather intricate perspec-
tives from which to view the world, per-
spectives that vary in determinate ways as
interests and goals shift, but do not yield
any notion of reference in the technical
sense. Similarly, ordinary usage seems to
have no such terms as 'reference', or
'meaning' understood as mode of fixing
reference, in anything like the Fregean
sense - which is why Frege had to invent
technical concepts.
The fact that the philosophical theories
depart from ordinary usage is not a criti-
cism; so does rational inquiry generally, for
good and understandable reasons. In this
case, however, the theoretical systems that
have been devised do not seem to apply to
natural language. Study of reference and
meaning in natural language, it seems,
should follow a different course: asking how
the representations at the interface enter
into performance systems and relate to
other systems of the mind/brain, much as
we proceed in the study of articulation and
perception. Similar questions arise more
generally with regard to questions of
INTENTION ALITY, and attempts to under-
stand people and what they do, their
thought and action.
Even if correct, such conclusions would
166
not imply that the Fregean project is mis-
guided insofar as it is concerned with a
'logically perfect language' designed for the
formal sciences and perhaps the natural
sciences as well. It could be argued that
naturalistic inquiry seeks to construct a
symbolic system with Fregean properties: a
store of common thoughts; a common
public symbolic system in which these are
expressed, incorporating calculus, or formal
arithmetic, or something else that is remote
from natural language in its basic proper-
ties; and that the goal is for the terms of
this system to denote in the technical sense,
relating to what we take to be things in the
world and the kinds of nature. If so, that is
a fact about the science-forming faculty.
But that would tell us nothing about
language, or commonsense understanding
(,folk science') generally. Theories of lan-
guage and thought will have this natur-
alistic goal as well, but, obviously, we
cannot impute properties of the theory to
its subject matter.
The technical concepts 'broad content'
(see EXTERNALISM/INTERNALISM) and 'per-
ceptual content', which have played a
leading role in recent thinking, also have no
obvious place here. That too could turn out
to be a merit. Insofar as these concepts rely
on notions of community and norm. they
suffer from serious problems; irremediable
ones, I believe. Insofar as they are based on
a relation of reference that holds between
expressions of natural language and things
(natural kinds, etc.), or perhaps between the
terms of a 'language of thought' and things,
that seems problematic. It is also not clear
that these technical notions have counter-
parts in folk psychology, but if they did, it
would not be of much significance for
theoretical inquiry into language, meaning,
and thought. No doubt it is by virtue of
facts of the world that statements and
beliefs are true. desires fulfilled, and so on.
But we seem far short of any framework of
theoretical understanding in which such
facts can be accommodated and interpreted.
This only skims the surface of very large
topics. I do not want to leave the impression
that I think a naturalistic approach along
the lines sketched deals with classic ques-
tions of language and thought. It definitely
does not. For one thing, it omits entirely the
creative aspect of language use, the best
evidence for the existence of other minds,
for the Cartesians. It does not reach as far
as the study of groups and institutions in
which people take part, though I see no
issues here, apart from choice of topic of
inquiry. The relevant question is whether
there is a more promising path towards the-
oretical understanding of the range of ques-
tions that do seem amenable to naturalistic
inquiry in the sense of this discussion. Not
to my knowledge, at least.
See also COGNITIVE PSYCHOLOGY; CONTENT;
DENNETT; DRETSKE; FODOR; LANGUAGE OF
THOUGHT; LEWIS; QUINE; RATIONALITY;
SEARLE.
BIBLIOGRAPHY
Chomsky. N. 1966. Cartesian Linguistics. New
York: Harper & Row.
-- 1965. Aspects of the Theory of Syntax.
Cambridge. MA: MIT Press.
-- 1968. Language and Mind. New York:
Harcourt Brace Jovanovich. Enlarged
edition 1 972.
-- 1975. Reflections on Language. New York:
Pantheon.
-- 1980. Rules and Representations. New
York: Columbia University Press.
-- 1986. Knowledge of Language. New York:
Praeger.
-- 1988. Language and Problems of Knowl-
edge. Cambridge, MA: MIT Press.
-- 1992. Language and interpretation: Phi-
losophical reflections and empirical inquiry.
In Inference. Explanation and Other Philoso-
phical Frustrations. ed. J. Earman. Berkeley:
University of California Press.
-- 1993. Explaining language use. Philoso-
phical Topics.
Cohen, B.I. 1987. The Newtonian scientific
revolution and its intellectual significance:
A tercentenary celebration of Isaac New-
ton's Principia. Bulletin, The American
Academy of Arts and Sciences, XLI: 3.
De La Mettrie. J.O. 1966. L'Homme Machine. In
my Cartesian Linguistics. ch. 1.
Dummett. M. 1986. Comments on Davidson
COGNITIVE PSYCHOLOGY
and Hacking. In Truth and Interpretation. ed.
E. Lepore. Oxford: Basil Blackwell.
-- 1991. The Logical Basis of Metaphysics.
Cambridge, MA: Harvard University Press.
Frege. G. 1956. On sense and reference. In
Philosophical Writings of Gottlob Frege, ed.
and trans. M. Black and P. T. Geach.
Oxford: Basil Blackwell.
Jespersen. O. 1924. Philosophy of Grammar.
London: George Allen & Unwin. 1924.
Kuhn, T. 1959. The Copernican Revolution.
New York: Vintage.
Newton. I. 1693. Letter cited by J. Yolton in
Thinking Matter. Minnesota: University of
Minnesota Press.
Quine. W.V.O. 1960. Word and Object. Cam-
bridge, MA: MIT Press.
-- 1972. Methodological reflections on
current linguistic theory. In Semantics of
Natural Language. ed. D. Davidson and G.
Harman. New York: Humanities Press.
Rosenfield. L.c. 1941. From Beast-Machine to
Man Machine. Oxford University Press.
Searle. J. 1990. Consciousness. explanatory
inversion. and cognitive science. Behavioral
and Brain Sciences. 13(4).
Strawson, P.F. 1952. Introduction to Logical
Theory. London: Methuen.
Turing. A.M. 1950. Computing machinery
and intelligence. Mind. LIX, 236.
Von Humboldt. W. 1836. Uber die Verschie-
denheit des Menschlichen Sprachbaues.
Wellman, K. 1992. De la Mettrie: Medicine.
Philosophy. and Enlightenment. Duke.
Yolton, J. 1983. Thinking Matter. Minnesota:
University of Minnesota Press.
NOAM CHOMSKY
cogito see HISTORY.
cognitive psychology The Oxford English
Dictionary gives the everyday meaning of
cognition as 'the action or faculty of
knowing'. The philosophical meaning is the
same, but with the qualification that it is to
be 'taken in its widest sense, including
sensation. perception, conception, etc., as
distinguished from feeling and volition'.
Given the historical link between psychol-
ogy and philosophy. it is not surprising that
'cognitive' in 'cognitive psychology' has
something like this broader sense, rather
167
than the everyday one. Nevertheless, the
semantics of 'cognitive psychology', like
that of many adjective-noun combinations,
is not entirely transparent. Cognitive psy-
chology is a branch of psychology, and its
subject matter approximates to the psycho-
logical study of cognition. Nevertheless, for
reasons that are largely historical, its scope
is not exactly what one would predict.
THE SCOPE OF COGNITIVE PSYCHOLOGY
The table of contents of a student textbook
on cognitive psychology provides a good
indication of the topics studied by cognitive
psychologists. There are many such texts,
but their authors are agreed on the broad
divisions of the subject. For example, all the
following were covered in four popular texts:
sensory memory, pattern recognition, atten-
tion, MEMORY proper, learning, MENTAL
REPRESENTATIONS (including schemata,
IMAGER Y, cognitive maps, CONCEPTS), or-
ganization of knowledge, language, think-
ing, reasoning, problem solving, decision
making. More recent books are also likely to
discuss the relation between cognition and
EMOTION, reflecting a new-found interest in
cognitive theories of emotion on the part of
cognitive psychologists.
Cognition, Action and Emotion
It should come as no surprise to philoso-
phers that the traditional three-fold distinc-
tion between cognition, conation, and
affection, alluded to in the OED definition, is
not important in contemporary psychology.
Cognitive psychology is not defined by its
contrast with conative and affective psy-
chology. Indeed, psychologists who study
ACTION and those who study emotion
recognize that both have strong cognitive
components. The study of action is frag-
mented. Skilled performance is studied in a
different framework from decision making,
for example. Decision making itself stands in
an uncomfortable relation to the rest of cog-
nitive psychology. In the 1950s and 1960s
it was studied primarily by economists and
mathematical psychologists. Since that
time, its cognitive aspects have been
168
increasingly recognized, though its proper
integration into cognitive psychology
remains a distant prospect. Ironically, work
on other aspects of reasoning has begun to
suggest a strong link to practical reasoning.
For example, 85 per cent of university stu-
dents routinely fail to solve the abstract
version of the notorious Wason selection
task (figure 1). Problems with parallel struc-
ture, but with concrete content, can be
much easier. Manktelow and Over (1991)
and others have argued that the facilitatory
effect of such content is explained by the
fact that it elicits practical, rather than
theoretical, reasoning skills.
The study of emotions in psychology is
less fragmented, though nevertheless riddled
with controversy. One major point of dis-
agreement is whether there are any basic
emotions. One view is that six basic emo-
tions can be identified from facial expres-
sions: happiness, surprise, sadness, anger,
disgust, fear (e.g. Ekman, 1982, though
Ekman's own views have changed since
1982). However, even if there are a small
number of biologically based emotions, the
other emotions are more complex, and have
interpretative components that are cognitive
in nature.
HISTORICAL CONSIDERATIONS
The many idiosyncrasies of contemporary
cognitive psychology are largely the
product of a complex set of historical pro-
cesses. Its boundaries have been determined
partly by the development of psychology
since it became an independent empirical
science in the 1870s. For example, low-
level perceptual processes, corresponding
roughly to 'sensation' in the OED definition,
have from the early days of psychology
been studied separately from cognition.
Given what has been learned about these
processes, the detailed links that have been
made with the underlying physiology, and
the tendency towards specialization that
occurs as science advances, it is both un-
likely and undesirable that the study of low-
level perception will be reintegrated with
cognitive psychology.
Figure 1. The Wason Selection Task.
(a) Abstract Version. Each of the four cards below has a letter on one side and a number on the
other side. Which cards need to be turned over to see if the following statement is true of the
cards? If a card has a vowel on one side it has an even number on the other side.
(b) Practical Reasoning Version (Manktelow and Over, 1991). Imagine that you are a boy who has
been given a rule by your mother about keeping the house tidy while she is at work. The rule is: if
you tidy your room you may go out to play. You have not been seeing eye to eye with your
mother lately because you think that sometimes she makes rules like this but doesn't keep to
them. So you decide to record on cards what she does on the first few days after she introduces
the rule. You record on one side of each card whether you tidied up your room or not, and on the
other side whether your mother let you go out to play or not. Which cards would show whether
the mother had broken the rule?
I tidied
my room
I did not
tidy my room
Cognitive Psychology and Social and
Developmental Psychology
A more problematic set of divisions within
psychology, from the point of view of the
integrity of cognitive psychology, recog-
nizes DEVELOPMENTAL PSYCHOLOGY and
social psychology as distinct branches of
the discipline, even though there are
cognitive components to social behaviour,
and even though one of the central con-
cerns of developmental psychology is cog-
nitive development. For example, common-
sense suggests that language use is pri-
marily a social behaviour. However, cogni-
tive psychologists believe that many, if not
all, aspects of language processing are best
studied away from the complexities of
social interaction, and that they are best
explained in terms of structures and pro-
cesses in the mind of an individual
language user.
The gulf between social psychology and
cognitive psychology is wider than that be-
tween developmental psychology and cogni-
tive psychology. Part of the reason has been
the antagonism between some members of
the social and cognitive psychological com-
munities. Some social psychologists argue
that laboratory testing not only changes the
Mother let me
go out to play
Mother did not let
me go out to play
behaviour it is supposed to be investigating,
but that it makes it impossible to draw
useful conclusions about that behaviour.
Some cognitive psychologists argue that the
alternative methods of study suggested by
social psychologists are not rigorous enough
to produce scientific knowledge. Other cog-
nitive psychologists, particularly in memory
research, have themselves identified pro-
blems with laboratory-based studies, but
argue for rigorous experimental studies in
ecologically valid settings. However, general
considerations about scientific method
provide no support for the idea that phe-
nomena are best studied in their natural
settings.
The call for ecologically valid empirical
methods should not be confused with
arguments for what is known as 'ecolog-
ical psychology'. Ecological psychology,
which takes a variety of forms, claims that
environmental factors should playa greater
role in theories of human behaviour.
Ecological psychology is antipathetic to
those types of cognitive psychology that
place too much emphasis on what happens
in the mind. However, there need be no real
disagreement between the two approaches,
provided that cognitive psychologists are
169
sensitive to the richness of environmental
variations, which must be mentally repre-
sented if they are to influence behaviour.
Returning to social psychology, there is,
within that subdiscipline, a strong experi-
mental tradition. Indeed, there have been
bitter methodological divisions within social
psychology. Furthermore, there is a flour-
ishing tradition of cognitive social psychology,
which emphasizes the role of individuals'
mental processes in social behaviour.
Nevertheless, even though cognitive psy-
chology and cognitive social psychology are
both responses to criticisms of behaviour-
ism, there is little interaction, between cog-
nitive psychologists and cognitive social
psychologists, and cognitive social psychol-
ogy is not regarded as part of cognitive
psychology.
The link between developmental psychol-
ogy and cognitive psychology is closer.
Developmental psychology was for a long
time dominated by the ideas of Jean Piaget,
whose primary concern was a theory of
cognitive development (his own term was
'genetic epistemology'). Furthermore, like
modern-day cognitive psychologists, Piaget
was interested in the mental representa-
tions and processes that underlie cognitive
skills. However, Piaget's genetic epistemol-
ogy never coexisted happily with cognitive
psychology, though Piaget's idea that rea-
soning is based on an internalized version
of predicate calculus has influenced
research into adult thinking and reasoning.
One reason for the lack of a closer interac-
tion between genetic epistemology and
cognitive psychology was that, as cognitive
psychology began to attain prominence,
developmental psychologists were starting
to question Piaget's ideas. Many of his
empirical claims about the abilities, or more
accurately the inabilities, of children of
various ages were discovered to be con-
taminated by his unorthodox, and in retro-
spect unsatisfactory, empirical methods.
And many of his theoretical ideas were
seen to be vague, or uninterpretable, or
inconsistent. Despite the welcome turning-
away from what was often an uncritical
acceptance of everything Piaget said, devel-
170
opmental psychology has retained a strong
cognitive component. However, if anything
it has been more strongly influenced by
ideas from the philosophy of mind than by
ideas from cognitive psychology. This is
particularly true in work on children's
theory of mind, which rose to prominence
in the late 1980s. One reason why work in
the philosophy of mind is important in this
field is that it is often hard to determine, or
even to decide how one might determine,
what a child believes or how a child con-
ceptualizes the world, for example. In the
face of these difficulties, philosophical ana-
lysis is sometimes useful.
Information Theory and Human Experimental
Psychology
From the 1920s to the beginning of the
1950s, BERA VIOURISM was the dominant
tradition in Anglo-Saxon psychology, and
the primary focus of research was on laws
of learning. These laws were assumed to
apply to all 'organisms', and so were studied
in rats and pigeons, which are easy to keep
in the laboratory. By the beginning of the
1950s, however, ideas from the nascent
study of complex computing machines
began to influence psychologists. In parti-
cular, Claude Shannon's information theory
suggested to Colin Cherry and Donald
Broadbent new ways of thinking about
information encoding and selective atten-
tion. The same theory suggested to George
Miller new ideas about constraints on short-
term memory, and about how language
allows communication between one person
and another. Behaviourists also realized the
importance of language - which they
termed 'verbal behaviour' - as an aspect of
human behaviour that presented a chal-
lenge to the generality of the laws of learn-
ing. Skinner, the most influential
behaviourist of the period, devoted a book
(called Verbal Behavior, 1957) to an attemp-
ted demonstration of how behaviourist
principles could be extended to language.
The use of information theory, together
with other influences, led to a substantial
body of research, which was originally
dubbed 'human experimental psychology'.
For reasons to be described below, people
whose primary interest was in the processes
of language understanding were shortly
diverted in another direction. Meanwhile,
work on sensory memory, pattern recogni-
tion, memory and attention continued
apace. It was this work that formed the
basis of Ulric Neisser's (1967) book Cognitive
Psychology, which popularized the term as
an alternative to 'human experimental
psychology', and which bears much of the
responsibility for its current ubiquity.
COGNITIVE PSYCHOLOGY AND
GENERATIVE GRAMMAR
George Miller, under the influence of the
linguist Noam CHOMSKY, had become dis-
satisfied with the application of informa-
tion theory to language processing.
Chomsky, in addition to introducing new
methods for describing regularities in the
syntactic systems of natural languages,
proposed a cognitive interpretation of
linguistic theory. Rules discovered by
linguists were held to be in the minds of
language users, and used to understand
and produce utterances and texts. Miller
proposed a psychological theory, the
Derivational Theory of Complexity, which
suggested one way in which Chomsky's
rules might be used in comprehension.
Although Chomsky is recognized as one
of the main forces in the overthrow of
behaviourism and in the initiation of the
'cognitive era' - his review of Skinner's
Verbal Behavior had a huge rhetorical
impact - the relation between psycho-
linguistics and cognitive psychology has
always been an uneasy one. Indeed the
term 'psycho linguistics' is often taken to
refer primarily to psychological work on
language that is influenced by ideas from
linguistic theory. Mainstream cognitive
psychologists, for example when they write
textbooks, often prefer the term 'psychol-
ogy of language'. The difference is not,
however, merely in a name. In a discus-
sion of the psychology of language one is
likely to encounter notions such as seman-
tic network and spreading activation that are
common in other areas of cognitive psy-
chology. There may be an active hostility
toward, and often an insensitivity to, the
issues that interest linguistic theorists. Fur-
thermore, a discussion of the psychology
of language is likely to emphasize com-
monalities between language processing
and other cognitive processes. Psycholin-
guists are more likely to endorse FODOR'S
modularity hypothesis, and to see lan-
guage processing as drawing on a special
and encapsulated set of linguistic know-
ledge.
THE INFLUENCE OF ARTIFICIAL
INTELLIGENCE
Information theory was the first set of ideas
from computer science to have an impact
on what was to become cognitive psychol-
ogy. ARTIFICIAL INTELLIGENCE (AI) has
had a more profound, and more complex,
relation with the discipline. One strand of
this relationship can be traced back to the
first main wave of AI research, Newell,
Shaw and Simon's studies of heuristic
problem-solving techniques, which were
intended to suggest models of human abil-
ities. Newell et al.'s (1957) first important
program, the Logic Theorem Machine (LT),
scored a remarkable success when it pro-
duced a proof of a theorem in Whitehead
and Russell's Principia Mathematica that was
shorter and more elegant than the original.
Principles embodied in LT and in Newell et
al.'s chess playing programs were general-
ized in the General Problem Solver (GPS)
and in later production system-based
models of human problem solving (Newell
and Simon, 1972). GPS's lack of success (in
solving problems, rather than in elucidating
principles) was attributed to the later recog-
nized importance of domain-specific know-
ledge in problem solving, an idea that is
readily incorporated within the production
system framework, in particular in expert
systems.
Within cognitive psychology Newell and
Simon's work has had its greatest impact in
research on problem solving. Newell (1990)
subsequently proposed that the theory of
171
problem solving can be generalized to a
unified theory of cognition. In particular he
argued for two ideas. First, almost any cog-
nitive 'task', and also its subtasks, can be
thought of as a heuristically guided search
through a space of possibilities (as problem
solving can). This idea is supported by work
in AI where, for example, parsing has been
conceptualized as a search through the set
of possible analyses for sentences for the
analysis of the sentence currently being
processed. Newell's second idea is that,
when an impasse is reached in searching
for a solution, weaker but more generally
applicable techniques should be used to
resolve it. Once the impasse has been
resolved, the solution (in the form of a pro-
duction rule), but not the details of how the
solution was discovered, can be stored in
long-term memory. Thus Soar, as Newell
calls his 'architecture for cognition', con-
tains both a problem-solving and a learning
mechanism.
COGNITIVE PSYCHOLOGY AND COGNITIVE
SCIENCE
In the late 1970s the relation between cog-
nitive psychology and AI was further com-
plicated by the emergence of cognitive
science. Cognitive science is an inter-
disciplinary approach to cognition that
draws primarily on ideas from cognitive
psychology, artificial intelligence, linguistics
and logic. It attempts to wed the concern for
formal. well-specified, testable theories from
the last three of these disciplines with the
attention to properly psychological methods
and psychological data that characterizes
cognitive psychology. For many, language
is the central concern of cognitive science,
particularly those of its adherents who
accept Fodor's notion of a LANGUAGE (or
languages) OF THOUGHT - a language-like
represen tational system (or systems) for
encoding and utilizing information in the
mind. However, the paradigmatic ally suc-
cessful piece of research in cognitive science
remains David Marr' s (1982) study of visual
processing. Indeed, so important is this
work, that it has been claimed for AI and
172
for connectionism (see below) as well as for
cognitive science. One of the most import-
ant aspects of Marr's work is his identifica-
tion of three levels at which a cognitive
system must be analysed. First, a task ana-
lysis leads to a computational theory of
what the system does, and why it does it.
Second, details of the algorithm and (system
of) representation used to make the compu-
tations specified by the computational
theory must be determined. Third, the
neural implementation has to be specified -
details of the machinery on which the
computations are carried out. Neurophysiol-
ogists, AI researchers and cognitive psy-
chologists all, according to Marr, tend to be
guilty of ignoring the all important level of
computional theory.
In vision, unlike in some other branches
of cognitive psychology, there is a real pos-
sibility, as Marr showed, of integrating
detailed information about the underlying
neurophysiology into a broadly psycholo-
gical theory (though Marr's greatest success
in this respect was at the lowest levels of
visual processing which, at least arguably,
lie outside the domain of cognitive psychol-
ogy). In the psychology of language, for
example, almost nothing is known about
neural mechanisms at the level of individual
cells. However, much is known about the
neuropsychology (as opposed to neuro-
physiology) of language. Traditional neuro-
psychology focuses on brain lesions, and
tends to describe the cognitive impairments
they produce in relatively unsophisticated
terms. Modern cognitive approaches some-
times eschew information about lesion sites
altogether, but take a more analytic
approach to patterns of deficit, and in parti-
cular focus on double dissociations, in
which each member of a pair of cognitive
functions is, in different patients, destroyed
while the other is preserved. Cognitive
neuropsychology has been particularly
important in informing models of word
identification, and it may yet prove the
downfall of the connectionist models (see
below) that have proved so successful in
explaining word identification in normal
adults.
COGNITIVE PSYCHOLOGY AND
CONNECTIONISM
CONNECTIONISM is another development of
the 1980s whose relation to cognitive psy-
chology remains problematic. Connectionist
theories model the learning and perform-
ance of psychological tasks (not neces-
sarily cognitive) in terms of interactions
between a large number of simple inter-
connected neuron-like units that exchange
activation until they settle into a stable
state, which is interpreted as a response to
the current state of the world. Learning in
connectionist systems is a process of
making very many small adjustments to
the strengths of the interconnections
between the units, using a technique called
back propagation. The adjustments are made
on the basis of the difference between the
system's actual response to an input and
the correct response. There are connec-
tionist models of processes, such as word
identification, that are traditionally studied
in cognitive psychology, though some
would argue that, with its emphasis on
COMPUTATIONAL MODELS, connectionism is
cognitive science (or even AI). Indeed con-
nectionism has been described as challen-
ging the traditional symbolic paradigm of
AI.
Connectionist models respond to statisti-
cal regularities in the world. They may
appear to follow rules, but because they are
basically associationist, it has been argued
that they cannot operate successfully in
complex domains, such as natural language
syntax, where the descriptive power of (at
least) a phrase-structure grammar is
needed. On this view, connectionist systems
cope well with, for example, mappings
between the spellings and sounds of English
words because those mappings are just sets
of statistical regularities. They have pro-
blems, for example, in parsing English sen-
tences because sentence structure is
described by rules, not statistical regular-
ities.
Even if cognitive science has not resul-
ted in the interdisciplinary research teams
that were once envisaged, much of the
best work in cognitive psychology draws
on ideas from related disciplines in model-
ling the mental mechanisms, representa-
tions and processes that underlie our
cognitive abilities. And there is a trend
toward better formulated theories, though
this is partly a reflection of the maturity of
the discipline. The tendency remains,
however, for cognitive psychologists to be
relatively weak on theory and over-zealous
on data collection. There is, for example,
almost nobody who could reasonably be
described as a theoretical cognitive psy-
chologiSt. Before the advent of connection-
ism, many theories in cognitive psychology
were presented in the form of box dia-
grams, in which neither the contents of
the boxes nor the nature of the inter-
connections was well-specified. Further-
more, theoretical ideas that have been
borrowed from other disciplines have often
proved problematic. Notoriously, the
concept of a schema (originally borrowed
by AI from psychologyl) has been used to
'explain' the fact that structured know-
ledge in long-term memory is vitally
important in many cognitive tasks. Unfor-
tunately, the notion of a schema is so
vague that it imposes almost no con-
straints on how knowledge is organized.
Furthermore, many researchers have failed
to distinguish between describing the
structure of particular bits of information
in memory, and contributing to a theory
of how knowledge is organized.
The lack of theoretical sophistication,
together with the fact that some - though
a diminishing proportion - of findings in
cognitive psychology are obvious or
commonsensical, has led to its positive
achievements being underestimated in
neighbouring disciplines. The best cognitive
psychologists have an eye for a good
problem and for a good experimental
paradigm, even if there are, often lengthy,
disputes about how data should be inter-
preted. However, one of the problems with
good paradigms in cognitive psychology is
that once they are discovered they are
overworked, with rapidly diminishing
returns.
173
THE METHODOLOGY OF COGNITIVE
PSYCHOLOGY
Most cognitive psychologists prefer to inves-
tigate cognitive processes using tightly con-
trolled experiments rather than, say,
naturalistic observational studies. Depend-
ing on the aspect of cognition under investi-
gation, subjects will be set a task to perform,
their responses will be recorded (classified as
right or wrong, if appropriate), and perhaps
timed. More controversially, the subjects
might be asked to 'think aloud' as they
perform the task. The methodology of cog-
nitive psychology can be illustrated with
two pieces of work widely regarded as
important within the diScipline. However, it
should be borne in mind that these studies
are only illustrative, and that the amount of
research carried out in cognitive psychology
is enormous.
In the first piece of work, Roger Shepard
(Shepard and Metzler, 1971) showed pairs
of pictures of simple 3-D objects to people
and asked them to judge whether the objects
were identical or mirror images. He found
that the time to make the judgment was
directly proportional to the angle through
which one of the images had to be rotated so
that it could be superimposed (in the case of
identical objects) on the other. This finding
has obvious implications for the kinds of
representations and processes used by sub-
jects to perform such tasks. However, the
exact sense in which people perform 'mental
rotations' remains unclear. In the second
piece of work, Anne Treisman (Treisman
and Gelade, 1980) asked subjects to find
specified items in two-dimensional displays
containing up to 30 items. If they had to
pick out a blue letter from green letters or an
's' from 'T's, the time was approximately
constant, no matter how many items were
in the display. However, if the subjects had
to pick out a blue'S' from a mixed display of
green 'S's and green and blue 'T's, the time
taken increased in an approximately linear
manner with the number of other items in
the display. These results suggest that the
simple discriminations can be carried out by
processes that work on the whole display at
174
once, no matter how many items are in it,
whereas the more complicated discrimina-
tions have to be carried out by examining
each item in turn.
Although the mechanics of these experi-
ments is easy to grasp, the point of them
may not be, without the requisite back-
ground knowledge. Philosophers (and lin-
guists) often misunderstand what cognitive
psychologists are trying to do. Some of the
reasons why can be illustrated using a dif-
ferent example: pronoun interpretation.
Philosophers find little of interest in a 'text'
such as:
John went into the room. He sat down.
In their study of pronouns they focus on
cases in which there are known difficulties
in providing a systematic account of
meaning, such as the infamous 'donkey
sentences', for example:
Every farmer who owns a donkey beats
it.
Such sentences may be very rare, and they
may, for all philosophers know, be very diffi-
cult to understand. All that is important is
that clever people who think about them for
long enough can agree on what their
meaning should be. Philosophers can ignore
sentences of the first kind, because they can
see how to assign them meaning system-
atically. However, even if psychologists can
also think of a mechanism that can assign
meanings to such sentences, they have to
determine whether it is psychologically plau-
sible, and whether it is compatible with what
is known, empirically, about how pronouns
are interpreted. Furthermore, philosophers
may dismiss as ill-formed 'texts' such as:
Young Toby fights every playtime, but
they never lead to injuries.
Psychologists, on the other hand, have to
ask if they cause comprehension problems;
if, for example, they are more difficult to
understand than corresponding well-formed
texts:
Young Toby gets into fights every play-
time, but they never lead to injuries
or what judgments unsophisticated subjects
make about them. There has been, since the
advent of cognitive science in the late
1970s, a narrowing of the gap between the
interests of formal semanticists and cogni-
tive psychologists. A comparison of the
concerns of Hans Kamp's (1981) discourse
representation theory, or Jon Barwise and
John Perry's (1983) situation semantics,
and the mental models theory of text com-
prehension (Garnham, 1987; Johnson-
Laird, 1983) makes this point. Nevertheless,
psychological concerns and philosophical
ones do not always coincide, and without
an understanding of why this lack of co-
incidence arises, philosophers and psycholo-
gists are likely to dismiss each other's work
unnecessarily and to the detriment of pro-
gress in understanding cognition.
PHILOSOPHY OF COGNITIVE PSYCHOLOGY
Some philosophers may be cognitive scien-
tists, others concern themselves with the
philosophy of cognitive psychology and cog-
nitive science. Indeed, since the inaugura-
tion of cognitive science these disciplines
have attracted much attention from certain
philosophers of mind. The attitudes of these
philosophers and their reception by psychol-
ogists vary considerably. Many cognitive
psychologists have little interest in philoso-
phical issues. Cognitive scientists are, in
general, more receptive.
Fodor, because of his early involvement
in sentence processing research, is taken
seriously by many psycho linguists. His
modularity thesis is directly relevant to
questions about the interplay of different
types of knowledge in language under-
standing. His innateness hypothesis,
however, is generally regarded as unhelpful.
and his prescription that cognitive psychol-
ogy is primarily about PROPOSITIONAL
ATTITUDES is widely ignored. DENNETT'S
recent work on CONSCIOUSNESS treats a
topic that is highly controversial. but his
detailed discussion of psychological research
findings has enhanced his credibility among
psychologists. In general, however, psychol-
ogists are happy to get on with their work
without philosophers telling them about
their 'mistakes'.
Connectionism has provoked a somewhat
different reaction among philosophers.
Some - mainly those who, for other
reasons, were disenchanted with traditional
AI research - have welcomed this new
approach to understanding brain and be-
haviour. They have used the success,
apparent or otherwise, of connectionist
research, to bolster their arguments for a
particular approach to explaining be-
haviour. Whether this neuro-philosophy will
eventually be widely accepted is a difficult
question. One of its main dangers is suc-
cumbing to a form of reductionism that
most cognitive scientists, and many philoso-
phers of mind, find incoherent.
See also INNATENESS; MODULARITY; PSY-
CHOLOGY AND PHILOSOPHY.
BIBLIOGRAPHY
Barwise, J., and Perry, J. 1983. Situations and
Attitudes. Cambridge, MA: MIT Press/Brad-
ford Books.
Ekman, P. 1982. Emotion in the Human Face.
Cambridge University Press.
Garnham, A. 1987. Mental Models as Repre-
sentations of Discourse and Text. Chichester,
Sussex: Ellis Horwood.
Johnson-Laird, P.N. 1983. Mental Models:
Towards a Cognitive Science of Language, Infer-
ence, and Consciousness. Cambridge Uni-
versity Press.
Kamp, H. 1981. A theory of truth and seman-
tic representation. In Formal Methods in the
Study of Language, ed. J. Groenendijk, T.
Janssen, and M. Stokof. Amsterdam: Math-
ematical Centre Tracts, pp. 255-78.
Marr, D. 1982. Vision: A Computational Investi-
gation into the Human Representation and Pro-
cessing of Visual Information. San Francisco:
Freeman.
Manktelow, K.I., and Over, D.E. 1991. Social
roles and utilities in reasoning with deontic
conditionals. Cognition, 39, 85-105.
Neisser, U. 1967. Cognitive Psychology. New
York: Appleton-Century-Crofts.
175
COMPUTATIONAL MODELS OF MIND
Newell, A. 1990. Unified Theories of Cognition:
The 1987 William James Lectures. Cam-
Newell, A., Shaw, J.C., and Simon, H.A. 1957.
Empirical explorations with the Logic
Theory Machine: A case study in heuristics.
Proceedings of the Western Joint Computer
Conference, 15,218-239.
Newell, A., and Simon, H.A. 1972. Human
Problem Solving. Englewood Cliffs, N.J.:
Prentice-Hall.
Shepard, R.N., and Metzler, J. 1971. Mental
rotation of three-dimensional objects.
Science, 171. 701-703.
Skinner, B.P. 1957. Verbal Behavior. New
York: Appleton-Century-Crofts.
Treisman, A., and Gelade, G. 1980. A feature-
integration theory of attention. Cognitive
Psychology, 12,97-136.
ALAN GARNHAM
computational models of mind The
hypothesis driving most of modern cognitive
science is simple enough to state: the mind is
a computer. What are the consequences for
the philosophy of mind? This question
acquires heightened interest and complexity
from new forms of computation employed in
recent cognitive theory.
Cognitive science has traditionally been
based upon symbolic computation: systems of
rules for manipulating structures built up of
tokens of different symbol types. (This
classical kind of computation is a direct
outgrowth of mathematical logic; see AR TI-
FICIAL INTELLIGENCE; TURING). Since the
mid-1980s, however, cognitive theory has
increasingly employed connectionist computa-
tion: the spread of numerical activation
across interconnected networks of abstract
processing units (see CONNECTIONISM).
Symbolic and connectionist computation
respectively constitute computational
accounts of idealized mental and neuronal
processes; thus the interaction of these two
forms of computation in contemporary cog-
nitive theory provides, at the very least,
food for thought concerning the mind-body
problem.
How can the mind perform in accordance
with epistemic constraints? How can it
176
possess unbounded competence, in reason-
ing, arithmetic, grammar? How can
abstract mental elements arise from a phy-
sical substrate, and interact causally with
the physical world? Approaches to answer-
ing these fundamental questions can be
derived from the hypothesis that the mind/
brain is a computer: for computers are well
understood, physically realized abstract
systems that can be endowed with unboun-
ded competence, and whose function can
respect epistemic constraints. The answers
to the fundamental questions differ, how-
ever, depending on the kind of computer the
mind/brain is alleged to be: a symbolic com-
puter, a connectionist computer, or some
mixture of the two.
(1) VARIETIES OF COMPUTATION:
ALTERNATIVE COMPUT ATIONAL
ABSTRACTIONS
Different types of computation provide dif-
ferent kinds of computational abstractions
to serve the role of mentalist theoretical
constructs. These abstractions, which will
be denoted '\I.', can be considered in light of
the following criteria for assessing their ad-
equacy in meeting the needs of cognitive
theory.
Desiderata for Mentalist Abstracta \I.
(1) Computational sufficiency: \I. must be part
of a system with sufficient computa-
tional power to compute classes of func-
tions possessing the basic characteristics
of cognitive functions. Frequently at-
tested such characteristics include:
unbounded competence or 'productiv-
ity'; recursion; systematicity; Turing
Universality; structure-sensitivity; con-
cept and language learn ability; statist-
ical sensitivity.
(2) Empirical adequacy: \I. must enable the
construction of accounts of human
competence and performance which
explain the empirical facts of human
cognition acknowledged by COGNI-
TIVE PSYCHOLOGY, linguistics (see
CHOMSKY), etc.
(3) Physical [neural] realizability: rx must be
realizable in a physical [neural] system
These criteria together pose the central
paradox of cognition. On the one hand, espe-
cially in higher cognitive domains like rea-
soning and language, cognition seems
essentially a matter of structure processing.
On the other hand, in perceptual domains,
and in the details and variance of human
performance, cognition seems very much a
matter of statistical processing; and the
underlying neural mechanisms seem thor-
oughly engaged in numerical processing.
Thus these criteria seem on the one hand to
demand computational abstractions rx which
provide symbolic structure processing, and
on the other, abstractions rx which provide
numerical and statistical processing. And,
indeed, contemporary cognitive theory relies
on two types of computation, one addressing
the structural, the other the numerical/
statistical, side of the central paradox.
Symbolic Computation and the Classical
Computational Theory of Mind
In symbolic computation, the abstractions
provided are symbols (rx
sym
) and rules (rxrul).
According to the classical computational
theory of mind, MENTAL REPRESENTATIONS
are symbol structures, and mental processes
consist in the manipulation of these rep-
resentations according to symbolic algor-
ithms, based on symbolic rules.
The symbolic computational abstractions
rxsvm&rul fare admirably with respect to the
structural side of the criteria of computa-
tional sufficiency and empirical adequacy.
Physical computers demonstrate the phy-
sical realizability of symbolic computation.
But the statistical side of computational
sufficiency and empirical adequacy are
essentially ignored by pure symbolic com-
putation, and little if any insight has arisen
from the symbolic theory concerning
whether, and how, it can be effectively
neurally realized.
Local Connectionist Computation
'Local connectionist models' are connec-
tionist networks in which each unit cor-
responds to a symbol or symbol structure
(perhaps an entire proposition). Typically,
the connections in such networks, and the
computations performed in each connec-
tionist unit, are carefully hand-constructed
to implement specific symbolic rules.
The abstractions provided by this compu-
tational framework are the individual unit's
activity (rxind.act) and the individual connec-
tion (rxind.cnx)' These function very much
like the symbol and rule of symbolic compu-
tation. Indeed, this kind of connectionism
has strong affinities to traditional symbolic
computation. It is the kind of network pro-
posed by McCulloch and Pitts (1943; re-
printed in Boden, 1990) which historically
contributed to the development of symbolic
computation; it is intimately related to con-
temporary research in the design of algor-
ithms with 'fine-grained parallelism'. It
offers some of the same kinds of computa-
tional sufficiency as symbolic computation,
but with significant limitations. These
networks often employ units which are
discrete-state machines, with discrete mes-
sages passed along connections. However,
genuine numerical processing can also be
used; certain kinds of statistical inference
can then be performed.
Neural realizability has been a main
driving force in much of this research,
through Jerome Feldman's '100 step rule':
these networks are typically designed to
perform their computations in roughly the
same number of steps or 'cycles' available to
neurons - 100. The plausibility of these
networks as simplified neural models
depends on the hypothesis that symbols and
rules are more or less localized to individual
neurons and synapses (see Feldman's article
in Nadel et aI., 1989, which includes refer-
ences on local connectionist models).
PDP Connectionist Computation and an
Eliminativist Theory
A different style of connectionist modelling is
the 'parallel distributed processing (PDP),
approach. Here, the computational abstrac-
tions offered to replace symbols and rules are
patterns of numerical activity over groups of
units (rxpat.act), and patterns of weights
177
over groups of connections (expat.cnx)'
Crucially, a given unit or connection parti-
cipates in many such patterns, which may
be simultaneously present (multiple patterns
being superimposed upon each other).
Whereas local connectionist models typi-
cally rely explicitly on symbolic rules and
representations which are implemented in
individual units and connections, PDP cog-
nitive models typically strive to eliminate
such symbolic elements. As a result, such
models usually do not provide much (if
anything) in the way of structure-sensitive
processing. Rather than implementing ex-
plicitly stated rules through hand-designed
connections and units, PDP modellers typi-
cally put the necessary knowledge into the
connection weights by training the
network, using a PDP learning procedure,
on examples of the desired performance.
The resulting weights generally constitute
some sort of complex statistical analysis of
the training data. PDP algorithms are
heavily numerical.
The strengths and weaknesses of the
abstractions expat.act&cnx of PDP computation
are almost exactly the mirror image of those
of symbolic computation. Computational
and empirical sufficiency is greatest with
respect to the statistical. and weakest with
respect to the structural. side of the central
paradox. With respect to most of the specific
criteria listed in 'computational sufficiency',
PDP computation (in the basic form con-
sidered now) is quite limited. On the other
hand, while much remains in obscurity
with regard to the neural realizability of
PDP models, this form of computation has
for the first time provided a bridge for fruit-
ful interaction between computational cog-
nitive modelling and neuroscience.
An Integrated Connectionist/Symbolic
Cognitive Architecture
The complementary strengths and weak-
nesses of symbolic and PDP connectionist
computation have increaSingly led cognitive
scientists to attempt to combine the two
computational frameworks. Most often this
takes the form of 'hybrid architectures' in
which the overall computational system has
178
two separate components, one symbolic, the
other connectionist. The relation between
these components is that of being two parts
of a common whole.
In an alternative, more integrative,
approach, a symbolic and a PDP connec-
tionist component are each two descriptions
of a single, unitary, computational system;
the symbolic and the connectionist are
related as higher- and lower-level descrip-
tions of a common system (see SUPER-
VENIENCE). This is achieved by structuring
PDP networks so that the computational
abstractions expat.act&cnx which they provide
- activation and connection patterns - are
Simultaneously describable as the computa-
tional abstractions exsym&rul of symbolic com-
putation - symbol structures and rules.
This 'in tegrated connectionist/symbolic
(ICS), computational framework is presented
in Smolensky, Legendre, and Miyata (1994).
The structuring of PDP computation which
enables the higher-level structure of
activation and connection patterns to be
analysable as symbol structures and rules is
achieved by applying the mathematics of
tensor calculus. The distributed pattern of
activity s realizing a symbol structure is a
superposition (sum) of distributed patterns
Cj, each of which realizes one of the symbolic
constituents of the structure: s = Ejcj. The
pattern Cj which realizes each constituent
has the following structure: it is a dis-
tributed pattern fj encoding a symbolic part
multiplied by another distributed pattern rj
encoding the role played by the part in the
structure as a whole. The multiplication
used is the tensor product ; Cj = fj rj
[Given two activity patterns f = (0.2, -0.3)
and r = (0.1, 0.5), their tensor product is
simply f r = (0.02, 0.05, -0.03, -0.15):
the elements of the pattern f r are all
numerical products of one element of f and
one element of r.J
Such 'tensor product representations'
enable distributed patterns of activity to
realize much of symbolic representation,
including recursion. These representations
can be processed by simple connectionist
networks, 'tensor networks', in which the
connections have a higher-level tensorial
Representation Processing Combined
IXsyro = symbol structure IXrul = symbolic rule
IXsyro&rul
IXpat.act&cnx
IXind.act&cnx
IXpat.act = pattern of activity
IXind.act = individual unit activity
IXpat.cnx = pattern of connection weights
IXind.cnx = individual connection weight
Table 1. Summary of Computational Abstractions
structure which comports with that of the
representations. These networks can prov-
ably compute specific symbolic functions.
including recursive functions. Since it will
be referred to several times below. a tensor
network which computes a symbolic func-
tion f will be denoted N
f
. Such a network
can be described at multiple computational
levels (d. 2). At the lowest level. it is a set
of units and connections. and processing is
simple spread of activation. The distributed
activity and connections are globally struc-
tured in such a way that they can be given
a formal higher-level description using
tensor calculus. The elements of this
description. in turn. can be precisely re-
described as symbols and rules. The func-
tion f computed (but. as discussed shortly.
not the means of computing it) can be given
a formal symbolic specification.
Connectionist computation is not used in
ICS merely to implement symbolic compu-
tation (cf. 3). For example. ICS provides a
new formalism for grammar. 'Harmonic
Grammar'. according to which the linguis-
tic structures of a language that are
grammatical are exactly those that opti-
mally satisfy a set of conflicting parallel soft
constraints of varying strength: these soft
constraints constitute the grammar. The
strengths of constraints may be numerical.
or the constraints may simply be ranked by
strength. each constraint stronger than all
weaker constraints combined. This latter
grammar formalism. Optimality Theory.
developed by Prince and Smolensky (1994).
turns out to Significantly strengthen the
theory of the phonological component of
grammar: it becomes possible to eliminate
delicately ordered sequences of symbolic
rewrite-rules in favour of parallel soft con-
straints of great generality. These con-
straints often turn out to be universal
(applying in all languages): languages
differ principally in the relative strengths
they assign the constraints. By formally
realizing aspects of symbolic computation in
lower-level PDP connectionist computation.
ICS leads to new grammatical theories
which strengthen the theory of universal
grammar. enhancing the explanatory ade-
quacy of linguistic theory.
In addition to this evidence for its empiri-
cal adequacy. a number of formal results
argue the computational adequacy of the
ICS framework. ICS unifies the computa-
tional abstractions of symbolic and PDP
computation. summarized in Table 1. in
order to confront effectively both the struc-
tural and the numerical/statistical sides of
the central paradox of cognition.
Other important abstractions IX which can
be treated computationally but which are
not discussed here include propositions and
goals/desires (see PROPOSITIONAL ATTI-
TUDES) and images (see IMAGER Y).
(2) VARIETIES OF COMPUTATION ALISM:
ALTERNATIVE ROLES FOR
COMPUTATIONAL ABSTRACTIONS
Having distinguished several different types
of computational abstractions IX which are
made available by various kinds of compu-
tation. we now turn to consideration of
various roles a given computational
abstraction IX may play in a theory of mind.
Computation is a complex notion. and
different aspects of it can figure in formulat-
ing the hypothesis cognition is computation.
Elaborating on a notion of computational
level articulated by Marr (1982). we can
179
Computational Levell
Computationalist Hypothesis at level 1 concerning
abstractions rt. (and cognitive process P)
Description Function CHfun(rt.): P can be described as instantiating a
function J. mapping inputs to outputs.
specifiable in terms of rt.
Explanation Proof C H p r ~ rt.): That P instantiates J can be explained
by proofs based on properties of P
stated in terms of rt.
Causation Algorithm CHa1g(rt.): P can be described via an algorithm
Realization Implementation
A. stated in terms of rt.. which computes J
CHimp(rt.): The spatiotemporal structure of A.
stated over rt.. corresponds to that of its
physical realization
Table 2. Summary of Computationalist Hypotheses
distinguish different members of a family of
computationalist hypotheses. summarized in
table 2. Each computationalist hypothesis
CHj(rt.) asserts that some cognitive process P
can be given a computational account at
level 1 in terms of a class of abstract ele-
ments rt.. We focus on rt. = rt.sym&rulo and
abbreviate CH
j
( rt.sym&rul) by CHj.
CH
fun
: Description oj cognitive behaviour via
function specification
The weakest hypothesis. CHfun(rt.). asserts
only that the cognitive process P can be
analysed (using rt.) as a function mapping
certain inputs to certain outputs. If the
abstract elements rt. in question are rt.sym&rul.
as we assume for the rest of 2. then
CHfun(rt.sym&rul) = CH
fun
claims that P is a
process which takes symbol structures as
inputs. that P produces symbol structures as
outputs. and that the function J from input
structure to output structure that P
instantiates can be specified. somehow.
using symbol manipulation rules.
For example. CH
fun
holds of the following
inference process P (see RATIONALITY): an
input is a REPRESENT A TION (as formulae) of
a set of propositions. and an output is a cor-
responding representation of a specified set
of propositions which follow from the input
propositions. according to some criteria
specified over the formulae.
180
To suggest how CH
fun
might fail. consider
a perceptual process P. where the inputs are
sensory stimuli and the output is a percept.
Such inputs and outputs might well fail to
be satisfactorily analysed as structures of
discrete symbols. in which case the corre-
sponding function J could not possibly be
specified via symbol manipulation rules.
(Certain 'cognitive linguistic' theories claim
that for linguistic processes. input and/or
output representations must be analysed as
non-discrete structures: something more
like images.)
CHpr/ Explanation oj cognitive behaviour via
proof
Whereas CH
fun
makes the purely descriptive
claim that P instantiates a function J speci-
fied using symbols and rules. CH
prf
further
claims that some properties p of P can be
characterized using symbols and rules such
that it can be proved from p that the output
resulting from a given input i is indeed f(i):
these properties provide some explanation of
how it is that P instantiates f.
The network N
f
introduced in l illus-
trates one way CHprf can hold. P is realized
in a connectionist network whose connec-
tion weights possess tensor-algebraic struc-
ture capturable in some mathematical
property p; and from p it follows that when
an input activity pattern is given to the
network which is tensor-product realization
of a symbol structure i. then an output
pattern is produced which realizes I(i)
(where I is a symbolically specified func-
tion). The property p explains how P
instantiates f.
CH
prf
might fail, even when CH
fun
holds
(this is a kind of EPIPHENOMENALISM with
respect to (lsym&rul)' Suppose a connectionist
network has been trained on appropriately
encoded examples of input/output pairs of a
symbolically specified function I, and that
the net eventually learns I (Le. CH
fun
holds).
Typically, (a) no analysis of the connection
weights can identify any properties from
which it may be proved that the net com-
putes I; and (b) it cannot be proved that I is
guaranteed to be learned by the connec-
tionist learning algorithm. Then there is no
evidence that the special situation CHprf
demands actually holds.
CH
a1g
: Abstract causal account 01 cognitive
processing via algorithms
In classical symbolic cognitive theory, CH
prf
is typically known to hold because in fact a
stronger hypothesis, CH
aig
holds: a symbolic
algorithm characterizes p, and this algor-
ithm A can be proved to compute I.
For example, suppose P can be char-
acterized as following the steps of some
theorem-proving algorithm A: then it can
be proved that the function I which P
instantiates maps input propositions to
valid consequences. Or suppose P follows
the steps of a rewrite-rule grammar for a
language L: by sequentially rewriting
strings of sym-bols, the rules provably gen-
erate all and only the strings ('sentences')
of L, so P provably instantiates a function I
which (say) takes as input a phrase cate-
gory label and generates as output an
example string of that category. Such
examples could be multiplied almost
without bound, since traditional symbolic
cognitive science has for decades been suc-
cessfully producing symbolic algorithms A
for computing interesting cognitive func-
tions I; each of these provides a process for
which CH
aig
holds. (See Newell and Simon,
1976, reprinted in Boden, 1990.)
CH
aig
can fail even though CH
fun
holds.
The network N
f
takes an input activity
pattern which realizes a symbol structure i
and, in one step, directly computes the
output activity pattern which realizes the
structure I(i). 'In one step' means that the
input activity flows from the input units
through a single set of connections to the
output units, which immediately then
contain the output activation pattern. There
is no symbolic algorithm which describes
how N
f
computes its output structure from
its input structure. There exists a traditional
symbolic algorithm for computing I which
sequentially performs primitive symbolic
operations such as extracting elements from
lists or trees, comparing tokens to see if they
are of the same type, and concatenating
smaller lists or trees together to form larger
structures. But such algorithms do not
describe how N
f
performs its computation:
CH
aig
fails. That I is computed by N
f
is not a
mystery, however: this is provable from the
structure of N/s connection weights, as dis-
cussed above: CH
prf
holds.
CHimp: Physical realization 01 cognitive
systems via implementation
The distinction between the algorithmic
level and the implementation level can be
stated in terms of the properties of P avail-
able at each level of description. The level of
physical implementation involves properties
such as the actual time required to compute
output from input; the actual physical loca-
tion of computational elements; the actual
material of which the computing device is
constructed, and the physical states and
causal processes which realize the computa-
tional states and processes. The algorithmic
level provides a more abstract causal
account of P: instead of actual time to
compute the output, we have the number of
primitive steps of computation required;
instead of the actual physical location of
computational elements (e.g. symbols), we
have 'locations' in an abstract data struc-
ture; instead of physical causation leading
from one physical state to the next, we have
abstract causation in which primitive steps
in an algorithm lead - by unexplained
181
COMPUT A TION AL MODELS OF MIND
means - from one computational state to
the next.
When CHalg(lX) holds, the stronger
hypothesis CHimp(lX) can be formulated: the
abstract causal structure of the algorithm A
corresponds to the physical causal structure
of the device realizing the process P. This
means that there is a direct one-to-one cor-
respondence between primitive steps in an
algorithm and physical events, between
abstract locations and physical locations,
etc.
Traditional electronic computers running
a program for a symbolic algorithm provide
an example in which CHimp holds: at any
given moment during computation, it can
be said of each symbol and each rule in the
algorithm exactly which physical location
in the computer's memory 'holds' the
symbol - exactly what part of the physical
state of the hardware physically realizes the
symbol or rule.
To illustrate a case when CHimp fails, but
CHaig holds, consider a direct physical
implementation of 'Tensor Product Produc-
tion System' (Dolan and Smolensky, 1989),
a connectionist network which realizes a
symbolic algorithm in which production
rules are sequentially applied to manipulate
a working memory storing triples of
symbols. The reason CHimp fails is that the
symbols in working memory are all super-
imposed on top of each other, so that a
given connectionist processing unit parti-
cipates in the representation of many
symbols. There is no one-to-one correspond-
ence between abstract locations of symbols
in structures and physical locations in the
device. (While CH
imp
( IXsym&rul) fails, CHimp
(lXind.act&cnx) holds, as we now discuss.)
Level Symbolic Local Connectionist
CH
fun IXsym&rul IXsym&rul,lXind.act&cnx
CH
prf IXsym&rul IXsym&rul,lXind.act&cnx
CHalg IXsym&rul IXsym&rul,lXind.act&cnx
CHimp IXsym&rul,lXind.act&cnx
(3) IMPLICATIONS FOR THE PHILOSOPHY
OF MIND
Table 3 summarizes the consequences of the
various computational theories from 1 for
the various computationalist hypotheses
CHI from 2: shown are the abstractions IX
for which each CHI holds in each theory.
Symbolic and Local Connectionist Theory: The
Success of Symbolic Computationalism
Commitment to CHaig - the existence of
algorithms stated as rules for manipulating
symbol structures - is the heart of the sym-
bolic theory of mind. In philosophical terms,
the symbolic story is that elements of a
LANGUAGE OF THOUGHT - symbolic con-
stituents of beliefs, desires, and the like -
have causal roles in the functioning of the
mind: there is a causal story to be told, an
algorithm, about how these symbolic con-
stituents 'push each other around', through
the mediation of rules. For instance, if the
rules are taken to be valid rules of inference,
then their operation on symbolic belief
structures will be truth-preserving. If the
rules are ones for generating grammatical
sentences, or for arithmetic manipulation of
numerals, then their operation will provide
unbounded grammatical or arithmetic com-
petence. In many parts of cognitive science
outside philosophy - notably artificial intel-
ligence, psychological modelling, and com-
putational linguistics - the main theoretical
activity of symbolic theory has been the
development of rule-based symbol-
processing algorithms to compute cognitive
functions.
With the validity of CHaig in this theory
comes the validity of the weaker hypotheses
PDP Connectionist Integrated (ICS)
Input/output: IXpat.act
IXsym&rul,lXpat.act&cnx
Function:-
IXsym&rul,lXpat.act&cnx
lXind.act&cnx lXind.act&cnx
lXind.act&cnx lXind.act&cnx
Table 3. Summary of Computational Theories and Computationalism
182
CH
fun
and CH
prf
: from the existence of
(provably correct) symbolic algorithms
come the existence of proofs that the com-
putational system actually computes the
cognitive function f(CH
prf
) as well as a sym-
bolic way of specifying f directly via the
algorithm. if no more perspicuous specifica-
tion can be given.
That symbolic computation can be phys-
ically realized is important. but a commit-
ment to CHimp is not part of the symbolic
theory of mind: there is no claim that the
brain. like a traditional computer. will have
separate locations for each symbol and rule.
More than an absence of commitment. in
fact. the symbolic theory is essentially silent
on the matter of its physical realization in
the human case.
Local connectionist theory. like symbolic
theory. entails that CHaig holds. provided we
admit the corresponding parallel algorithms
into the class of symbolic algorithms. Since
individual units and connections correspond
quite directly to symbol structures and
rules. these parallel algorithms provide a
new sort of rule-based symbol manipula-
tion. As in symbolic theory. CH
fun
and CH
prf
typically also hold. In addition. under the
typical assumption that a unit corresponds
to a physical neuron (or small distinct
group of neurons). CHimp also holds. Thus
every computationalist hypothesis holds of
Q(sym&rul. as well as Q(ind.act&cnx'
PDP Connectionist Theory: The Failure of
Symbolic Computationalism
At its most eliminativist. the PDP connec-
tionist theory of mind denies that any of the
computationalist hypotheses are valid with
respect to the abstractions Q(sym&rul' Even the
inputs and outputs of cognitive functions.
the extreme position would claim. must be
treated as graded. non-discrete. patterns of
activity: even the weakest hypothesis.
CH
fun
. fails with respect to discrete symbolic
structures.
With respect to non-symbolic computa-
tional abstractions. however. certain of the
computationalist hypotheses hold in the
PDP theory. although the story is more
complicated than for the symbolic view.
Semantically interpretable inputs and
outputs of cognitive functions are defined
over distributed patterns of activity (Q(pat.act).
A typical PDP network does not allow a
semantic characterization of the function f
computed: CHfun(Q(pat.act) fails. Except for the
Simplest (feedforward) networks. the func-
tion f cannot even be specified in terms of
individual unit activities: CH
fun
( Q(ind.act) fails.
The failure of CHfun(Q() entails the failure of
CHprr{Q(). PDP networks are algorithms
defined over individual activities and con-
nections Q(ind.act&cnx. a lower level of descrip-
tion than the pattern level Q(pat.act where
semantic interpretation occurs: CHaig
(Q(ind.act&cnx) holds. The most straightfor-
ward neural realization of a PDP network
identifies each unit with a neuron and each
connection with a synapse; in this case.
CH
imp
( Q(ind.act&cnx) also holds. The same is
true under the more typical assumption
that each connectionist unit corresponds to
a (distinct) group of neurons.
The failure of PDP theory to satisfy
CHprr{Q() for any kind of abstraction Q( can be
viewed as a failure of that theory to offer
strong explanations of cognitive behaviour.
ICS Theory: A New Explanatory Strategy
In the ICS theory. networks such as N
f
provide a story intermediate between the
symbolic and the eliminativist PDP. With
respect to Q(sym&rul. CH
fun
and CH
prf
both
hold: the inputs. outputs. and function f
computed all have a symbolic characteriza-
tion. and proofs that the network computes
f rely crucially on symbolic abstractions.
However. CHaig (and CHimp) fail: at the level
of algorithm (or implementation). symbolic
abstractions do not figure. Instead. the story
is the same as for PDP theory: CHalg(Q() (and
CHimp(Q()) hold for Q( = Q(ind.act&cnx. And.
because of the tensor-algebraic structure of
activity and connection patterns in ICS.
CHfun(Q() and CHprf(Q() hold not just of
Q( = Q(sym&rul. but also of Q( = Q(pat.act&cnx:
symbols and rules are realized as distributed
patterns of activity and connections.
While both the symbolic and ICS stories
explain how cognitive functions get com-
puted. they differ crucially. The symbolic
183
story derives its explanation from the causal
roles of symbols and rules, captured in
algorithms: CHalg(Clsym&rul) holds. In the ICS
story, however, CH
a1g
( Clsym&rul) fails: it is not
in virtue of algorithmic or causal structure
that symbolic functions get computed.
Rather, it is a more abstract tensorial struc-
ture present in the network which enables
proofs and explanations of how f gets com-
puted. This structure resides in the global
properties of activity patterns and connec-
tions, not in the spatio-temporal interac-
tions which are manifest in algorithms.
In other words, in ICS, symbols and rules
have crucial explanatory roles, but no causal
roles. There is no magic here: symbols and
rules are realized in lower-level activities
and connections which themselves have
causal roles: CH
a1g
( Clind.act&cnx) holds. But
these lower-level processing algorithms
cannot be 'pulled up' to the higher level:
symbols and rules have no causal roles in
their own right - in stark contrast to the
symbolic theory.
Other Issues
Besides computationalism, several other
issues are at stake in the question of which
kinds of computational abstractions provide
appropriate models for the mind/brain.
One is the problem of the semantic rela-
tion between internal mental representa-
tions and their external referents (see
CONTENT; INTENTIONALITY). If mental
representations are built of continuous con-
nectionist activity or connection patterns
rather than discrete symbols or rules, then
the relevant referents ought to be con-
tinuous properties quite different from tradi-
tional discrete propositions (see Haugeland's
chapter in Ramsey et aI., 1991). The func-
tional roles in the overall computational
system of connectionist as opposed to sym-
bolic representations seem quite different
(see FUNCTIONALISM). If representations are
constructed through connectionist learning,
then their functional roles may need to be
understood through their adaptational roles
(Le. in evolutionary terms).
A number of questions concerning cogni-
tion can be phrased: how can we explain
184
the goodness of fit between adaptive, intelli-
gent agents and their environments? The
intricate structure of the linguistic environ-
ment of the child is remarkably matched by
ultimate adult linguistic competence; the
rational adult's behaviour is remarkably
attuned to the truth-relations between pro-
positions about the environment. Tradi-
tional symbolic cognitive theory has been
focused on those environmental regularities
that are heavily structural. PDP theory
redirects attention to environmental
regularities that are statistical. and con-
tributes to understanding how adaptive
agents can become successfully attuned to
such regularities. ICS theory, in turn, chal-
lenges us to address new regularities where
structure and statistics, discrete and con-
tinuous, merge.
Finally, discussions of the mind-body
problem, especially those concerning
various IDENTITY THEORIES, often pre-
suppose a high degree of discord between
the natural kinds of mentalist psychology
and those of neuroscience. This presump-
tion may be quite mistaken. Computational
cognitive theory proposes to analyse both
the mind and the brain as kinds of com-
puters. Contemporary research such as the
ICS work discussed above suggests that
there may well be a natural fit between the
computational abstractions characterizing
mentalist psychology and their counterparts
in neuroscience.
See also COGNITIVE PSYCHOLOGY; MOD-
ULARITY.
BIBLIOGRAPHY
Boden, M.A., ed. 1990. The Philosophy of Arti-
ficial Intelligence. Oxford University Press.
Clark, A., ed. 1992. Special issue: Philosoph-
ical issues in connectionist modeling Connec-
tion Science, 4,171-381.
Cummins, R. 1989. Meaning and Mental Repre-
sentation. Cambridge, MA: MIT Press/Brad-
ford Books.
Dolan, C.P. and Smolensky, P. 1989. Tensor
product production system: a modular
architecture and representation. Connection
Science, 1. 53-68.
Horgan, T., and Tienson, J. 1991. Connection-
ism and the Philosophy of Mind. Dordrecht:
Kluwer Academic Publishers.
Kirsh, D., ed. 1991. Special volume: Founda-
tions of artificial intelligence. Artificial Intel-
ligence, 47,1-346.
Legendre, G., Miyata, Y., and Smolensky, P.
1994. Principles for an Integrated Connec-
tionist/Symbolic Theory of Higher Cognition.
Hillsdale, New Jersey: Lawrence Erlbaum
Associates.
Marr, D. 1982. Vision. San Francisco, Calif.:
Freeman.
Nadel, L., Cooper, L.A., Culicover, P., and
Harnish, R.M., eds. 1989. Neural Connec-
tions, Mental Computation. Cambridge, MA.:
MIT Press/Bradford Books.
Partridge, D. and Wilks, Y., eds. 1990. The
Foundations of Artificial Intelligence: A Source-
book. Cambridge University Press.
Prince, A., and Smolensky, P. 1994. Optim-
ality Theory: Constraint interaction in gen-
erative grammar. Cambridge, MA.: MIT
Press/Linguistic Inquiry Monograph Series.
Pylyshyn, Z. 1984. Computation and Cognition:
Toward a Foundation for Cognitive Science.
Cambridge, MA.: MIT Press/Bradford Books.
Ramsey, W., Stich, S.P., and Rumelhart, D.E.
1991. Philosophy and Connectionist Theory.
Hillsdale, New Jersey: Lawrence Erlbaum
Associates.
Smolensky, P. 1994. Constituent structure
and explanation in an integrated connec-
tionist/symbolic cognitive architecture. In
The Philosophy of Psychology: Debates on
Psychological Explanation, ed. C. G. Mac-
donald and G. Macdonald. Oxford: Basil
Blackwell.
Smolensky, P., Legendre, G., and Miyata, Y.
1994. Principles for an Integrated Connec-
tionist/Symbolic Theory of Higher Cognition.
Cambridge, MA: MIT Press.
PAUL SMOLENSKY
concepts The notion of a concept, like the
related notion of meaning, lies at the heart of
some of the most difficult and unresolved
issues in philosophy and psychology. The
word 'concept' itself is applied to a bewilder-
ing assortment of phenomena commonly
thought to be constituents of THOUGHT.
These include internal ME NT AL REPRES
CONCEPTS
ENT A TIONS, IMAGES, words, stereotypes,
senses, PROPER TIES, reasoning and dis-
crimination abilities, mathematical func-
tions. Given the lack of anything like a
settled theory in this area, it would be a
mistake to fasten readily on anyone of
these phenomena as the unproblematic
referent of the term. One does better to
survey the geography of the area and gain
some idea of how these phenomena might
fit together, leaving aside for the nonce just
which of them deserve to be called 'con-
cepts' as ordinarily understood.
There is a specific role that concepts are
arguably intended to play that may serve as
a point of departure. Suppose one person
thinks that capitalists exploit workers, and
another that they don't. Call the thing that
they disagree about 'a proposition', e.g.
[Capitalists exploit workers]. It is in some
sense shared by them as the object of their
disagreement, and it is expressed by the
sentence that follows the verb 'thinks that'
(mental verbs that take such sentence com-
plements as direct objects are called verbs of
'PROPOSITION AL ATTITUDE'. We won't be
concerned here with whether propositions
are ultimate objects, or whether talk of
them can be reduced to talk of properties or
predicates). Concepts are the constituents of
such propositions, just as the words 'capi-
talists', 'exploit' and 'workers' are con-
stituents of the sentence. Thus, these people
could have these beliefs only if they had,
inter alia, the concepts [capitalist], [exploit],
[workers] (I shall designate concepts by
enclosing in square brackets the words that
express them).
Propositional attitudes, and thus con-
cepts, are constitutive of the familiar form of
explanation (so-called 'intentional explana-
tion') by which we ordinarily explain the
behaviour and states of people, many
animals, and perhaps some machines (see
INTENTIONALITY, REASONS AND CAUSES).
In the above example, the different thoughts
that people have about capitalism might
explain their different voting behaviour. By
and large, philosophers and psychologists
such as Fodor (1975, 1991) or Peacocke
(1992), interested in intentional explana-
185
CONCEPTS
tion take themselves to be committed to the
existence of concepts; whereas those wary
of this form of explanation, e.g. Quine
(1960), tend to be sceptical of them.
Just which sentential constituents express
concepts is a matter of some debate. The
central cases that are discussed tend to be
the concepts expressed by predicates or
general terms, such as 'is a capitalist' or 'x
exploits y', terms true of potentially many
different individual things. But there are
presumably concepts associated with logical
words ('and', 'some', 'possibly'). And some
philosophers have argued for the import-
ance of individual concepts, or concepts of
individual things, e.g. [Rome], [25], both in
psychology, and in logic and mathematics.
Since concepts as constituents of thought
are shareable, both by different people, and
by the same person at different times, they
need to be distinguished from the particular
ideas, images, sensations that, consciously
or unconsciously, pass through our minds
at a particular time. Just what kind of
shareable object a concept might be is a
matter of considerable difference between
theorists. In much of the psychological lit-
erature, where the concern is often with an
agent's system of internal representation,
concepts are regarded as internal repres-
entation types that have individual ideas as
their specific token see TYPE/TOKEN (in the
way that the type word 'cat', can have
many different inscriptions as tokens). But
many philosophers take the view that these
internal representation types would no
more be identical to concepts than are the
type words in a natural language. One
person might express the concept [city] by
the word 'city', another by the word 'ville';
still another perhaps by a mental image of
bustling boulevards; but, for all that, they
might have the same concept [city]: one
could believe and another doubt that cities
are healthy places to live. Moreover, differ-
ent people could employ the same repres-
entation to express different concepts: one
person might use an image of Paris to
express [Paris], another to express [France].
Now it might be supposed that the
common object of people's thoughts are
186
simply the referents of their terms, that is,
the objects in the world picked out by the
terms or internal representations, for
example, in the case of 'city', all the parti-
cular cities in the world (this is a view
defended by some proponents of theories of
'direct reference'). There are a number of
difficulties with this view. At least in the
case of general terms (or predicates), there
are standardly at least three different candi-
dates for their referents: (1) the extension, or
set of actual objects that satisfy the predicate
(e.g. the particular cities: New York,
Paris, ... ); (2) the intension, or function
from POSSIBLE WORLDS to sets of possible
objects that satisfy the predicate in a world
(e.g. [city] would be the function that takes
us in the real world to the set containing
New York, Paris, etc., and in another world
to a set of possible cities, e.g. North Polis);
and (3) the causally efficacious property (e.g.
cityhood) that all the (possible) objects have
in common. Extensional logicians like Quine
(1960), eschewing all talk of non-actual
worlds or onto logically suspicious 'proper-
ties', prefer the first option; modal logicians
and formal semanticists like Montague
(1974), interested in accounting for the
semantics of natural languages, tend to
prefer the second; and many philosophers of
mind like Dretske (1987), Millikan (1984)
and Fodor (1991), interested in causal
interactions between animals and the
world, tend to prefer the third.
Moreover, in addition to the referent of a
general term, many (following Frege, 1966)
have argued that there exists its 'sense', or
'mode of presentation' (sometimes 'inten-
sion' is used here as well). After all, 'is an
equiangular triangle' and 'is an equilateral
trilateral' pick out the same things not only
in the actual world, but in all possible
worlds, and so refer - insofar as they are
taken to refer to any of these things at all -
to the same extension, same intension and
(arguably from a causal point of view) the
same property; but they differ in the way
these referents are presented to the mind:
it's one thing to think of something as an
equilateral triangle, another to think of it as
an equilateral trilateral (which is why the
proof that they're necessarily coextensive is
interesting). For some (e.g. Peacocke, 1992)
concepts might be senses so understood. But
we then need a theory of senses. Some phi-
losophers look to an ability and/or a rule
that prescribes a particular inferential (or
'conceptual') role that a representation plays
in an agent's thought, from stimulation
through intervening states to behaviour:
e.g. stimuli and inferences that lead to the
application of a term, and from it to the
application of other terms and action (see
CONCEPTU AL ROLE SEMANTICS). Thus, the
counting of angles might lead to 'triangle'
and back again; or the sight of crowds
might lead to 'crowded', which might in
turn lead to 'ten or more' and perhaps
certain (dispositions to) behaviour, e.g. to
say 'Ten's a crowd', and these patterns of
cause and effect (or causal roles) might be
taken to be constitutive of the concepts [tri-
angle] or [crowd].
In respecting all of the above distinctions,
it is important to be especially careful with
the peculiar idiom 'concept of x', as in 'the
child's' or the 'ancient Greek's concept of
causality', an idiom that figures promi-
nently in DEVELOPMENTAL PSYCHOLOGY
and studies in the history of science. This
could mean merely the concept [causality],
which the child has (as do most adults); or
it could mean the child's ability to deploy
the concept in reasoning and discrimina-
tion; or it could mean any of the extension,
intension, or rule that children associate
with the English word 'causality' and its
related forms; or it could mean (as in fact it
very often does mean) the representation
and/or standard beliefs (what I prefer to call
the conception) that children associate with
the extension, intension, rule or ability
[causality]. Which of these candidates is
intended would all depend upon what entity
one thinks of as the concept and what a
mere accompaniment of it. What can't be
seriously intended is the suggestion that a
child has a concept [causality] that is both
identical to but different from the adult's.
But choosing what one thinks of as a
concept and what a mere accompaniment
depends upon one's theory of concepts and
CONCEPTS
of what explanatory role they are being
asked to play, to which issues we now turn.
THEORIES ABOUT CONCEPTS
The Classical View
Historically, a great deal has been asked of
concepts. As shareable constituents of the
objects of attitudes, they presumably figure
in cognitive generalizations and explana-
tions of animals' capacities and behaviour.
They are also presumed to serve as the
meanings of linguistic items, underwriting
relations of translation, definition, syno-
nymy, antinomy, and semantic implication
(see Katz, 1972). Much work in the seman-
tics of natural languages (e.g. Jackendoff,
1983) takes itself to be addreSSing con-
ceptual structure.
Concepts have also been thought to be
the proper objects of 'philosophical analy-
sis', the activity practised by Socrates and
twentieth-century 'analytic' philosophers
when they ask about the nature of justice,
knowledge or piety, and expect to discover
answers by means of a priori reflection
alone (see e.g. Chisholm, 1957).
The expectation that one sort of thing
could serve all these tasks went hand in
hand with what has come to be called the
'Classical View' of concepts, according to
which they have an 'analysis' consisting of
conditions that are individually necessary
and jointly sufficient for their satisfaction,
which are known to any competent user of
them. The standard example is the espe-
cially simple one of [bachelor], which seems
to be identical to [eligible unmarried male].
A more interesting, but problematic one has
been [knowledge], whose analysis was tra-
ditionally thought to be [justified true
belief].
This Classical View seems to offer an illu-
minating answer to a certain form of meta-
physical question: in virtue of what is
something the kind of thing it is - e.g. in
virtue of what is a bachelor a bachelor? -
and it does so in a way that supports coun-
terfactuals: it tells us what would satisfy the
concept in situations other than the actual
ones (although all actual bachelors might
187
CONCEPTS
turn out to be freckled, it's possible that
there might be unfreckled ones, since the
analysis doesn't exclude that). The View
also seems to offer an answer to an epis-
temological question of how people seem to
know a priori (or independently of experi-
ence) about the nature of many things, e.g.
that bachelors are unmarried: it is con-
stitutive of the competency (or possession)
conditions of a concept that they know its
analysis, at least on reflection.
Empiricism and Verificationism
The Classical View, however, has always
had to face the difficulty of primitive con-
cepts: it's all well and good to claim that
competence consists in some sort of mastery
of a definition, but what about the primitive
concepts in which a process of definition
must ultimately end? Here the British
Empiricism of the seventeenth century
began to offer a solution: all the primitives
were sensory. Indeed, they expanded the
Classical View to include the claim, now
often taken uncritically for granted in dis-
cussions of that view, that all concepts are
'derived from experience': 'every idea is
derived from a corresponding impression'.
In the work of Locke, Berkeley and Hume
this was often thought to mean that
concepts were somehow composed of intro-
spectible mental items - 'images', 'impres-
sions' - that were ultimately decomposable
into basic sensory parts. Thus, Hume ana-
lysed the concept of [material object] as
involving certain regularities in our sensory
experience, and [cause] as involving spatio-
temporal contiguity and constant conjunc-
tion.
Berkeley noticed a problem with this
approach that every generation has had to
rediscover: if a concept is a sensory impres-
sion, like an image, then how does one dis-
tinguish a general concept [triangle] from a
more particular one - say, [isoceles triangle]
- that would serve in imagining the general
one. More recently, Wittgenstein (1953)
called attention to the multiple ambiguity of
images. And, in any case, images seem
quite hopeless for capturing the concepts
associated with logical terms (what is the
188
image for negation or possibility?). What-
ever the role of such representations, full
conceptual competence must involve some-
thing more.
Indeed, in addition to images and impres-
sions and other sensory items, a full
account of concepts needs to consider issues
of logical structure. This is precisely what
the Logical Positivists did, focusing on logi-
cally structured sentences instead of sensa-
tions and images, transforming the
empiricist claim into the famous 'Verifia-
bility Theory of Meaning': the meaning of a
sentence is the means by which it is con-
firmed or refuted, ultimately by sensory
experience; the meaning or concept asso-
ciated with a predicate is the means by
which people confirm or refute whether
something satisfies it.
This once-popular position has come
under much attack in philosophy in the last
fifty years. In the first place, few, if any,
successful 'reductions' of ordinary concepts
(like [material object], [cause]) to purely
sensory concepts have ever been achieved
(see e.g. Ayer, 1934 for some proposals, and
Quine 1953, and Chisholm 1957 for criti-
cism). Our concept of material object and
causation seem to go far beyond mere
sensory experience, just as our concepts in a
highly theoretical science seem to go far
beyond the often only meagre evidence we
can adduce for them.
Moreover, there seemed to be a pattern to
the failures. Taking a page from Pierre
Duhem, Quine (1951) pointed out that 'our
beliefs confront the tribunal of experience
only as a corporate body': litmus paper
turning red confirms that a solution is
acidic only in conjunction with a great deal
of background chemical and physical
theory, indeed, many have argued, only in
conjunction with the whole of a person's
system of beliefs (a view called 'confirmation
holism'). Hence, if a concept is to be ana-
lysed as its verification conditions, its
meaning would be similarly holistic
('meaning HOLISM'). Given that no two per-
son's beliefs are likely to be precisely the
same, this has the consequence that no two
people ever share precisely the same con-
cepts - and no one could. strictly speaking.
remember the same thing over any amount
of time that included a change of belief!
'Sameness of concept' for Quine becomes by
and large an 'indeterminate' issue. The best
one might hope for is a similarity of infer-
ential role between symbols in different the-
ories or symbol systems. a conclusion
enthusiastically endorsed by such diverse
writers as Kuhn (1962). Harman (1972)
and Block (1986).
Fodor and LePore (1992) have recently
argued that the arguments for meaning
holism are. however. less than compelling.
and that there are important theoretical
reasons for holding out for an entirely
atomistic account of concepts. On this view.
concepts have no 'analyses' whatsoever:
they are simply ways in which people are
directly related to individual properties in
the world, ways which might obtain for
someone for one concept but not for any
other one: in principle. someone might have
the concept [bachelor] and no other con-
cepts at all. much less any 'analysis' of it.
Such a view goes hand in hand with
Fodor's rejection of not only verificationist.
but any empiricist account of concept learn-
ing and construction: indeed. given the
failure of empiricist constructions. Fodor
(1975. 1979) notoriously argues that con-
cepts are not constructed or 'derived from
experience' at all. but are (nearly enough)
all innate (see INN ATENESS).
Non-classical Approaches: Prototypes
WITTGENSTEIN (1953) raised a different
issue of whether a concept actually need
have any Classical analysis at all. Certainly.
people are seldom very good at producing
adequate definitions of terms that they are
nonetheless competent to use. Wittgenstein
proposed that. rather than classical defini-
tions that isolated what. for example. all
games had in common. the different uses of
the word 'game' involved a set of over-
lapping and criss-crossing 'family resem-
blances. This speculation was taken
seriously by Rosch (1973) and Smith and
Medin (1981) as a testable psychological
hypothesis. They showed that people
CONCEPTS
respond differently (in terms of response
time and other measures) to questions
about whether. for example. penguins as
opposed to robins are birds. in a fashion
that suggested that concept membership
was a matter not of possessing a Classical
analysis; but of 'distance' from a 'prototype'
or typical 'exemplar'. Thus. a robin satisfies
many more of the features of a typical bird
than does a penguin and so is a 'better'
member of the category; and a malicious lie
is a better case of a lie than a well-inten-
tioned one.
It has not always been clear precisely
what sort of thing a prototype or examplar
might be. One needs to resist a strong temp-
tation to import into the mind or brain pro-
cedures. such as comparing one actual bird.
or even a picture of one. with another. that
make sense only outside of it. Presumably it
is some sort of list of selected properties.
perhaps accompanied by a mental image.
and a metric for determining the distance of
a candidate from that list. Some writers
have proposed exploiting the resources of
'fuzzy set theory' to capture the intended
structure, whereby membership in a cat-
egory is not understood as an all-or-none
affair. but a matter of degree: everything
satisfies every concept to some degree.
however small.
But proto typicality. which presumably
involves distances among a complex cluster
of diverse properties. must be distinguished
from both vagueness and guessing. It is a
commonplace that nearly every concept
that applies to things in space and time is to
some extent 'vague' in that there are
always 'hard cases' in which it is not clear
whether or not the concept applies. There
are. for example. plenty of people for whom
it seems to be indeterminate whether or not
they are bald. but this is no objection to the
Classical View that [bald] is analysable as
[lacking cranial hair]. since. whatever
vagueness is involved in applying [bald]
might well be involved in applying that
analysis.
Guessing. like prototypicality and vague-
ness. also comes in degrees. Unlike them.
however. it is nota metaphysical issue of
189
CONCEPTS
the actual conditions something must
satisfy in order to satisfy the concept, but
rather an epistemological one concerning
the belief or epistemic probability that some-
thing satisfies the conditions, given certain
evidence. The sight of someone with a toupe
may mean that there is a 90% probability
that he's actually 50% bald, or a 40% prob-
ability that he's actually 95% bald. The
question of whether [bald] is Classical or
prototypical is untouched by this issue as
well.
This distinction between the metaphysical
question about how things are and the epi-
stemic one about evidence is, however, par-
ticularly difficult to enforce with regard to
concepts (ordinary English can often en-
courage running the two questions together:
the metaphysical question can be phrased
'What determines what's what?' and can
then be confused with the epistemological
one 'How does someone determine what's
what?'). As Rey (1983, 1985) pointed out,
concepts are too often expected to answer
both sorts of question, and we shouldn't
suppose without argument that they can
perform both sorts of work. For example,
what metaphysically determines whether
something is male or female is presumably
facts about its reproductive capacities or
genes; but neither of these are the way we
standardly guess a person's gender, which
usually involves accidental features of
name, hair and dress. Indeed, in view of the
aforementioned confirmation holism, epi-
stemic procedures may well involve almost
anything whatsoever, since there may be
no limit to the ingenuity of a person in
exploiting roundabout evidence and infer-
ences - or just relying on what other people
say. But surely the fact that people may
exploit anything in finding out what's what
doesn't entail that just anything is part of a
concept's proper analysis. In any case, the
fact that people may be slower to agree that
penguins rather than robins are birds is no
reflection upon the status of penguins as
bona fide birds even in the minds of the
subjects of these experiments. Most people,
after all, know that the stereotypes they
have of things are not entirely reliable and
190
can be transcended on a little reflection.
Indeed, there is no reason to suppose that
the Classical analyses of a concept need be
employed in 'on-line' reasoning except in
the unusually demanding tasks of 'philoso-
phical analysis'.
The empiricist tradition, particularly in its
verificationist mood, made a point,
however, of connecting the metaphysical
with the epistemic: the defining conditions
for a concept were to be stated in terms of
experiential evidence. If that tradition is to
avoid the move from confirmation to
meaning holism, it would need to distin-
guish among the ways in which concepts
are related to experience those ways that
are due to the genuine analysis of the
concept and those that are due merely to
beliefs an agent may have involving it. The
aforementioned 'inferential role' accounts of
concepts hope to begin to do this, distin-
guishing inferences from 'bachelor' to
'unmarried' from 'bachelor' to 'freckled'.
Those interested in defining concepts in
terms of their roles in theories are pre-
sumably making just such an appeal.
However, Quine's (1953) famous attack
on the analytic/synthetic distinction has
seemed to many to show that no such dis-
tinction among inferences can generally be
drawn (see QUINE). He argued that the
notions of analyticity, meaning, synonymy
and possibility form a vicious circle of
notions insusceptible to empirical test. Quite
apart from this issue, moreover, there is the
difficulty of drawing a principled limit to
how deviant people can be about the infer-
ences they draw with concepts they seem
nonetheless entirely competent to use: some
creationists believe that people aren't
animals, idealists that material objects are
ideas, and nominalists that numbers are
numerals. These all seem like possible cog-
nitive states, and so present a prima facie dif-
ficulty for a theory of concepts that claims
their identity involves speCific connections
to other concepts or experiences.
Non-classical: Causal Approaches
The work of Kripke (1972), Putnam (1975)
and Burge (1979) suggests another strat-
egy. They argue that the meanings of
words, particularly of proper names and
natural kind terms, don't involve definitions
known to users of them, but rather causal
relations with their actual referents and/or
the social community in which the term is
used. (Whether there nevertheless exist
definitions that users may not know is an
issue neither they nor many others
address.) PUTNAM (1975), in particular,
imagined there to be a planet, TWIN-EARTH,
exactly like the earth in every way except
for having in place of H
2
0, a different, but
superficially similar chemical XYZ. He
argued that the word 'water' in Twin-
Earthling's mouths would mean something
different from what it means in ours, and
that they would have a different concept
[twater] from the concept [water] that we
on earth normally employ, despite the fact
that everything about the organization of
the brains of Twin-Earthlings would be (ex
hypothesi) indistinguishable from that of our
own.
While these intuitions provide an inter-
esting challenge to the Classical View, they
sacrifice its account of conceptual com-
petence. If competence doesn't consist in a
grasp of a definition, what makes it true on
this view that someone has one concept
rather than another? What makes it true
that a child or an adult has the concept
[cause], or [knowledge], if she can't define
it? Merely causally interacting in a certain
community and environment can't be
enough, since surely not every sentient
being in New York City has all the concepts
of a Columbia University physicist.
So a number of writers have proposed
varieties of counterfactual causal links: x
has the concept y iff some state of x did/
would causally co-vary with y (Le. x did/
would discriminate instances of y) under
certain (ideal, normal, evolutionarily sig-
nificant) conditions, as a matter of nomolo-
gical necessity. Thus, someone has the
concept [horse] iff she could under certain
conditions tell the horses from the non-
horses. This is the idea behind 'informa-
tional' (or 'covariational') theories of the
sort proposed by Dretske (1980), Millikan
CONCEPTS
(1984) and Fodor (1991). In a way, it is
simply a development of an idea that
empiricists often implicitly presumed (and
that was explicit in the work of B. F.
Skinner) for the sensory primitives: someone
had the concept [red] iff she could dis-
criminate red things. Informational theories
simply extend this idea to non-sensory-
terms as well.
Unfortunately, this solution doesn't seem
quite adequate. There is first of all the sub-
stantial difficulty of specifying the appro-
priate conditions for the covariation in a
non-circular fashion. Many suspect that this
will fall afoul of 'Brentano's Thesis' of 'the
irreducibility of the intentional': spelling out
the appropriate conditions would involve
mentioning other intentional/semantic/con-
ceptual conditions, such as that the agent is
paying attention, doesn't believe her percep-
tual experience is misleading, wants to
notice what's going on, etc. This potential
circle is particularly troubling for those con-
cerned with 'naturalizing' talk of concepts,
Le. of fitting it into theories of the rest of
nature (biology, physics).
But there are also a number of more
specific problems that have been Widely
discussed, but not decisively solved in the
literature: transitivity, disjunction, coex-
tensivity and co-instantiation. Transitivity is
the problem that, whereas covariation
(and other causal 'information') is a transi-
tive notion, conceptual representation is
not (if A covaries with Band B with C,
then A covaries with C; but 'A' doesn't
thereby express both [B] and [C]). For
example, the symbol 'smoke' might covary
under the appropriate conditions with
smoke, and smoke, itself, might covary
with fire; and so 'smoke' would covary
with fire as well; but 'smoke' means
[smoke] and not [fire].
Disjunction is the problem of distinguish-
ing the misapplication of a concept [A] to Bs,
from the correct application of the dis-
junctive concept [A or B]: e.g., a representa-
tion that covaries with horses and is
misapplied to cows on a dark night is a
representation that could be taken to co-
vary with horses or cows on dark nights
191
CONCEPTS
(Fodor (1991) discusses these latter pro-
blems in detail; see also the related problem
of determining which arithmetic function
someone is computing, discussed in Kripke,
1982).
We already noticed the problem of neces-
sary coextensiveness in the case of triangles
versus trilaterals: covariation of a symbolic
state with one would necessarily be co-
variation with the other. The problem of
coinstantiated concepts is different. Quine
(1960, ch. 2) first raised it with his famous
discussion of translating a foreign sentence
'Gavagai', tokens of which he assumes
covary with the presence of rabbits. The
problem is that whenever there is a rabbit
present there are both undetached rabbit parts
and time slices of rabbits. The respective con-
cepts in each of these cases are not even
coextensive (rabbits aren't parts or slices of
rabbits) much less identical. Yet because of
their necessary coinstantiation, any state
that covaried with one would covary with
the other. So something else is needed to
pin particular concepts down.
TENTATIVE CONCLUSION
It is tempting to return to some feature of
the Classical View and select some feature of
inferential role that would disambiguate
these cases: patterns of inference would cer-
tainly seem to be what distinguish [smoke]
from [fire], [horse] from [horse or cow on a
dark night], [triangle] from [trilateral],
[rabbit] from [undetached rabbit parts].
Another possibility, explored most extens-
ively by Millikan (1984) and Dretske
(1987), has been to appeal to the tele-
ological functions related to a discrimina-
tion ability (perhaps there is or was a
biological advantage to discriminating
rabbits rather than their undetached parts)
that might play a role in explaining why a
trait was selected. Probably the best hope is
in some subtle amalgamation of the various
views that have been sketched here - for
example, of both a type-representation,
inferential role, and some covariational
view. These are the issues of continuing
research.
192
We might summarize the present situa-
tion with regard to candidates for 'con-
cepts' that have been discussed here as
follows: there is the token representation in
the mind or brain of an agent, types of
which are shared by different agents.
These representations could be words,
images, definitions, or 'prototypes' that play
specific inferential roles in an agent's cogni-
tive system and stand in certain causal and
covariant relations to phenomena in the
world. By virtue of these facts, such rep-
resentations become associated with an
extension in this world, possibly an inten-
sion that determines an extension in all
possible worlds, and possibly a property
that all objects in all such extensions have
in common. Which of these (italicized)
entities one selects to be concepts depends
on the explanatory work one wants con-
cepts to perform. Unfortunately, there is as
yet little agreement on precisely what that
work might be.
See also BELIEF; CONTENT; DRETSKE; FODOR;
HISTOR Y; RATIONALITY.
BIBLIOGRAPHY
Ayer, A.J. 1934 Language, Truth and Logic.
New York: Dover.
Block, N. 1986. Advertisement for a semantics
for psychology. In Studies in the Philosophy of
Mind, ed. P. French, T. Vehling and H.
Wettstein. Minneapolis: University of Min-
nesota Press.
Burge, T. 1979. Individualism and the mental.
In Studies in Metaphysics, ed. P. French, T.
Vehling and H. Wettstein. Minneapolis: Uni-
versity of Minnesota Press.
Carnap, R. 1969. The Logical Structure 0/ the
World and Pseudo-problems in Philosophy.
Berkeley: University of California Press.
Chisholm, R. 1957. Perceiving: a Philosophical
Study. Ithaca: Cornell University.
Dretske, F. 1980. Knowledge and the Flow 0/
In/ormation. Cambridge MA.: MIT Press.
Dretske, F. 1987. Explaining Behavior. Cam-
bridge MA.: MIT Press.
Fodor, J. 1975., The Language o/Thought. New
York: Crowell.
Fodor, J. 1979. 'The present status of the
innateness controversy.' In Representations;
Essays on the Foundations of Cognitive Science.
Cambridge MA.: MIT Press.
Fodor, J. 1991. A Theory of Content. Cambridge
MA.: MIT Press.
Frege, G. 1966. Translations from the Philo-
sophical Writings of Gottlob Frege, trans. P.
Geach and M. Black. Oxford: Blackwell.
Harman, G. 1972. Thought. Princeton Uni-
versity Press.
J ackendoff, R. 1983. Semantics and Cognition.
Cambridge MA.: MIT Press.
Katz, J. 1972. Semantic Theory. New York:
Harper and Row.
Kripke, S. 1980. Naming and Necessity. Cam-
bridge MA.: Harvard University Press.
Kripke, S. 1982. Wittgenstein on Rules and
Private Language. Cambridge MA.: Harvard
University Press.
Kuhn, T. 1962. The Structure of Scientific Revo-
lutions. Chicago: University of Chicago
Press.
Millikan, R. 1984. Language, Thought and Other
Biological Categories. Cambridge MA.: MIT
Press.
Montague, R. 1974. Formal Philosophy. New
Haven: Yale Univeristy Press.
Peacocke, C. 1992. Concepts. Cambridge MA.:
MIT Press.
Putnam, H. 1975. The meaning of 'meaning'.
In Language, Mind and Knowledge. ed. K.
Gunderson. Minneapolis: University of Min-
nesota Press.
Quine, W.V. 1953. Two dogmas of empiri-
cism. In From a Logical Point of View and
other Essays. Cambridge MA.: Harvard Uni-
versity Press.
Quine, W.V. 1960. Word and Object. Cam-
Rey, G. 1985. Concepts and conceptions.
Cognition, 19,297-303.
Rey, G. 1983. Concepts and stereotypes. Cog-
nition, 15, 237-62.
Rosch, E. 1973. On the internal structure of
perceptual and semantic categories. In Cog-
nitive Development and Acquisition of Lan-
guage. ed. T. E. Moore. New York: Academic
Press.
Smith, E., and Medin, D. 1981. Categories and
Concepts. Cambridge MA.: Harvard Uni-
versity Press.
tions. New York: Macmillan.
GEORGES REY
CONCEPTUAL ROLE SEMANTICS
conceptual role semantics Conceptual
(sometimes computational, cognitive, causal
or functional) role semantics (CRS) entered
philosophy through the philosophy of lan-
guage, not the philosophy of mind. The core
idea behind CRS in the philosophy of lan-
guage is that the way linguistic expressions
are related to one another determines what
the expressions in a language mean. These
relations constitute the role of an expression
in a language. This core idea goes back in
philosophy at least as far as Wilfrid Sellars.
Its most vigorous defenders at present in the
philosophy of mind are Ned Block, Michael
Devitt, Gilbert Harman, Brian Loar, and
William Lycan.
There is a considerable affinity between
CRS and structuralist semiotics that has
been influential in linguistics. According to
the latter, languages are to be viewed as
systems of differences; the basic idea is that
the semantic force (or 'value') of an utter-
ance is determined by its position in the
space of possibilities that one's language
offers (Saussure, 1983). CRS also has
affinities with what the ARTIFICIAL INTEL-
LIGENCE researchers call 'procedural
semantics'. The essential idea here is that
providing a compiler for a language is
equivalent to specifying a semantic theory
for that language; semantics consists of pro-
cedures that a computer is instructed to
execute by a program (Woods, 1981).
POSITIVE REASONS FOR CRS
Frege-Problems
In order to get a grip on this idea it's useful
to keep in mind that traditionally the
meaning of an expression was thought to
involve some symbol-world relation. So, the
expression 'the World Trade Towers' differs
in meaning from the expression 'the Empire
State Building' because the former is about
the World Trade Towers, whereas the latter
is not. CRS got its footing in the philosophy
of language because of a perceived defi-
ciency in this idea that meaning is somehow
determined by word-world relations.
Meaning, according to this criticism, can't
simply be a word-world relation because, in
193
the classic example, 'Hesperus' and 'Phos-
phorus' are both attached to the same non-
linguistic thing, viz., to Venus. Yet someone
can, without self-contradiction, assert that
the Hesperus shines, while denying that the
Phosphorus shines. According to CRS,
'Hesperus' and 'Phosphorus' mean different
things, despite their both being attached to
Venus, because they have different roles in
the (English) language.
CRS and Logical Connectives
On a more constructive side, practically
everyone believes that CRS is most plausible
for an account of the meaning of the logical
constants. Indeed, beginning as early as
with the logician Gentzen some philoso-
phers and logicians have characterized a
logical constant as an expression whose
meaning is fixed by the ordered pair of the
set of inferences you can validly infer from,
and the set of inferences you can validly
infer to, a sentence dominated by it. For
example, consider what would be involved
for a symbol '#' in a language to express the
material conditional: the '#' in 'Bob is a
man # Bob is human' expresses the material
conditional only if when (tokenings of) this
sentence interacts appropriately with
(tokenings of) 'Bob is a man', the result is a
tendency to token 'Bob is human' (ceteris
paribus); or only if when it interacts with
'Bob is not human', the result is a tendency
not to token 'Bob is a man' but rather a
tendency to token 'Bob is not a man'. And
so on. Plainly, talk of tendency to token
here requires that it is tokens of datable,
placeable inscriptions, sounds, etc., that
have conceptual roles. (See Field, 1977, and
Harman, 1987 for further discussion.)
The Language of Thought
What has any of this to do with the philo-
sophy of mind? Philosophers of mind are
interested in the nature of thought.
Thoughts have contents (meanings) in
much the same way that linguistic expres-
sions do. My thought that the World Trade
Towers were bombed differs in content
(meaning) from my thought that the Empire
State Building was bombed. The former is
194
true and the latter false; the former is about
the World Trade Towers and the latter is
not. Since, as many philosophers maintain,
it is the content (or the meaning) of a
thought that determines what that thought
is about, its INTENTIONALITY, it is natural
to think that content (or meaning) involves
some symbol-world relation, where the
symbols here are expressions in, if you like,
a LANGUAGE OF THOUGHT - say, mentalese.
But now notice how the same problems that
apparently threaten the classical referential
model of meaning for natural language
expressions, also apparently threaten a clas-
sical referential story about the content of
thoughts.
If you think that there is no apparent
contradiction in believing that Hesperus
shines without believing that Phosphorus
shines; or in believing that water quenches
thirst. without believing that H
2
0 quenches
thirst, or in my believing that I'm thirsty
without my believing that Ernie Lepore is
thirsty, say, because I simply forgot that I'm
Ernie Lepore or am deluded in thinking that
I'm David Hume, then the content of these
thoughts cannot be determined solely by
the relationship between these thoughts and
what they are about in the non-linguistic
world. Each pair of thoughts is about
exactly the same thing, and apparently no
mind-world relation will distinguish them.
And thus entered CRS into the philosophy
of mind: it is the way mentalese expressions
are related to one another that fixes what
they mean. These relations constitute the
role of a mentalese expression.
CRS as Semantics versus CRS as Metaphysics
In describing the relations between expres-
sions, we say (depict systematically) what
each expression means. In this sense, a CRS
is in much the same business as any other
semantic theory. It aims to assign to each
meaningful expression in a language L -
spoken, written or thought - a meaning,
viz., for CRS theorists. a conceptual role.
However, some CRS theorists defend not a
semantic thesis about what expressions of a
language - mental or public - mean, but
instead a thesis about the nature of
meaning. For these philosophers, CRS
theorists are in much the same business as
are those philosophers who are either
intention-based Griceans or information-
based semanticists. Griceans tell us that at
least public linguistic expressions have their
meaning in virtue of the intentions with
which they are spoken (H. P. Grice, Jona-
than Bennett, and, at one time, Brian Loar
and Stephen Schiffer). Information-based
semanticists tell us that expressions have
their meanings in virtue of certain causal or
nomological relations that obtain between
tokens of expressions and features of the
non-linguistic world (Jerry FODOR, Fred
DRETSKE, Ruth Millikan). Some CRS theor-
ists intend to answer the non-semantical
metaphysical question, In virtue of what
does, say, an expression e of language L
mean m? The following, though not equiva-
lent, are sometimes used interchangeably
with this 'in virtue of' question: what deter-
mines that expression e in language L means
m?; what is the meaning m of expression e
constituted by?; what is it metaphysically
dependent upon?; what does e's meaning
supervene on? Any philosopher who asks
these questions probably supposes that
semantic facts cannot be brute, primitive,
facts, but must somehow be explicable in
terms of more basic facts.
As a metaphysician, a CRS theorist might
hold that the meaning of a thought is a
function from possible worlds to truth
values (a la Montague), or that the meaning
of a thought is its truth condition (a la
DAVIDSON), or a structured interpreted tree
(a la Katz), but that it's in virtue of e's
having a certain conceptual role that a
thought has the semantic properties it does.
Semantic theories do not, normally, wear
their metaphysical commitments on their
sleeves.
Methodological Solipsism
A more sophisticated reason, squarely in
the philosophy of mind, for adopting CRS
other than the Frege-problem derives from
Jerry Fodor's worries about methodological
solipsism. In a series of highly influential
papers in the 1970s, Hilary Putnam, David
Kaplan, Saul Kripke, and Tyler Burge, each
drew attention to the ways in which the
meanings of many terms depend crucially
upon the environment of the speaker.
Putnam, for example, asks us to imagine
two planets, Earth and TWIN EARTH, and
two of their residents, say, Harry and Twin-
Harry. Twin Earth is almost a physical
replica of Earth. The only difference is that
on Twin Earth the clear liquid the twin-
people drink, that fills their oceans, and that
they call 'water', is composed not of H
2
0
molecules but of XYZ molecules. According
to Putnam, the expression 'water' on Earth
refers to the stuff composed of H
2
0 and not
composed of XYZ. It is exactly the reverse
for the same expression on Twin Earth.
Thus, two subjects can be molecule for
molecule alike inside their heads and still
have intuitively different beliefs. Harry's
belief that water is wet is about H
2
0, while
Twin-Harry's belief 'that water is wet' is
about XYZ, and their truth and referential
conditions differ accordingly.
What such examples are suppose to show
is that the contents of thoughts, in the
referential sense, do not per se figure in the
explanation of brute physical behaviour,
since people who are brain state type-
identical will behave alike regardless of
what their beliefs are about. Methodological
solipsism is Fodor's idea that psychologists,
in explaining behaviour, must rely only on
what is in the head. So, if semantics is to
help in explaining behaviour, it would seem
that purely referential theories are insuffi-
cient for this task. The mind (and its com-
ponents) has no way of recognizing the
reference or truth conditions of the repres-
entations it operates on.
Enter CRS, the referential theories' solips-
istically motivated rival. Since, according to
CRS, the meaning of a thought is deter-
mined by the thought's role in a system of
states, to specify a thought is not to specify
its truth or referential conditions, but to
specify its role. Harry's and Twin-Harry's
thoughts, though different in truth and
referential conditions, share the same con-
ceptual role, and it is by virtue of this com-
monality that they behave type-identically.
195
If Harry and Twin-Harry each has a belief
that he would express by 'water quenches
thirst' , CRS can explain and predict their
dipping their cups into H
2
0 and XYZ
respectively. Thus CRS, it would seem
(though not to Fodor, who rejects CRS for
both external and internal problems - see
below), is better suited to predicting and
explaining what someone decides or does,
so long as it ignores information about the
external world.
But, if, as Fodor contends, thoughts have
recombinable linguistic ingredients, then, of
course, for a CRS theorist, questions arise
about the role of expressions in this lan-
guage of thought as well as in the public
language we speak and write. And, accord-
ingly, CRS theorists divide not only over
their aims, but also about CRS's proper
domain. Two options avail themselves.
Some hold that public meaning is somehow
derivative (or inherited) from an internal
mental language (mentalese), and that a
mentalese expression has autonomous
meaning (partly) in virtue of its conceptual
role (Block, 1986). So, for example, the
inscriptions on this page require for their
understanding translation, or at least trans-
literation, into the language of thought;
representations in the brain require no such
translation or transliteration. Others hold
that the language of thought just is public
language internalized and that it is expres-
sions of public language that have autono-
mous (or primary) meaning in virtue of
their conceptual role (Sellars, 1963;
Harman, 1987; Devitt, 1981). On this
option, English is the language of thought
for English speakers.
What are Conceptual Roles?
After one decides upon the aims and the
proper province of CRS, the crucial question
remains of exactly which relations among
expressions - public or mental - constitute
their conceptual roles. Because most CRS
theorists leave the notion of a role iI). CRS as
a blank cheque, the options are open-ended.
The conceptual role of a (mental or public)
expression might be its causal associations:
any disposition to token (for example, utter
196
or think) one expression e when tokening
another e', or a an ordered n-tuple < e' e",
... >, or vice versa, can count as the con-
ceptual role of e. A more common option is
to characterize conceptual role not causally,
but inferentially (these need not be incom-
patible, contingent upon one's attitude
about the naturalization of inference): the
conceptual role of an expression e in L
might consist of the set of actual and poten-
tial inferences to e or the set of actual or
potential inferences from e, or, as is more
common, the ordered pair consisting of
these two sets. Or, if it is sentences which
have non-derived inferential roles, what
would it mean to talk of the inferential role
of words? Some have found it natural to
think of the inferential role of a word as
represented by the set of inferential roles of
the sentences in which the word appears
(Block,1986).
No matter how broadly inferential role is
characterized, even if it is unpacked caus-
ally, some philosophers argue it is a mistake
to identify conceptual role with inferential
role. They want to include non-inferential
components, for example,sensory input and
behavioural output, as contributing to the
conceptual role of an expression. Sellars
speaks of 'language entry and exit rules' in
his specification of conceptual role. Harman
distinguishes between purely inferential
accounts of conceptual role (which he dubs
solipsistic CRS) and a broader notion of con-
ceptual role (which he dubs non-solipsistic
CRS). Obviously, these non-solipsistic CRS
theorists do not respect Fodor's methodolo-
gical solipsism.
Unlike most semanticists, the majority of
CRS theorists argue not for specific semantic
theories, but for a whole class of them.
Sellars, Harman, and Block provide only
frameworks for a future CRS, and some
arguments why this framework is the only
adequate one. A 'possible' exception is Field,
who has developed a more detailed account
in terms of subjective probability.
Field characterizes conceptual role in
terms of a subjective probability function
defined over all the sentences of a person's
language. It specifies a person's commit-
ments concerning how he will change
degrees of belief when he acquires new
information. The probability function, by
specifying inductive and deductive relations,
characterizes the conceptual roles of ex-
pressions. A and B are said to have the
same conceptual role if, and only if,
P(A/C) = P(B/C), for all sentences C in the
language. On this account 'Tully orates'
and 'Cicero orates' may have different con-
ceptual roles for a person, since there may
be an S for which P(,Tully orates' IS)
#- P(,Cicero orates'/S). The conceptual role
of non-sentential expressions is specified in
terms of the conceptual roles of all the sen-
tences in which it appears. There may be a
simple characterization of the conceptual
roles of some expressions. For example, the
role of negation is specified by the prob-
ability laws involving negation.
Field's account is but a 'possible' excep-
tion because Field writes that he is giving a
notion of sameness of meaning, not
meaning itself. And also, since his account
is based on subjective probabilities, what-
ever criterion of sameness of meaning he
comes up with is intrapersonal, not inter-
personal. He does not think it makes sense
to compare different individuals' conceptual
roles.
REASONS AGAINST CRS
Whatever version of CRS one adopts, one
must meet a couple of principled objections,
both external and internal.
Epistemological and Ontological Problems with
CRS
External criticisms of CRS impugn its con-
sequences for epistemology and ontology,
rather that its actual coherence. Regardless
of how one characterizes conceptual role, if
the meaning of an expression is (or super-
venes upon) its role in a language or a
mind, this invites the inference that expres-
sions that belong to different languages -
public or mentalese - are ipso facto different
in meaning. Once you start to identify the
meaning of an expression e with its con-
ceptual role in a language L, it's hard to
avoid (short of question begging) proceeding
to identify the meaning of e with its whole
role in L, and having gone that far it's hard
to avoid the conclusion that if languages
differ at all with respect to the propositions
they express, they differ completely with
respect to all the propositions they endorse
(Fodor and Lepore, 1992). And if languages
differ at all in respect of the propositions
they can express, then they differ entirely in
respect of the propositions they can express.
For example, Field's characterization of con-
ceptual role is holistic. In characterizing the
conceptual role of a sentence one must
simultaneously characterize the conceptual
roles of all other sentences. Any change in
the probability function - even just extend-
ing its domain to a new vocabulary -
results in a change in conceptual role for
every sentence (because two people will
seldom assign the same conceptual role to
expressions ).
Ultimately, it looks as if the CRS theorist
must wind up endorsing a number of
relativistic, idealistic and solipsistic con-
sequences: that no two people ever share a
belief; that there is no such relation as
translation; that no two people ever mean
the same thing by what they say; that no
two time slices of the same person ever
mean the same thing by what they say;
that no one can ever change his mind; and
that no two statements or beliefs can ever
be contradicted (to say nothing of refuted).
A number of philosophers, linguists and AI
theorists are prepared to bite these big
bullets.
CRS and Truth/Reference
Independently of these external problems,
there are also internal questions about
whether CRS can capture key aspects of
meaning. If you believe, as do many philo-
sophers, that meaning determines truth and
reference conditions, then meaning can't be
conceptual role alone (Field 1977; Lepore
and Loewer, 1987; Lycan 1984). If some-
one knows the meaning of a sentence and is
omniscient regarding physical facts (that is,
is omniscient about all the non-semantic/
197
non-intentional facts}, then he knows
whether the sentence is true. But Twin
Earth cases show that conceptual role does
not satisfy this condition. One can know the
conceptual role of 'this is water' and also
know just which things are H
2
0 and which
are XYZ without thereby knowing whether
'this is water' is true in a given context,
because conceptual role does not distinguish
water from XYZ.
Dual Aspect Semantics
A number of philosophers have responded
to this argument not by rejecting CRS in
toto but by constructing two-factor (also
called two-tiered and dual-aspect) semantic
theories (Block, 1986; Field, 1977; Lycan,
1984). According to these accounts, a
theory of meaning for a language L consists
of two distinct components. One com-
ponent, usually, a theory of truth, is inten-
ded to provide an account of the relations
between language and the world: truth,
reference, satisfaction, etc. The other,
usually CRS, is supposed to provide an
account of understanding and cognitive sig-
nificance, that is, whatever is entirely 'in
the head' (what Block calls 'narrow
meaning').
Gilbert Harman, as mentioned above,
advocates a semantic theory with only a
single factor, viz., conceptual role. Yet he
tries to avoid the criticism that an adequate
semantic theory must not ignore truth and
referential conditions by having his notion
of conceptual role 'reach out into the world
of referents'. Block speaks of Harman's con-
ceptual roles as 'long-armed', extending into
the world, in opposition to the 'short-
armed', stopping at the skin, conceptual
roles of two-factor theorists like himself
(Block, 1986, p. 636). Block argues that
Harman's non-solipsistic CRS is equivalent
to the two-factor account he endorses; that
difference between Harman's non-solipsistic
account and Block's two-factor account 'is
merely verbal'. Loar (1981) criticizes Har-
man's non-solipsistic account by arguing
that it is ad hoc.
The two-factor approach can be regarded
as making a conjunctive claim for each sen-
198
tence: what its conceptual role is and what
its (say) truth conditions are. Two-factor
theories of meaning incur the obligation of
saying which conceptual roles pair up with
which contextually determined aspects of
meaning, which truth conditions, to con-
stitute specific meanings. I know of no
reason to think that this obligation can be
met (see Fodor and Lepore, 1992, p. 170).
Still, it would seem that two-factor theories
are prey to the one external objection
above. On the two-factor view, if two sen-
tences differ in their conceptual roles, then
they differ in the propositions they each
express. Since conceptual role is holistic (on
the two-factor view), it follows that if two
sentences differ in conceptual roles in differ-
ent languages, then no two sentences in the
two languages will ever express the same
proposition.
CRS and Compositionality
Another internal problem is the following
(Fodor and Lepore, 1992): you can't iden-
tify meanings with conceptual roles tout
court, since unlike meanings conceptual
roles tout court aren't compositional. That is,
the meanings of syntactically complex
expression (for example, mental sentences)
are a function of their syntactic structures
and the meanings of their lexical con-
stituents. Some inferences, for example, the
inference from 'that's a rattling snake' to
'that's rattling', is compositional in the
sense that it follows just from the linguistic
principles which connect the meanings of
syntactically complex English expressions
with the meanings of their syntactic con-
stituents. However, the thought 'that's a
rattling snake' licenses the inference 'that's
dangerous'. Intuitively, that it does depends
not just on the meaning of 'rattling' and
'snake' but also on a (presumed) fact about
the world; viz., the fact that rattling snakes
are dangerous. So, this inference is non-
compositional (as are, by parity of argu-
ment, all other synthetic inferences). In
short, it appears that some, but not all, of
the inferential potential of 'that's a rattling
snake' (some of its role in one's language or
belief system), is determined by the respect-
ive inferential potential of 'rattling' and
'snake: the rest being determined by one's
'real world' beliefs about rattling snakes.
How bad is this?
Productivity is the thesis that every lan-
guage. natural or mental, can express an
open-ended set of propositions. And sys-
tematicity the thesis that every language,
natural or mental, that can express the pro-
position that P will also be able to express
many propositions that are semantically
close to P. If, for example, a language can
express the proposition that aRb, then it
can express the proposition that bRa; if
it can express the proposition that P -+ Q,
then it can express the proposition that
Q -+ P, and so forth (see Fodor and Lepore,
1992, ch. 6). Ifcompositionality is the only
available explanation of productivity, and
these are both pervasive features of natural
languages and of the mental life, then you
cannot identify meaning with conceptual
role (Fodor and Lepore, 1992, pp. 175-6).
You still might identify meanings with con-
ceptual roles in (non-structural) analytic
inferences; viz., inferences that you must
accept if you mean F by (the expression) 'F'.
The contrast between structural and non-
structural analyticities is illustrated by the
following examples respectively: 'rattling
snake' -+ 'rattling' and 'rattling snake' -+
'rattling reptile'. The former inference is
guaranteed by the linguistic principles that
effect the construction 'rattling snake'
alone, whereas the latter turns on the
lexical inventory of its premise as opposed
to its linguistic structure.
Let's suppose that just these inferences are
meaning determinate; the others semant-
ically irrelevant. So, for example, though the
inference from 'that's a rattling snake' to
'that's dangerous' is not meaning deter-
minate, the non-structural inference from
'that's a snake' to 'that's a reptile' is mean-
ing determinate, that is. is analytic. is defi-
nitional.
In the context of characterizing the
nature of meaning. it may be question
begging to invoke the notion of an 'analytic'
inference. But also. obviously. the cost of
identifying meanings with roles in these so-
called analytic inferences is to buy into the
analytic/synthetic distinction. So the cost of
CRS is buying into the analytic/synthetic
distinction. But many philosophers. if not
most. think that there is no notion of
analyticity that a CRS theorist. qua CRS
theorist. can legitimately invoke to preserve
compositionality. If. then. you believe that
compositionality is non-negotiable. it
follows that either the analytic/synthetic
distinction is in fact principled. contrary to
what Quine is thought to have established.
or CRS is internally flawed.
In conclusion: there is ample reason for
more attention to detailed versions of CRS;
however. before evaluating the prospects of
these prospective CRS theories. some serious
objections require response.
See also CONCEPTS; CONTENT; FUNCTIONAL-
ISM; THOUGHT; THOUGHT AND LANGUAGE.
BIBLIOGRAPHY
Block. N. 1986. Advertisement for a semantics
for psychology. In Midwest Studies in Philo-
sophy. vol. 10. Minneapolis: University of
Minnesota Press.
de Saussure. F. 1983. Course in General Lin-
guistics. ed. C. Bally and A. Sechehaye. tr.
and annotated by R. Harris. London: Duck-
worth.
Devitt, M. 1981. Designation. New York:
Columbia University Press.
Field. H. 1977. Logic, meaning and con-
ceptual role. Journal of Philosophy. 69. 379-
408.
Fodor. J .. and Lepore, E. 1992. Holism: A shop-
per's guide. Oxford: Basil Blackwell.
Harman. G. 1987. (Non-solipsistic) conceptual
role semantics. In New Directions in Seman-
tics. ed. Lepore. E. London: Academic Press.
Lepore. E.. and Loewer, B. 1987. Dual aspect
semantics. In New Directions in Semantics.
ed. E. Lepore. London: Academic Press.
Loar. B. 1981. Mind and Meaning. Cambridge
University Press.
Lycan. W. 1984. Logical Form in Natural Lan-
guage. Cambridge, MA. MIT Press.
Putnam. H. 1975. The meaning of'meaning'.
In Mind, Language and Reality. ed. H.
Putnam. Cambridge University Press.
Sellars. W. 1963. Some reflections on lan-
199
CONNECTIONISM
guage games. In Science. Perception. and
Reality. London: Routledge & Kegan Pau!'
An article based on Sellar's reading of Witt-
genstein's Philosophical Investigations.
Woods. W. 1981. Procedural semantics as a
theory of meaning. In Elements of Discourse
Understanding. ed. A. Joshi. B. Webber and I.
Say. Cambridge University Press.
What preceded benefited from suggestions by
Johannes Brandl. Jerry Fodor and Sam Gutten-
plan.
ERNIE LEPORE
connectionism Theorists seeking to
account for the mind's activities have long
sought analogues to the mind. In modern
cognitive science, these analogues have pro-
vided the bases for simulation or modelling
of cognitive performance (see COGNITIVE
PSYCHOLOGY). Simulation is one way of
testing theories of the mind: if a simulation
performs in a manner comparable to the
mind, that offers support for the theory
underlying that simulation. The analogue
upon which the simulation is based,
however, also serves a heuristic function,
suggesting ways in which the mind might
operate.
In cognitive science, two analogues have
provided the basis for most of the simulation
activity. On the one hand, the digital com-
puter can be used to manipulate symbols;
insofar as it became possible to program the
symbol processing computer to execute
tasks that seemed to require intelligence.
the symbol processing computer became a
plausible analogue to the mind and numer-
ous cognitive science theorists have been
attracted to the proposal that the mind itself
is a symbol processing device (see COMPUT A-
TIONAL MODELS). The other analogue was
the brain. As techniques for analysing the
anatomy and physiology of the brain were
developed in the first half of the twentieth
century. the view that the brain consisted of
a network of simple electrical processing
units which stimulated and inhibited each
other became popular. Researchers such as
McCulloch and Pitts (1943) began to
analyse how networks built out of such pro-
200
cessing units could perform computations
such as those of sentential logic. Other
researchers. such as Rosenblat (1962) and
Selfridge (1959). explored the usefulness of
networks in more perceptually oriented
tasks.
The second approach came to be known
as connectionism. although some theorists.
often those coming from the neurosciences.
prefer the term neural networks and others
prefer the term parallel distributed processing.
While at the outset connectionism was a
serious competitor to the symbol processing
approach to simulating cognition. by the
time cognitive science began to take shape
as a multidisciplinary research cluster in the
1970s. the connectionist approach had lost
much of its appeal. In part this was due to
the success of symbol processing approaches
in developing plausible models of perfor-
mance on higher cognitive tasks (Newell
and Simon. 1972; J. R. Anderson. 1976). In
part it was due to the apparent limitations of
networks; Minsky and Papert's (1969)
mathematical analysis of Rosenblat's per-
ceptrons. for example. was often taken as
showing the inherent limitations of network
approaches. As a result. use of networks to
simulate cognitive activities did not expand
during the 1970s. while simulations within
the symbol processing tradition mush-
roomed. There were. however. some ex-
tremely important theorists who persevered
during this time: J. A. Anderson (1972).
Grossberg (1982). and Kohonen (1972).
Interest in connectionism was rekindled in
the 1980s (two influential publications were
Hinton and Anderson. 1981 and McClelland
and Rumelhart. 1981). In part this new
interest stemmed from growing frustration
with the dominant symbol processing
approach. It was further due to advances in
network design. some stemming from the
influential physicist John Hopfield (1982.
1984). others from psychologists David
Rumelhart. Jay McClelland. and their collea-
gues (see Rumelhart. Hinton. and Williams.
1986). The latter collaborators. then both at
the University of California San Diego.
developed an influential research group (the
PDP Research Group); publication of their
two-volume work, Parallel Distributed Proces-
sing: Explorations in the Microstructure of
Cognition, was one of the galvanizing
developments in the emergence of the new
connectionist movement. By 1990 connec-
tionism had once again emerged as a cred-
ible competitor to the symbol processing
approach to simulating cognition.
THE MECHANICS OF CONNECTIONIST
MODELS
To see how connectionist networks operate,
let us begin with a relatively simple network
which nonetheless simulates an interesting
cognitive task: generating the proper pho-
nemic representation of the past tense of
CONNECTIONISM
English verbs. This task is challenging since,
although most verbs form their past tense
in a regular manner by adding either ledl
(add-+ added) , Idl (play-+played), or It I
(walk-+walked), some verbs form their past
tense in an irregular fashion (is-+was).
Rumelhart and McClelland (1986) simu-
lated this task with a network of only two
layers, an input layer and an output layer,
with connections going from each input
unit to each output unit. While their
network exhibited some significant features
of the pattern exhibited by humans learning
the English past tense, it was severely criti-
cized by Pinker and Prince (1988). (See
Bechtel and Abrahamsen, 1991. for discus-
sion and evaluation of that controversy.)
OUTPUT LAYER: PHONOLOGICAL FEATURE REPRESENTATION OF PAST TENSE FORM
First Phoneme Second Phoneme Third Phoneme
I clv voi manner place I 1 c/v voi manner place II c/v voi manner place II
... 1 C_/V __ vo_i __ m_a_n_n_e_r _p_la_c_e-J1 ... 1 C_/V_V_O_i _m_a_nn_e_r __ p_la_c_e-J11 c/v voi
manner place
First Phoneme Second Phoneme Third Phoneme
INPUT LAYER: PHONOLOGICAL FEATURE REPRESENTATION OF VERB STEM
C)
c
'00
(/)
Q)
u
e
0..
Suffix
Figure 1. Three-layer feedforward network used in Plunkett and Marchman's (1991) simulation
of learning the past tense of English verbs. The input units encode representations of the three
phonemes of the present tense of the artificial words used in this simulation, while the output
units encode the three phonemes employed in the past tense form and the suffix (/d/, ledl, or It/)
used on regular verbs. Each input is connected to each of the 30 hidden units and so spreads acti-
vation to those units in proportion to its activation and the weight on the connection. Likewise,
all hidden units are connected to every output unit.
201
CONNECTIONISM
Subsequently. Plunkett and Marchman
(1991) developed an alternative simulation
that addressed many of the objections to the
earlier work and provides a clear example of
how multi-layer feedforward networks
operate. They developed an artificial vocab-
ulary of 500 three-phoneme words. Most of
the vocabulary items formed their past tense
in one of the three regular ways. Some.
however. were assigned irregular forms
(either arbitrary mappings such as found in
the English go-+went. identity mappings as
in hit-+hit. and vowel change mappings as
in run -+ ran.
In the network simulation each phoneme
is represented on 6 of the 18 input units;
these units encode whether each phoneme
is a vowel or consonant. is voiced or
unvoiced. as well as the manner and place
of articulation (figure 1). From these input
units activation is passed along weighted
connections to 30 hidden units. and from
them to 20 output units. On the output
units the network generates a representa-
tion of the past tense of the verb supplied on
the input units; 18 of the output units serve
Netinput to Unit 4:
(1.5) (1) + (0) (1) + (0.5) (-1)
= 1.0
Weight
= 1.5
the same function as they do on the input
layer. while the remaining two units encode
the possible regular endings.
The ability of the network to determine
the correct past tense is governed by the
weights on the connections. These weights
determine how much a unit on one layer
excites or inhibits a unit on a subsequent
layer. For example. the input from a given
input unit to a hidden unit is obtained by
multiplying the activation of the input unit
by the weight of the connection leading to
the hidden unit; the corresponding values
from all the input units are summed to
determine the netinput to the hidden unit
(figure 2). The hidden unit's activation is
then determined from the netinput by using
the logistic activation function. The process
is repeated to determine the activations of
the output units.
One of the attractions of connectionist
networks is that there are procedures that a
network can use for determining the appro-
priate weights on the various connections
in it so that these connections do not have
to be hand set. These procedures are known
Activation of Unit 4 =
1/1 + e-netinput
= 0.73
Weight
= 0.5
Figure 2. An illustration of processing in a connectionist network. The activation levels of the four
units are shown beneath their labels. The weights on the three connections leading to Unit 4 are
also shown. The netinput to Unit 4 is determined by multiplying the activation of each feeding
unit by the weight on the connection and summing across the three feeding units. The activation
of Unit 4 is then determined according to the logistic activation function.
202
as learning rules; a general principle applied
to learning rules is that they must invoke
only information locally available at the
units so that application of a learning rule
does not require an external homunculus
overseeing the behaviour of the network.
Plunkett and Marchman employed one of
the most commonly used learning rules.
backpropagation (Rumelhart. et al.. 1986).
To apply this rule. the network starts with a
set of random weight assignments. is sup-
plied with the phonemic representation of a
verb on its input layer. and is allowed to
compute the values for the various output
units. The activation generated on a given
output unit is compared with the target
activation for that unit: the difference con-
stitutes a measure of error. The derivative of
the error with respect to the activation of
the unit is then used to guide changing the
weights on the connections feeding into
that unit so as to generate weights that will
reduce the error in the future; this proce-
dure is recursively applied to connections at
lower levels in the network.
Plunkett and Marchman demonstrated
not only that their network could learn to
generate the proper past tense for the words
in their vocabulary. but also that the learn-
ing showed some important similarities to
the way children learn the English past
tense. The network was initially trained on
a set that included a large number of irreg-
ular words. Then additional words. includ-
ing a smaller percentage of irregulars. was
gradually added to the training set. Like
children. the network first learned the
correct form of the irregulars in its corpus.
but subsequently tended to overgeneralize
the regular form and applied it to some of
the irregular forms (thus producing. using
an English example. corned instead of came).
before learning the correct past-tense of all
verbs in its corpus.
A network such as this is considered feed-
forward because connections carry activa-
tions in only one direction. from the layer of
input units to the layer of output units.
Connectionists have also explored a variety
of network designs. An important kind of
design employs inhibitions (connections
CONNECTIONISM
with negative weights) between units in a
layer so that. while processing cycles
between units of the layer. one unit or
group of units becomes active and sup-
presses the activation of all the other units
(Kohonen. 1988). Another strategy is to
permit activations to flow in both directions
between layers. allowing units in later units
that become active to further excite those
input units that have excited them (McClel-
land and Rumelhart, 1981). An additional
strategy is to do away with the notion of
layers and allow connections in both direc-
tions between units (Hopfield. 1982;
Ackley. Hinton. and Sejnowski. 1985).
As exciting as some of these alternative
designs are. feedforward networks continue
to be the most widely used. perhaps because
it is far more easy for researchers to keep
track of and interpret processing in them.
However. researchers also recognize that in
order for connectionist networks to simulate
interesting cognitive tasks. more structured
networks are required. Some interesting
variations in the standard feedforward net-
work have appeared in recent years (these
are described more fully in Bechtel. 1993).
One strategy has been to modularize net-
works by allowing. for example. different
networks (often with different numbers of
layers and units) to process the same input
(see MODULARITY). In this design. a further
network serves as a gating network. deter-
mining which network is permitted to
respond to a given input. Jacobs. Jordan.
Nowlan. and Hinton (1991) have shown
that the gating network can be trained by
similar algorithms as used for the other net-
works to select the network that has given
the best answers to particular classes of
input. The result is a set of expert networks.
each trained to respond to particular input
sets. By allowing separate experts to handle
different tasks. modularization overcomes
one of the most severe problems confronting
standard feedforward networks. that later
learning can cause catastrophic interference
with material learned earlier (McCloskey
and Cohen. 1989).
Another limitation of feedforward net-
works is that all information to which the
203
CONNECTIONISM
network is to respond must be presented at
once on the input units; the network has no
memory of what has been recently pro-
cessed. This makes it very difficult for a
network to process serially encoded infor-
mation such as that of natural language.
Elman (1990). adapting a design of Jordan
(1986). has proposed' an alternative
approach. In what he calls recurrent net-
works. the activations produced on hidden
units on previous cycles of processing are
copied back onto a set of special input units
called context units. These are fed into the
network along with new input patterns on
subsequent cycles of processing. Elman has
shown that such networks can become sen-
sitive to such things as syntactical depen-
dencies in language. The recurrent network
design has been employed in a more elab-
orate network by st. John and McClelland
(1990) which has demonstrated remarkable
abilities to arrive at semantic interpretations
of a significant variety of active and passive
English sentences.
CONNECTIONIST COGNITIVE
ARCHITECTURES: PRO AND CON
On the basis of this brief overview. let us
turn now to the usefulness of connectionist
networks for simulating human perform-
ance. Perhaps the first thing to note about
connectionist systems is that they point to a
very different conception of cognition than
do symbolic simulations. The symbol strings
that are stored and manipulated in symbolic
simulations are often propositions and the
rules for manipulating are inspired by
symbolic logic. The result is that symbolic
simulations have proven quite useful in
modelling cognitive performance that seems
to involve logical or quasi-logical reasoning.
Connectionist simulations. on the other
hand. operate on very different principles.
In feedforward networks. the connections
allow the network to respond to patterns
activated on the input units by activating
patterns on the output units. Networks can
be trained so that a variety of different
patterns all result in activation of the same
output pattern. while other collections of
204
patterns generate different output patterns.
As a result. such networks can readily be
interpreted as performing activities of
pattern recognition and categorization.
These networks. in fact. exhibit some rather
useful properties for pattern recognition; for
example. patterns similar to but not ident-
ical to those on which the network was
trained will generally result in generation of
output patterns similar to those used in
training.
For many cognitive scientists. and espe-
cially for philosophers (see Rosenberg.
1990). however. the term cognition refers
not to such basic processes as pattern
recognition. but to processes involving rea-
soning (see RATIONALITY). SO far. connec-
tionist networks have proved less adept at
modelling reasoning tasks (except for ones
involving satisfaction of multiple soft-
constraints. which is a strength of connec-
tionist systems). Thus. one might be inclined
to dismiss connectionism as irrelevant to
cognition proper. In the face of this sort of
objection. two sorts of responses can be
advanced. First. it appears possible to train
connectionist networks to use their pattern
recognition capacities to carry out logical
operations and even construct natural
deductions (Bechtel and Abrahamsen. 1991;
see also Touretzky. 1990). thereby showing
how to ground symbol processing on con-
nectionist pattern recognition. Second. one
might construe it as an advance to refocus
attention in cognitive science away from
reasoning to these more basic cognitive pro-
cesses that have been overlooked. Two of the
most important aspects of our lives are our
ability to navigate our world and to identify
objects in it. This requires recognizing
patterns in our environment and inaugurat-
ing patterns of response. These are abilities
we share with many other creatures. But
these are precisely the abilities that have
been most troubling for investigators inter-
ested in robotics. In part. the difficulty in
accounting for these processes may be that
they require pattern recognition and this
process needs to be taken more seriously.
The idea that pattern recognition and
categorization are basic to cognition has
been advanced by a number of theorists.
Margolis (1987) argues that thinking is
simply a process of pattern recognition;
sequential thinking involves recognizing
one pattern and using that recognition as
part of the input for recognizing another
pattern. Dreyfus and Dreyfus (1986)
contend that skilled or expert performance
in a domain such as chess rests on 'holistic
discrimination and association'; one recog-
nizes a situation to be like a previous situa-
tion and applies similar strategies to the
new situation as worked in the old one.
Paul Churchland (1989), drawing on the
notion of prototype which has become
central in recent psychological research on
categorization, argues that the most basic
form of cognitive life consists in activating
perceptual prototypes and transforming
them into motor prototypes.
Van Gelder (1992) has challenged this
proposal to reinterpret cognition in terms of
pattern recognition. Claiming that what is
central to pattern recognition or categoriza-
tion is putting items into similarity classes
to which like responses can be given, van
Gelder argues that in at least two domains,
motor control and language comprehen-
sion, what is required of a cognitive system
is continuously varying responses, not dis-
crete responses, and the capacity to recog-
nize new inputs as distinctive, not instances
of previously encountered categories. Each
time we reach for an object, we typically
must reach to a new location, and each
sentence we hear is generally different than
those encountered previously. There is some
irony in van Gelder's objection: in order to
get networks to give categorical responses,
special strategies need to be employed;
otherwise, networks tend to give similar but
slightly varying responses to new inputs
that are similar to previous inputs. More-
over, research since Rosch (1975; see Bar-
salou, 1992, for a review) shows that
people readily make prototypicality judg-
ments when assigning items to categories,
suggesting that categorization is not an all-
or-nothing matter. Perhaps, then, van
Gelder is employing a common but inap-
propriate concept of categorization.
CONNECTIONISM
However, the broader thrust of van Gel-
der's complaint is clearly correct: categor-
ization cannot be all there is to cognition.
Processing novel sentences is not simply a
matter of responding as one did to the most
similar sentence in the training set. At
minimum, the cognizer must be sensitive to
the grammatical structure of sentences and
the constraints these place on interpreta-
tion. One way more complex processing is
achieved in connectionist networks is to
employ multiple layers of units so that each
layer can recategorize the outputs of cat-
egorization by lower layers. This allows
connectionist networks to accomplish more
than simple association. But is it enough?
Fodor and Pylyshyn (1988; see also Fodor
and McLaughlin, 1990) argue that it is not.
They appeal to properties of language such
as productivity (the fact that additional sen-
tences can always be added to the corpus of
a language) and systematicity (the fact that
for any sentence of a language there are
systematically related expressions that are
also sentences of the language) and argue
that these are also characteristics of
thought (see FODOR). Thus, focusing on sys-
tematicity, they argue that any cognitive
system that can comprehend a sentence
such as
Joan loves the florist
can also understand the systematically
related sentence
The florist loves Joan.
They argue that classical symbol systems,
like languages, naturally exhibit systemati-
city since they employ operations upon
symbolic representations which are built up
via a compositional syntax that supports
compositional semantics. Connectionist
architectures do not employ symbolic repre-
sentations with a compositional syntax
(although they can be used to implement
such systems) and hence are themselves
inadequate as cognitive models.
Fodor and Pylyshyn's critique has pro-
vided a host of responses by advocates of
205
CONNECTIONISM
connectionism. One is to deny that most
thought, especially infra-human thought, is
systematic (Bechtel, 1990; Dennett, 1991;
Groschke and Koppelberg, 1991). While
this might reduce the challenge facing con-
nectionists, since it allows that much cogni-
tion might have a different character than
Fodor and Pylyshyn assume, connectionists
must still explain how adult human cogni-
tion, at least, often does manifest systemati-
city. Some connectionists have attempted to
demonstrate ways in which connectionist
networks, without employing explicitly syn-
tactically compositional representations
such as those found in natural language,
can employ functionally compositional
representations, that is, representations
from which component structure can be
extracted but is not present on the surface
(van Gelder, 1990). Two of the best known
examples of this approach are the tensor
product networks of Smolensky (1990) and
the RAAM (recursive auto-associative
memory) networks of Pollack (1988). The
claim of those pursuing these models is that
while connectionists may have to employ
structured internal representations, the
structuring principles might be fundamen-
tally different from those employed in
natural language or symbolic models of
cognition. Yet a different strategy is to
argue that the systematicity that does
appear in human cognition is due to the
fact that humans have learned natural lan-
guages, which themselves employ a compo-
sitional syntax and semantics (Clark, 1989;
Bechtel and Abrahamsen, 1991; Dennett,
1991; see Clark and Karmiloff-Smith (1993)
for a critical response). On this view, net-
works must become sensitive to the struc-
ture of language so as to be able to extract
information from linguistic inputs and
produce linguistic outputs, but this does not
entail building up an internal replica of the
linguistic structure for use in internal pro-
cessing. Simulations such as those of St.
John and McClelland (1990) are suggestive
as to how this might be accomplished.
It is unclear as yet whether any of these
approaches are sufficient to account for the
kind of cognition found in adult humans. It
206
may be that ultimately some version of the
classical symbolic approach will be required.
Some connectionists are disposed to such a
possibility and are developing hybrid models
that employ both connectionist principles
and symbolic processing principles (Dyer,
1991; Sun, 1991). Even if a hybrid
approach is required, though, the connec-
tionist elements in such models serve to
extend the conception of cognition beyond
the traditional perspective of logical reason-
ing. In helping create this expansion of our
understanding of the cognitive, connection-
ism has thus made a contribution to cogni-
tive science.
PHILOSOPHICAL IMPLICATIONS OF
CONNECTIONISM
Insofar as it expands or transforms our
understanding of cognition, connectionism
will necessarily have implications for philo-
sophy of mind. Two areas in particular on
which it is likely to have impact are the
analysis of the mind as a representational
system and the analysis of intentional
idioms (see CONTENT; INTENTIONALITY;
REPRESENT ATION).
Fodor (1980) distinguishes the computa-
tional theory of mind from the representa-
tional theory of mind. The representational
theory holds that systems have mental
states by virtue of encoding representations
and standing in particular relations to
them. The computational theory adds that
cognitive activity consists of formal opera-
tions performed on these representations.
Fodor and Pylyshyn's arguments against
connectionism noted above fault it for
failing to endorse the computational theory
(connectionist processing does not respect
the formal, compositional structure of repre-
sentations). But Fodor and Pylyshyn con-
strue connectionist models as nonetheless
representational and so potentially conform-
ing to the representational theory of mind.
This is because connectionists routinely
interpret the activations of units or groups
of units as representing contents. This is
most obviously the case for input and
output units; in order to supply a cognitive
interpretation of a network's activity, a the-
orist must treat the input as a representa-
tion of a problem and the output as
representing the answer. Connectionists
also tend to interpret the activations of units
within a network. Sometimes this is done
unit by unit: a given unit is found to be
activated by inputs with certain features
and so is interpreted as representing those
features. Other times a more holistic analy-
sis is used: a cluster analysis may reveal
that similar patterns of activation are gener-
ated by common features in the input. This
is interpreted as showing that the network
has differentiated inputs with those features
from inputs with different features. This
suggests that connectionist systems can
indeed be understood as embodying the
representational theory of mind: each layer
of units in a network generates a different
representation of the input information
until the output pattern is produced.
Even if connectionist networks exemplify
the representational theory of mind, they
are significantly different from more tradi-
tional exemplars of the representational
theory. First, it is not clear that we can
always give an interpretation of what units
in a connectionist network are representing
in natural language terms. Even when it
appears that we can do so, there seems to
be considerable noise in the representation
so that the units or patterns do not seem to
be picking out precisely what we designate
in natural languages. Second, the repre-
sentations that are constructed are not dis-
crete but distributed or superimposed (van
Gelder, 1991). That is, the same units and
same connections sub serve many different
representational roles rather than employ-
ing one representation per role. This distin-
guishes connectionist representations from
those humans have previously designed.
Third, it needs to be emphasized that the
patterns of activations on hidden units in
connectionist systems are the product of the
learning the system has undergone. The
interpretations assigned to these units are
not arbitrary, as they often seem to be in
symbolic systems, but are analyses of how
the network has solved the problem it was
CONNECTIONISM
confronting. Especially in cases in which the
network is connected to real sensory inputs,
and not supplied inputs by the modeller, the
intentionality of these representations is
genuine, not merely a product of the the-
orist's interpretation.
The intentional representations found
within connectionist networks are typically
not representations of propositions. When
philosophers focus on the intentional states
of cognitive systems, however, they tend to
focus on states characterized in terms of
PROPOSITION AL ATTITUDES (e.g. the belief
that the storm will soon cease). Proposi-
tional attitudes comprise the core of what
philosophers call 'POLK PSYCHOLOGY', since
it is in terms of propositional attitudes that
ordinary people typically characterize their
own and other peoples' mental states. A
number of philosophers attracted to con-
nectionism have argued that connection-
ism, if correct, would show that folk
psychology is wrong (this claim is often
referred to as 'ELIMINATIVISM'). The reason
for this is that, due to the distributed and
superpositionaI nature of connectionist
representations, connectionist models do
not seem to have internal states that could
be discretely identified as particular inten-
tional states (Churchland, 1989). Ramsey,
Stich, and Garon (1991), for example,
argue that in a network they trained to
report the truth values of 16 propositions
there are not functionally discrete, semant-
ically interpretable states that can play dis-
tinct causal roles. Whereas according to folk
psychology we may claim that a person did
something for one reason and not another,
we cannot do so with a connectionist
network since all units and connections are
always involved. Thus, they claim that if
connectionism is true, folk psychology must
be false and should be eliminated.
Not all philosophers have concurred with
this judgment of the relation of connection-
ism to folk psychology. Lycan (1991), for
example, argues that even if connectionist
networks only employ units representing
microfeatures of what is represented at the
propositional attitude level, propositional
representations and consequently proposi-
207
CONNECTIONISM
tional attitudes may supervene upon these.
Even though the presence or absence of a
representation of a proposition about which
we may have a belief may not directly
figure in the causal machinery of the
system. the differences between a system on
which that propositional representation
supervenes and one on which it does not
will suffice to make the relatively imprecise
causal claims we commonly make in the
folk idiom. Another approach. largely com-
patible with Lycans. is inspired by
DENNETT'S (1978) reasons for instrumental-
ism about intentional attitudes. While
Dennett acknowledges that intentional atti-
tudes provide important information about
cognitive systems. they do not pick out
internal states of those systems. This
permits one to advance a realist interpreta-
tion of intentional attitudes as long as such
attitudes are assigned to cognitive agents.
not their sub-personal operations. The inter-
nal operations must make the whole system
adhere to the folk psychological character-
ization. but the folk vocabulary need not
refer to discrete internal states. On this con-
strual as well. connectionism offers no
threat to folk psychology (Bechtel and Abra-
hamsen. 1993).
Whether or not connectionism undercuts
folk psychology. however. it clearly chal-
lenges any philosophical view that assumes
that cognition is primarily a matter of
storing propositions and processing them
according to rules of logic. If. for example.
an epistemology construed knowledge as
involving the internal representation of a
proposition and the justificatory argument
for its truth. that epistemology would not be
compatible with connectionism. However. if
all that is required is the capacity to produce
a proposition and. if required. to identify evi-
dence for it. then a linguistically trained
connectionist network is subject to episte-
mological analysis. However. connectionism
might also have further implications for
epistemology. It might. for instance. inspire
us to consider evidential relations less strong
than logical argument and the possibility
that knowledge might be exhibited in abil-
ities to act (what Ryle referred to as knowing
208
how) as well as in propositions that can be
recited. Thus. connectionism might also
broaden our perspective not just in our the-
ories of mind. but also in our theories in
domains such as epistemology.
BIBLIOGRAPHY
Ackley. D.H .. Hinton. G.E .. and Sejnowski, T.J.
1985. A learning algorithm for Boltzmann
machines. Cognitive Science. 9. 147-69.
Anderson. J.A. 1972. A simple neural network
generating an interactive memory. Math-
ematical Biosciences. 14. 197-220.
Anderson. J.R. 1976. Language. Memory. and
Thought. Hillsdale. NJ: Lawrence Erlbaum.
Barsalou. L.W. 1992. Cognitive Psychology: An
Overview for Cognitive Science. Hillsdale. NJ:
Lawrence Erlbaum.
Bechtel, W. 1990. Multiple levels of inquiry in
cognitive science. Psychological Research. 52.
271-81.
Bechtel. W. 1993. Currents in connectionism.
Minds and Machines 3.
Bechtel. W .. and Abrahamsen. A. 1991. Con-
nectionism and the Mind. An Introduction to
Parallel Processing in Networks. Oxford: Basil
Blackwell.
Bechtel. W .. and Abrahamsen. A. 1993. Con-
nectionism and the future of folk psychol-
ogy. Minds: Natural and Artificial. ed. R.
Burton. Albany. NY: SUNY Press.
Churchland. P.M. 1989. A Neurocomputational
Perspective. The Nature of Mind and the Struc-
ture of Science. Cambridge. MA: MIT Press.
Clark. A. 1989. Microcognition: Philosophy.
Cognitive Science. and Parallel Distributed Pro-
cessing. Cambridge. MA: MIT Press.
Clark. A .. and Karmiloff-Smith. A. 1993. The
cognizer's innards: A philosophical and
developmental perspective on human
thought. Mind and Language. 6. 487-579.
Dennett. D.C. 1978. Brainstorms. Cambridge.
MA: MIT Press.
Dennett. D.C. 1991. Mother nature versus the
walking encyclopedia. In Philosophy and
Connectionist Theory. ed. W. Ramsey. S. P.
Stich. and D. E. Rumelhart. Hillsdale. NJ:
Lawrence Erlbaum.
Dreyfus. H.L.. and Dreyfus. S.E. 1986. Mind
over Machine: The Power of Human Intuition
and Expertise in the Era of the Computer. New
York: Harper & Row.
Dyer, M. 1991. Symbolic neuroengineering for
natural language processing: A multilevel
research approach. In Advances in Connec-
tionist and Neural Computational Theory, ed. J.
Barnden and J. Pollack. Norwood, NJ:
Ablex.
Elman, J.L. 1990. Finding structure in time.
Cognitive Science, 14, 179-212.
Fodor, J.A. 1980. Methodological solipsism
considered as a research tradition in
cognitive psychology. Behavioral and Brain
Sciences, 3, 63-73.
Fodor, J.A., and McLaughlin, B.P. 1990. Con-
nectionism and the problem of systemati-
city: Why Smolensky's solution doesn't
work. Cognition, 35, 183-204.
Fodor, J.A., and Pylyshyn, Z.W. 1988. Con-
nectionism and cognitive architecture: A
critical analysis. Cognition, 28, 3-7l.
Groschke, T., and Koppelberg, D. 1991. The
concept of representation and the repre-
sentation of concepts in connectionist
models. In Philosophy and Connectionist
Theory, ed. W. Ramsey, S. P. Stich, and
D. E. Rumelhart. Hillsdale, NJ: Lawrence
Erlbaum.
Grossberg, S. 1982. Studies of Mind and Belief.
Dordrecht: Reidel.
Hinton, G.E., and Anderson, J.A. 1981. Paral-
lel Models of Associative Memory. Hillsdale,
NJ: Erlbaum.
Hopfield, J.J. 1982. Neural networks and
physical systems with emergent collective
computational abilities. Proceedings of the
National Academy of Sciences, 79,2554-8.
Hopfield, J.J. 1984. Neurons with graded
response have collective computational
properties like those of two state-neurons.
Proceedings of the National Academy of
Sciences, 81, 3088-92.
Jacobs, R.A., Jordan, M.l., Nowlan, S.J. and
Hinton, G.E. 1991. Adaptive mixtures of
local experts. Neural Computation, 3, 79-87.
Jordan, M. 1986. Attractor dynamics and
parallelism in a connectionist sequential
machine. Proceedings of the Eighth Annual
Conference of the Cognitive Science Society.
Hillsdale, NJ: Lawrence Erlbaum.
Kohonen, T. 1972. Correlation matrix mem-
ories. IEEE Transactions on Computers C-21
353-9.
Kohonen, T. 1988. Self Organization and Asso-
ciative Memory. New York: Springer-Verlag.
CONNECTIONISM
Lycan, W.G. 1991. Homuncular functionalism
meets PDP. In Philosophy and Connectionist
Theory, ed. W. Ramsey, S. P. Stich, and
D. E. Rumelhart. Hillsdale, NJ: Lawrence
Erlbaum.
Margolis, H. 1987. Patterns, Thinking, and Cog-
nition. University of Chicago Press.
McClelland, J.L., and Rumelhart, D.E. 1981.
An interactive activation model of context
effects in letter perception: Part 1. An
account of basic findings. Psychological
Review, 88, 375-407.
McClelland, J.L. Rumelhart, D.E., and the PDP
Research Group 1986. Parallel Distributed
Processing: Explorations in the Microstructure
of Cognition, Vol. 2: Psychological and Biologi-
cal Models. Cambridge, MA.: MIT Press.
McCloskey, M., and Cohen, N.J. 1989. Cata-
strophic interference in connectionist net-
works: The sequential learning problem. In
The Psychology of Learning and Motivation,
vol. 24, ed. G. H. Bower. New York: Aca-
demic Press.
McCulloch, W.S., and Pitts, W.H. 1943. A
logical calculus of the ideas immanent in
nervous activity. Bulletin of Mathematical
Biophysics, 5, 115-33.
Minsky, M.A., and Papert, S. 1969. Percep-
tions. Cambridge, MA: MIT Press.
Newell, A., and Simon, H.A. 1972. Human
Problem Solving. Englewood Cliffs, NJ: Pren-
tice Hall.
Pinker, S., and Prince, A. 1988. On language
and connectionism: Analysis of a parallel
distributed processing model of language
acquisition. Cognition, 28, 73-193.
Plunkett, K., and Marchman, V. 1991 U-
shaped learning and frequency effects in a
multi-layered perception: Implications for
child language acquisition. Cognition, 38, 1-
60.
Pollack, J. 1988. Recursive auto-associative
memory: Devising compositional distributed
representations. Proceedings of the 10th
Annual Conference of the Cognitive Science
Society. Hillsdale, NJ.: Erlbaum.
Ramsey, W., Stich, S.P., and Garon, J. 1991.
Connectionism, eliminativism, and the
future of folk psychology. Philosophy and
Connectionist Theory, ed. W. Ramsey, S. P.
Stich, and D. E. Rumelhart. Hillsdale, NJ:
Lawrence Erlbaum.
Rosenberg, J.F. 1990. Treating connectionism
209
CONSCIOUSNESS
properly: Reflections on Smolensky. Psycho-
logical Research, 52, 163-74.
Rosenblat, P. 1962. Principles of Neurody-
namics: Perceptrons and the Theory of Brain
Mechanisms. Washington, DC: Spartan
Books.
Rosch, E. 1975. Cognitive representation of
semantic categories. Journal of Experimental
Psychology: General, 104, 192-233.
Rumelhart, D.E., and McClelland, J.L. 1986.
On learning the past tense of English verbs.
In Parallel distributed processing: Explorations
in the microstructure of cognition, Vol. 2: Psy-
chological and Biological Models, ed. J. L.
McClelland, D. E. Rumelhart and the PDP
Research Group. Cambridge, MA.: MIT
Press.
Rumelhart, D.E., McClelland, J.1., and the PDP
Research Group 1986. Parallel Distributed
of Cognition, Vol. 1: Foundations. Cambridge,
MA.: MIT Press.
Rumelhart, D.E., Smolensky, P., McClelland,
J.L., and Hinton, G.E. 1986. Schemas and
sequential thought processes in PDP models.
In Parallel distributed processing: Explorations
in the microstructure of cognition, Vol. 2: Psy-
chological and Biological Models, ed. J. L.
McClelland, D. E. Rumelhart, and the PDP
Press.
Selfridge, O.G. 1959. Pandemonium: A para-
digm for learning. Symposium on the Mechan-
ization of Thought Processes. London: HMSO.
Smolensky, P. 1990. Tensor product variable
binding and the representation of symbolic
structures in connectionist systems. Artificial
Intelligence, 46, 159-216.
St. John, M.P., and McClelland, J.L. 1990.
Learning and applying contextual con-
straints in sentence comprehension. Artifi-
cial Intelligence, 46, 217-57.
Sun, R. 1991 Connectionist models of rule-
based reasoning. Proceedings of the 13th
Annual Conference of the Cognitive Science
Society. Hillsdale, NJ: Erlbaum.
Touretzky, D.S. 1990. BoitzCONS: Dynamic
symbol structures in a connectionist
network. Artificial Intelligence, 46, 5-46.
van Gelder, T. 1990. Compositionality: A con-
nectionist variation on a classical theme.
van Gelder, T. 1991. What is the '0' in 'PDP'?
210
A survey of the concept of distribution, In
Philosophy and Connectionist Theory, ed. W.
Ramsey, S. P. Stich, and D. E. Rumelhart.
van Gelder, T. 1992. Is cognition categoriza-
tion?
WILLIAM BECHTEL
consciousness T. H. Huxley (1866) said
'How it is that anything so remarkable as a
state of consciousness comes about as a
result of irritating nervous tissue, is just as
unaccountable as the appearance of Djin
when Aladdin rubbed his lamp.' This is the
famous 'explanatory gap'. We have no con-
ception of our physical or functional nature
that allows us to understand how it could
explain our subjective experience. This fact
(in a form expressed by Nagel, 1974; the
term 'explanatory gap' comes from Levine,
1983) has dominated the last 20 years of
discussion of consciousness. Francis Crick
and Christoff Koch (1990) have famously
hypothesized that the neural basis of con-
sciousness is to be found in certain phase-
locked 40 Hz neural oscillations. But how
does a 40 Hz neural oscillation explain what
it's like (in Nagel's memorable phrase) to be
us? What is so special about a 40 Hz oscil-
lation as opposed to some other physical
state? And why couldn't there be creatures
with brains just like ours in their physical
and functional properties, including their
40 Hz oscillation patterns, whose owners'
experiences were very unlike ours, or who
had no subjective experiences at all? One
doesn't have to suppose that there really
could be creatures with brains just like ours
who have different experiences or no experi-
ences to demand an account of why not?
But no one has a clue about how to answer
these questions. This is the heart of the
mind-body problem.
Consciousness in the sense discussed is
phenomenal consciousness. 'What's that?',
you ask. There is no non-circular definition
to be offered; the best that can be done is
the offering of synonyms, examples and one
or another type of pointing to the phenom-
enon (Goldman, forthcoming). For example,
I used as synonyms 'subjective experience'
and 'what it is like to be us'. In explaining
phenomenal consciousness. one can also
appeal to conscious properties or qualities.
e.g. the ways things seem to us or immedi-
ate phenomenological qualities. Or one can
appeal to examples: the ways things look or
sound. the way pain feels and more gen-
erally the experiential properties of sensa-
tions. feelings and perceptual experiences. I
would also add that thoughts. wants and
emotions often have characteristic con-
scious aspects. and that a difference in
representational content can make a phe-
nomenal difference. Seeing something as a
cloud differs from seeing it as a part of
a painted backdrop (see PERCEPTUAL
CONTENT). What it is like to hear Bulgarian
spoken depends on whether one under-
stands the language.
We gain some perspective on the explan-
atory gap if we contrast the issue of the
physical/functional basis of consciousness
with the issue of the physical/functional
basis of thought. In the case of thought. we
do have some theoretical proposals about
what thought is. or at least what human
thought is. in scientific terms. Cognitive sci-
entists have had some success in explaining
some features of our thought processes in
terms of the notions of representation and
computation (see Block. 1990). There are
many disagreements among cognitive scien-
tists: especially notable is the disagreement
between connectionists and classical
'LANGUAGE OF THOUGHT' theorists (see COG-
NITIVE PSYCHOLOGY; CONNECTIONISM).
However. the notable fact is that in the case
of thought. we actually have more than one
substantive research programme. and their
proponents are busy fighting it out. compar-
ing which research programme handles
which phenomena best. But in the case of
consciousness. we have nothing - zilch -
worthy of being called a research pro-
gramme. nor are there any substantive pro-
posals about how to go about starting one.
(See Baars. 1988. for an indication of how
what passes for a research programme
about phenomenal consciousness is just
more cognitive psychology - actually a
CONSCIOUSNESS
theory of a different notion of conscious-
ness. access-consciousness. to be described
below.) Researchers are stumped. There
have been many tantalizing discoveries
recently about neuropsychological syn-
dromes in which consciousness seems to be
in some way missing or defective (see
Young. 1994; Baars, 1988). but no one has
yet come up with a theoretical perspective
that uses these data to narrow the explana-
tory gap. even a little bit.
PERSPECTIVES ON THE EXPLANATORY
GAP
Needless to say. there are many different
attitudes towards this problem. but five of
them stand out. First. we might mention
ELIMINATIVISM. the view that conscious-
ness as understood above simply does not
exist (Churchland. 1983; Dennett. 1988;
Rey. 1983). So there is nothing for there to
be an explanatory gap about. Second. we
have various forms of reductionism. notably
FUNCTIONALISM and PHYSICALISM. Accord-
ing to these views. there is such a thing as
consciousness. but there is no singular
explanatory gap. that is. there are no mys-
teries concerning the physical basis of con-
sciousness that differ in kind from run of the
mill unsolved scientific problems about the
physical/functional basis of liquidity. inheri-
tance or computation. On this view. there is
an explanatory gap. but it is unremarkable.
A third view is what Flanagan (1992) calls
the new mysterianism. Its most extreme
form is transcendentalism (White. 1991). the
view that consciousness is simply not a
natural phenomenon and is not explainable
in terms of science at all. A less extreme
form of new mysterianism is that of McGinn
(1991). which concedes that consciousness
is a natural phenomenon but emphasizes
our problem in understanding the physical
basis of consciousness. McGinn argues that
there are physical properties of our brains
that do in fact explain consciousness. but
though this explanation might be available
to some other type of being. it is cognitively
closed off to us. A fourth view that has no
well-known name (see Nagel. 1974; Flana-
211
CONSCIOUSNESS
gan. 1992; Searle. 1992). holds that
though there may be important differences
between a naturalistic explanation of con-
sciousness and naturalistic explanations of
other phenomena. there is no convincing
reason to regard consciousness as non-
natural or unexplainable in naturalistic
terms (see NATURALISM). This view is sug-
gested by Nagel's remark that we are like
the person ignorant of relativity theory who
is told that matter is a form of energy but
who does not have the concepts to appreci-
ate how there could be chains of reference-
links leading from a single phenomenon to
both 'matter' and 'energy'. The explanatory
gap exists - and we cannot conceive of how
to close it - because we lack the scientific
concepts. But future theory may provide
those concepts. A fifth view could be des-
cribed as deflationist about the explanatory
gap. The gap is unclosable. but not because
we cannot find the right physical concepts.
Rather. it is unclosable because reductive
explanation requires an a priori analysis of
the phenomenon to be explained. and no
such analysis can be given of our concepts
of conscious experience. (See QUALIA for
more on this view.)
Dennett (1988) argues for eliminativism.
He uses a thought experiment about two
coffee tasters. Chase and Sanborn. Both
liked Maxwell House coffee when they
started. and both dislike it now. But they
tell very different stories about what hap-
pened. Chase says that he has become more
sophisticated; he used to like the taste of
Maxwell House. but he no longer does.
Sanborn is equally certain that he can
remember what the coffee used to taste like.
but he still thinks that that original taste is
a great taste: the source of his change. he
says. is a change in his perceptual appara-
tus. perhaps his taste buds. He no longer
gets that nice old taste when he drinks
Maxwell House. Dennett points out that
either Chase or Sanborn (or both) might be
wrong because of changes in their memories
of the tastes. For example. perhaps Maxwell
House coffee still tastes exactly the same to
Sanborn as it always did. but his memory of
what it used to taste like has slowly. imper-
212
ceptibly changed over time. So Sanborn's
current dislike of Maxwell House could be
traced to a memory change plus a change
in his standards with no change at all in his
perceptual machinery. Further. Dennett
points out. their reports are consistent with
a variety of combinations of partial or total
changes in memory. aesthetic standards
and perceptual machinery.
It is not easy to see how this example is
supposed to support eliminativism. but I
would reconstruct the main argument as
follows:
(1) Suppose phenomenal consciousness
exists. that is. there are real phenomen-
ally conscious properties.
(2) Then there can be a fact of the matter
as to whether Chase and Sanborn are
right.
(3) But if there are real phenomenally con-
scious properties. they are transparent.
(4) However. only an expert. say a neuro-
physiologist who can examine the
brains of Chase and Sanborn. could tell
whether their memories. aesthetic stan-
dards and current conscious qualities
on drinking Maxwell House have
changed.
(5) The fact that we cannot rely on the
testimony of Chase and Sanborn them-
selves shows that phenomenally con-
scious qualities are not transparent.
(6) From (3) and (5) we can deduce that
there are no real phenomenally con-
scious qualities.
Once we actually set out the argument. it
is easy to see what is wrong with it: for (3)
and (5) to both be true. there must be an
equivocation on 'transparent'. The fact that
it is possible that the stories Chase and
Sanborn believe are wrong shows only that
their memories could be wrong. and that an
expert might be able to tell them that. But
no advocate of transparency of phenomenal
consciousness ought to suppose that mem-
ories of conscious states are literally incor-
rigible. There are a variety of ways of
understanding 'transparent' in which it
plausibly applies to phenomenally conscious
qualities of my states when I am having those
states, and these senses would make (3) true
without making (5) true. (See Flanagan,
1992, for reconstructions and rebuttals of
other arguments for eliminativism.)
OTHER CONCEPTS OF CONSCIOUSNESS
Thus far I have been talking about phenom-
enal consciousness. But there are other con-
cepts of consciousness - cognitive or
intentional or functional concepts of con-
sciousness - that are often not distinguished
from it, and it is common for defiationists or
reductionists about phenomenal conscious-
ness to tacitly slide from phenomenal con-
sciousness to one or another of these
cognitive or intentional or functional con-
cepts (see Dennett, 1991, and my review,
Block, 1993). I will mention three such
concepts of consciousness: self-conscious-
ness, monitoring-consciousness and access-
consciousness.
(1) Self-consciousness is the possession of
the concept of the SELF and the ability to use
this concept in thinking about oneself. There
is reason to think that animals or babies
can have phenomenally conscious states
without employing any concept of the self.
To suppose that phenomenal consciousness
requires the concept of the self is to place an
implausible intellectual condition on phe-
nomenal consciousness. Perhaps phenom-
enally conscious states have a non-
conceptual content that could be described as
'experienced as mine', but there is no
reason to think that this representational
aspect of the state exhausts its phenomenal
properties. After all, if both my experience
as of blue and my experience as of red are
experienced as mine, we still need to explain
the difference between the two experiences;
the fact, if it is a fact, that they are both
experienced as mine will not distinguish
them. (The 'as of' terminology is intended
to preclude cases in which red things don't
look red.)
(2) Monitoring-consciousness takes many
forms. (The source of these ideas in the
philosophical literature is Armstrong, 1968,
1980; see also Lycan, 1987; Rosenthal,
1986; Carruthers, 1989.) One form is
CONSCIOUSNESS
'internal scanning', but it would be a
mistake to confiate internal scanning with
phenomenal consciousness. As Rey (1983)
notes, ordinary laptop computers are
capable of internal scanning, but it would
be silly to think of one's laptop as conscious.
Rey favours supposing that internal scan-
ning is sufficient for consciousness, if there
is such a thing, and so he concludes that
consciousness is a concept that both
includes and precludes laptop computers
being conscious, and hence that the concept
of consciousness is incoherent. But even if
we acknowledge 'internal scanning con-
sciousness', we should drop the idea that
internal scanning is sufficient for phenomenal
consciousness, and so we get no inco-
herence.
Another form of monitoring conscious-
ness is that of accompaniment by a higher-
order thought. That is, a conscious state is
one that is accompanied by a thought
(grounded non-inferentially and non-
observationally) to the effect that one is in
that state. I favour a liberal terminological
policy, and so I have no objection to this
idea as a concept of consciousness. But I do
object to the idea (Rosenthal, 1986) that
phenomenal consciousness should be identi-
fied with higher-order-thought conscious-
ness. One way to see what is wrong with
that view is to note that even if I were to
come to know about states of my liver non-
inferentially and non-observationally - as
some people just know what time it is - that
wouldn't make the states of my liver phe-
nomenally conscious (see Dretske, 1993).
Another objection is that phenomenal con-
sciousness does not require the intellectual
apparatus that is required for higher-order
thought. Thus, the identification of phe-
nomenal consciousness with higher-order
thought shares the over-intellectualism of
the identification of phenomenal conscious-
ness with self-consciousness. Dogs and
babies may have phenomenally conscious
pains without thoughts to the effect that
they have those pains.
A distinction is often made between state
consciousness, or intransitive consciousness -
and consciousness of, or transitive conscious-
213
CONSCIOUSNESS
ness (Rosenthal, 1986). For example, if I
say I'm nauseous, I ascribe a kind of intran-
sitive consciousness to myself, and if I say
I am now seeing something as a mosquito,
I ascribe transitive consciousness. The
higher-order thought view purposely col-
lapses these notions. According to the
higher-order thought view, a conscious
state (intransitive consciousness) of mine is
simply a state that I am conscious of (tran-
sitive consciousness), and consciousness of
is simply a matter of accompaniment by a
thought to the effect that I am in that state.
So what it is for a state of mine to be con-
scious (intransitively) is for it to be accom-
panied by a thought that I am in that state
(see Rosenthal, 1986; Carruthers, 1989).
This intentional conflation has an
element of plausibility to it, which can be
seen by comparing two dogs, one of which
has a perceptual state whereas the other
has a similar perceptual state plus a repre-
sentation of it. Surely the latter dog has a
conscious state even if the former dog does
not! Quite so, because consciousness of
brings consciousness with it. But it is the
converse that is problematic. State con-
sciousness makes less in the way of intellec-
tual demands than consciousness of, and so
the first dog could be conscious without
being conscious of anything.
(3) Access-consciousness does not make
the intellectual demands of self-conscious-
ness or higher-order-thought consciousness,
and for that reason, reductionists about
phenomenal consciousness would do better
to identify phenomenal consciousness with
access-consciousness. A state is access-
conscious if, in virtue of one's having the
state, a representation of its content is (a)
inferentially promiscuous, i.e. freely avail-
able as a premise in reasoning, and (b)
poised for rational control of action and (c)
poised for rational control of speech. One
can speak of both states and their contents
as access-conscious. My claims about
access-consciousness have been criticized in
Flanagan (1992, p. 145-6), in Searle
(1990; 1992, p. 84), and improvements
suggested in Davies and Humphreys
(1993b). There are three main differences
214
between access-consciousness and phenom-
enal consciousness that ought to be
acknowledged by those of us who are
realists about a non-intentional, non-func-
tional, non-cognitive notion of conscious-
ness. First, it is in virtue of its phenomenal
content (or the phenomenal aspect of its
content) that a state is phenomenally con-
scious, whereas it is in virtue of its repre-
sentational content or the representational
aspect of its content that a state is access-
conscious. Second, access-consciousness is a
functional notion, but phenomenal con-
sciousness is not. If you are a functionalist
about phenomenal consciousness, it would
be very natural to identify it with access-
consciousness. Note that I deny that the
concept of phenomenal consciousness is
functional, but I acknowledge the empirical
possibility that the scientific essence of
phenomenal consciousness is something to
do with information processing (see Loar,
1990). Third, access-consciousness applies
to state tokens, or rather tokens at times,
but phenomenal consciousness is best
thought of as a feature of state types. Let me
explain. The following inscription, 'teeth',
that you just read contains five letter
tokens, but of only three letter types. There
is a token of the type dog in my office, but
the type dog itself is an abstract object that
doesn't exist anywhere in space-time. Here
is why access is a matter of tokens at times:
a single token state might be access-con-
scious at one time but not another, because
of changes in information flow in the
system, just as my keys are accessible at
some times but not others (when I lose
them). (And any actually accessible token
might not have been accessible.) But a
token of a phenomenal type is necessarily
phenomenally conscious - it can't become
non-phenomenally conscious without dis-
appearing altogether. In other words,
access-consciousness - but not phenomenal
consciousness - is a functional notion - and
a single token can change function (botp
actually and counterfactually).
A good way to- see the distinction is to
note cases of one without the other. Con-
sider a robot with a computer brain that is
behaviourally and computationally identical
to ours. The question arises as to whether
what it is like to be that robot is different
from what it is like to be us, or, indeed,
whether there is anything at all that it is
like to be that robot. If there is nothing it is
like to be that robot, the robot is a zombie. If
zombies are conceptually possible, they
certainly illustrate access-consciousness
without phenomenal consciousness. But
there is widespread opposition to the con-
ceptual coherence of zombies (see Shoe-
maker, 1975, 1981; Dennett, 1991). So for
illustrating access-consciousness without
phenomenal consciousness, I would rather
rely on a very limited sort of partial zombie.
Consider blindsight, a neurological syn-
drome in which subjects seem to have
'blind' areas in their visual fields. If the
experimenter flashes a stimulus to one of
those blind areas, the patient claims to see
nothing at all. But if the experimenter
insists that the subject guess, and the
experimenter supplies a few alternatives, the
blindsight patients are able to 'guess' reli-
ably about certain features of the stimulus,
features having to do with motion, location,
direction, and they are able to discriminate
some simple forms (Weiskrantz, 1986;
Young, 1994). Consider a blindsight patient
who 'guesses' that there is an 'X' rather
than an '0' in his blind field. The patient
has no access-consciousness of the stimulus
(because, until he hears his own guess, he
cannot use the information freely in reason-
ing or in rational control of action), and it is
plausible that he has no phenomenal con-
sciousness of it either. Now imagine some-
thing that does not exist, what we might
call super-blindsight. A real blindsight patient
can only guess when given a choice among
a small set of alternatives ('X' /,0', hori-
zontal/vertical. etc.). But suppose (appar-
ently contrary to fact) that a blindsight
patient could be trained to prompt himself
at will, guessing what is in the blind field
without being told to guess. Visual informa-
tion from the blind field simply pops into his
thoughts the way that solutions to problems
sometimes pop into ours or (to use an
example given earlier) the way some people
CONSCIOUSNESS
just know what time it is without any
special perceptual experience. The super-
blindsight patient says there is something
it is like to see an 'X' in his sighted field,
but not in his blind field, and we believe
him. This would be a case of access-
consciousness without phenomenal con-
sciousness, a sort of partial zombie.
Here is an example of the converse of the
zombie cases, namely phenomenal con-
sciousness without access-consciousness. It
appears that some areas of the brain spe-
cialize in reasoning and rational control of
action, whereas other areas subserve sensa-
tion. If a person's brain has the former
areas destroyed, he is unable to use the
deliverances of the senses to rationally
control action, to reason or to report sen-
sibly, but he can still have experiences.
Such a person has phenomenal conscious-
ness without access-consciousness.
Here is a different sort of example.
Suppose that you are engaged in intense
thought when suddenly at midnight you
realize that there is now and has been for
some time a deafening pounding noise
going on. You were aware of the noise all
along, but only at midnight were you con-
sciously aware of it. That is, you were phe-
nomenally conscious of the noise all along,
but only at midnight did you become
access-conscious of it. The period before
midnight illustrates phenomenal conscious-
ness without access-consciousness. 'Con-
scious and 'aware' are roughly
synonymous, so it is natural to use one for
the period before midnight, and both for the
period after midnight when there are two
kinds of consciousness present.
Another illustration of phenomenal con-
sciousness without access-consciousness
derives from a famous experiment by
George Sperling. Sperling (1960) flashed
arrays of letters (e.g. 3-by-3) to subjects for
50 milliseconds. Subjects typically said that
they could see all of the letters, but typically
could report only about half of them. Were
the subjects right in saying that they could
see all of the letters? Sperling tried signalling
the subjects with a tone. A high tone meant
that the subject was to report the top row, a
215
CONSCIOUSNESS
medium tone indicated the middle row, etc.
If the tone was given immediately after the
stimulus, the subjects could usually get all
the letters in the row, no matter which row,
but once they had named those letters they
could get no others. The experiment is
taken by psychologists to indicate some sort
of raw visual storage, the 'visual icon'. But I
have a different interest, the question of
what it is like to be a subject in this experi-
ment. My own experience is that I see all or
almost all the letters - other subjects report
the same thing (Baars, 1988, p. 15). And I
would say I see them as 'N', 'J', 'B', etc., as
specific letters, not just as blurry or vague
or non-specific letters. I (and others) cannot
report much more than a single row. So
subjects are phenomenally conscious of all
(or almost all) the letters at once, but not
access-conscious of all of them at once. In
sum, one can be phenomenally conscious of
more than one is access-conscious of.
The two cases I've mentioned of phenom-
enal consciousness without access-con-
sciousness are also counterexamples to the
higher-order thought theory of phenomenal
consciousness. If the subject has no access
to the phenomenal state, he can't think
about it either. Before midnight, I have a
phenomenally conscious state caused by the
noise but no thought to the effect that I am
in such a state. And in the Sperling experi-
ment, I am phenomenally conscious of all
or almost all the letters, but since I can't
access all the letters at once, I can't repre-
sent them all at once (except as letters).
Akins (1993) has argued against the dis-
tinction between a phenomenal and a
representational aspect of experience. She
keys her discussion to Nagel's (19 74) claim
that we cannot know what it is like to be a
bat, challenging the reader to imagine that
what it is like to be a bat is just what it is
like to be us - only all those experiences
represent totally different things. Correctly,
she says that you cannot imagine that. That
is because, as I mentioned earlier, repre-
sentational differences of a certain sort make
a phenomenal difference. What it is like to
hear a sound as coming from the left is dif-
ferent from what it is like to hear a sound as
216
coming from the right. And what it is like to
hear Bulgarian depends on whether one
speaks the language (see Goldman, 1993;
Davies, forthcoming; Peacocke, 1992; Tye,
forthcoming a). But from the fact that some
representational differences make a phe-
nomenal difference, one should not con-
clude that there is no distinction between
the representational and the phenomenal.
Note, for example, that representational dif-
ferences of the sort that obtain between me
and my twin on Putnam's TWIN EARTH
needn't make a phenomenal difference.
Further, there are phenomenal states that
aren't at all representational, orgasm for
example. Further, two distinct phenomenal
contents can overlap representation ally, for
example a visual and a kinaesthetic repre-
sentation of one's hand moving. Note that
the point is not just that there is a repre-
sentational overlap without a corresponding
phenomenal overlap (as is said, for example,
in Pendlebury, 1992). That would be com-
patible with the following story: phenom-
enal content is just one kind of
representational content, but these experi-
ences overlap in non-phenomenal repre-
sentational content. The point, rather, is
that the phenomenal qualities are them-
selves representational; the as of motion is
part of the two phenomenal contents, two
phenomenal contents that themselves
overlap representationally, but the two phe-
nomenal contents represent the same thing
via different phenomenal qualities.
IS 'CONSCIOUSNESS' AMBIGUOUS?
I have distinguished a number of different
concepts of consciousness, phenomenal con-
sciousness, self-consciousness, monitoring
consciousness and access-consciousness.
Am I saying that the word 'conscious' is
ambiguous? I don't think that the different
concepts of consciousness I have mentioned
indicate a straightforward ambiguity; for
example, you won't find a straightforward
distinction among any of these concepts in a
dictionary. (Though some dictionaries
mention self-consciousness separately.) I
would rather say that 'conscious' (together
with 'aware', 'experience' and other words
similarly used) should be ambiguous. An
analogy: Kuhn (1964) points out that
Aristotle failed to distinguish between
average velocity and instantaneous velocity,
and made a crucial mistake because of this
conflation. There is no ambiguity in 'velo-
city' or 'speed' in ordinary English, but the
seeds are there, and a distinction is needed
for some purposes. My own view on the
empirical linguistic question as to what
'consciousness' means is that it is a cluster
concept. Consider the concept of a religion,
a cluster concept that involves belief in a
supreme being, a moral code, rituals, sacred
objects, a special plane of experience, etc.
There can be (and often are) religions
that lack anyone or two of these (see
Alston, 1967). The cluster in the case of
'consciousness' involves phenomenal con-
sciousness, access-consciousness, self-
consciousness, monitoring-consciousness,
and perhaps other elements. The problem is
that these elements are so different from one
another that failure to distinguish among
the different elements can lead to serious
confusion. An undifferentiated 'conscious'
works well for most purposes, but for
serious thinking we need conceptual clar-
ification. (Of recent books on consciousness,
the ones I've noticed that distinguish phe-
nomenal consciousness from other notions
are Lycan, 1987; Flanagan, 1992; and
Davies and Humphreys, 1993a.)
CONFLATIONS
There are two notable sorts of trouble
writers get into by not making these distinc-
tions. One is to be found in Jaynes (1976)
and Dennett (1986, 1991). These authors
allege that consciousness is a cultural con-
struction - Jaynes even gives its invention a
date: between the events reported in the
Oddysey and the Iliad. They seem to be
talking about phenomenal consciousness,
but if one accepts a notion of phenomenal
consciousness as distinct from the cognitive
and functional notions I have described, the
idea that consciousness was invented by the
ancient Greeks is ludicrous. If there is such
CONSCIOUSNESS
a thing as phenomenal consciousness as
distinct from the cognitive and functional
notions I have described, surely it is a basic
biological feature of us. The same is true for
access-consciousness, which is the best
guess as to what Dennett is usually talking
about. Obviously, our ability to access infor-
mation from our senses is genetically pro-
grammed. And I would say the same for
most forms of monitoring. What Jaynes and
Dennett ought to be saying is that there is
no such thing as phenomenal consciousness
as distinct from the other consciousnesses.
They ought to be reductionists or elim-
inativists about consciousness. The confla-
tion is especially silly in Jaynes, where it is
obvious that 'consciousness' in the sense in
which it is supposed to have been invented
by the Greeks is something like a theory of
consciousness in roughly the phenomenal
sense. (See Dennett, 1986, for a defence of
Jaynes.)
Another type of problem has nothing to
do with reductionism or eliminativism. Con-
sider, for example, Searle's (1992) reason-
ing about a function of consciousness.
Searle mentions Penfield's description of
petit mal epilepsy patients who are 'totally
unconscious', but nonetheless continue
their activities of walking or driving home
or playing a piano piece, but in an inflexible
and uncreative way. Searle says that the
lack of consciousness explains the lack of
flexibility and creativity, and so one of the
functions of consciousness is to add powers
of flexibility and creativity. Searle is talking
about the function of phenomenal conscious-
ness, but he gives no reason to think that
the petit mal patients lack that kind of con-
sciousness. The patients don't cope with
new situations very well, but they show
every sign of normal sensation. For
example, Searle describes the epileptic
walker as threading his way through the
crowd. Isn't there something it is like for
him to see, say, a red sign, which he knows
to turn right at? What is most obviously
deficient in these patients is not phenom-
enal consciousness, but cognitive and func-
tional consciousnesses. Penfield (1975) does
say that the patients are 'totally uncon-
217
CONSCIOUSNESS
scious', but he appears to have some sort of
monitoring consciousness in mind, not phe-
nomenal consciousness. In another publica-
tion, Searle (1990) criticizes the idea that
there is any access-sense of consciousness,
saying that the idea that there is such a
thing confuses 'what I would call peripheral
consciousness or inattentiveness with total
unconsciousness. It is true, for example,
that when I am driving my car "on auto-
matic pilot", I am not paying much atten-
tion to the details of the road and the traffic.
But it is simply not true that I am totally
unconscious of these phenomena. If I were,
there would be a car crash' (p. 635). Note
the contradiction. In one place, he says that
a 'totally unconscious' epileptic can drive
home, and in another place he says that if a
driver were 'totally unconscious' the car
would crash. In the first place, he is think-
ing of total lack of phenomenal conscious-
ness (combined with some degree of access-
consciousness), whereas in the second, he is
thinking of a total lack of access-conscious-
ness (or both types of consciousness).
A variant on this mistake is found in the
writings of many authors, including Flana-
gan (1992), who argues that since a thirsty
blindsight patient will make no move
towards a water fountain in his blind field,
and since phenomenal consciousness of the
water fountain is missing, phenomenal con-
sciousness must have a function in initiat-
ing action. But in the blindsight patient,
access-consciousness and monitoring con-
sciousness of events in the blind field are
also missing. We can explain why the blind-
sight patient does not move towards the
water fountain by noting that with the
missing access-consciousness, the content of
the perceptual state is not available for
rational control of action. (I don't have the
space to argue here that in general we need
not appeal to the missing monitoring con-
sciousness to explain this phenomenon.)
There is no need to appeal to the missing
phenomenal consciousness, even if Flana-
gan is right that phenomenal consciousness
is missing. (And if phenomenal conscious-
ness is not missing in blindsight, we have a
different fallacy-like the Searle mistake just
218
pointed out.) The main error here is to
transfer by conflation an obvious function
of access-consciousness to phenomenal con-
sciousness.
See also CONTENT; An Essay on Mind section
2.2; PAIN; SENSATIONS; SUBJECTIVITY.
BIBLIOGRAPHY
Akins, K. 1993. A bat without qualities. In
Davies and Humphreys, 1993a.
Alston, W. 1967. Religion. In The Encyclopedia
of Philosophy. New York: Macmillan.
Armstrong. D.M. 1968. A Materialist Theory of
Mind. London: Humanity Press.
Armstrong, D.M. 1980 'What is conscious-
ness?' In The Nature of Mind. Ithaca: Cornell
University Press.
Baars, 8.J. 1988. A Cognitive Theory of Con-
sciousness. Cambridge University Press.
Block, N. 1990. The computer model of the
mind. In An Invitation to Cognitive Science:
Thinking. ed. D. Osherson and E. Smith.
Block. N. 1993. Review of D. Dennett. Con-
sciousness Explained. The Journal of Philoso-
phy, XC, 4: 181-93.
Carruthers. P. 1989. Brute experience. Journal
of Philosophy. 86. 258-69.
Chalmers. D.J. 1993. Toward a Theory of Con-
sciousness. University of Indiana Ph.D.
thesis.
Churchland. P.S. 1983. Consciousness: the
transmutation of a concept. Pacific Philoso-
phical Quarterly. 64. 80-93.
Crick. F. and Koch. C. 1990. Towards a neu-
robiological theory of consciousness. Sem-
inars in the Neurosciences. 2. 263-75.
Davies, M. and Humphreys. G. 1993a. Con-
sciousness. Oxford. Basil Blackwell.
Davies. M. and Humphreys. G. 1993b. Intro-
duction. In Davies and Humphreys 1993a.
Davies. M. forthcoming. Externalism and
experience.
Dennett. D. 1986. Julian Jaynes' software
archaeology. Canadian Psychology. 27: 2.
149-54.
Dennett. D. 1988. Quining Qualia. In Con-
sciousness in Contemporary Society. ed. A.
Marcel and E. Bisiach. Oxford University
Press.
Dennett. D. 1991. Consciousness Explained.
New York: Little Brown.
Dretske. F. 1993. Conscious experience. Mind.
102: 406. 263-84.
Flanagan. O. 1992. Consciousness Reconsidered.
Cambridge. MA.: MIT Press.
Goldman. A. 1993. The psychology of folk
psychology. The Behavioral and Brain
Sciences. 16:1. 15-28.
Goldman. A. forthcoming. Consciousness. folk
psychology and cognitive science. Conscious-
ness and Cognition.
Humphreys. N. 1992. A History of the Mind.
New York: Simon & Schuster.
Huxley. T.H. 1866. Lessons in Elementary Phy-
siology. Quoted in Humphrey (1992).
Jaynes. J. 1976. The Origin of Consciousness in
the Breakdown of the Bicameral Mind. Boston:
Houghton-Millin.
Kuhn. T. 1964. A function for thought experi-
ments. In Melanges Alexandre Koyre. vol. 1.
Hermann.
Levine. J. 1983. Materialism and qualia: the
explanatory gap. Pacific Philosophical Quar-
terly. 64. 354-61.
Levine. J. 1993. On leaving out what it is like.
In Davies and Humphreys. 1993a.
Loar. B. 1990. Phenomenal properties. In Phi-
losophical Perspectives: Action Theory and
Philosophy of Mind. ed. J. Tomberlin. Atasca-
dero. CA.: Ridgeview.
Lycan. W. 1987. Consciousness. Cambridge.
MA.: MIT Press.
McGinn. C. 1991. The Problem of Conscious-
ness. Oxford: Basil Blackwell.
Nagel. T. 1974. What is it like to be a bat?
Philosophical Review. 83.435-50.
Peacocke. C. 1992. A Study of Concepts. Cam-
bridge. MA.: MIT Press.
Pendlebury. M. 1992. Experience. theories of.
In A Companion to Epistemology. ed. J. Dancy
and E. Sosa. Oxford: Basil Blackwell.
Penfield. W. 1975. The Mystery of the Mind: A
Critical Study of Consciousness and the Human
Brain. Princeton University Press.
Rey. G. 1983. A reason for doubting the exist-
ence of consciousness. In Consciousness and
Self-Regulation. Vol. 3. ed. R. Davidson. G.
Schwartz. D. Shapiro. New York: Plenum.
Rosenthal. D. 1986. Two concepts of con-
sciousness. Philosophical Studies. 49. 329-
59.
Searle. J. 1990. Who is computing with the
brain? Behavioral and Brain Sciences. 13:4.
632-42.
CONTENT (1)
Searle. J. 1992. The Rediscovery of the Mind.
Shoemaker. S. 1975. Functionalism and
quaiia. Philosophical Studies. 27. 291-315.
Shoemaker. S. 1981. Absent quaiia are impos-
sible - a reply to Block. The Philosophical
Review. 90: 4.581-99.
Sperling. G. 1960. The information available
in brief visual presentations. Psychological
Monographs. 74. 11.
Tye. M. forthcoming a. Does pain lie within
the domain of cognitive psychology? In
Philosophical Perspectives. 8. ed. J. Tomberlin.
Atascadero. CA.: Ridgeview.
Tye. M. forthcoming b. How to become
puzzled about phenomenal consciousness?
Weiskrantz. 1. 1986. Blindsight. Oxford Uni-
versity Press.
White. S. 1991. Transcendentalism and its
discontents. In The Unity of the Self. Cam-
Young. A.W. 1994. Neuropsychology of
awareness. In Consciousness in philosophy
and cognitive neuroscience. ed. M. Kappinen
and A. Revonsuo. New York: Erlbaum.
NED BLOCK
content (1) Mental events. states or pro-
cesses with content include seeing the door
is shut; believing you are being followed;
and calculating the square root of 2. What
centrally distinguishes states. events or pro-
cesses - henceforth. simply states - with
content is that they involve reference to
objects. properties or relations. A mental
state with content can fail to refer. but there
always exists a specific condition for a state
with content to refer to certain things.
When the state has a correctness or fulfil-
ment condition. its correctness is deter-
mined by whether its referents have the
properties the content specifies for them.
This highly generic characterization of
content permits many subdivisions. It does
not in itself restrict contents to con-
ceptualized content; and it permits contents
built from Frege's senses as well as Russel-
lian contents built from objects and proper-
ties. It leaves open the possibility that
unconscious states. as well as conscious
219
CONTENT (1)
states, have content. It equally allows the
states identified by an empirical. computa-
tional psychology to have content. A correct
philosophical understanding of this general
notion of content is fundamental not only
to the philosophy of mind and psychology,
but also to the theory of knowledge and to
metaphysics.
ASCRIBING CONTENTS
A widely discussed idea is that for a subject
to be in a certain set of content-involving
states is for attribution of those states to
make the subject as rationally intelligible as
possible, in the circumstances. In one form
or another, this idea is found in the writings
of Davidson (1984), Dennett (1987), Mc-
Dowell (1986), Putnam (1988) and Sellars
(1963). Perceptions make it rational for a
person to form corresponding beliefs. Beliefs
make it rational to draw certain inferences.
Beliefs and desires make rational the
formation of particular intentions, and the
performance of the appropriate actions.
People are frequently irrational of course,
but a governing ideal of this approach is
that for any family of contents, there is some
minimal core of rational transitions to or
from states involving them, a core that a
person must respect if his states are to be
attributed with those contents at all (see
RATIONALITY).
This approach must deal with the point
that it seems metaphysically possible for
there to be something that in actual and
counterfactual circumstances behaves as if
it enjoys states with content, when in fact it
does not. If the possibility is not denied, this
approach must add at least that the states
with content causally interact in various
ways with one another, and also causally
produce intentional action (see REASONS
AND CAUSES). The existence of such causal
links could well be written into the minimal
core of rational transitions required for the
ascription of the contents in question.
It is one thing to agree that the ascription
of content involves a species of rational
intelligibility; it is another to provide an
explanation of this fact. There are compet-
220
ing explanations. One treatment regards
rational intelligibility as ultimately depen-
dent upon what we find intelligible, or on
what we could come to find intelligible in
suitable circumstances (McDowell, 1986).
This is an analogue of classical treatments
of secondary qualities, and as such is a form
of subjectivism about content. An alter-
native position regards the particular condi-
tions for correct ascription of given contents
as more fundamental. This alternative states
that interpretation must respect these par-
ticular conditions. In the case of conceptual
contents, this alternative could be developed
in tandem with the view that concepts are
individuated by the conditions for posses-
sing them (Peacocke, 1992). These posses-
sion conditions would then function as
constraints upon correct interpretation. If
such a theorist also assigns references to
concepts in such a way that the minimal
rational transitions are also always truth-
preserving, he will also have succeeded in
explaining why such transitions are correct.
Under an approach that treats conditions
for attribution as fundamental. intelligibility
need not be treated as a subjective property.
There may be concepts we could never
grasp because of our intellectual limitations,
as there will be concepts that members of
other species could not grasp. Such concepts
have their possession conditions, but some
thinkers could not satisfy those conditions.
Ascribing states with content to an actual
person has to proceed simultaneously with
attribution of a wide range of non-rational
states and capacities. In general. we cannot
understand a person's reasons for acting as
he does without knowing the array of emo-
tions and sensations to which he is subject;
what he remembers and what he forgets;
and how he reasons beyond the confines of
minimal rationality. Even the content-
involving perceptual states, which play a
fundamental role in individuating content,
cannot be understood purely in terms relat-
ing to minimal rationality. A perception of
the world as being a certain way is not (and
could not be) under a subject's rational
control. Though it is true and important
that perceptions give reasons for forming
beliefs, the beliefs for which they funda-
mentally provide reasons - observational
beliefs about the environment - have con-
tents which can only be elucidated by refer-
ring back to perceptual experience. In this
respect (as in others), perceptual states differ
from those beliefs and desires that are indi-
viduated by mentioning what they provide
reasons for judging or doing; for frequently
these latter judgments and actions can be
individuated without reference back to the
states that provide reasons for them.
What is the significance for theories of
content of the fact that it is almost certainly
adaptive for members of a species to have a
system of states with representational con-
tents which are capable of influencing their
actions appropriately? According to tele-
ological theories of content, a constitutive
account of content - one which says what
it is for a state to have a given content -
must make use of the notions of natural
function and teleology. The intuitive idea is
that for a belief state to have a given
content p is for the belief-forming mechan-
isms which produced it to have the function
(perhaps derivatively) of producing that
state only when it is the case that p (Milli-
kan, 1986; Papineau, 1987). One issue this
approach must tackle is whether it is really
capable of associating with states the classi-
cal. realistic, verification-transcendent con-
tents which, pretheoretically, we attribute
to them. It is not clear that a content's
holding unknowably can influence the
replication of belief-forming mechanisms.
But even if content itself proves to resist elu-
cidation in terms of natural function and
selection, it is still a very attractive view
(one also developed in Millikan, 1984) that
selection must be mentioned in an account
of what associates something - such as a
sentence - with a particular content, even
though that content itself may be individu-
ated by other means.
Contents are normally specified by
'that .. .' clauses, and it is natural to
suppose that a content has the same kind of
sequential and hierarchical structure as the
sentence that specifies it. This supposition
would be widely accepted for conceptual
CONTENT (1)
content. It is, however, a substantive thesis
that all content is conceptual. One way of
treating one sort of PERCEPTUAL CONTENT
is to regard the content as determined by a
spatial type, the type under which the
region of space around the perceiver must
fall if the experience with that content is to
represent the environment correctly. The
type involves a specification of surfaces and
features in the environment, and their dis-
tances and directions from the perceiver's
body as origin (Peacocke, 1992). Such con-
tents prima facie lack any sentence-like
structure at all. Supporters of the view that
all content is conceptual will argue that the
legitimacy of using these spatial types in
giving the content of experience does not
undermine the thesis that all content is
conceptual. Such supporters will say that
the spatial type is just a way of capturing
what can equally be captured by conceptual
components such as 'that distance', or 'that
direction', where these demonstratives are
made available by the perception in ques-
tion. Friends of non-conceptual content will
respond that these demonstratives them-
selves cannot be elucidated without men-
tioning the spatial types which lack
sentence-like structure.
CONTENT AND EXTERNALISM
The actions made rational by content-
involving states are actions individuated in
part by reference to the agent's relations to
things and properties in his environment.
Wanting to see a particular movie and
believing that that building over there is a
cinema showing it makes rational the
action of walking in the direction of that
building. Similarly, for the fundamental case
of a subject who has knowledge about his
environment, a crucial factor in making
rational the formation of particular attitudes
is the way the world is around him. One
may expect, then, that any theory that links
the attribution of contents to states with
rational intelligibility will be committed to
the thesis that the content of a person's
states depends in part upon his relations to
the world outside him. We can call this
221
CONTENT (1)
thesis the thesis of externalism about
content.
Externalism about content should steer a
middle course. On the one hand, it should
not ignore the truism that the relations of
rational intelligibility involve not just things
and properties in the world, but the way
they are presented as being - an externalist
should use some version of Frege's notion of
a mode of presentation (see CONCEPTS). On
the other hand, the externalist for whom
of rational intelligibility are
pertinent to the individuation of content is
likely to insist that we cannot dispense with
the notion of something in the world - an
object, property or relation - being pre-
sented in a certain way. If we dispense with
the notion of something external being pre-
sented in a certain way, we are in danger of
regarding attributions of content as having
no consequences for how an individual
relates to his environment, in a way that is
quite contrary to our intuitive under-
standing of rational intelligibility.
Externalism comes in more and less
extreme versions. Consider a thinker who
sees a particular pear, and thinks a thought
'that pear is ripe', where the demonstrative
way of thinking of the pear expressed by
'that pear' is made available to him by his
perceiving the pear. Some philosophers,
including Evans (1982) and McDowell
(1984), have held that the thinker would be
employing a different perceptually based
way of thinking were he perceiving a differ-
ent pear. But externalism need not be com-
mitted to this. In the perceptual state that
makes available the way of thinking, the
pear is presented as being in a particular
direction from the thinker, at a particular
distance, and as having certain properties.
A position will still be externalist if it holds
that what is involved in the pear's being so
presented is the collective role of these com-
ponents of content in making intelligible in
various circumstances the subject's rela-
tions to environmental directions, distances
and properties of objects. This can be held
without commitment to the object-depend-
ence of the way of thinking expressed by
'that pear'. This less strenuous form of
222
externalism must, though, address the epi-
stemological arguments offered in favour of
the more extreme versions, to the effect that
only they are sufficiently world-involving.
Externalism about content is a claim
about dependence, and dependence comes
in various kinds. In the discussions of the
1970s, the writings of Kripke (1980),
Putnam (1975) and Burge (1979) moved
the contents of beliefs to the forefront of dis-
cussion. The apparent dependence of the
content of beliefs on factors external to the
subject can be formulated as a failure of
supervenience of belief content upon facts
about what is the case within the bound-
aries of the subject's body (see SUPER-
VENIENCE). To claim that such super-
venience fails is to make a modal claim: that
there can be two persons the same in
respect of their internal physical states (and
so in respect of those of their dispositions
that are independent of content-involving
states), who nevertheless differ in respect of
which beliefs they have. Putnam's cele-
brated example of a community on TWIN
EAR TH, where the water-like substance in
lakes and rain is not H
2
0 but some different
chemical compound XYZ - twater - illus-
trates such failure of supervenience. A
molecule-for-molecule replica of you on
twin earth has beliefs to the effect that
twater is thus-and-so. Those with no chemi-
cal beliefs on twin earth may well not have
any beliefs to the effect that water is thus-
and-so, even if they are replicas of persons
on earth who do have such beliefs. Burge
emphasized that this phenomenon extends
far beyond beliefs about natural kinds.
In the case of content-involving percep-
tual states, it is a much more delicate
matter to argue for the failure of super-
venience. The fundamental reason for this is
that attribution of perceptual content is
answerable not only to factors on the input
side - what in certain fundamental cases
causes the subject to be in the perceptual
state - but also to factors on the output side
- what the perceptual state is capable of
helping to explain amongst the subject's
actions. If differences in perceptual content
always involve differences in bodily-
described actions in suitable counterfactual
circumstances, and if these different actions
always have distinct neural bases, perhaps
there will after all be supervenience of
content-involving perceptual states on inter-
nal states. But if this should turn out to be
so, that is not a refutation of externalism for
perceptual contents. A different reaction to
this situation is that the elaboration of the
relation of dependence as one of super-
venience is in some cases too strong. A
better characterization of the dependence in
question is given by a constitutive claim:
that what makes a state have the content it
does is certain of its complex relations to
external states of affairs. This can be held
without commitment to the modal separ-
ability of certain internal states from
content-involving perceptual states.
Attractive as externalism about content
may be, it has been vigorously contested,
notably by Jerry Fodor (1981). Fodor en-
dorses the importance of explanation by
content-involving states, but holds that
content must be narrow, constituted by
internal properties of an individual (Fodor,
1991). One influential motivation for
narrow content is a doctrine about explana-
tion, that molecule-for-molecule counter-
parts must have the same causal powers.
Externalists have replied that the attribu-
tions of content-involving states presuppose
some normal background or context for the
subject of the states, and that content-
involving explanations commonly take the
presupposed background for granted. Molec-
ular counterparts can have different pre-
supposed backgrounds, and their content-
involving states may correspondingly differ.
Presupposition of a background of external
relations in which something stands is found
in other sciences outside those that employ
the notions of content, including astronomy
and geology, as Davies (1986) and Burge
(1986) have respectively noted.
A more specific concern of those sympa-
thetic to narrow content is that when
content is externally individuated, the
explanatory principles postulated in which
content-involving states feature will be a
priori in some way that is illegitimate. For
CONTENT (1)
instance, it appears to be a priori that
behaviour is intentional under some
description involving the concept water will
be explained by mental states that have the
externally individuated concept water in
their content. The externalist about content
will have a twofold response. First, explana-
tions in which content-involving states are
implicated will also include explanations of
the subject's standing in a particular rela-
tion to the stuff water itself, and for many
such relations, it is in no way a priori that
the thinker's so standing has a psycho-
logical explanation at all. Some such cases
will be fundamental to the ascription of
externalist content on treatments that tie
such content to the rational intelligibility of
actions relation ally characterized. Second,
there are other cases in which the identifi-
cation of a theoretically postulated state in
terms of its relations generates a priori
truths, quite consistently with that state
playing a role in explanation. It arguably is
a priori that if a gene exists for a certain
phenotypical characteristic, then it plays a
causal role in the production of that char-
acteristic in members of the species in ques-
tion. Far from being incompatible with a
claim about explanation, the characteriza-
tion of genes that would make this a priori
also requires genes to have a certain causal-
explanatory role.
If anything, it is the friend of narrow
content who has difficulty accommodating
the nature of content-involving explanation.
States with narrow content are fit to explain
bodily movements, provided they are not
characterized in environment-involving
terms. But we noted that the characteristic
explananda of content-involving states,
such as walking towards the cinema, are
characterized in environment-involving
terms. How is the theorist of narrow
content to accommodate this fact? He may
say that we merely need to add a descrip-
tion of the context of the bodily movement,
which ensures that the movement is in fact
a movement towards the cinema. But
adding a specification of a new environ-
mental property of an event to an explana-
tion of that event does not give one an
223
CONTENT (1)
explanation of the event's having that
environmental property, let alone a content-
involving explanation of the fact. The bodily
movement may also be a walking in the
direction of Moscow, but it does not follow
that we have a rationally intelligible ex-
planation of the event as a walking in the
direction of Moscow. Perhaps the theorist of
narrow content would at this point add
further relational properties of the internal
states, of such a kind that when his expla-
nation is fully supplemented, it sustains the
same counterfactuals and predictions as
does the explanation that mentions exter-
nally individuated contents. But such a fully
supplemented explanation is not really in
competition with the externalist's account.
It begins to appear that if such extensive
supplementation is adequate to capture the
relational explananda, it is also sufficient to
ensure that the subject is in states with
externally individuated contents (Peacocke,
1993). This problem affects not only treat-
ments of content as narrow, but any
attempt to reduce explanation by content-
involving states to explanation by neuro-
physiological states.
One of the tasks of a subpersonal com-
putational psychology is to explain how
individuals come to have beliefs, desires,
perceptions and other personal-level
content-involving properties. If the content
of personal-level states is externally indi-
viduated, then the contents mentioned in a
subpersonal psychology that is explanatory
of those personal states must also be extern-
ally individuated. One cannot fully explain
the presence of an externally individuated
state by citing only states that are internally
individuated. On an externalist conception
of subpersonal psychology, a content-
involving computation commonly consists
in the explanation of some externally
individuated states by other externally indi-
viduated states.
This view of subpersonal content has,
though, to be reconciled with the fact that
the first states in an organism involved in
the explanation of a particular visual
experience - retinal states in the case of
humans - are not externally individuated.
224
The reconciliation is effected by the pre-
supposed normal background, whose im-
portance to the understanding of content
we have already emphasized. An internally
individuated state, when taken together
with a presupposed external background,
can explain the occurrence of an externally
individuated state.
An externalist approach to subpersonal
content also has the virtue of providing a
satisfying explanation of why certain per-
sonal-level states are reliably correct in
normal circumstances. If the sub personal
computations that cause the subject to be in
such states are reliably correct, and the
final computation is of the content of the
personal-level state, then the personal-level
state will be reliably correct. A similar point
applies to reliable errors, too, of course. In
either case, the attribution of correctness
conditions to the subpersonal states is
essential to the explanation.
Externalism generates its own set of
issues that need resolution, notably in the
epistemology of attributions. A content-
involving state may be externally individu-
ated, but a thinker does not need to check
on his relations to his environment to know
the content of his beliefs, desires and per-
ceptions. How can this be? A thinker's judg-
ments about his own beliefs are rationally
responsive to his own conscious beliefs. It is
a first step to note that a thinker's beliefs
about his own beliefs will then inherit
certain sensitivities to his environment that
are present in his original (first-order)
beliefs. But this is only the first step, for
many important questions remain. How can
there be conscious externally individuated
states at all? Is it legitimate to infer from the
content of one's states to certain general
facts about one's environment, and if so
how, and under what circumstances?
Ascription of attitudes to others also
needs further work on the externalist treat-
ment. In order knowledgeably to ascribe a
particular content-involving attitude to
another person, we certainly do not need to
have explicit knowledge of the external rela-
tions required for correct attribution of the
attitude. How then do we manage it? Do we
have tacit knowledge of the relations on
which content depends. or do we in some
way take our own case as primary. and
think of the relations as whatever underlies
certain of our own content-involving states?
If the latter. in what wider view of other-
ascription should this point be embedded?
Resolution of these issues. like so much else
in the theory of content. should provide us
with some understanding of the conception
each one has of himself as one mind
amongst many. interacting with a common
world which provides the anchor for the
ascription of contents.
See also ACTION; BELIEF; CONTENT (2);
DAVIDSON; FOLK PSYCHOLOGY; INTENTION;
TUDES; REPRESENTATION; THOUGHT.
BIBLIOGRAPHY
Burge. T. 1979. Individualism and the Mental.
Midwest Studies in Philosophy. Vol. 4. 73-
121.
Burge. T. 1986. Individualism and Psychol-
ogy. Philosophical Review. XCV(I). 3-45.
Davidson. D. 1984. Inquiries into Truth and
Interpretation. Oxford: Clarendon Press.
Davies. M. 1986. Externality. Psychological
Explanation and Narrow Content. Proceed-
ings of the Aristotelian Society Supplementary
Volume. 60. 263-83.
Dennett. D. 1987. The Intentional Stance. Cam-
Evans. G. 1982. The Varieties of Reference.
Fodor. J. 1981. Methodological SolipSism con-
sidered as a Research Strategy in Cognitive
Psychology. In Representations. Hassocks:
Harvester Press.
Fodor. J. 1991. A Modal Argument for Narrow
Content. Journal of Philosophy. LXXXVIII(I).
5-26.
Kripke. S. (1980). Naming and Necessity.
McDowell. J. 1984. De Re Senses. Philosophical
Quarterly. xxxiv. 283-94.
McDowell. J. 1986. Functionalism and Anom-
alous Monism. In Actions and Events: Per-
spectives on the Philosophy of Donald
Davidson. ed. E. Lepore and B. McLaughlin.
CONTENT (2)
Millikan. R. 1984. Language. Thought and Other
Biological Categories. Cambridge. MA.: MIT
Press.
Millikan. R. 1986. Thoughts without Laws:
Cognitive Science with Content. Philosophical
Review. 95.47-80.
Papineau. D. 1987. Reality and Representation.
Peacocke. C. 1993. Externalist Explanation.
Proceedings of the Aristotelian Society. XCIII.
203-30.
Putnam. H. 1975. The Meaning of 'Meaning'.
In Mind. Language and Reality. Cambridge
University Press.
Putnam. H. 1988. Representation and Reality.
Sellars. W. 1963. Empiricism and the Philoso-
phy of Mind. In Science. Perception and
Reality. London: Routledge & Kegan Paul.
CHRISTOPHER PEACOCKE
content (2) A central assumption in
much current philosophy of mind is that
PROPOSITIONAL ATTITUDES like BELIEFS
and DESIRES playa causal or explanatory
role in mediating between PERCEPTION
and behaviour (see REASONS AND CAUSES).
This causal-explanatory conception of pro-
positional attitudes. however. casts little
light on their representational aspect. The
causal-explanatory roles of beliefs and
desires depend on how they interact with
each other and with subsequent actions.
But the representational contents of such
states can often involve referential rela-
tions to external entities with which thin-
kers are causally quite unconnected. These
referential relations thus seem extraneous
to the causal-explanatory roles of mental
states. It follows that the causal-explana-
tory conception of mental states must
somehow be amplified or supplemented
if it is to account for representational
content.
INTERPRETATION AL SEMANTICS
Much research in ARTIFICIAL INTELLI-
GENCE aims to develop computational
225
CONTENT (2)
systems which can simulate the working of
other systems. For example. in arithmetical
calculators the causal relations between dif-
ferent calculator states simulate arithmet-
ical relations between numbers; in weather-
forecasting computers the causal relations
between computer states simulate meteoro-
logical interactions between weather condi-
tions; and so on (see COMPUTATIONAL
THEORIES OF MIND).
This has led some philosophers to suggest
that representation is a matter of interpreta-
tional semantics. by which they mean that
the states in a computational system can be
taken to represent any system of objects
whose relations are isomorphic to the
causal relations between the computational
states (see Cummins. 1989. ch. 8).
However. while this is arguably a satis-
factory account of the notion of representa-
tion used in artificial intelligence. it is
inadequate as an account of mental rep-
resentation. This is because. according to
interpretational semantics. every cognitive
system will represent an indefinite number
of other systems in addition to its 'intended
interpretation'. Thus. for example. an
adding machine can be interpreted. not
only as representing sums of integers. but
also as representing sums of even integers.
sums of multiples of three sums of money
accumulated in a bank account. sums of
distances travelled on train journeys. and so
on. ad infinitum.
In practice. of course. the beliefs of the
human users of the adding machine will
determine one of these as the appropriate
interpretation. But this only serves to
bring out the point that something beyond
interpretational semantics must determine
the representational features of human
beliefs. For if the beliefs of human users
can fix determinate representational con-
tents for computer states. then they will
need to have determinate contents them-
selves. And this cannot be because some
further interpreter fixes those contents -
for this would only set us off on a regress
- but because human beliefs have determi-
nate representational contents in their
own right.
226
CAUSAL SEMANTICS
A natural alternative suggestion is to
explain the representational contents of
human beliefs as those external conditions
that cause those beliefs. This suggestion in
effect adds an external dimension to the
causal-explanatory conception of mental
states. Instead of thinking of causal-explan-
atory roles purely in terms of those mental
interactions that occur inside the agent. we
also include interactions between mental
states and conditions external to the agent.
However. there is an obvious problem
facing this simple causal strategy. It has dif-
ficulty accounting for misrepresentation. that
is. for false beliefs. Take the belief that there
is a cow in front of you. say. This belief will
on occasion be caused by things other than
cows. such as gnus. plastic cows. holograms
of cows. and so on. But this then means
that. on the present suggestion. this belief
should stand. not just for the presence of a
cow. but rather for the disjunctive condi-
tion: cow-or-gnu-or-plastic-cow-or ... and
so on. for all the other possible causes of the
belief. And this would then make it impos-
sible for the belief to be falsely held. since
anything that can cause the belief will
thereby be counted as part of its disjunctive
truth condition (see CONCEPTS).
This 'disjunction problem' would be
soluble if we could somehow distinguish a
privileged set of 'typical' or 'ideal' circum-
stances for the formation of beliefs. For then
we could equate the truth conditions of
beliefs specifically with their causes in such
ideal circumstances. and thus leave room
for beliefs to be false when they arise from
other possible causes in non-ideal circum-
stances. It is doubtful. however. whether
there is any non-question-begging way of
picking out such ideal circumstances. For
there seems no principled way of specifying
such circumstances. except as those where
people form true beliefs. But this last
mention of representational truth will then
make the causal account of representational
content circular (see Cummins. 1989. ch. 4;
Fodor. 1990. ch. 2).
DRETSKE (1981) aims to solve the dis-
junction problem by identifying truth condi-
tions specifically with those conditions that
cause beliefs during learning. That is,
Dretske argues that truth conditions are
those conditions with which beliefs are asso-
ciated while we are acquiring the ability to
form them: other causes which operate after
learning is over are excluded from truth
conditions, and can hence give rise to false
beliefs. There are obvious difficulties facing
this theory, however: first, there is no clear
point at which learning stops; second, it
does not seem that a child will be dis-
qualified from learning the concept cow just
because it is shown a few gnus in the learn-
ing process (or, for that matter, just because
it is shown some pictures of cows); third, this
suggestion will not apply to any types of
beliefs that are innate (see INNA TENESS).
FODOR (1987, 1990) suggests a different
way of dealing with the disjunction
problem. He argues that our concept cow
stands for cows, and not cows-or-gnus,
because cows would still cause cow-
thoughts even if gnus didn't, but gnus
wouldn't cause cow-thoughts if cows didn't.
Fodor's idea is that the occurrence of cow-
thoughts depends on cows in a way it
doesn't depend on gnus. This suggestion has
been subject to a number of detailed objec-
tions (see Fodor, 1990, ch. 4). A more
general worry is that it implicitly pre-
supposes what it ought to be explaining:
while it is true that cow-thoughts depend
asymmetrically on cows rather than gnus,
this is more naturally seen as a consequence
of the representational significance of cow-
thoughts, rather than as the basis of it.
Causal theories of representation share an
important corollary with both traditional
verificationist theories of meaning and with
Donald DAVIDSON'S philosophical seman-
tics. This is because causal theories imply
that in general believers will tend to have
true beliefs. For if the truth conditions of
beliefs are those conditions that (typically)
cause those beliefs, then it follows that the
beliefs that humans are (typically) caused to
have will be true.
Verificationism and Davidson arrive at
similar conclusions by different routes.
CONTENT (2)
Verificationism analyses representation in
terms of the rules governing judgments,
rather than their causes. And Davidson
argues that interpretation should be gov-
erned by a 'principle of charity'. But all
three approaches take it that the truth con-
ditions of beliefs should be equated, roughly
speaking, with the conditions under which
those beliefs are held.
Because of this, they all agree that there
are principled reasons, flowing from the
nature of representation, why true beliefs
are the norm, and false beliefs the excep-
tion.
Some philosophers regard this as an
attractive consequence, since it reduces the
space available for sceptical arguments to
challenge our claims to knowledge.
However, other philosophers regard such
verificationist consequences as intrinsically
implausible, notwithstanding any desirable
epistemological consequences they may
have, and would prefer an analysis of rep-
resentation that avoids them.
TELEOLOGICAL SEMANTICS
As we have seen, causal theories of content
are stymied by the 'disjunction problem':
they find it difficult to distinguish, among
all the possible causes of a given belief,
those that are part of its truth condition
from those that are not. A number of philo-
sophers have sought to remedy this defect
by introducing teleological considerations
(Millikan, 1984; Papineau, 1984). This
account views beliefs as states with biologi-
cal purposes, and analyses their truth con-
ditions specifically as those conditions that
they are biologically supposed to covary
with. Our cow-beliefs stand for cows, and
not gnus, according to this teleological
account, because it is their purpose to be
held when cows, but not gnus, are present.
This teleological theory of representation
needs to be supplemented with a philo-
sophical account of biological teleology.
Defenders of the teleological theory of rep-
resentation generally favour a natural
selectionist account of biological purpose,
according to which item F has purpose G if
227
CONTENT (2)
and only if it is now present as a result of
past selection by some process which
favoured items with G. So a given belief type
will have the purpose of covarying with p
say, if and only if some mechanism has
selected it because it has covaried with p in
the past.
On this account of biological purposes,
the paradigm purposive traits are genes
whose effects have led to increases in their
frequency over generations. But such inter-
generational selection of genes is not the
only selection process that can give rise to
biological purposes. For learning is also a
natural selection process, namely a process
that favours items that give rise to reward-
ing results. So an item that has been
learned will for that reason have a biologi-
cal purpose. Most types of beliefs are learned
rather than innate, and the teleological
theory will therefore take their purposes to
derive from processes of learning rather
than inter-generational genetic selection.
When combined with the natural selec-
tion account of teleology, the teleological
theory of representation does have a
counterintuitive consequence. Imagine that
I suddenly acquire an 'accidental replica',
which coagulates out of passing atoms by
cosmic happenstance, but which is mol-
ecule-for-molecule identical to myself. Since
this being has no past, its states are not the
result of any selection processes, and so,
according to the teleological theory, will not
have any representational contents. Despite
its physical identity to myself, it won't share
my thoughts.
Though this consequence is unquestion-
ably counterintuitive, a number of re-
sponses are open to defenders of the
teleological theory. First, they can observe
that recent debates about 'broad contents'
provide independent grounds for doubting
that representational contents are always
fixed by ('supervene on') internal physical
make-up (see EXTERNALISM/INTERNALISM).
However, this is only part of an answer, for
the accidental replica implies a far more
radical failure of SUPERVENIENCE than the
familiar examples of broad contents.
Second, defenders of the teleological theory
228
can seek some alternative account of bio-
logical purpose, which does not depend on
past histories of natural selection, and
which would thereby detach the teleological
theory of representation from the natural
selection account of teleology. This would
allow them to argue that the accidental
replica's states do have biological purposes,
and hence representational contents, despite
their lack of selectional history. However,
prospects for a non-selectional account of
biological purposes seem poor. Third, defen-
ders of the teleological theory can argue
that their theory is intended as a theoret-
ical. rather than a conceptual, reduction of
the everyday notion of representation. As
such it is no objection that their theory has
counterintuitive consequences, any more
than it is an objection to the atomic theory
of matter that water's being made of hydro-
gen and oxygen runs counter to common
sense.
A more substantial objection to the tele-
ological theory is that it fails to solve the
disjunction problem. Consider the state in a
frog's brain which registers that a fly is
crossing its field of vision. This state also
responds to any small black dots. The tele-
ological theory holds that this state rep-
resents flies, rather than small black dots,
because it is the biological purpose of this
state to covary with flies. But, as Fodor has
asked (1990, ch. 3), what shows that the
biological purpose of this state is to covary
with flies, rather than with flies-or-any-
small-black-dots? After all, the frog's cog-
nitive system is not malfunctioning when it
responds to a black dot rather than a fly. So
why not say that state's purpose is to
respond to any small black dots (noting in
addition that this is a useful purpose in
environments where most small black dots
are flies)?
However, Fodor's thinking is here overly
influenced by the causal theory of repres-
entation. The purpose of a belief, like the
purpose of any other biological item, is
primarily a matter of the results produced
by the belief, not its causes. We need to ask
what benefit the belief provides, once it is
present, not what leads to the presence of
the belief in the first place. More specifically,
we need to focus on those conditions that
enable the belief to produce the advanta-
geous results which led to its selection. For
the frog-state, these conditions specifically
involve the presence of flies, not just any
black dots, since it is specifically when flies
are present that the frog-state leads the frog
to behave in ways that have advantageous
results.
Note that when the teleological theory is
interpreted in this way, the verificationist
thesis that beliefs generally tend to be true
no longer follows. Truth conditions are now
a matter of output - when will the belief
lead to advantageous results? - rather than
input - what conditions give rise to the
belief? - and there is no reason to suppose
that beliefs will normally, or even often, be
present in those conditions where they lead
to advantageous results. True, they must
sometimes have led to advantageous results
in the past, in order to have been favoured
by natural selection. But selection does not
require that advantageous results have a
high frequency, provided that sufficient
advantage accrues when they do occur. In
addition, the environment may have
changed since the initial selection in such a
way as to further reduce the frequency with
which advantageous results occur.
For example, if the frog-belief represents
flies, then it is true that, according to the
teleological theory, it must sometimes have
been co-present with flies in the past, other-
wise it could not have been selected for the
advantageous results it then produced. But
this still allows that the frog-belief may
nearly always occur in the absence of flies,
and so nearly always be false. Non-fly black
dots may always have triggered the belief
more often than flies, for even if the normal
outcome of the belief was a mouthful of leaf
or grit, rather than a fly, the state would
still have been selected if sufficient nutrition
accrued on those few occasions when a fly
was caught. And maybe the environmental
frequency of flies has fallen even further
since the state was selected, with the result
that the frog-belief is now triggered by non-
fly black dots even more often. To have the
CONTENT (2)
truth condition fly, the belief must have
advantageous results when it is copresent
with flies. But there is no reason why it
should usually have advantageous results,
and so no reason why it should usually be
true.
SUCCESS SEMANTICS
Some beliefs have biological purposes that
do not require them to be true, thus
casting doubt on the teleological theory's
equation of truth conditions with the
conditions under which beliefs serve their
biological purposes. Consider the not
uncommon belief, found among those
facing some imminent trial of violence, that
they will not be injured in the ensuing
conflict. It is arguable that humans are bio-
logically prone to form this belief, in order
to ensure that they will not flinch in battle.
But this purpose, of stopping them flinch-
ing, will be fulfilled even in cases where
they do get injured, in apparent conflict
with the teleological theory of representa-
tion, which equates the truth condition of
the belief - that they won't get injured -
with the condition required for it to fulfil its
purpose.
This objection requires that we look more
closely at the different ways in which beliefs
can give rise to advantageous results. It will
turn out that the teleological theory of
representation needs to be supplemented by
a 'success semantics', which equates the
truth conditions of beliefs specifically with
the conditions under which beliefs will lead
to the satisfaction of desires.
The primary way in which a belief can
lead to advantageous results is by com-
bining with a desire to generate an ACTION
that succeeds in satisfying that desire
because the belief is true. But some beliefs
also have secondary purposes, apart from
their primary purpose of enabling the satis-
faction of desires, to which their truth is
irrelevant. The belief about injury is a case
in point: it has the secondary purpose of
getting people to fight effectively, whether or
not they desire to, and it will do this inde-
pendently of whether it is true. (The reason
229
CONTENT (2)
that some beliefs have two biological pur-
poses is that there are some circumstances
in which normal human desires, like the
desire to avoid injury at all costs, are bio-
logically inappropriate. So natural selection
has compensated by inclining us to form
beliefs in those circumstances that will
make us act in ways which won't satisfy
those desires, but will satisfy our biological
needs. Cf. Papineau, 1993, ch. 3, 4.)
So, in response to the objection that some
beliefs have purposes that don't require
truth, defenders of the teleological theory
can say they are concerned specifically with
primary purposes, and can correspondingly
equate truth conditions specifically with
those conditions under which beliefs enable
the satisfaction of desires. That some beliefs
have other secondary purposes, to which
truth is irrelevant, does not discredit this
equation of truth conditions with conditions
that ensure desire satisfaction.
Some philosophers maintain that such a
'success semantics' offers a theory of repre-
sentation that is independent of the tele-
ological theory (cf. Whyte, 1990). Why not
simply analyse truth conditions, they ask,
as those conditions that ensure the satisfac-
tion of desires, without bringing in bio-
logical purposes?
The difficulty facing this position,
however, is to give some account of the
notion of satisfaction for desires. For satisfac-
tion is itself a representational notion,
which stands to desires as truth stands to
beliefs. So an account of representation that
simply helps itself to the notion of satisfac-
tion will to that extent be incomplete.
The teleological theory fills this gap by
saying that the satisfaction condition of a
desire is that result that it is the biological
purpose of the desire to produce. However,
it may be possible to fill this gap in other,
non-teleological ways. One possibility would
be to analyse the satisfaction conditions of
desires as those results that extinguish the
desire when they are known to obtain (cf.
Whyte, 1991). Another would be to analyse
satisfaction conditions as those results that
reinforce the performance of actions
prompted by the desire (cf. Dretske, 1988).
230
These alternatives all agree that truth
conditions should be analysed as those con-
ditions that ensure that actions will succeed
in satisfying desires, and differ only in their
accounts of desire satisfaction. This com-
mon commitment to 'success semantics'
raises a number of issues which call for
more detailed discussion (see Papineau,
1993, ch. 3, 6). In favour of this approach,
it should be noted that 'success semantics'
yields a natural explanation of the existence
of 'broad contents'. For which conditions
will satisfy an agent's desire will, according
to all the above suggestions, often depend
on the agent's environment and history as
well as the agent's physical make-up; and
so it follows that the truth conditions of
beliefs - that is, the conditions under which
actions issuing from those beliefs will
succeed in satisfying desires - will similarly
often depend on the agent's environment
and history.
See also CONTENT (1); An Essay on Mind
section 2.1; FOLK PSYCHOLOGY; REPRE-
SENTATION.
BIBLIOGRAPHY
Cummins, R. 1989. Meaning and Mental
Representation. Cambridge, MA. MIT Press.
Dretske, F. 1981. Knowledge and the Flow of
Information. Cambridge, MA: MIT Press.
Dretske, F. 1988. Explaining Behaviour. Cam-
MA. MIT Press.
Fodor, J. 1990. A Theory of Content. Cam-
Millikan, R. 1984. Language, Thought and Other
Biological Categories. Cambridge, MA.: MIT
Press.
Papineau, D. 1984. Representation and expla-
nation. Philosophy of Science, 51, 550-72.
Papineau, D. 1993. Philosophical naturalism
Whyte, J. 1990. Success semantics. Analysis,
50,149-57.
Whyte, J. 1991. The normal rewards of
success. AnalYSiS, 51, 65-73.
DAVID PAPINEAU
D
Davidson, Donald There are no such
things as minds, but people have mental
properties, which is to say that certain psy-
chological predicates are true of them.
These properties are constantly changing,
and such changes are mental events. Ex-
amples are: noticing that it is time for
lunch, seeing that the wind is rising,
remembering the new name of Cambodia,
deciding to spend next Christmas in Bots-
wana, or developing a taste for Trollope.
Mental events are, in my view, physical
(which is not, of course, to say that they are
not mental). This is a thesis that follows
from certain premises, all of which I think
are true. The main premises are:
(1) All mental events are causally related
to physical events. For example, BELIEFS
and DESIRES cause agents to act, and
ACTIONS cause changes in the physical
world. Events in the physical world often
cause us to alter our beliefs, INTENTIONS
and desires.
(2) If two events are related as cause and
effect, there is a strict law under which they
may be subsumed. This means: cause and
effect have descriptions which instantiate a
strict law. A 'strict' law is one which makes
no use of open-ended escape clauses such as
'other things being equal'. Thus such laws
must belong to a closed system: whatever
can affect the system must be included in it.
(3) There are no strict psychophysical
laws (laws connecting mental events under
their mental descriptions with physical
events under their physical descriptions).
Take an arbitrary mental event M. By (I),
it is causally connected with some physical
event P. By (2), there must be a strict law
connecting M and P; but by (3), that law
cannot be a psychophysical law. Since only
physics aims to provide a closed system gov-
erned by strict laws, the law connecting M
and P must be a physical law. But then M
must have a physical description - it must
be a physical event.
The three premises are not equally plaus-
ible. (1) is obvious. (2) has seemed true to
many philosophers; Hume and Kant are
examples, though their reasons for holding
it were very different. It has been questioned
by others. A detailed defence would be out
of place here, but the defence of (2) would
begin by observing that physics is defined
by the aim of discovering or devising a
vocabulary (which among other things
determines what counts as an event) which
allows the formulation of a closed system of
laws. I shall offer considerations in favour of
(3) below. It should be noted that (3) rules
out two forms of reductionism: reduction of
the mental to the physical by explicit defini-
tion of mental predicates in physical terms
(some forms of BEHA VIOURISM suggest such
a programme), and reduction by way of
strict bridging laws - laws which connect
mental with physical properties.
(1)-(3) do, however, imply ontological
reduction, since they imply that mental
entities do not add to the physical furniture
of the world. The result is ontological
monism coupled with conceptual dualism:
this is in many ways like Spinoza's meta-
physics. (However, Spinoza apparently re-
jected (1).) Because I deny that there are
strict psychophysical laws, I call my position
ANOMALOUS MONISM (Davidson, 1970).
What is the reason for the irreducibility of
mental concepts? Why can't there be strict
psychophysical laws? In trying to answer
these questions I first call attention to the
HOLISM of the mental. PROPOSITIONAL
ATTITUDES, in terms of which mental
events are characterized, cannot exist in
231
DAVIDSON, DONALD
isolation. Individual beliefs, intentions,
doubts and desires owe their identities in
part to their position in a large network of
further attitudes: the character of a given
belief depends on endless other beliefs;
beliefs have the role they do because of their
relations to desires and intentions and per-
ceptions. These relations among the atti-
tudes are essentially logical: the content of
an attitude cannot be divorced from what it
entails and what is entailed by it. This
places a normative constraint on the correct
attribution of attitudes: since an attitude is
in part identified by its logical relations, the
pattern of attitudes in an individual must
exhibit a large degree of coherence. This
does not, of course, mean that people may
not be irrational. But the possibility of ir-
rationality depends on a background of
rationality; to imagine a totally irrational
animal is to imagine an animal without
thoughts (Davidson, 1975, 1982).
We typically identify attitudes by using
sentences like 'Gertrude Stein thinks that
Ezra Pound is a village explainer', or 'Allen
Alker Read claimed that native speakers of a
language cannot make linguistic mistakes'.
An utterance' of such a sentence must, for
seman tical reasons, be analysed as rela-
tional: it relates a person to an entity that is
specified by uttering a contained sentence
('Ezra Pound is a village explainer', etc.).
This entity is often unhelpfully called a pro-
position. It is unhelpful, that is, until it is
explained exactly how the words in the con-
tained sentence manage to name or describe
a proposition, and here the proffered expla-
nations (including Frege's) lack conviction.
The reason for choosing a proposition as the
appropriate entity is that propositions sound
like the sort of things that can be 'before the
mind', 'entertained', or 'grasped'.
Aside from misleading grammar, there is,
however, no good reason to suppose that
having a propositional attitude requires an
entity which the mind entertains or grasps.
Having an attitude is just being in a certain
state; it is a modification of a person. There
need not be any 'object' in or before the
mind for the person to be thinking, doubt-
ing, intending or calculating. The object to
232
which an attitude attribution relates the
holder of the attitude must, of course, be
known, but it is only the attributor who
must know it. Such objects serve much the
same function as numbers serve in keeping
track of temperature or weight. There are
no such things as weights or temperatures;
'This box weighs 9 pounds' relates the box
to a number on the pound scale, but the
number is an abstract object unknown to
the box (Davidson, 1986, 1989a).
Anyone capable of attributing an attitude
has at his or her command an infinite set of
abstract objects suited to keeping track of
the attitudes of others: the sentences of his
or her language. (An alternative is to take
the relevant objects as the actual utterances
of sentences rather than the sentences;
there are advantages and disadvantages
either way. I explore the utterance option in
Davidson (1968). Success in interpretation
is always a matter of degree: the resources
of thought or expression available to an
interpreter can never perfectly match the
resources of the interpreted. We do the best
we can. It is always possible, of course, to
improve one's understanding of another, by
enlarging the data base, by adding another
dose of sympathy or imagination, or by
learning more about the things the subject
knows about. This is the process of radical
interpretation. There is no further court of
appeal, no impersonal objective standard
against which to measure our own best
judgments of the rational and the true.
Here lies the source of the ultimate differ-
ence between the concepts we use to
describe mental events and the concepts we
use to describe physical events, the differ-
ence that rules out the existence of strict
psychophysical laws. The physical world
and the numbers we use to calibrate it are
common property, the material and abstract
objects and events that we can agree on
and share. But it makes no sense to speak of
comparing, or coming to agree on, ultimate
common standards of rationality, since it is
our own standards in each case to which
we must turn in interpreting others. This
should not be thought of as a failure of
objectivity, but rather as the point at which
'questions come to an end'. Understanding
the mental states of others and under-
standing nature are cases where questions
come to an end at different stages. How we
measure physical quantities is decided inter-
subjectively. We cannot in the same way go
behind our own ultimate norms of ration-
ality in interpreting others. Priority is not
an issue. We would have no full-fledged
thoughts if we were not in communication
with others, and therefore no thoughts
about nature; communication requires that
we succeed in finding something like our
own patterns of thought in others (David-
son, 1991b).
I have been stressing rationality as coher-
ence, the fitting of one thought to another.
The need to find the thoughts of others
more or less coherent (by our own stan-
dards, it goes without saying) in order to
acknowledge and identify them as thoughts
is sometimes called a principle of charity.
The term is misleading, since there is no
alternative if we want to make sense of the
attitudes and actions of the agents around
us. The principle of charity has another
application, again with no implication of
goodness of heart. It is plain that we learn
what many simple sentences, and the terms
in them, mean through ostension. 'This is
green', 'That is thyme', 'It's raining', are
often learned in this way (perhaps first as
one-word sentences). It is my view that
such situations establish what the learner
correctly takes to be the meaning of these
sentences as spoken by the teacher. It is
irrelevant whether the teacher is speaking
as he was taught, or as others in the neigh-
bourhood or profession or family speak; as
long as the learner comes to associate sen-
tence with situation as the teacher does, he
is on the way to understanding that much
of what the teacher says. Similarly, commu-
nication between teacher and learner does
not depend on the learner speaking as the
teacher does; if the learner comes con-
sistently to utter his own same sentence in
situations in which the teacher utters his
own same sentence, communication has
taken hold (Davidson, 1986a).
I question, then, two familiar claims. One
DAVIDSON. DONALD
is that what someone means by what he
says depends only on what is in or on his
mind, and that the situations in which
words are learned merely constitute evi-
dence of what those words mean, rather
than conferring meaning on them. (To hold
that the situations in which words are
learned confers meaning on them is to
embrace a form of externalism (see EXTERN-
ALISM/INTERNALISM).) The other is that
what someone means depends, at least in
part, on what others in his linguistic com-
munity mean by the same words, even if
the speaker is ignorant of current or
'correct' usage. In this matter I am with
Allen Alker Read.
If everyone spoke in his or her own way
(as in fact they do to some degree), does this
suggest that there is no answer to Wittgen-
stein's question how there can be a differ-
ence between following a rule and thinking
one is following a rule? How can there be a
rule if only one person follows it? I'm not
sure the concept of a rule is idoneous, but
there certainly must be a way of distin-
guishing between correct and incorrect uses
of a sentence, cases where it is true and
where it is false. What is required, I think, is
not that people speak alike, though that
would serve. What is required, the basis on
which the concepts of truth and objectivity
depend for application, is a community of
understanding, agreements among speakers
on how each is to be understood. Such
'agreements' are nothing more than shared
expectations: the hearer expects the speaker
to go on as he did before; the speaker
expects the hearer to go on as before. The
frustration of these expectations means that
someone has not gone on as before, that is,
as the other expected. Given such a diver-
gence there is no saying who is wrong; this
must depend on further developments or
additional observers. But the joint expecta-
tions, and the possibility of their frustration,
do give substance to the idea of the differ- .
ence between being right and being wrong,
and to the concept of objective truth. They
therefore provide an answer to Wittgen-
stein's problem about 'following a rule'
(Davidson, 1992).
233
DAVIDSON, DONALD
For a speaker to follow a rule is, as I am
interpreting it, for the speaker to go on as
before; and this in turn means for the
speaker to go on as his audience expects,
and as the speaker intends his audience to
expect. (A finer analysis must allow for
cases in which the speaker goes on in a way
the audience does not anticipate, but in
which the audience nevertheless detects the
anomaly as intended by the speaker.) How,
though, can shared expectations be the
basis of the concept of objective truth?
All creatures are born making distinc-
tions. An infant from the start reacts differ-
entially to loud noises, the breast, and soon
to individual people and to certain facial
expressions. Similarities are not marked by
nature; it is we who find loud noises rele-
vantly similar, and who classify the re-
sponses of the infant as similar. If we ask
what, exactly, the infant is responding to,
the answer is that it is those objects or
events we naturally class together that are
best correlated with the responses of the
infant that we naturally class together. In
the end, we must ask this notion of what
comes naturally to do serious work. For
how do we decide whether the infant is
responding to the noise, or rather to the
vibrations of its eardrum, or to the signals
from the inner ear to the brain? It hardly
matters when we are in a position to specify
an appropriate stimulus at any of various
points along the causal chain from noise
source to brain. But if we think of responses
to the mother, most of us have no idea what
class of neural stimuli touch off the rele-
vantly similar responses; the best we can do
is to say it is the class of stimuli (sense data,
appearances, etc.) caused by the mother.
This is why, when we have taught the child
to say 'Mama' when stimulated by the
mother, we conclude that the child means
that its mother is present (rather than that
it is receiving a certain neural input).
The learning which confers meaning on
the most basic sentences necessarily
involves, then, three elements: the 'teacher'
(which may be a community of speakers
with no pedagogical intentions), the
'learner' (who may be entering a first lan-
234
guage, or consciously trying to decipher
another), and a shared world. Without the
external world shared through ostension,
there is no way a learner could discover how
speech connects with the world. Without a
'teacher', nothing would give content to the
idea that there is a difference between
getting things right and getting them wrong.
Only those who thus share a common world
can communicate; only those who commu-
nicate can have the concept of an inter-
subjective, objective world.
A number of things follow. If only those
who communicate have the concept of an
objective world, only those who commu-
nicate can doubt whether an external world
exists. Yet it is impossible seriously (con-
sistently) to doubt the existence of other
people with thoughts, or the existence of an
external world, since to communicate is to
recognize the existence of other people in a
common world. Language, that is, commu-
nication with others, is thus essential to
propositional thought. This is not because it
is necessary to have the words to express a
thought (for it is not); it is because the
ground of the sense of objectivity is inter-
subjectivity, and without the sense of objec-
tivity, of the distinction between true and
false, between what is thought to be and
what is the case, there can be nothing
rightly called thought (Davidson, 1990a,
1991a).
It is characteristic of mental states that
people usually know, without appeal to evi-
dence or inference, that they are in them.
(Let us call this characteristic 'PIRST-
PERSON AUTHORITY'.) The existence of first-
person authority is not an empirical dis-
covery, but rather a criterion, among
others, of what a mental state is. Among
others; so it can happen that we concede
error on occasion. But exceptions do not
throw in doubt the presumption that we
know our own minds. What accounts for
this presumption? INTROSPECTION offers no
solution, since it fails to explain why one's
perceptions of one's own mental states
should be any more reliable than one's per-
ceptions of anything else. The suggestion,
perhaps derived from Wittgenstein, that
mental predicates just have the property
that they are applied to oneself without
benefit of evidence. but are applied to others
on the basis of evidence. merely deepens the
mystery: for why should what we believe
without evidence be more certain than
what we believe on the basis of evidence?
And given that the grounds for attributing a
mental state to others are so different from
the grounds for attributing a mental state to
ourselves. why should we think they are the
same sort of state? I think the answer to this
conundrum is simple: we must interpret the
thoughts of others on the basis of evidence;
interpreting ourselves does not (aside from
special cases) make sense. It is. in theory. a
difficult empirical question how I know that
your sentence (and the thought it may
express) 'Snow is white' is true if and only if
snow is white. But if I am right about how
anyone comes to be interpretable. then I
must in general be right in thinking things
like this: my sentence 'Snow is white' is true
if and only if snow is white. The difference
between the two cases is that when I inter-
pret you. two languages are involved. yours
and mine (the same words may mean differ-
ent things in your language and mine). In
the second case. only one language is
involved. my own; interpretation is there-
fore not (exceptional instances aside) in the
picture (Davidson. 1984b).
The explanation of first-person authority
also shows why the external determinants
of meaning do not threaten our knowledge
of the character of our own thoughts. What
we mean by what we say is. if my account
of the role of learning in conferring
meaning is right. partly fixed by events of
which we may be ignorant. It does not
follow that we do not know what we mean.
for the content of what we think we mean
is determined by exactly the same circum-
stances that determine what we mean
(Davidson. 1987).
The topics I have discussed above are. in
my view. inseparable from many further
subjects. These include:
the philosophy of language. which
includes the concepts of truth and refer-
DAVIDSON. DONALD
ence. and has strong implications for
ontology (Davidson. 1986b).
the theory of action. which belongs with
the study of the mental because what
makes an event an action is that it is
intentional under some description. and
intentionality is defined and explained by
its relations to beliefs. affects. and inten-
tions (Davidson. 1980);
irrationality. for example WEAKNESS OF
THE WILL. wishful thinking. SELF-
DECEPTION. The existence of irrationality
creates puzzles for the account of prac-
tical reasoning and intention (Davidson.
1974. 1986b);
radical interpretation: the study of how it
is possible. given the interdependence of
mental states. for one person to come to
understand another (Davidson. 1990b).
See also An Essay on Mind section 3.6.2;
IDENTITY THEORIES; INTENTION ALITY; PHY-
sIcALIsM; QUINE; NATURALISM; REASONS
AND CAUSES; SUPERVENIENCE; THOUGHT;
BIBLIOGRAPHY
Davidson. D. 1968. On saying that. Synthese.
19. 130-46.
-- 1970. Mental events. In Experience and
Theory. ed. L. Foster and J. W. Swanson.
The University of Massachussetts Press and
Duckworth. Reprinted in Davidson. 1980.
-- 1974. Paradoxes of irrationality. In
Freud: A Collection of Essays. ed. R. Woll-
heim. New York: Doubleday Press.
-- 1975. Thought and talk. In Mind and
Language. ed. S. Guttenplan. Oxford Uni-
versity Press. Reprinted in Davidson. 1984.
-- 1980. Essays on Actions and Events.
-- 1982. Rational animals. Dialectica. 36.
317-27.
-- 1984a. First person authority. Dialectica.
38.101-111.
-- 1984b. Inquiries into Truth and Interpreta-
tion. Oxford University Press.
-- 1986a. A nice derangement of epitaphs.
In Philosophical Grounds of Rationality. ed. R.
235
DENNETT, DANIEL C.
Grandy and R. Warner. Oxford University
Press.
-- 1986b. Deception and division. In The
Multiple Self, ed. J. Elster. Cambridge Uni-
versity Press.
-- 1987. Knowing one's own mind. Pro-
ceedings and Addresses of the American Philo-
sophical Association, 441-58.
-- 1989a. The myth of the subjective. In
Relativism: Interpretation and Confrontation,
ed. M. Krausz. University of Notre Dame
Press.
-- 1989b. What is present to the mind? In
The Mind of Donald Davidson, ed. J. Brandl
and W. Gombocz. Grazer Philosophische
Studien Band, 36, 3-18.
-- 1990a. A coherence theory of truth and
knowledge. In Reading Rorty, ed. A. Mali-
chowski. Oxford: Basil Blackwell.
-- 1990b. The structure and content of
truth. Journal of Philosophy, 87, 279-328.
-- 1991a. Epistemology externalized. Dia-
lectica, 45, 19l-202.
-- 1991b. Three varieties of knowledge. In
A. J. Ayer Memorial Essays, ed. A. P. Grif-
fiths. Cambridge University Press.
-- 1992. The second person. Midwest
Studies in Philosophy, Vol 17.
DON ALD DA VIDSON
Dennett, Daniel C. In my opinion, the two
main topics in the philosophy of mind are
CONTENT and CONSCIOUSNESS. As the title
of my first book, Content and Consciousness
(1969) suggested, that is the order in which
they must be addressed: first, a theory of
content or INTENTIONALITY - a phenom-
enon more fundamental than consciousness
- and then, building on that foundation, a
theory of consciousness. Over the years I
have found myself recapitulating this basic
structure twice, partly in order to respond
to various philosophical objections, but
more importantly, because my research on
foundational issues in cognitive science led
me into different aspects of the problems.
The articles in the first half of Brainstorms
(1978a) composed in effect a more detailed
theory of content, and the articles in the
second half were concerned with specific
problems of consciousness. The second re-
236
capitulation has just been completed, with a
separate volume devoted to each half: The
Intentional Stance (1987a) is all and only
about content; Consciousness Explained
(1991 a) presupposes the theory of content
in that volume and builds an expanded
theory of consciousness.
BEGINNINGS AND SOURCES
Although quite a few philosophers agree
that content and consciousness are the two
main issues confronting the philosophy of
mind, many - perhaps most - follow tradi-
tion in favouring the opposite order: con-
sciousness, they think, is the fundamental
phenomenon, upon which all intentionality
ultimately depends. This difference of per-
spective is fundamental. infecting the intu-
itions with which all theorizing must begin,
and it is thus the source of some of the
deepest and most persistent disagreements
in the field. It is clear to me how I came by
my renegade vision of the order of depen-
dence: as a graduate student at Oxford, I
developed a deep distrust of the methods I
saw other philosophers employing, and
decided that before I could trust any of my
intuitions about the mind, I had to figure
out how the brain could possibly accom-
plish the mind's work. I knew next to
nothing about the relevant science, but I
had always been fascinated with how
things worked - clocks, engines, magic
tricks. (In fact, had I not been raised in
a dyed-in-the-wool 'arts and humanities'
academic family, I probably would have
become an engineer, but this option would
never have occurred to anyone in our
family.) So I began educating myself, always
with an eye to the curious question of how
the mechanical responses of 'stupid'
neurons could be knit into a fabric of activ-
ity that actually discriminated meanings.
Somehow it had to be possible, I assumed,
since it was obvious to me that DUALISM
was a last resort, to be postponed indefi-
nitely.
So from the outset I worked from the
'third-person point of view' of science, and
took my task to be building - or rather
sketching the outlines of - a physical struc-
ture that could be seen to accomplish the
puzzling legerdemain of the mind. At the
time - the mid-60s - no one else in philoso-
phy was attempting to build that structure,
so it was a rather lonely enterprise, and
most of the illumination and encourage-
ment I could find came from the work of a
few visionaries in science and engineering:
Warren McCulloch, Donald MacKay,
Donald Hebb, Ross Ashby, Allen Newell,
Herbert Simon, and J. Z. Young come to
mind. Miller, Galanter and Pribram's 1960
classic, Plans and the Structure of Behaviour,
was a dimly understood but much appre-
ciated beacon, and Michael Arbib's 1964
primer, Brains, Machines and Mathematics,
was very helpful in clearing away some of
the fog.
Given my lack of formal training in any
science, this was a dubious enterprise, but I
was usually forgiven my naivete by those
who helped me into their disciplines, and
although at the time I considered myself
driven by (indeed defined by) my disagree-
ments with my philosophical mentors,
QUINE and R YLE, in retrospect it is clear
that my deep agreement with both of them
about the nature of philosophy - so deep as
to be utterly unexamined and tacit - was
the primary source of such intellectual
security as I had.
The first stable conclusion I reached, after
I discovered that my speculative forays
always wandered to the same place, was
that the only thing brains could do was to
approximate the responsivity to meanings
that we presuppose in our everyday menta-
listic discourse. When mechanical push
came to shove, a brain was always going to
do what it was caused to do by current,
local, mechanical circumstances, whatever
it ought to do, whatever a God's-eye view
might reveal about the actual meanings of
its current states. But over the long haul,
brains could be designed - by evolutionary
processes - to do the right thing (from the
point of view of meaning) with high relia-
bility. This found its first published expres-
sion in Content and Consciousness (1969, 9,
'Function and Content') and it remains the
DENNETT. DANIEL C.
foundation of everything I have done since
then. As I put it in Brainstorms (19 78a),
brains are syntactic engines that can mimic
the competence of semantic engines. (See also
the thought experiment - a forerunner of
Searle's Chinese Room - about being locked
in the control room of a giant robot, in
1978b.) Note how this point forces the
order of dependence of consciousness on
intentionality. The appreciation of meanings
- their discrimination and delectation - is
central to our vision of consciousness, but
this conviction that I, on the inside, deal
directly with meanings turns out to be
something rather like a benign 'user illu-
sion'. What Descartes thought was most
certain - his immediate introspective grasp
of the items of consciousness - turns out to
be not even quite true, but rather a meta-
phorical by-product of the way our brains
do their approximating work. This vision
tied in beautifully with a doctrine of Quine's
that I had actually vehemently resisted as
an undergraduate: the indeterminacy of
radical translation. I could now see why, as
Quine famously insisted, indeterminacy was
'of a piece with' Brentano's thesis of the
irreducibility of the intentional, and why
those irreducible intentional contexts were
unavoidably a 'dramatic idiom' rather than
an expression of unvarnished truth. I could
also see how to re-interpret the two philoso-
phical works on intentionality that had had
the most influence on me, Anscombe's
Intention (1957) and Taylor's The Explana-
tion of Behaviour (1964).
If your initial allegiance is to the physical
sciences and the third-person point of view,
this disposition of the issues can seem not
just intuitively acceptable, but inevitable,
satisfying, natural. If on the other hand
your starting point is the traditional philo-
sophical allegiance to the mind and the
deliverances of introspection, this vision can
seem outrageous. Perhaps the clearest view
of this watershed of intuitions can be
obtained from an evolutionary perspective.
There was a time, before life on earth, when
there was neither intentionality nor con-
sciousness, but eventually replication got
under way and simple organisms emerged.
237
DENNETT. DANIEL C.
Suppose we ask of them: Were they con-
scious? Did their states exhibit intention-
ality? It all depends on what these key terms
are taken to mean, of course, but under-
neath the strategic decisions one might
make about pre-emptive definition of terms
lies a fundamental difference of outlook.
One family of intuitions is comfortable
declaring that while these earliest ancestors
were unconscious automata, not metaphy-
sically different from thermostats or simple
robotic toys, some of their states were
nevertheless semantically evaluable. These
organisms were, in my terms, rudimentary
intentional systems, and somewhere in the
intervening ascent of complexity, a special
subset of intentional systems has emerged:
the subset of conscious beings. According to
this vision, then, the intentionality of our
unconscious ancestors was as real as inten-
tionality ever gets; it was just rudimentary.
It is on this foundation of unconscious
intentionality that the higher-order com-
plexities developed that have culminated in
what we call consciousness. The other
family of intuitions declares that if these
early organisms were mere unconscious
automata, then their so-called intentionality
was not the real thing. Some philosophers
of this persuasion are tempted to insist that
the earliest living organisms were conscious
- they were alive, after all - and hence their
rudimentary intentionality was genuine,
while others suppose that somewhere
higher on the scale of complexity, real con-
sciousness, and hence real intentionality,
emerges. There is widespread agreement in
this camp, in any case, that although a
robot might be what I have called an inten-
tional system, and even a higher-order
intentional system, it could not be con-
scious, and so it could have no genuine
intentionality at all.
In my first book, I attempted to cut
through this difference in intuitions by pro-
posing a division of the concept of con-
sciousness into awarenessl, the fancy sort of
consciousness that we human beings enjoy,
and awareness20 the mere capacity for
appropriate responsivity to stimuli, a capa-
city enjoyed by honey bees and thermostats
238
alike. The tactic did not work for many
thinkers, who continued to harbour the
hunch that I was leaving something out;
there was, they thought, a special sort of
sensitivity - we might call it animal con-
sciousness - that no thermostat or fancy
robot could enjoy, but that all mammals
and birds (and perhaps all fish, reptiles,
insects, molluscs .... ) shared. Since robotic
devices of considerably greater behavioural
and perceptual complexity than the simplest
of these organisms are deemed unconscious
by this school of thought, it amounts to
some sort of latter-day vitalism. The more
one learns about how simple organisms
actually work, the more dubious this hunch
about a special, organic sort of sensation
becomes, but to those who refuse to look at
the science, it is a traditional idea that is
about as comfortable today as it was in the
seventeenth century, when many were hor-
rified by Descartes's claims about the
mechanicity of (non-human) animals. In
any event, definitional gambits are ineffec-
tive against it, so in later work I dropped the
tactic and the nomenclature of 'awarel' and
'aware2' - but not the underlying intuitions.
My accounts of content and conscious-
ness have subsequently been revised in
rather minor ways and elaborated in rather
major ways. Some themes that figured
heavily in Content and Consciousness lay
dormant in my work through the 70s and
early 80s, but were never abandoned, and
are now re-emerging, in particular the
theme of learning as evolution in the brain
and the theme of content being anchored in
distributed patterns of individually ambig-
uous nodes in networks of neurons. The
truth is that while I can fairly claim to have
seen the beauty, and indeed the inevit-
ability, of these ideas in Content and Con-
ciousness (see also Dennett, 1974), and to
have sketched out their philosophical impli-
cations quite accurately, I simply couldn't
see how to push them further in the scien-
tific domain, and had to wait for others -
not philosophers - to discover these ideas for
themselves and push them in the new direc-
tions that have so properly captured recent
philosophical attention. My own recent dis-
cuss ions of these two themes are to be found
in Dennett (1986, 1987b, 1991a, 1991b,
1991c, 1992a).
CONTENT: PATTERNS VISIBLE FROM THE
INTENTIONAL STANCE
My theory of content is functionalist (see
FUNCTIONALISM): all attributions of content
are founded on an appreciation of the func-
tional roles of the items in question in the
biological economy of the organism (or the
engineering of the robot). This is a specifi-
cally 'teleological' notion of function (not
the notion of a mathematical function or of
a mere 'causal role', as suggested by David
LEWIS and others). It is the concept of func-
tion that is ubiquitous in engineering, in the
design of artefacts, but also in biology. (It is
only slowly dawning on philosophers of
science that biology is not a science like
physics, in which one should strive to find
'laws of nature', but a species of engineer-
ing: the analysis, by 'reverse engineering',
of the found artefacts of nature - which are
composed of thousands of deliciously com-
plicated gadgets, yoked together opportunis-
tically but elegantly into robust, self-
protective systems.) These themes were all
present in Content and Consciousness, but
they were clarified in 'Intentional Systems'
(1971) when I introduced the idea that an
intentional system was, by definition, any-
thing that was amenable to analysis by a
certain tactic, which I called the intentional
stance. This is the tactic of interpreting an
entity by adopting the presupposition that it
is an approximation of the ideal of an opti-
mally designed (Le. rational) self-regarding
agent. No attempt is made to confirm or dis-
confirm this presupposition, nor is it neces-
sary to try to specify, in advance of specific
analyses, wherein consists RATIONALITY.
Rather, the presupposition provides leverage
for generating specific predictions of beha-
viour, via defeasible hypotheses about the
content of the control states of the entity.
My initial analysis of the intentional
stance and its relation to the design stance
and physical stance was addressed to a tra-
ditional philosophical issue - the problem of
DENNETT. DANIEL C.
free will and the task of reconciling
mechanism and responsibility (1973). The
details, however. grew out of my reflections
on practices and attitudes I observed to be
ubiquitous in ARTIFICIAL INTELLIGENCE.
Both Allen Newell (1982) and David Marr
(1982) arrived at essentially the same
breakdown of stances in their own reflec-
tions on the foundations of cognitive
science. The concept of intentional systems
(and particularly, higher-order intentional
systems) has been successfully exploited in
clinical and developmental psychology,
ethology, and other domains of cognitive
science, but philosophers have been reluc-
tant to endorse the main metaphysical
implications of the theory (see COGNITIVE
PSYCHOLOGY; DEVELOPMENTAL PSY-
CHOLOGY).
In particular, I have held that since any
attributions of function necessarily invoke
optimality or rationality assumptions, the
attributions of intentionality that depend on
them are interpretations of the phenomena -
a 'heuristic overlay' (1969), describing an
inescapably idealized 'real pattern' (1991d).
Like such abstracta as centres of gravity and
parallelograms of force, the BELIEFS and
DESIRES posited by the highest stance have
no independent and concrete existence. and
since this is the case, there would be no
deeper facts that could settle the issue if -
most improbably - rival intentional inter-
pretations arose that did equally well at
rationalizing the history of behaviour of an
entity. Quine's thesis of the indeterminacy
of radical translation carries all the way in,
as the thesis of the indeterminacy of radical
interpretation of mental states and pro-
cesses.
The fact that cases of radical indetermi-
nacy. though possible in principle, are
vanishingly unlikely ever to confront us is
small solace, apparently. This idea is deeply
counterintuitive to many philosophers, who
have hankered for more 'realistic' doctrines.
There are two different strands of 'realism'
that I have tried to undermine:
(1) realism about the entities purportedly
described by our everyday mentalistic
239
DENNETT, DANIEL C.
discourse - what I dubbed FOLK-
PSYCHOLOGY (1981) - such as beliefs,
desires, pains, the self;
(2) realism about content itself - the idea
that there have to be events or entities
that really have intentionality (as
opposed to the events and entities that
only behave as if they had intention-
ality),
Against (1), I have wielded various argu-
ments, analogies, parables. Consider what
we should tell the benighted community of
people who speak of 'having fatigues' where
we speak of being tired, exhausted, etc.
(1978a). They want us to tell them what
fatigues are, what bodily states or events
they are identical with, and so forth. This is
a confusion that calls for diplomacy, not
philosophical discovery; the choice between
an 'eliminative materialism' (see ELIM-
INATIVISM) and an 'IDENTITY THEORY' of
fatigues is not a matter of which 'ism' is
right, but of which way of speaking is most
apt to wean these people of a misbegotten
feature of their conceptual scheme.
Against (2), my attack has been more
indirect. I view the philosophers' demand
for content realism as an instance of a
common philosophical mistake: philoso-
phers often manreuvre themselves into a
position from which they can see only two
alternatives: infinite regress versus some
sort of 'intrinsic' foundation - a prime
mover of one sort or another. For instance,
it has seemed obvious that for some things
to be valuable as means, other things must
be intrinsically valuable - ends in them-
selves - otherwise we'd be stuck with a
vicious regress (or circle) of things valuable
only as means. It has seemed similarly
obvious that although some intentionality is
'derived' (the aboutness of the pencil marks
composing a shopping list is derived from
the intentions of the person whose list it is),
unless some intentionality is original and
underived, there could be no derived inten-
tionality.
There is always another alternative,
which naturalistic philosophers should look
on with favour: a finite regress that peters
240
out without marked foundations or thresh-
olds or essences. Here is an easily avoided
paradox: every mammal has a mammal for
a mother - but this implies an infinite gen-
ealogy of mammals, which cannot be the
case. The solution is not to search for an
essence of mammalhood that would permit
us in principle to identify the Prime
Mammal. but rather to tolerate a finite
regress that connects mammals to their
non-mammalian ancestors by a sequence
that can only be partitioned arbitrarily. The
reality of today's mammals is secure
without foundations.
The best known instance of this theme in
my work is the idea that the way to explain
the miraculous-seeming powers of an intel-
ligent intentional system is to decompose it
into hierarchically structured teams of ever
more stupid intentional systems, ultimately
discharging all intelligence-debts in a fabric
of stupid mechanisms (1971; 1974, 1978a,
1991a). Lycan (1981) has called this view
homuncular functionalism. One may be
tempted to ask: are the subpersonal compo-
nents real intentional systems? At what
point in the diminution of prowess as we
descend to simple neurons does real inten-
tionality disappear? Don't ask. The reasons
for regarding an individual neuron (or a
thermostat) as an intentional system are
unimpressive, but not zero, and the security
of our intentional attributions at the highest
levels does not depend on our identifying a
lowest-level of real intentionality. Another
exploitation of the same idea is found in
Elbow Room (1984): at what pOint in evolu-
tionary history did real reason-appreciators,
real selves, make their appearance? Don't
ask - for the same reason. Here is yet
another, more fundamental. version: at
what point in the early days of evolution
can we speak of genuine function, genuine
selection-for and not mere fortuitous pre-
servation of entities that happen to have
some self-replicative capacity? Don't ask.
Many of the most interesting and important
features of our world have emerged, gradu-
ally, from a world that initially lacked them
- function, intentionality, consciousness,
morality, value - and it is a fool's errand to
try to identify a first or most-simple instance
of the 'real' thing. It is for the same reason
a mistake to suppose that real differences in
the world must exist to answer all the ques-
tions our systems of content attribution
permit us to ask. Tom says he has an older
brother living in Cleveland and that he is an
only child (1975b). What does he really
believe? Could he really believe that he had
a brother if he also believed he was an only
child? What is the real content of his mental
state? There is no reason to suppose there is
a principled answer.
The most sweeping conclusion I have
drawn from this theory of content is that
the large and well-regarded literature on
PROPOSITION AL ATTITUDES (especially the
debates over wide versus narrow content,
'de re versus de dicto' attributions, and what
Pierre believes about London) is largely a
disciplinary artefact of no long-term impor-
tance whatever, except perhaps as history's
most slowly unwinding unintended reductio
ad absurdum. By and large the disagree-
ments explored in that literature cannot
even be given an initial expression unless
one takes on the assumptions I have argued
are fundamentally unsound (see especially
1975b, 1978a, 1982, 1987b, 1991d):
strong realism about content, and its con-
stant companion, the idea of a LANGUAGE
OF THOUGHT, a system of mental representa-
tion that is decomposable into elements
rather like terms, and larger elements
rather like sentences. The illusion that
this is plausible, or even inevitable, is
particularly fostered by the philosophers'
normal tactic of working from examples of
'believing-that-p' that focus attention on
mental states that are directly or indirectly
language-infected, such as believing that
the shortest spy is a spy, or believing that
snow is white. (Do polar bears believe that
snow is white? In the way we do?) There
are such states - in language-using human
beings - but they are not exemplary or
foundational states of belief; needing a term
for them, I call them opinions (,How to
Change your Mind', in 1978a; see also
1991c). Opinions playa large, perhaps even
decisive, role in our concept of a person, but
DENNETT, DANIEL C.
they are not paradigms of the sort of cogni-
tive element to which one can assign
content in the first instance. If one starts, as
one should, with the cognitive states and
events occurring in non-human animals,
and uses these as the foundation on which
to build theories of human cognition, the
language-infected states are more readily
seen to be derived, less directly implicated in
the explanation of behaviour, and the chief
but illicit source of plausibility of the doc-
trine of a language of thought. Postulating
a language of thought is in any event a
postponement of the central problems of
content ascription, not a necessary first
step. (Although a few philosophers - espe-
cially Millikan, Robert STALNAKER, Stephen
White - have agreed with me about large
parts of this sweeping criticism, they have
sought less radical accommodations with
the prevailing literature.)
CONSCIOUSNESS AS A VIRTUAL MACHINE
My theory of consciousness has undergone
more revisions over the years than my
theory of content. In Content and Conscious-
ness the theory concentrated on the role of
language in constituting the peculiar but
definitive characteristics of human con-
sciousness, and while I continue to argue for
a crucial role of natural language in gen-
erating the central features of consciousness
(our kind), my first version overstated the
case in several regards. For instance, I went
slightly too far in my dismissal of mental
imagery (see the corrections in 1978a,
1991a), and I went slightly too fast .- but
not too far! - in my treatment of colour
vision, which was unconvincing at the time,
even though it made all the right moves, as
recent philosophical work on colour has
confirmed, in my opinion. But my biggest
mistake in Content and Consciousness was
positing a watershed somewhere in the
brain, the 'awareness line', with the follow-
ing property: revisions of content that
occurred prior to crossing the awareness
line changed the content of consciousness;
later revisions (or errors) counted as post-
experiential tamperings; all adjustments of
241
DENNETT, DANIEL C.
content, veridical or not, could be located, in
principle, on one side or the other of this
postulated line. The first breach of this intu-
itive but ultimately indefensible doctrine
occurred in 'Are Dreams Experiences?'
(1975a), in which I argued that the distinc-
tion between proper and improper entry into
memory (and thence into introspective
report, for instance) could not be sustained
in close quarters. Related arguments
appeared in 'Two Approaches to Mental
Imagery' (in 1978a) and 'Quining Qualia'
(1988), but only in Consciousness Explained
(1991a) and 'Time and the Observer'
(Dennett and Kinsbourne. 1992) was an
alternative positive model of consciousness
sketched in any detail. the Multiple Drafts
model.
The best way to understand this model is
in contrast to the traditional model. which I
call the Cartesian Theatre. The fundamental
work done by any observer can be char-
acterized as confronting something 'given'
and taking it - responding to it with one
interpretive judgment or another. This
corner must be turned somehow and some-
where in any model of consciousness. On
the traditional view. all the taking is de-
ferred until the raw given. the raw materi-
als of stimulation, have been processed in
various ways and sent to central head-
quarters. Once each bit is 'finished' it can
enter consciousness and be appreciated for
the first time. As C. S. Sherrington (1934)
put it:
The mental action lies buried in the
brain. and in that part most deeply
recessed from outside world that is
furthest from input and output.
In the Multiple Drafts model. this single
unified taking is broken up in cerebral space
and real time; the judgmental tasks are
fragmented into many distributed moments
of micro-taking (Dennett and Kinsbourne.
1992). Since there is no place where 'it all
comes together'. no line the crossing of
which is definitive of the end of pre-
conscious processing and the beginning of
conscious appreciation. many of the familiar
242
philosophical assumptions about the deni-
zens of human phenomenology turn out to
be simply wrong. in spite of their traditional
obviousness.
For instance. from the perspective pro-
vided by this model one can see more
clearly the incoherence of the absolutist
assumptions that make QUALIA seem like a
good theoretical idea. It follows from the
Multiple Drafts model that 'inverted spec-
trum' and 'absent qualia' thought experi-
ments. like the thought experiments
encountered in the propositional attitude
literature (Twin Earth. what Pierre believes.
beliefs about the shortest spy), are funda-
mentally misbegotten, and for a similar
reason: the 'common sense' assumption of
'realism' with regard to the mental items in
question - beliefs. in the first instance.
qualia. in the second - is too strong.
OVERVIEW
The intermediate ontological position I
recommend - I call it 'mild realism' - might
be viewed as my attempt at a friendly
amendment to Ryle's (1949) tantalizing but
unpersuasive claims about category mis-
takes and different senses of 'exist' (see espe-
cially 1969. ch. 1; 1991d). What do you
get when you cross a Quine with a Ryle? A
Dennett. apparently. But there is a novel
texture to my work. and an attitude. which
grows primarily. I think. from my paying
attention to the actual details of the sciences
of the mind - and asking philosophical
questions about those details. This base
camp in the sciences has permitted me to
launch a host of differently posed argu-
ments, drawing on overlooked considera-
tions. These arguments do not simply add
another round to the cycle of debate. but
have some hope of dislodging the traditional
intuitions with which philosophers pre-
viously had to start. For instance. from this
vantage point one can see the importance of
evolutionary models (1969. 1974. 1978a,
1983. 1984a. 1990b. 1991f) and. con-
comitantly. the perspective of cognitive
science as reverse engineering (1989. 1991.
1992a). which goes a long way to over-
coming the conservative mindset of pure
philosophy. The idea that a mind could be a
contraption composed of hundreds or thou-
sands of gadgets takes us a big step away
from the overly familiar mind presupposed
by essentially all philosophers from Des-
cartes to the present.
Something else of mine that owes a debt
to Quine and Ryle is my philosophical style.
No sentence from Quine or Ryle is ever dull.
and their work always exhibits the import-
ance of addressing an audience of non-philo-
sophers. even when they know that
philosophers will be perhaps 95% of their
actual and sought-for audience. They also
both embody a healthy scepticism about the
traditional methods and presuppositions of
our so-called discipline. an attitude to which
I have always resonated. I have amplified
these points. attempting to follow their
example in my own writing. But I have also
been self-conscious about philosophical
methods and their fruits. and presented my
reflections in various meta-level digressions.
in particular about the role of intuition
pumps in philosophy (1980. 1984a.
1991a). and about the besetting foible of
philosophers: mistaking failures of imagina-
tion for insights into necessity.
My insistence on the need for philoso-
phers to stoke up on the relevant science
before holding forth. and my refusal to
conduct my investigations by the traditional
method of definition and formal argument.
have made me a distinctly impure philoso-
pher of mind. Moreover. on both main
topics. content and consciousness. I main-
tain a 'radical' position. which in a rather
lonely and implausible fashion declares that
much of the work at the presumed cutting
edge is beyond salvage. I thus cut myself off
from some of the controversies that capture
the imaginations of others in the field. but
the philosophical problems that arise
directly in non-philosophical research in
cognitive science strike me as much more
interesting. challenging. and substantive. So
I concentrate on them: the frame problem
(1984b. 1991e). problems about mental
imagery and 'filling in' (1992b). the binding
problem and the problem of temporal
DENNETT, DANIEL C.
anomalies (1991a; Dennett and Kinsbourne.
1992). I take these to be the real. as opposed
to artefactual. problems of mental repres-
entation. and I encourage philosophers of
mind to contribute to their solution.
BIBLIOGRAPHY
Basil Blackwell.
Dennett. D.C. 1969. Content and Consciousness.
London: Routledge & Kegan Paul.
-- 1971. Intentional systems. Journal of Phi-
losophy, 8. 87-106.
-- 1973. Mechanism and responsibility. In
Essays on Freedom of Action. ed. T. Honder-
ick. London: Routledge & Kegan Paul.
-- 1974. Why the law of effect will not go
away. Journal of the Theory of Social Beha-
viour, 5. 169-87.
-- 1975. Are dreams experiences? Philoso-
phical Review. 73. 151-71.
-- 1975b. Brain writing and mind reading.
In Language, Mind, and Meaning. Minnesota
Studies in Philosophy of Science. Vol. 7, ed. K.
Gunderson. Minneapolis: University of Min-
nesota Press.
-- 1978a. Brainstorms: Philosophical Essays
on Mind and Psychology. Montgomery. VT:
Bradford.
--- 1978b. Current issues in the philosophy
of mind. American Philosophical Quarterly.
15. 249-61.
-- 1980. The milk of human intentionality.
Behavioral and Brain Sciences, 3, 428-30.
-- 1982. Beyond belief. In Thought and
Object, ed. A. Woodfield. Oxford University
Press.
-- 1983. Intentional systems in cognitive
ethology: the 'Panglossian paradigm' de-
fended. Behavioral and Brain Sciences, 6,
343-90.
-- 1984a. Elbow Room: The Varieties of Free
Will Worth Wanting. Cambridge. MA: MIT
Press.
-- 1984b. Cognitive wheels: the frame prob-
lem of AI. In Minds, Machines and Evolution.
ed. C. Hookway. Cambridge University Press.
-- 1986. The logical geography of compu-
tational approaches: A view from the East
Pole. In The Representation of Knowledge and
Belief. ed. R. Harnish and M. Brand. Tucson:
University of Arizona Press.
243
DESIRE
--1987a. The Intentional Stance. Cambridge.
MA.: MIT Press.
-- 1987b. Evolution. Error and Intention-
ality. In 1987a.
-- 1988. Quining qualia. In Consciousness in
Contemporary Science. ed. A. Marcel and E.
Bisiach. Oxford University Press.
-- 1990a. Memes and the exploitation of
imagination. Journal of Aesthetics and Art
Criticism. 48.127-35.
-- 1990b. The interpretation of texts.
people and other artifacts. Philosophy and
Phenomenological Research. SO. 177-94.
-- 1991a. Consciousness Explained. Boston:
Little Brown.
-- 1991b. Mother Nature versus the
walking encyclopedia. In Philosophy and
Connectionst Theory. ed. W. Ramsey. S. Stich
and D. Rumelhart. Hillsdale. NJ: Erlbaum.
-- 1991c. Two contrasts: folk craft versus
folk science and belief versus opinion. in The
Future of Folk Psychology: Intentionality and
Cognitive Science. ed. J. Greenwood. Cam-
bridge University Press.
-- 1991d. Real patterns. Journal of Philoso-
phy. 89. 27-51.
-- 1991e. Producing future by telling
stories. In The Robot's Dilemma Revisited: The
Frame Problem in ArtifiCial Intelligence. ed.
K. M. Ford and Z. Pylyshyn. Norwood. NJ:
Ablex.
-- 1991f. Ways of establishing harmony. In
Dretske and his Critics. ed. B. McLaughlin.
-- 1992a. Cognitive science as reverse
engineering: Several senses of 'top-down'
and 'bottom-up'. In Proc. of the 9th Interna-
tional Congress of Logic. Methodology and Phi-
losophy of Science. ed. D. Prawitz. B. Skyrms.
and D. Westerstahl. North-Holland.
-- 1992b. Filling in versus finding out: A
ubiquitous confusion in cognitive science. In
Cognition. ed. H. Pick et aI.. Washington DC:
American Psychological Association.
Dennett. D.C .. and Kinsbourne. M. 1992. Time
and the observer: The where and when of
consciousness in the brain. Behavioral and
Brain Sciences. IS. 183-201.
Lycan. W.G. 1981. Form. function. and feel.
Journal of Philosophy. 78. 24-49.
Marr. D. 1982. Vision. San Francisco:
Freeman.
Miller. G .. Galanter. E .. and Pribram. K. 1960.
244
Plans and the Structure of Behavior. New
York: Holt. Rinehart and Winston.
Newell. A .. 1982. The knowledge level. Artifi-
cial Intelligence. 18.81-132.
Ryle. G. 1949. The Concept of Mind. London:
Hutchinson.
Sherrington. C.S. 1934. The Brain and Its
Mechanism. London: Hamilton.
Taylor. C. 1964. The Explanation of Behaviour.
London: Routledge & Kegan Paul.
DANIEL C. DENNETT
desire We contrast what we want to do. to
begin with. with what we must do -
whether for reasons of morality or duty. or
even for reasons of practical necessity (to
get what we wanted in the first place).
Accordingly our own desires have seemed
to be the principal source of the ACTIONS
that most fully express our own individual
natures and will. and those for which we
are personally most responsible. But desire
has also seemed to be a principle of action
contrary to and at war with our better
natures. as rational and moral agents. For it
is principally from our own differing per-
spectives upon what would be good. that
each of us wants what he does. each point
of view being defined by one's own interests
and pleasures. In this. the representations of
desire are like those of sensory PERCEPTION.
similarly shaped by the perspective of the
perceiver and the idiosyncrasies of the per-
ceptual. or appetitive. apparatus. So the phi-
losophical dialectic about desire and its
object recapitulates that of perception and
sensible qualities. The strength of desire. for
instance. varies with the state of the subject
more or less independently of the character.
and the actual utility. of the object wanted.
Such facts cast doubt on the 'objectivity' of
desire. and on the existence of a correlative
property of goodness. inherent in the objects
of our desires. and independent of them.
Perhaps. as Spinoza put it. it is not that we
want what we think good. but that we
think good what we happen to want - the
'good' in what we want being a mere
shadow cast by the desire for it (Spinoza.
1677. III. prop. 9. scholium). (There is a
parallel Protagorean view of belief, similarly
sceptical of truth.) The serious defence of
such a view, however, would require a sys-
tematic reduction of apparent facts about
goodness to facts about desires. and an ana-
lysis of desire which in turn makes no refer-
ence to goodness. While that is yet to be
provided, moral psychologists have sought
to vindicate an idea of objective goodness.
for example as what would be good from all
points of view, or none; or, in the manner of
Kant, to establish another principle (THE
WILL, or PRACTICAL REASON) conceived as
an autonomous source of action, indepen-
dent of desire or its object; and this tradition
has tended to minimize the role of desire in
the genesis of action.
In the general philosophy of mind, and
more recently, desire has received new
attention from those who would understand
mental states in terms of their causal or
functional role in the determination of
rational behaviour, and in particular from
philosophers trying to understand the
semantic content or intentional character of
mental states in those terms (see FUNCTION-
ALISM; CONCEPTIONAL ROLE SEMANTICS).
The RATIONALITY of an action is deter-
mined by the contents of the BELIEFS and
desires that explain why the agent does it: if
something is done because the agent wants
something that he believes he can get by
doing that thing, the action is prima facie
reasonable. This is a truism; but perhaps an
account of the INTENTIONALITY of these
mental states can be extracted from it,
answering such questions as these: What is
it for a desire to be for the particular object
it's for? What determines that its object is the
particular thing it is? Does a desire have
some sense or CONTENT that determines
what its object is, rather as the meaning of
an expression determines its referent?
Of course. the object of the desire, Le. what
one wants, is the thing the possession of
which (or the state of affairs the obtaining
of which) would satisfy that desire. The
object of the desire is referred to by the
grammatical object of a sentence that attri-
DESIRE
butes the desire to some subject - a sen-
tence of the form 'She wants a cow'.
Generally, that phrase ('a cow') does not, on
the face of it, express a proposition - as the
grammatical object of 'She believes cows
give milk' ('cows give milk') does do. But in
fact the grammatical object of the desire
sentence is also a clause or sentence: she
wants it to be that she has a cow, as it
might be put. (An ambiguity introduced by
adverbs superficially modifying 'desire' pro-
vides grammatical evidence of this under-
lying subordinate clause with the un-
expressed verb 'have', which on one reading
of the sentence the adverb must be taken to
modify: 'She wants a cow today' might
mean that her desire today is for a cow,
whereas yesterday it was a goat she
wanted, or it might mean that she wants to
have it today. And of course if she wants a
cow her desire is that she should have one.)
The fact that sentences ascribing desires
have grammatical objects that express propo-
sitions does not entail that the states them-
selves are 'PROPOSITIONAL ATTITUDES',
which have those propositions as their
INTENSION AL objects. That conclusion is
drawn instead from the fact that the truth
of that proposition is what satisfies the
desire, and that the desire that P makes it
reasonable for the agent to act so as to
bring it about that P - specifically that state
of affairs, but not necessarily other, for
example coextensive, states of affairs that the
action also brings about. In any case, we
may distinguish (1) the grammatical object
of 'desire', from (2) the propositional object
of the desire, or its 'objective', and that
again, from (3) the representational content
of the desire, and (4) the object or state the
desire represents.
Now, what makes it the case that a given
desire has the object that it does? Here
'logical behaviourism' made a suggestive, if
flawed, beginning (see BEHAVIOURISM):
mental states are dispositions to behave in
appropriate ways, and they may be identi-
fied and distinguished by specifications of
the wayan agent in a given state is therein
disposed to behave, with respect to some-
thing 'external'. A desire might be a desire,
245
DESIRE
specifically, for an apple - that might be its
object - because a desire is a disposition to
behave in such a way as to get one: to try
to get an apple. (Anscombe 1957, p. 67:
'The primitive sign of wanting is trying to
get'.) A real behaviourist might have
thought this trying to get something was
behaviour recognizable as such without
assumptions about other inner states, but
that idea has long since been given up.
Someone's shaking a tree qualifies as an
attempt to get an apple, and as behaviour
'appropriate' to the desire for one, only if he
believes what he's doing might get him an
apple (Geach, 1957, p. 8). And that belief is
a disposition to do such things, but only
assuming he wants an apple.
Beliefs and desires make an essentially
coordinate contribution to the production of
rational behaviour, and the behavioural
effects of states of either kind can be identi-
fied by (though only by) reference to states
of the other kind; more precisely, to the
semantic character - the truth or the satis-
faction-condition - of causally operative
states of the other kind. Thus the actions a
desire is a disposition to perform are those
that would satisfy that desire provided the
agent's operative beliefs were true - those
being the beliefs (whatever they may be)
which, together with those desires, cause
the action in question, or comprise the
reasons for which he does it. And a belief, in
turn, is a state which, together with a rele-
vant desire, would dispose the agent to do
what would tend to satisfy that desire in
case the belief is true (Stalnaker, 1984, ch.
1). Thus the characterization of desire
depends essentially upon an interlocking
definition of belief. It may seem that this
approach to the object of desire is objection-
ably 'circular'. Be that as it may, the char-
acterization of the desire in terms of its
effects must be completed by a reference to
its typical causes - and this will acknowl-
edge factors that determine the objects of
desire independently of the beliefs referred
to in the definition. This reference to causes
is required, arguably, to do justice to the
status of desires as reasons (see REASONS
AND CAUSES).
246
For a desire isn't merely a disposition to
do certain things, it is a reason to do those
things. Irritability is also a disposition, to
react irritably - or solubility, a disposition to
dissolve - but it is not a reason to do so.
The desire is a reason for which we do what
it causes us to do. It must be explained how
this is so. A belief (that P) also comprises a
reason to act accordingly, and there, to
state what is believed, its 'propositional
content' (P), is to specify that reason. The
desire for an apple is likewise a reason to
act accordingly, but here, to cite what is
wanted - an apple; or that one have an
apple - is not to cite a reason for doing any-
thing. It is not to cite a fact, or an apparent
fact, that might figure as a reason to act. It
is not clear why wanting that object is a
reason to behave accordingly, until (beyond
its propositional object, or objective) the
representational content of the desire is made
clear. The belief that P contains a reason to
act accordingly because, in belief, it is re-
presented as being the case that P. The
desire that P is also a reason to act accord-
ingly, but what does it represent as being
the case? It does not, certainly, represent it
as being the case that P (see RE-
PRESENT A TION).
Some philosophers have denied that there
is any representational content in desire. If
no such representational content is found, it
may still be possible to identify the object of
the desire with the state which tends
(through action) to result from the desire,
specified more closely as a result which in
turn has the effect of reinforcing the ten-
dency of that desire to produce the beha-
viour that produced that result. (cf. Dretske,
1984, ch. 5.) This view yields an account of
the causal efficacy of the desire's intentional
character. If the desire's having an object 0
is identified with its being such that getting
o will reinforce the desire's tendency to
cause the behaviour that got 0, then it is its
being a desire for 0 that explains such
behaviour. But is the object of the desire
defined by the properties of 0 that reinforce
that desire's capacity to cause behaviour, or
instead by the properties of 0 that would
apparently benefit the agent, and thus
rationalize the behaviour it would cause?
(See below on the intensionality of the
context'S wants 0' read 'opaquely'.) It may
be questioned, then, whether this does
justice to the status of desires as reasons for
behaving in ways that would satisfy them.
It was Hume who most famously denied
that a desire (or any such 'passion') con-
tains ' ... any representative quality, which
renders it a copy of any other existence or
modification' (Hume, 1739). When I want
something, as when I am angry, 'I am actu-
ally possest with the passion, and in that
emotion have no more a reference to any
other object, than when I am thirsty, or
sick, or more than five foot high'. Thus
while a desire based on some unreasonable
beliefs may be described, metonymically, as
'unreasonable', strictly, and in its own
right, no desire can be contrary to reason,
or conformable to it. The premise from
which these dramatic pronouncements
were derived is that desire represents
nothing as being the case, and thus 'has no
reference to' truth. Whereas the belief that
there's a pot of gold at the end of the
rainbow is true, in case what is believed is
true, the desire for the pot of gold is not
'true' in case what is wanted (that one gets
that pot of gold) is true.
But if one comes to have the pot of gold,
the desire will come true; it will be satisfied,
and the desire's being satisfied (which it is if
and only if the propositional complement of
'desire' is true) is plainly parallel to the
belief's being 'true', in case the proposi-
tional complement of 'believes' is true.
That's one sort of 'reference' the desire has
to truth, though it is truth by the name
'satisfaction'. (The paradigm causal relation
between a desire and its satisfaction condi-
tion differs, of course, from that between a
belief and its truth condition; while the
belief tends to result from states in which its
truth condition holds, for example through
the medium of perception, a desire tends to
result in states in which its satisfaction con-
dition holds, through the medium of
action.)
It is true that the desired pot of gold
needn't be an 'existing reality' and neither
DESIRE
is the envisioned possession of it an existing
state of affairs; nor are they represented as
such in the desire to have a pot of gold. But
the claim that desire is not a representation
of any existing reality, or state of affairs, in
the way that a perceptual belief may be, is
irrelevant to the question whether a desire
might be reasonable. A belief is reasonable,
roughly, if one has reason to believe that
things are as they are believed to be, or
therein represented as being. And a desire is
reasonable if one has reason to want the
thing wanted, that is, to want the desired
state of affairs (wherein one has a pot of
gold) to be 'an existing reality'. A reason to
want the thing would be that it would be
good to have it. Then why should that not
be what a desire represents as being the
case? If so, then there is something the
desire represents, or misrepresents
namely, the benefit or utility, if you like, of
having a pot of gold. And we may say that
the object of a desire (whether it exists
or not) is therein represented as something
the possession of which would be good. In the
desire that p, it is represented as being the
case that it would be good were it the case
that P. ('Good' drops out as redundant in
the context'S wants [= mentally represents
it as good that] p', as 'true' drops out as
redundant in'S believes [mentally repres-
ents it as true that] P'.) Of course it needn't
be true that it would be good if P were the
case, in order for someone to want it to be
so. If desires are, in effect, utility indicators,
they need not be, and they are not, entirely
reliable ones. (There are other views of the
representational content of desire, including
the important Socratic view: the desire that
P represents its being the case that P as
what would be best. Cf. Plato's Gorgias,
467c-468e; Lysis, 279d-282a.)
Nor is it obvious that such representa-
tional content can belong to desire only via
some belief. A desire might result 'directly'
from its being the case that it would be
good to have the thing wanted (or its being
such as to make it seem good), rather in the
way a perception may have representa-
tional content without intervening belief.
Something can look green to someone
247
DESIRE
without the perceiver's believing it is green
or even that it looks green. Further, that
perception is in itself a reason to believe the
thing is green, regardless of one's contrary
beliefs. Similarly, perhaps, in the desire for
sleep, it seems as if it would be good to sleep,
and that, in itself, is a reason to sleep,
regardless what one believes or even knows
about the merits of sleeping, and even if one
has better reason to stay awake (Stampe,
1987). This argues for a certain 'modu-
larity' in the generation of desires.
The treatment of desire as a kind of per-
ceptual state may be pressed further, identi-
fying some state of the subject - for
example, some state of deprivation or need -
as the object of consciousness which causes
it to seem as if the thing wanted would be
good, as the object that is seen causes it to
look to one as if such and such is the case.
Here, then, through its connection with its
causes, desire gains an infusion of semantic
content independent of belief, breaking the
circle in which the contents of one are
defined in terms of contents of the other.
And at this point, finally, we may deny
what Hume asserts: the desire that it be the
case that P does represent an actually
'existing reality', sc. some state of the
subject that makes it seem to him as if it
would be good if it were the case that P.
In his contention that 'Reason is and
ought only to be the slave of the pas-
sions .. .' - desire being a passion - Hume
meant to tweak an orthodox rationalism
that pitted reason against desire, as if
reason were an autonomous principle of
action. A more egalitarian view is Aris-
totelian. Reason is not something on the
side of judgment (or belief), but is rather a
principle operative alike as beliefs give rise
to further beliefs, or when desires give rise
to intention and thus to action. In either
case, the rational inference secures some
relevant semantic value to its product: it
guarantees the truth of beliefs validly
derived from true beliefs, i.e. provided things
are as they were therein represented as
being; and it guarantees similarly that the
action deriving from a desire will be good if
what one would thereby get is as it was
248
represented as being, in that desire: as
something that it would be good to have. A
deSire, like a belief, is itself a reason for the
person to do things that would satisfy that
desire. Of course, that desire apart, it may
be madness to do such things, and that's
why the desire, while it is a reason, may not
be a good reason, to act accordingly. A
representational view of desire accounts for
that fact. It is reasonable to act in accord-
ance with the belief that P because in that
belief it is represented as being the case that
P; by parity of reasoning, if it is reasonable
to do what would make it the case that p,
as one wants, that might likewise be owing
to the way the object of desire is repres-
ented, in that state of mind. If, in a desire,
its object (or one's having the thing
wanted), is represented as something good,
this falls into place.
(But there are apparent differences
between the principles of sound practical
reasoning and those of 'theoretical' (deduc-
tive or inductive) reasoning. Kenny (1966)
suggested that the decision to do something
believed merely sufficient (and not neces-
sary) to satisfy some desire is, relative to
that desire, a logically valid decision. Thus,
unlike a valid deductive inference, a valid
practical inference need not draw a conclu-
sion that is necessary if its premises are
true; instead, the conclusion must be suffi-
cient for the truth of a premise representing
some desire - that is, expressing the desire's
propositional object. This turns on differ-
ences between the semantic values con-
ferred on the conclusion by valid inferences
of one kind or the other - that is, between
truth on the one hand, and goodness or
satisfactoriness on the other.)
Further support for a representational
view of desire concerns the 'intensionality'
of the terms designating what is wanted in
sentences used to describe desires. '(CEdipus
wanted to marry the Queen' says something
true, but while the Queen was one and the
same person as his own mother, we cannot
substitute the phrase 'his own mother' for
the term 'the Queen'. Why should this be
so? Evidently, the term designating the
object of this desire is doing something
other than, or more than, referring to the
thing wanted. Frege held that the phrase
instead refers, here, to the meaning or sense
the phrase would ordinarily have, but this
leaves the question how a reference to the
sense of a phrase might serve to character-
ize the desire. An alternative view is that
the phrase here serves, as usual, to refer to
the person herself (thus to say who (CEdipus
wanted to marry) but in addition it may be
used, in this context, to provide a further
specification of the desire; and it may do this
by specifying the reason the desire com-
prises for acting so as to satisfy it. (It makes
sense that a desire should be identified in
this way if a desire is a reason for acting so
as to satisfy it.) Thus 'the Queen' identifies
or refers to the woman by mentioning prop-
erties of her the possession of which com-
prises a reason the subject would therein
have to act so as to satisfy the desire
ascribed to him. While her being the Queen
comprises such a reason, other properties
that distinguish her, such as her being his
own mother, are reasons for not marrying
the woman. Accordingly the substitution of
that phrase, though it refers to the same
woman, may change the statement from
true to false.
(We may ask, however, whether in iden-
tifying desires, co-referential expressions
may be freely substituted when they refer to
the object in terms of those of its properties
that reinforce the state's merely causal effi-
cacy, in the production of action - or must
they describe it in terms of properties of the
object that also make those actions reason-
able? Consider someone who likes to experi-
ence the heights of euphoria, and therefore
periodically wants to inject the most eu-
phoric drug there is, and does so. Suppose
that drug happens also to be the most pow-
erfully addictive of drugs, so that property of
it has reinforced the tendency of this desire
to cause the action that satisfies it. Now, he
wants to take the drug because - as a result
of the fact that - it is addictive and because
- Le. for the reason that - it is so euphoric.
But, narrowly specified, what is it that he
wants? Is it to take the drug that will most
powerfully addict him to its use? Or instead
DESIRE
to take the drug that will produce the great-
est euphoria? The issue is aired in Stampe
(1990) where Dretske responds.)
It seems that the causal power of a desire
may outrun its capacity to rationalize
actions done to satisfy it. There is a notion
of the strength of a desire, which is some-
thing not determined solely by the quality of
the reason the desire affords, so that there
may be such a thing as a desire that is
excessively powerful or insufficiently strong.
I want to smoke, which is a reason to
smoke, and I want not to offend my non-
smoking companion, which is a reason not
to smoke. The comparative quality of my
reasons to smoke and not to smoke depends
partly on how badly I want to do the one
thing or the other, but largely, also, on
factors external to those desires: on what
other desires I have, and beliefs about how
their satisfaction will be affected by my
satisfying the desires in question. I have a
better reason not to smoke, than to smoke,
if I also want a job, and my companion is
interviewing me for that job. Even so, the
desire to light up may overwhelm 'my
better judgment', if it is the more powerful
or forceful of the two desires.
It may be questioned whether such refer-
ences to the 'force' of a desire are genuinely
explanatory, and thus whether there really
is such a thing. Such doubts may be pro-
voked by the dogmatic assertion of 'the law
of preponderant desire', which says that the
more powerful desire will necessarily prevail
0. S. Mill, 1867, ch. 26.) This, it is objec-
ted, is no law of psychology, but an empty
tautology, for the 'stronger' desire is being
merely defined as the one that actually pre-
vails.
It would be a mistake to infer that the
idea of the strength of desire cannot be
given explanatory content. Other definitions
of such a magnitude are possible, which
would make the proposition that the
stronger desire must prevail synthetic, in
principle falsifiable, and, apparently, in fact
false. For instance, the desire's power, rela-
tive to other desires, might be measured by
the relative probability of each producing
actions intended to satisfy it, ceteris paribus -
249
DEVELOPMENTAL PSYCHOLOGY
that is, where other factors are the same,
such as the relative cost or difficulty of the
acts, the relative probability of their success,
their incidental benefits or drawbacks, etc.
The stronger desire is the one that will more
probably produce action, ceteris paribus, but
where other factors are not the same, the
weaker desire, thus defined, may prevail. If
they cost the same, the desire to eat lobster
would prevail over the desire to eat cod; it's
the stronger desire. Since they don't cost the
same, the weaker desire ordinarily prevails.
This is principally because other desires
(regarding money) come into play. (The
strongest navy is the one that is stronger
than any other single navy; it needn't be
stronger than two others in alliance.)
The excessively powerful desire is one,
intuitively, which may prevail though other
desires could be satisfied with greater
benefit. Acting upon that desire is un-
reasonable, since the reasonable action
would maximize the satisfaction of all one's
desires. It is an effect of rational deliberation
that the power of any relevant desire is
brought to bear on the decision, lending its
power to the efficacy of the resulting inten-
tion. Ideally, the power of a desire would be
proportional to the benefits of its satisfac-
tion, and there is another mechanism (other
than reason) which tends, if imperfectly, to
produce that correlation: for the satisfaction
of a desire provides positive reinforcement,
augmenting the tendency of states of that
type to produce behaviour of that type. But
the principal cause of excessive strength in
a desire also lies here. For the probability of
a desire's being acted upon may be raised
when its satisfaction coincides with the
satisfaction of some second desire, with the
result that the first desire's power is there-
after stronger, disproportionately to the
benefits that would regularly, and in other
contexts, attend its satisfaction. Thus the
child's desire to please, when it causes com-
pliant behaviour, is rewarded, so it inci-
dentally yields satisfaction of the desire for
candy, with the effect that the desire to
please becomes so strong that it is effective,
even when there is no candy in the offing,
and even when the benefits of compliant
250
behaviour are smaller than the benefits of
defiant behaviour that would satisfy other
desires.
See also FOLK PSYCHOLOGY; PRACTICAL
REASON; PSYCHOAN AL YTIC EXPLANATION;
REPRESENTATION; WEAKNESS OF WILL.
BIBLIOGRAPHY
Basil Blackwell.
Dretske, F. 1984. Explaining Behavior. Cam-
Geach, P. 1957. Mental Acts. London: Rout-
ledge & Kegan Paul.
Hume, D. (1739) 1960. A Treatise of Human
Nature, Book II, Section iii; Hume's Treatise,
ed. L. A. Selby-Bigge: Oxford University
Press.
Jowett, B.A., trans. 1937. The Dialogues of
Plato. New York: Random House.
Kenny, A. 1966. Practical inference. Analysis.
Marks, J. 1986. The Ways of Desire. Chicago:
Precedent.
Mill, J.S. 1867. An Examination of Sir William
Hamilton's Philosophy, ch. XXVI. London.
Spinoza, B. 1677. Ethics demonstrated in geo-
metrical order; The Ethics and Selected Letters,
trans. S. Shirley, ed. S. Feldman. Indianapo-
lis: Hackett.
Stalnaker, Robert C. 1984. Inquiry. Cam-
Stampe, Dennis W. 1987. The authority of
desire. The Philosophical Review, XCVI, 3,
335-81.
Stampe, Dennis W. 1990. Reasons and desires
- discussion notes on Dretske's Explaining
Behavior: Philosophy and Phenomenological
Research, vol. L.
DENNIS W. STAMPE
developmental psychology This entry is
about the developmental origins of natural
philosophical ideas about the mind.
Although there is some intriguing work on
children's explicit philosophizing - most of it
by the philosopher Gareth Matthews (1984)
- we will be concerned here with children's
philosophies of mind 'in action' (e.g. to
explain another's behaviour) rather than
their ability to construct philosophical argu-
ments. Indeed, in order to engage in human
interaction, to predict others' behaviour, to
understand their INTENTIONS/BELIEFS/
DESIRES, to interpret their STATEMENTS /
GESTURES/ACTIONS, to understand irony/
metaphor, to interpret utterances or facial
expressions which are discrepant from
actual feelings, to understand how point of
view and perception can influence belief for-
mation, and so forth, children have to learn
to become natural philosophers of mind. We
will focus on two essential questions: (a)
what do human beings bring into the world
to make the development of successful social
interaction possible?, and (b) what course
does the development take?
DOMAIN-GENERAL VERSUS DOMAIN-
SPECIFIC THEORIES OF MIND
We first need to introduce a distinction
which underpins two fundamentally differ-
ent ways of thinking about the child's con-
ceptions of mind. It will become our
chapter's motif. FODOR (1983) distinguished
between 'input systems' in the mental
architecture and 'central systems'. A simple
way to capture their difference is by think-
ing of the first as being rather like reflexes
and the second as being like thoughts, or of
the first as bottom-up/data-driven and the
second as top-down/hypo thesis-driven.
The input systems are 'domain-specific'
modules insofar as they perform sui generis
kinds of computations for each domain.
They are relatively unaffected by higher-
level, knowledge-based systems and by
current states of CONSCIOUSNESS. They
operate swiftly, mandatorily and indepen-
dently from other modules. They have fixed
neural architectures and a fixed timetable of
development. The central systems, by con-
trast, are domain-general and global.
Indeed, to use Fodor's metaphor, one can
see them as spreading horizontally above
and across the many vertical input systems
which package information from the exter-
nal world. Colour vision, face-processing,
movement perception, syntactic parsing are
examples of input systems. The central
systems cannot be listed so easily because
they do not naturally reduce to sets of com-
putational tasks. 'Belief fixation' (how we
acquire beliefs) and analogical reasoning
are central, because they are both domain-
general processes par excellence. That is to
say, there is no limit on the number of ways
we can come to hold a belief (no limit on
the input systems we recruit and how) and
no constraint on how we draw analogies.
In the light of the input system/central
distinction, what can we immediately say
about the development of children's philoso-
phies of mind - about what it is now cus-
tomary to call children's 'theory of mind'? Is
such a theory best regarded as a domain-
specific input system or as a domain-general
central system? The advantage of the first
option is that it allows us more easily to
assume that the theory is innately specified
(see INNATENESS). As Fodor aptly puts it,
'Here is what I would have done if I had
been faced with this problem in designing
Homo Sapiens. I would have made com-
monsense psychology innate; that way
nobody would have to spend time learning
it' (Fodor, 1987, p. 132). This certainly
would be an advantage because it is difficult
to see how the evidence could force upon us
the theory that there are beliefs and desires
which interact to cause behaviour. On the
other hand, can our conception of mind be
so neatly pre-packaged into an input
module? Do we not call on many domain-
general processes in our view of ourselves
and others as conscious fallible agents who
know some things and believe others?
The first modern account of theory of
mind development was given by Jean Piaget
(e.g. 1929, 1932). And it was the epitome
of a domain-general view. One way of
regarding Piaget's work is as an attempt to
explain how the child comes to forge a divi-
sion between subjectivity and objectivity,
between what is true of my mind and what
is true of reality. Piaget held that the basic
lineaments of thought (substance, causality,
space, time) evolved through sensorimotor
experience in infancy; that between about 2
and about 6 years these early sensorimotor
acquisitions gradually came to underpin
concrete verbal judgments; and that it was
251
DEVELOPMENT AL PS YCHOLOG Y
not until adolescence that children's
thought could move beyond coordinating
concrete judgments to coordinating hypoth-
eses and possibilities. One of the main theo-
retical tools Piaget used for describing
childish thought was egocentrism, by which
he meant the failure to distinguish a sub-
jective perspective (how it is to my mind)
from objective reality.
For Piaget, egocentrism was pervasive in
the thought of the child, in the sense that
the child's judgments of reality were sup-
posed to be infused with - and thus dis-
torted by - egocentric information about the
self's own physical, intellectual and social
perspective. Thus the failure of infants to
search for completely occluded objects was
explained in terms of their rule 'what I see
is what exists'; children's failure to judge
that a quantity is conserved across percep-
tual transformations (e.g. pouring it from a
short, fat to a tall, thin container) was
explained in terms of their egocentric fixa-
tion upon a single, striking unidimensional
change; children when asked to pick out the
view of three mountains that a person
sitting opposite them would actually see,
tend to pick out the view that they see
themselves. In other words, at different
ages, depending on the task, children tend
to overvalue the immediate evidence of their
own senses. According to this theory, ego-
centrism appears, fades and re-emerges
throughout development and across the
mental board, with the domain-specific
content of particular tasks having only a
minor role to play.
What had to evolve in the mental
journey towards a non-egocentric concep-
tion of reality? According to Piaget, the
child had to acquire a capacity for self-
regulation through mental action. She had
to regulate and thereby structure (Piaget
described this structuring by mathematical
group theory) the representations which
had been acquired in infancy. Restructuring
caused thinking to move to progressively
higher representational planes (from sensor-
imotor, to imagist, to verbal to hypothet-
ical). Egocentrism is overcome through a
central executive (not Piaget's term) reg-
252
ulating mental attention away from the
salient-to-self information, to objective
reality; and this regulation was supposed to
spread across all domains of knowledge.
Piaget was nothing if not a philosopher-
psychologist of the central systems.
What does a theory like this imply about
the young child's conception of mind? It
implies an overly pessimistic view. Indeed,
some of the claims that Pia get made about
failures to distinguish, say, speaking from
acting and dreaming from thinking till age 7
(Piaget, 1929) were certainly misguided. In
any event, the theoretical pendulum has
swung the other way, towards a domain-
specific conception of children's mental
philosophies. We will, in the next section,
describe this alternative view, followed by
some modern domain-general but non-
Piagetian views, before giving a brief recon-
sideration of the Piagetian conceptions at
the end.
THE PREREQUISITES TO THEOR Y OF MIND
IN INFANCY
If we adopt a domain-specific view of theory
of mind development, then we can expect to
find precursors in early infancy. What
might these be?
The first thing to say about them is that
they are essentially perceptual, as befits
input systems. Without the early perception
of the basic characteristics of humans (gait,
voice, face, etc.), it is difficult to see how
basic mental systems which support theory-
of-mind computations could ever get off the
ground (Karmiloff-Smith, 1992). Neonates
appear to be innately tuned to the gross fea-
tures of the primate face (Johnson and
Morton, 1991; Sargent and Nelson, 1992),
are capable of imitating facial gestures
(Meltzoff and Moore, 1977; Meltzoff, 1993)
and attending to changes in their emotional
expression (Field et ai., 1983). Similarly for
the human voice. At birth infants distin-
guish speech from other auditory input and
at 4 days they attend preferentially to their
native tongue, having already picked up
something about its prosodic structure
(Mehler et ai., 1986).
But there is nothing specifically mental
about human faces and voices. What about
infants' perception of the quintessentially
mental quality of agency? Leslie (1984) has
shown that 7 -month-olds can select out a
manual pick-up event as being different
from a non-manual means of raising
objects. Moreover, Premack (1990; see also
Dasser, Ulbaeck, and Premack, 1989), the
theorist who coined the phrase 'theory of
mind', has made the provocative claim that
in the earliest months of life we operate
with a purpose-built module for detecting
agency. This is done on the basis of the dis-
tinction between non self-propelled objects
and the self-propelling motion of biological
beings. This may be an automatic, some-
what mindless pick-up of information; but
without it, so the domain-specific theorist
argues, the child is unlikely to acquire a
conception of mental life - no matter how
well-regulated her 'mental actions'.
Beyond these initial attention-perception
biases, what aspects of very early interac-
tion could play a role in the development of
theory of mind? What kinds of behaviour
suggest that young children have some
inchoate conception of other minds? Com-
munication can plausibly be said to require
some form of 'mind reading', so perhaps
we should look towards prelinguistic com-
munication. Mutual eye-gaze or pointing to
a specific referent at about 9-10 months
act as a form of non-linguistic communica-
tion by directing the attention of the addres-
see to something of interest. It has been
shown that young infants become progres-
sively capable of joint attention via eye
contact. Note that we are talking of 'joint
attention', because eye contact alone can be
much like attending to inanimate objects.
Increasingly, infants make use of gaze alter-
nation (between the caretaker's eyes and a
coveted object or goal), to signal to the
caretaker that they wish to obtain a parti-
cular object. It is the coordination between
eye-contact and pointing gesture that leads
to ostensive communication. Interestingly,
several studies indicate that autistic children
- who often fail to develop a theory of mind
- are deficient in such joint attention co-
ordinations of eye contact and gesture
(Baron-Cohen, 1989; Dawson, Hill, Spencer,
Gal-pert, and Watson, 1990; Sigman,
Mundy, Sherman and Ungerer, 1986), sug-
gesting that this is a fundamental aspect of
the development of a theory of mind in
normal children. We will return to the case
of autism throughout the chapter.
What are the functions of these early
ostensive communications in the human
infant? They are of two types: so-called
'proto-imperatives' and 'proto-declaratives'
(Baron-Cohen, 1989). Proto-imperatives
involve the use of pointing or eye-gaze as
the infant's means of trying to obtain an
object by a non-verbal request directed at
the interlocutor who can reach the object
for the child. If the human infant were
mobile at birth as many species are, then
s/he would get the object herself or push
the adult towards it. But human immobility
is such that it forces the young infant into
finding other, interactional means of reach-
ing certain goals. The pointing gestures
therefore start out as instrumental requests.
These proto-imperatives rapidly become
proto-declaratives, i.e., a point becomes the
infant's means of making a non-verbal
comment about the state of the world (the
equivalent of, say, 'look, that's a nice dog')
rather than a request to obtain something.
Again, the difference between normal and
autistic children's proto-declarative pointing
to affect another's mental state is striking.
Autistic children neither use nor understand
proto-declarative pointing (Baron-Cohen,
1989); their competence is limited to proto-
imperative pointing to affect another's beha-
viour. They also rarely engage in so-called
'social referencing' (Hobson, 1989) in which
the child gauges how she should respond to
a novel event (is this danger or is it OK?)
from the mother's facial expression. Normal
children do this at around 12 months of age.
THE TODDLER'S THEORY OF MIND: A
DOMAIN-SPECIFIC MODULE FOR
MENT ALIZING?
Here is a typical conundrum in cognitive
development. By the time children reach the
253
end of their third year of life, they not only
use the language of mental states (think,
know, want, afraid, and so forth), but they
use these terms in a genuinely mentalistic
way (Bretherton and Beeghly, 1982; Shatz
et aI., 1983). Surely this ability cannot arise
out of language itself: the terms must be
gaining some cognitive foothold in a pre-
linguistic understanding of mental states.
Leslie (1987, 1990) has suggested that it is
an innately-specified module for computing
mental representations which provides this
foothold, a domain-specific module which
matures at around 18 months of age in
normal children. (see COMPUTATIONAL
MODELS; MODULARITY; REPRESENT ATION.)
This theory situates propositional attitude
competence in the realm of pretend play.
Leslie argues that children's pretend play
involves the same distinction between propo-
sitional CONTENT and PROPOSITIONAL
A TTITUDE as that found in the subsequent
use of mental state verbs. Because the psy-
chological structures of pretence are deemed
to be innately specified, the child can imme-
diately understand pretend acts when she is
first exposed to them. Leslie suggests that
the structure of young children's pretend
play should be understood as the computa-
tion of a 3-term relation between an agent
(usually the child herself), a primary repre-
sentation (the actual objects being played
with) and a decoupled, secondary repre-
sentation which represents the content of
the pretence. This contrasts sharply with
Piaget's arguments that young children
represent events as 'schemes' in which
agent, event and object form an undiffer-
entiated amalgam. For Leslie, it is the
notion of a decoupled representation that is
specific to theory of mind. The decoupling
allows the child to treat the pretend content
separately from the normal relations that
the representation of the real object/event
entertains. Thus, when a 3-year-old picks
up a block of wood and declares: 'Vroom,
vroom, vroom, toot, toot!', the pretend com-
putation involves: PRETEND [(Agent =
child)(Primary Representation = a mental
structure representing the fact that the
object on the table is a block of wood)
254
(Decou pled Representation = a copy of the
previous mental structure but cordonned off
from veridical descriptions and standing for
'the car')]. The primary and decoupled
representations involve different and sepa-
rate levels of processing and obey distinct
causal and logical inferential constraints.
Thus, pretending that a simple block of
wood has a steering wheel. a horn and four
wheels in no way detracts from toddlers'
understanding of the real properties of the
block of wood and of real cars, nor does it
change their representations of such proper-
ties. It is the decoupled (temporary) repre-
sentation that is 'tampered' with, not the
primary representations which continue to
entertain their normal representational rela-
tionships. And the decoupled representa-
tion, not necessarily linguistically encoded,
involves a distinction between a proposi-
tional attitude and the propositional content
on which it operates:
[I pretend that] [this block of wood] [it is a
car].
It is irrelevant to the truth value of the pro-
positional attitude PRETEND (and of
BELIEVE/THINK/CLAIM, etc.) that the block
of wood is not actually a car. Some proposi-
tional attitudes (e.g. KNOW/REMEMBER)
do, of course, entail the truth of their propo-
sitional contents.
DEVELOPING AN EXPLICIT BELIEF-DESIRE
PSYCHOLOGY
The claim we have just considered about
what pretence tells us about a child's
understanding of the mental realm is a con-
troversial one. Some would argue that there
is no real warrant for assuming that a pre-
tending (or pretence-observing) child is
mentally erecting a proposition (e.g. [it is a
car]) in relation to a mental attitude. All
that we can infer, the objection goes, is that
the child knows what it is to do something
pretendingly rather than for real. Pretending
may require fewer cognitive demands than
belief, but it seems to require more mental
insight than understanding what it means
to do something in earnest. However, that
does not force the conclusion that the child
has mentally separated attitudes from con-
tents - any more than understanding what
it means to do something in earnest forces
that reading of the child's behaviour. It may
therefore be preferable to be more cautious
and to suggest that the early conception of
pretence is sub-propositional. though it may
well ground the later understanding of the
relation between attitudes and contents.
In any event, if the pretending child has
some conception of what it means for a pro-
positional attitude to have a content, then it
is reasonable to expect that she also under-
stands the attitude of belief. Mental contents
are contents in the running for truth, with
the difference between pretending and
believing being that in the former case the
individual knows the content is false
whereas in the latter case the individual is
committed to its truth. But how are we to
assess the understanding of believing other
than by linguistic means?
At this point, then, we encounter another
developmental conundrum: how is one to
test for the pre-linguistic presence of an
ability which is - if what has just been said
is true - essentially propositional? What is a
non-verbal expression of belief? Are beliefs
only available to language-users, as some
philosophers would argue (DAVIDSON,
1974), or should we attribute beliefs pro-
miscuously? Whatever the answer to this
question, the assessment of the concept of
belief by developmental psychologists has
hitherto been only through verbal tests. In
general. the assumption is that a conception
of how false beliefs drive behaviour is a good
diagnostic test, because this captures both
the commitment to the truth of the proposi-
tion believed and the fact that it can be
false.
Take the following experimental set-up
for testing false belief, based on seminal
work by Heinz Wimmer and Josef Perner
(Wimmer and Perner, 1983). The child
watches a scene in which the experimenter
and a boy called Maxi are in a room to-
gether. The experimenter hides a piece of
chocolate under a box in front of Maxi.
Maxi then leaves the room momentarily
and, while he is absent, the experimenter
moves the chocolate to another hiding
place. The child is then asked where the
chocolate really is and, crucially for the
task, where Maxi will look for it upon her
return. In other words, the child has to dis-
tinguish between what she knows to be true
of the current state of the world (cf. Piaget's
'egocentrism') and what she knows to be
Maxi's current mental state. She also has to
know that Maxi's behaviour will be a func-
tion of his mental representations, not of
the physical reality.
This and other similar tasks (e.g. Perner,
Leekam and Wimmer, 1987) are simple, yet
stringent tests of the child's ability to
impute mental states with content to others.
Three-year-olds fail many false belief tasks
(at least when presented in this form) and
think that the protagonist will behave in
accordance with the real-world situation.
They do not seem to understand that he
will behave on the basis of his false belief.
Four-year-olds are successful. The minimal
criteria for possessing a theory of mind are,
according to DENNETT (1978), successfully
dealing with circumstances in which an
individual cannot rely on her own know-
ledge in order to assess another's mental
state. But in many tasks 3-year-olds do rely
solely on their own knowledge.
There are a number of theory-of-mind
tasks in addition to false belief, and they
share more than a family resemblance.
They all require a grasp of the representa-
tional nature of thought, they are typically
failed by 3-year-old children, and they cor-
relate well together (Moore and Furrow,
1991). So it does look - to some devel-
opmentalists at least (Leslie, 1987, 1990) -
that what we are seeing here is a special-
purpose, mental-state computing device
coming into operation. Taking a different
but related tack, Fodor (1992) has recently
argued that there is an innate belief-desire
theory but that it is not always accessed by
3-years-olds due to computational capacity
limitations that expand by age 4. But
perhaps the strongest reason for believing in
such a thing as an innate and specific
theory of mind module is the case of autism.
255
Recall that autistic children fail to show the
early markers of mentalizing ability such as
joint attention and proto-declarative point-
ing. They are also very unlikely to engage
spontaneously in pretend play; and they are
dramatically impaired on the false-belief and
other theory-of-mind tasks that normal chil-
dren pass at age 4 (Frith, 1989). All this is
consistent with the view first proposed by
Baron-Cohen, Leslie and Frith (1985) that
autism is the disorder which arises in chil-
dren who are born without the theory of
mind module, or with a module whose trig-
gering is delayed or deviant.
ALTERNATIVES TO DOMAIN-SPECIFIC
ACCOUNTS
The hypothesis that a theory of mind is not
a domain-general system but one similar to
syntactic parsing - formal, domain-specific
and circumscribed from 'general intelli-
gence' - is not only provocative but plaus-
ible. But does the autistic deficit prove that
theory of mind computations are modular,
i.e. encapsulated in a theory of mind
module as Leslie (1990) and Fodor (1983)
have argued? Frith (1989), for example,
locates autistic people's perceptual and
communicative difficulties in their domain-
general central processing. What we find in
autism, argues Frith, is lack of 'central
cohesion' - the ability to synthesize frag-
ments of information into meaningful
wholes.
Looking upon the human theory of mind
as a Fodorian module also presents us with
a form of dilemma familiar to students of
syntax development - a 'bootstrapping
problem'. As Pinker has argued (e.g. 1984),
knowing innately that there are (say) nouns
still leaves the learner with the problem of
determining what words in her native lan-
guage actually are nouns: the learner must
bootstrap herself from innate formal know-
ledge to a particular language. For Pinker,
the solution is to recruit some rough-and-
ready semantic generalizations (e.g. nouns
tend to be thing words). A similar problem
faces the novice philosopher of mind. It is
all very well to have the concept of belief,
256
desire, and so forth in one's native armoury;
but how is one to determine that this is a
case of believing, that this is a case of desir-
ing? Take pretence: a cognitive bridge of
some kind has to be built between the
innate representation of the pretending atti-
tude and the visual input of mother putting
a banana to her head and saying 'Oh Hi!
How nice of you to ringl' In a similar vein,
how could one recruit an innate representa-
tion of agency unless one had first-hand
experience of being an agent? (This is the
kind of point a Piagetian might make and,
indeed, Meltzoff has recently argued that
the pervasive human tendency to imitate
other agents from birth onwards is the
child's first entry into building up a theory
of mind (Meltzoff, 1993).)
There is, in fact, no shortage of altern-
atives to modular nativism about the child's
theory of mind (Astington, Harris and
Olson, 1988; see Russell, 1992, for a
review). Some draw an analogy between
child and scientist in that domain-general
theory development involves inferences on
the basis of unobservables (mental states
like belief), a coherent set of explanations of
causal links between mental states and
behaviour which are predictive of future
actions, and a growing distinction between
evidence and theory (Carey, 1985; Gopnik,
1993; Karmiloff-Smith, 1988; Perner,
1991; Wellman, 1990). Others reject
entirely the claim that our knowledge of
other minds has any theoretical structure,
arguing that we make judgments about the
beliefs of others by running a simulation of
our own beliefs, and attempt to explain 3-
year-olds' difficulties with the false belief
task in these terms (Harris, 1991; and see
special issue of Mind and Language, 1992).
Others, notably Perner (1991), maintain
that there is indeed a fundamental theoret-
ical shift at 4 years but that it is a domain-
general one in which a more adequate con-
ception of representation per se is develop-
ing, a conception that applies to pictures,
photographs, and words as well as to the
mind. Note, however, that autistic subjects
who fail theory of mind tasks have no
difficulties with equivalent tasks involving
photographs rather than other minds (Leslie
and Thaiss, 1992). This would clearly argue
against going all the way back to the
domain-generality of Piagetian theory.
'THE MIND ITSELF'. COMPARED TO
CHILDREN'S THEORIES OF IT
Developmental psychologists are fairly
unanimous that we should not return to
classic Piagetian theory which fails to give
modular mechanisms their due. But what
Piagetian theory can tackle is a different
question: How might a developing organism
set itself apart from reality and thus become
a mind? Piaget's answer, recall, was that it
does so by virtue of agency, by gaining
control over its perceptual (input-system?)
representations, by becoming progressively
less egocentric. It is by adapting to the con-
tours of reality that the mind emerges; and
such an entity exists over and above any
theories that the mind might have about
itself.
In a sense, then, the Piagetian student of
the central systems and the modular theor-
ist have somewhat different agendas when
they talk about development. But there is
an inevitable clash over empirical territory.
For Piagetians will have to say that com-
petence in mental agency (qua the ability to
suppress egocentrism) will determine the
child's efficient use of or access to the
theory. They might argue, for example, that
the child fails the false belief task when she
cannot suppress the mental salience of her
own knowledge that the object is no longer
where it was. In support of this conjecture,
3-year-olds have a similar kind of difficulty
with answering questions about which of
two objects a protagonist needs, referring
instead to the object which the child herself
needs (Russell, Jarrold and Moore, 1994).
With regard to autism, indeed, the Pia-
getian can even suggest a serious alter-
native to the theory-of-mind deficit view.
We know that people with autism have
clear difficulties with so-called 'executive
function tasks' (tasks that require the insti-
gation, control and monitoring of mental
actions). They have, for example, problems
in planning, shifting their attention to new
categories and performing means-end tasks
(e.g. Hughes and Russell, 1993, Ozonoff,
Pennington and Rogers, 1991). If this diffi-
culty exists from the outset, if there is a pro-
found impairment in determining which
experiences are generated by one's mental
action and which are generated by reality,
the sense of self will fail to develop ade-
quately - and thus the sense of others.
As in much of cognitive developmental
psychology, there is no elegant denouement
to the story of how the child's theory of
mind develops. Theoretical disputes are
intense, and the cognitive developmentalist
certainly needs the philosopher's help in
addressing the deeper questions. The child,
by contrast, needs only the minimum of
help to become a philosopher of mind. It's
natural.
See also COGNITIVE PSYCHOLOGY; LAN-
GUAGE OF THOUGHT.
BIBLIOGRAPHY
Astington, J.W., Harris, P.L., and Olson, D.R"
eds. 1988. Developing Theories of Mind. Cam-
bridge: Cambridge University Press.
Baron-Cohen, S. 1989. Perceptual role-taking
and proto-declarative pointing in autism.
British Journal of Developmental Psychology,
7,113-27.
Baron-Cohen, S., Leslie, A.M. and Frith, U.
1985. Does the autistic child have a 'theory
of mind'? Cognition, 21,37-46.
Bretherton, I., and Beeghly, M. 1982. Talking
about internal states: the acquisition of an
explicit theory of mind. Developmental Psy-
chology, 18, 906-21.
Carey, S. 1985. Conceptual Change in Childhood.
Chandler, M.J., and Boyes, M. 1982. Social-
cognitive development. In Handbook of Devel-
opment Psychology, ed. B. B. Wolman. Engle-
wood Cliffs, NJ: Prentice-Hall.
Dasser, V., Ulbaeck, L, and Premack, D. 1989.
The perception of intention. Science, 243,
365-7.
Davidson, D. 1974. Belief and the basis of
meaning. Synthese, 27, 309-29.
Dawson, G., Hill, D., Spencer, A., Galpert, L.,
and Watson, L. 1990. Affective exchanges
257
between young autistic children and their
mothers. Journal of Abnormal Child Psychol-
ogy, 18, 335-45.
Dennett, D.C. 1978. Brainstorms: Philosophical
Essays on Mind and Psychology. Mont-
gomery, VT: Bradford Books.
Field, T., Woodson, R., Greenberg, R., and
Cohen,'D. 1983. Discrimination and imita-
tion of facial expression by neonates.
Science, 218,179-81.
Fodor, J.A. 1983. The Modularity of Mind.
Fodor, J.A. 1987. Psychosemantics: The Problem
of Meaning in the Philosophy of Mind. Cam-
bridge, MA.: MIT Press/Bradford Books.
Fodor, J.A. 1992. A theory of the child's
theory of mind. Cognition, 44, 283-96.
Frith, U. 1989. Autism: Explaining the enigma:
Gopnik, A. 1993. How we know our minds:
The illusion of first-person knowledge of
intentionality. Behavioral and Brain Sciences,
16: 1, 1-l4.
Harris, P. 1991. The work of imagination. In
Natural Theories of Mind, ed. A. Whiten.
Hobson, P. 1989. On acquiring knowledge
about people and the capacity for pretense.
Psychological Review, 97, 114-21.
Hughes, C., and Russell, J. 1993. Autistic chil-
dren's difficulty with mental disengagement
from an object: Its implications for theories
of autism. Developmental Psychology, 29,
498-510.
Johnson, M.H., and Morton, J. 1991. Biology
and Cognitive Development: The Case of Face
Recognition. Oxford: Basil Blackwell.
Karmiloff-Smith, A. 1988. The child is a the-
oretician, not an inductivist. Mind and Lan-
guage, 3: 3, 183-95.
Karmiloff-Smith, A. 1992. Beyond Modularity:
A Developmental Perspective on Cognitive
Science. Cambridge, MA.: MIT Press/Brad-
ford Books.
Leslie, A.M. 1984. Infant perception of a
manual pickup event. British Journal of
Developmental Psychology, 2,19-32.
Leslie, A.M. 1987. Pretense and representa-
tion: The origins of 'Theory of Mind'. Psy-
chological Review, 94, 412-26.
Leslie, A.M. 1990. Pretense, autism and the
basis of 'Theory of mind'. The Psychologist,
3,120-3.
258
Leslie, A.M., and Thaiss, L. 1992. Domain
specificity in conceptual development:
neuropsychological evidence from autism.
Cognition, 43, 225-51.
Matthews, G.B. 1984. Dialogues with Children.
Cambridge, MA.: Harvard University Press.
Mehler, J., Lambertz, G., Jusczyk, P., and
Amiel-Tison, C. 1986. Discrimination de la
langue maternelle par Ie nouveau-ne. C.R.
Academie des Sciences, 303, S. III, 637-40.
Meltzoff, A.N. 1993. Imitation as a tool for
exploring the infant mind. Paper presented
at the Society for Research in Child Develop-
ment, New Orleans, March 1993.
Meltwff, A., and Moore, M. 1977. Imitation of
facial and manual gestures by human neo-
nates. Science, 198, 75-8.
Moore, C., and Furrow, D. 1991. The develop-
ment of the language of belief: The expres-
sion of relative certainty. In Children's
Theories of Mind, ed. D. Frye and C. Moore.
Ozonoff, S., Pennington, B.F., and Rogers, S.J.
1991. Executive function deficits in high-
functioning autistic individuals: Relation-
ship to theory of mind. Journal of Child Psy-
chology and Child Psychiatry, 32, 1081-105.
Perner, J. 1991. Understanding the Representa-
tional Mind. Cambridge, MA.: MIT Press/
Bradford Books.
Perner, J., Leekam, S., and Wimmer, H. 1987.
Three year olds' difficulty with false belief:
The case for a conceptual deficit. British
Journal of Developmental Psychology,S, 125-
37.
Piaget, J. 1929. The Child's Conception of the
World. London: Routledge & Kegan Paul.
Piaget, J. 1932. The Moral Judgement of the
Child. London: Kegan Paul. Trench Trubner.
Pinker, S. 1984. Language, Learnability and
Language development. Cambridge, MA.:
Harvard University Press.
Premack, D. 1990. Words: What are they, and
do animals have them? Cognition, 37, 197-
212.
Russell, J. 1992. The theory theory: So good
they named it twice? Cognitive Development,
7,485-519.
Russell, J., Jarrold, C. and Moore, C. 1994. The
'false need' task: evidence for an executive
transition between 3 and 4 years. Ms sub-
mitted for publication.
Sargent, P.L., and Nelson, C.A. 1992. Cross-
species recognition in infant and adult
humans: ERP and behavioral measures.
Poster presented at the International Con-
ference on Infant Studies, Miami Beach.
Shatz, M., Wellman, H.M., and Silber, S.
1983. The acquisition of mental verbs: A
systematic investigation of the child's first
reference to mental state. Cognition, 14,
301-21.
Sigman, M., Mundy, P., Sherman, T., and
Ungerer, J. 1986. Social interactions of
autistic, mentally retarded, and normal
children and their caregivers. Journal of Child
Psychology and Psychiatry, 27, 647-56.
Wellman, H.M. 1990. The Child's Theory of
Mind. Cambridge MA.: MIT Press.
Wimmer, H., and Perner, J. 1983. Beliefs
about beliefs. Representation and constrain-
ing function of wrong beliefs in young
children's understanding of deception. Cog-
nition, 13, 103-28.
ANNETTE KARMILOFF-SMITH
JAMES RUSSELL
Dretske, Fred An early interest in know-
ledge and PERCEPTION shaped my thinking
about BELIEF and experience. One cannot
think long about the relations, causal and
otherwise, underlying our perceptual
awareness of the world without reaching
definite - indeed, I would say irresistible -
conclusions about the nature of those inter-
nal states required for such awareness. If
perception of, and knowledge about, physi-
cal objects is possible, then, or so it seemed
to me, our experience of, and our beliefs
about, these objects must have a certain
character.
It is, I think, this epistemological perspec-
tive that most clearly distinguishes my own
work in the philosophy of mind from others.
In Dretske (1969, 1979), for example, I
argued for a distinction between what I
called epistemic and non-epistemic forms of
perception. To describe someone (or some
animal) as seeing a bug, for example, is to
describe a relationship between the person
and a bug which, like stepping on a bug,
does not require (though it is compatible
with) knowledge or belief about the bug.
The relationship is non-epistemic in the
DRETSKE. FRED
sense that it does not involve any con-
ceptual understanding of what is seen.
Seeing (of objects and events) is not, not
necessarily anyway, believing. This way of
seeing is to be contrasted with epistemic
forms: seeing that the bug is a bug or seeing
the bug as a bug (Le. recognizing it). This
idea, obvious though I (still) think it is, is
fundamental to understanding the basic dif-
ference between perception and conception
and, hence, the nature of, and differences
between, our experience of, and our beliefs
about, the world. A theory of the mind that
identifies all mental states, including experi-
ences, with some form of belief or judgment
(whether conscious or unconscious) has to
be wrong. If seeing is not believing in a
theory of knowledge, neither is it in the
philosophy of mind.
I have, as a result, always regarded
naturalistic theories in the philosophy of
mind (see NATURALISM) as confronting at
least two fundamentally different problems:
(1) those associated with the propositional
attitudes (belief and judgment) and related
processes (inference and thought); and (2)
those related to feelings and experiences,
the qualitative states constituting our
sensory life. No theory of the mind can be
complete unless it has a story to tell about
both aspects of our mental life. We need a
theory that tells us, not only what it is to
believe, think, or know that Judith is
playing the piano, but what the difference is
between seeing her play it and hearing her
play it.
Even more basic to my view of the mind
is the externalism I brought from my work
in epistemology (see EXTERNALISM/
INTERNALISM). Knowledge, or so I argued
in Dretske (1969, 1971), is not a matter of
justification, not a matter of getting your
beliefs secured by an evidential chain to a
foundational rock. It is, rather, a matter of
such beliefs being connected to the facts in
the right way, a relationship whose exist-
ence, because external or extrinsic to the
total system of beliefs, might be quite
unknown (perhaps even unknowable) to
the knowing mind. Sense perception is one
way, the most direct and reliable way, of
259
DRETSKE. FRED
getting oneself so connected. The wide-
spread use of measuring instruments in
science and elsewhere is merely a way of
extending this connection to more inacces-
sible affairs. It later seemed to me that
'information' was a useful word to describe
this external relation, and in Dretske (1981)
I tried to articulate an externalist theory of
knowledge by developing an account of
informational content. The conclusive
reasons of 1971 became the information of
1981. Since the notion of informational
content I developed was an intentional idea
(a signal could carry the information that
something was F without carrying the
information that it was G even though
every F was in fact G) I also hoped that this
general notion of information would be
useful in the philosophy of mind and in
semantics.
The naturalistic picture of knowledge and
perception that emerged from this study had
an obvious gap. Both perception and know-
ledge were identified with various mental
states (experiences and beliefs) standing in
appropriate external relations to objects and
facts, but until the nature of experience and
belief themselves were understood, the
account was incomplete. Without a natur-
alistic theory of experience I had no natur-
alistic theory of perception, and without a
naturalistic theory of belief I had no natur-
alistic theory of knowledge. My own
research took a turn away from epistemol-
ogy and toward the philosophy of mind
when I started worrying more about what
made something a belief than what made a
belief knowledge, more about what made
something an experience than how experi-
ences figured in perception.
Nonetheless, the externalism remained. If
one assumes that every belief aspires to be
knowledge and every experience aspires to
veridicality - assumes, that is. that the
function or purpose of experience and belief
is to inform and direct - then it is a short
step to supposing that what makes some
brain states into mental states (experiences
or beliefs, as the case may be) is related to
their information-carrying function. Just as
information can convert a belief into know-
260
ledge and an experience into perception, the
function of providing (in the case of percep-
tion) and using (in the case of belief) such
information can convert physical states into
mental states: an experience or a belief.
From both an evolutionary and a devel-
opmental standpoint, from both phylogeny
and ontogeny, it is reasonable to suppose
that sense experience and the beliefs nor-
mally consequent upon it have the job of
supplying and organizing the information
needed to coordinate behaviour with condi-
tions, both external and internal, on which
success in satisfying needs and desires
depends. This being so, mental states (at
least these mental states) might plausibly be
identified with whatever brain states have
(or service) the relevant functions. Experi-
ences and beliefs are merely those internal,
presumably physical, states of a system
having the function of providing informa-
tion (in the case of experience) and mobiliz-
ing it (in the case of belief) for use in the
control of behaviour. When they do their
job, the results are called perception and
knowledge; when they fail, we speak of illu-
sion, hallucination, false belief, and opinion.
On this picture, then, the mind is merely the
externalist face of the brain, that aspect of
biological activity having to do with the
provision, handling and use of information.
Just as externalism in epistemology says
that what converts a mental state - a belief,
say - into knowledge is outside the mind, so
externalism in the philosophy of mind says
that what converts a physical state - some
condition of the brain - into a belief is
outside the head.
This general picture of the mind (or that
part of it given over to cognitive affairs)
began to emerge in the final three chapters
of Dretske (1981). It was here that I began
to think of concepts as physical structure
types that acquired, in a process of learning,
a specific information-carrying function. In
speaking of a certain type of state as having
an information-carrying function I mean
(and meant) more than that these struc-
tures carry information, more than that
they. in fact, function this way. There are
many things (e.g. cloud formations) that
carry information (about weather) that do
not have this as their function. Not every-
thing that holds papers down is a paper-
weight - has this as its function.
If concepts of the simple (observational)
sort were internal physical structures t h a ~
had, in this sense, an information-carrying
function, a function they acquired during
learning, then instances of these structure
types would have a content that (like a
belief) could be either true or false. After
learning, tokens of these structure types,
when caused by some sensory stimulation,
would 'say' (Le. mean) what it was their
function to 'tell' (inform about). They
would, therefore, qualify as beliefs - at least
of the simple (observational) sort.
Any information-carrying structure
carries all kinds of information. If, for
example, it carries information A, it must
also carry the information that A or B. As ~
conceived of it, learning was supposed to be
a process in which a single piece of this
information is selected for special treatment,
thereby becoming the semantic content -
the meaning - of subsequent tokens of that
structure type. Just as we conventibnally
give artefacts and instruments information-
providing functions, thereby making their
activities and states - pointer readings,
flashing lights, and so on - representations
of the conditions in the world in which we
are interested, so learning converts neural
states that carry information - 'pointer
readings' in the head, so to speak - into
structures that have the function of provid-
ing some vital piece of the information they
carry (see REPRESENT A TION). When this
process occurs in the ordinary course of
learning, the functions in question develop
naturally. They do not, as do the functions
of instruments and artefacts, depend on the
intentions, beliefs, and attitudes of users. We
do not give brain structures these functions.
They get it by themselves, in some natural
way, either (in the case of the senses) from
their selectional history or (in the case of
thought) from individual learning. The
result is a network of internal representa-
tions that have (in different ways) the power
to represent: experiences and beliefs.
DRETSKE. FRED
It is important, inCidentally, to under-
stand that this approach to THOUGHT and
belief, the approach that conceives of them
as forms of internal representation, is not a
version of FUNCTION ALISM - at least not if
this widely held theory is understood, as it
often is, as a theory that identifies mental
properties with functional properties. For
functional properties have to do with the
way something, in fact, behaves, with its
syndrome of typical causes and effects. But,
as pointed out above, not everything that
functions as, say, a deterrent has deterrence
as its function. An informational model of
belief, in order to account for misrepresenta-
tion, a problem I struggled with in a pre-
liminary way in both (1981) and (1986).
needs something more than a structure that
provides information. It needs something
having that as its function. It needs some-
thing that is supposed to provide informa-
tion. As Sober (1985) so nicely puts it, for
an account of the mind we need functional-
ism with the function, the TELEOLOGY, put
back in it.
As I conceived of it in 1981, concept
learning of the simple (observational) sort
was supposed to be the process by means of
which certain neural states acquired an
information-providing function appropriate
to the classification of later tokens of that
type as beliefs. If, during learning, a state
acquired the function of indicating that
some perceptual object was F, then later
instances of that state, caused by some
perceptual object X, were beliefs (rep-
resentations), possibly false beliefs (mis-
repre-sentations), that X was F.
My (1981) account of this process was,
admittedly, sketchy. Nonetheless, though
short on details, the account was motivated
by a basic idea. one that has remained with
me up to the present and which formed the
basis of the (1988) theory. It is this: the sort
of conceptual states we invoke to explain
behaviour (belief and desire) really only find
a place in the world with animals capable of
learning. Even with these, beliefs only put
in an appearance with those of their beha-
viours that are in some way learned or. at
least capable of being modified by learning.
261
DRETSKE. FRED
We don't need beliefs, nor do we in fact
appeal to them, to explain reflexes and
other genetically determined behaviours.
Even if I have reasons to blink when
someone pokes at my eye, I do not blink for
these reasons. Even if I think I can protect
my eye by blinking, that isn't why I blink.
Instead, some story about my genes will
explain why I blink, and some story about
evolution will presumably explain why I
have such genes. Natural selection may, in
this way, explain a lot about the behaviour
of animals, including humans, but the
behaviours it helps explain are not the
behaviours that beliefs and desires explain.
This, indeed, is why we have less and less
use for the mind as we go down the phylo-
genetic scale, why it seems so pointless to
credit paramecia and sea slugs with
thoughts and intentions. Even when we
speak of them this way, the attribution has
a metaphoric, a stance-like, quality to it. as
long as the behaviour being explained is
neither the product of, nor modifiable by,
learning. This being so, what is more
natural than to suppose that beliefs (or,
better, the concepts needed to form beliefs)
are created in the very process in which are
created the behavioural dispositions (and,
thus, the behaviours) that beliefs are called
upon to explain. It is, so to speak, a package
deal: we get beliefs - at least the concepts
needed to have beliefs - in the very process,
the learning process, wherein is developed
the neural basis for those behaviours that
such beliefs are available to explain. This, in
turn, means that the content of beliefs and
desires, of fears and intentions, the repre-
sentational content that is featured in every
explanation of behaviour by reasons, must
derive from the development in learning of
those circuits that constrain and structure
voluntary action (see DEVELOPMENTAL
PSYCHOLOGY).
That, at least, was the guiding motivation
behind my 1981 effort to install the infor-
mational processes underlying concept
learning as the source of meaning and
content. The effort was, as I say, short on
details and it was criticized for this neglect.
Besides this, though, there was a criticism
262
by Jerry FODOR (1984) that stuck. This
problem, the so-called disjunction problem,
a problem for almost all extant naturalistic
theories, is that, according to Fodor, I was
making false belief possible only by an
'unprincipled' stipulation of what is relevant
to the learning situation. Only by artificially
restricting the circumstances. by ignoring
relevant counterfactuals. could I give an
internal structure one information-carrying
function (the content P. say) rather than
another (the content P or Q).
This criticism stuck because although I
had thought about this kind of problem.
and discussed similar sorts of examples at
great length (see 1981. pp. 222-31). I
really did not (as I did in 1988) have a
developed theory of representation. All I had
were robust intuitions that meaning. the
concepts required to hold beliefs, had to
develop there. Where else? My concept of a
martini. and hence my capacity for holding
beliefs about. and having desires for. marti-
nis. beliefs and desires that sometimes help
explain why, just before the dinner hour. I
go into the kitchen. surely cannot be
innate. Despite a tendency on the part of
some philosophers to classify everything
whose origin they do not understand as
innate or non-existent. that. surely. cannot
be the answer.
What I did not appreciate at the time.
and what only became clear to me later
when my interest shifted from epistemology
to the philosophy of mind. was that I did
not really understand the causal or explan-
atory role of information. The lacuna was
not serious as long as I was interested in
knowledge. but when (as in the final three
chapters of Dretske (1981). I turned to
belief itself. this problem became acute.
The problem became acute because if
information has to do with the nomic
dependencies between events, then for
information to do any real causal or explan-
atory work in the world these dependencies
have to do some causal or explanatory work
in the world. But how could they? And of
what? It is particular events. not the corre-
lations or dependencies between them -
hence, not information - that activate the
receptors and trigger belief. These particular
events may be correlated and dependent
events; they may, as a result, carry infor-
mation. Nonetheless, they are particular
events whose causal efficacy is confined to,
and hence explained by, their intrinsic
properties. It is not, therefore, information
itself that is causing belief, but, rather, the
events that carry information. I acknow-
ledged this point in 1981 but took it (I still
do) to be unimportant to epistemological
applications of information theory. What is
important for knowledge is that the proper-
ties of the signal in virtue of which it carries
information be the same as those that are
causally responsible for the belief. I can
come to know something by consulting a
gauge if the pointer position (which carries
the relevant information) causes the rele-
vant belief. It is not necessary that the
information itself (the relationship between
the pointer and what it carries information
about) cause the belief. It is information
that is necessary for knowledge, not the
causal efficacy of information.
Whether or not this is true, it appears to
leave information itself (as opposed to the
events and structures that carry it) in a
metaphysically precarious position. If it does
not cause or explain anything, who needs
it? The causal impotence of information
may be tolerable in epistemology where it is
generally conceded that knowledge does not
explain anything more about the behaviour
of a person who has it than does a true
belief. But in the philosophy of mind such
causal irrelevance smacks, ominously, of
EPIPHENOMEN ALISM. This threat is espe-
cially relevant when one is trying to erect,
as I was, a theory of psychological content,
a theory of belief, on an informational basis.
In Dretske (1980, 1981) I traced the
INTENTION ALlTY of cognitive states, includ-
ing belief, to the intentionality of informa-
tion, to the modality inherent in the
dependency relations constituting informa-
tion. But if information had no causal or
explanatory clout, what, then, was the
purpose or point of intentionality? How
could beliefs and desires be important in the
natural scheme of things if what we
DRETSKE. FRED
believed and desired, the intentional content
of these states, was explanatorily irrelevant
to everything we do? There is, I know, a
small industry dedicated to making FOLK
PSYCHOLOGY a convenient myth, but I have
always taken such sceptical results, as I
have taken scepticism in epistemology, as a
reductio of the views that lead to it.
Explaining Behavior (1988) was an
attempt to deal with this cluster of inter-
locking problems. Information had to be put
to work. If my intuitions were right, the
place for it to find useful employment was
in learning. And if it did, indeed, do some
useful work in learning, perhaps the work it
actually did could be described in a way
that illuminated the nature of mental
content and, thereby, the character of belief
and desire, those mental states that figure
most prominently in the determination of
behaviour. Experience was a tougher nut.
That, as we like to say, could wait until
later. (Note: Later has now arrived; this is a
current research project.)
I found that I could not begin to think in
any productive way about these issues
without first getting clear, or clearer, about
behaviour itself. Behaviour, after all, is what
beliefs and desires are supposed to explain.
Unless we understand what it is we are
asking X to explain, we cannot expect to
understand how it is that X explains it. And
if we do not understand how beliefs and
desires explain, we cannot hope to under-
stand what it is about them, their meaning
or content, that gives them their explana-
tory punch. One could as well try to under-
stand force and mass, the explanatory ideas
of Newtonian mechanics, without troubling
to understand weight and acceleration.
This investigation took me down unex-
pected paths. Behaviour, I concluded, could
not be identified with bodily movements.
Nor could it be identified with any other
event, state, or condition in which beha-
viour typically culminates. Instead, beha-
viour is a causal process having bodily
movement as a part. The difference between
a person moving her arm and her arm's
movement is that the latter is a part of the
former in the same way B is a part of A's
263
DRETSKE. FRED
causing B. Once this is accepted, the expla-
nation of behaviour takes on a new look.
Intentional action is no longer, a la David-
son, a bodily movement caused by reasons.
It is, rather, the causing of bodily move-
ments by reasons. This is not, at least not
merely, grammatical hair-splitting. It is the
metaphysically important difference be-
tween a process and its product, between a
caused event and the causing of it. Once the
product, bodily movement, is distinguished
from the process, the behaviour, the moving
of your body, reasons explain behaviour,
not by causing it, but, rather, by having
their content - what one believes and
desires - explain the causing.
This fundamental change in the explana-
tory relation between reasons and beha-
viour forces a basic revision in one's picture
of the role of learning in the explanation of
behaviour. Learning of the relevant kind is a
process in which the dependencies, the cor-
relations defining information, play the role
of what I call structuring causes: they help
reconstitute effector circuits and thereby
contribute, causally, to any future beha-
viour that depends on these reconstituted
circuits. Information thereby gets a real job
to do. Since concepts crystallize out of the
learning process (this is where states acquire
their information-carrying functions) at the
same time as the connections are being
created that underlie all future behaviour,
later instantiations of these concepts become
causally relevant to such behaviour.
I have always liked Ramsey's (1931)
image of beliefs as maps by means of which
we steer. As I now see it, beliefs become
maps, acquire representational powers, in
the same process, the learning process, as
that in which the information from which
they derive their content gets its hand on
the steering wheel. I see no other way for
content to become explanatorily relevant to
what we do. If the content of thought is, as
it seems it must be, an extrinsic fact about
the brain, if it is a relational affair, then the
only way to make content explanatorily
relevant to what we do is to make these
extrinsic facts, these relationships, active in
the process in which is formed the neural
264
substrate of the behaviour they will even-
tually explain. This, then, means that the
ultimate source of mental content, the sort
of content associated with thought, must be
learning.
This does, however, leave a question
about the role of the senses in this total
cognitive enterprise. If it is learning that, by
way of concepts, is the source of the repre-
sentational powers of thought, from whence
comes the representational powers of
experience? Or should we even think of
experience in representational terms? We
can have false beliefs, but are there false
experiences?
My current research is focused on a
representational account of sense experi-
ence. If beliefs are maps by means of which
we steer - representations that help guide
the ship - experience provides the informa-
tion needed to construct these maps. The
senses have the function, the biological func-
tion, of providing this information, and
experience, as the vehicle of this informa-
tion, thereby constitutes a representation of
the objects about which it carries informa-
tion. Experiences are not false, but they can
misrepresent. On this account, then, experi-
ence and thought are both representational.
The difference resides in the source of their
representational powers: learning in the
case of thought, evolution in the case of
experience. Or so I hope to show.
See also CONTENT; DA VlDSON; DENNETT.
BIBLIOGRAPHY
Dretske, F. 1969. Seeing and Knowing. Uni-
versity of Chicago Press.
-- 1971. Conclusive reasons. Australasian
Journal of Philosophy. 49, 1-22.
-- 1979. Simple seeing. In Body. Mind and
Method: Essays in Honor of Virgil Aldrich, ed.
D. F. Gustafson and B. 1. Tapscott. Dor-
drecht, Holland: D. Reidel.
-- 1980. The intentionality of cognitive
states. In Midwest Studies in Philosophy, Vol.
5. University of Minnesota Press, pp. 281-
94.
-- 1981. Knowledge and the Flow of Informa-
tion. Cambridge, MA.: MIT Press/Bradford
Books.
-- 1986. Misrepresentation. In Belie!, ed. R.
Bogdan. Oxford University Press.
-- 1988. Explaining Behavior. Cambridge,
MA.: MIT Press, Bradford Books.
Fodor, J. 1984. Semantics, Wisconsin style.
Synthese, 59, 231-50.
Ramsey, F.P. 1931. The Foundations o! Math-
ematics and Other Logical Essays. London:
Routledge & Kegan Paul.
Sober, E. 1985. Panglossian functionalism and
the philosophy of mind. Synthese, 64: 2,
165-93.
FRED DRETSKE
dualism The mind-body problem is the
question how the mind is attached to the
body. The materialist answer is that
because the mind is the body, or perhaps,
because of dead bodies, the body ticking
over in ways yet to be specified in the
glorious future of neuropsychology, each
person's mind is necessarily embodied. In
contrast, the dualist answer, as its name
suggests, is that each person's mind is at
least not identical with his body, so these
are two different things.
Plato and Descartes are probably the two
acknowledged philosophical masters who
were most explicitly dualists. and Descartes'
arguments for dualism remain live issues.
One such argument which some detect in
his text draws on central features of his
system. That system begins with the
method of systematic doubt; to begin with.
we are to count ourselves as knowing only
what we cannot doubt. The argument from
dreaming, that all our sense experience
could be hallucination because we have
been asleep and dreaming all our lives, and
the evil demon argument. that (in con-
temporary terms) a mad scientist could
have extracted our brains at birth and
wired them up to computers programmed to
simulate in us the hallucinatory sense
experience we have had, show that our
beliefs in the existence of matter can be
doubted. So we are not to begin with to
count ourselves as knowing that there is
matter. Since, in particular, our bodies
DUALISM
would be material, our bodies are at least
initially dubious. But a doubt that we doubt
is self-refuting, so we may be certain from
the start that we are thinking and thus that
we exist; this argument is called the cogito.
So our bodies are at least initially dubious.
but our minds are not, and thus by
LEIBNIZ'S LA w (things not sharing all their
properties are different) our minds are not
identical with our bodies.
Whether Descartes intended this argu-
ment or not. it is inadequate. In the first
place, psychological predicates do not
express properties sufficient to establish dif-
ferences by Leibniz's Law; some masked
man could be your father even though you
know who your father is but do not know
who the masked man is. So, since 'is at least
initially dubious' is a psychological pred-
icate, the present argument makes a falla-
cious use of Leibniz's Law.
The second objection goes to the heart of
dualism. Look at your left fist and open your
left hand. Your left hand is still there but
your left fist no longer is. So your left hand
and your left fist exist at different times, and
thus. by Leibniz's Law, are different things.
Yet surely a hand-fist dualism would be
preposterous, perhaps because your left fist
being your left hand clenched. the fist could
not exist without the hand. What the
mind-body dualist like Descartes means is
stronger than mere difference. He wants to
say that his mind is no less basic a thing
than his body (or, in the older terms of art.
that his mind is no less a substance than his
body). He means that his mind does not
depend for its existence on his body. This
notion of dependence is modal. That is, the
dualist means that he (his mind, the person
or self he is) could exist even if his body did
not. For short, dualism is a claim to pos-
sibility. the claim that you could be dis-
embodied. So taken, dualism yields
difference but. as the hand-fist example
shows, requires more. Since the argument
from dreaming and the cogito would at most
establish mere difference. it is. too weak for
dualism. But note that since dualism asserts
only the possibility of disembodiment, Des-
cartes need not claim that there actually are
265
DUALISM
any disembodied people about; he need not
enlist in the lunatic fringe. Similarly, mares
do not depend in any relevant way on stal-
lions for their existence, since there could be
mares (for a while) after stallions had been
killed off. This independence holds despite
the fact that mares are sired by stallions, so
the sort of independence of mind from body
that dualism requires is compatible with
causal commerce between the minds and
bodies there actually are.
But granted that dualism is a claim to
possibility, how might such a claim be
established? It seems plausible that mere
(non-actual) possibilities neither act on, nor
are acted on by actualities, so if knowledge
is to be understood naturally in terms of
interaction between knower and known,
then perhaps there is no knowledge of
objective possibility (that is, possibility
whose existence and nature is independent
of knowledge of it). In that case, the mind-
body problem might be ineluctably moot,
not because it misconceives mind or body,
but simply because it is modal. Such a con-
clusion seems so extreme, or even extrane-
ous to the problem, as to be a conclusion of,
at best, later resort. If we do allow modal
epistemology a run for its money, then we
might do well to start from the natural
history of modal belief. It seems likely that
anyone willing to take modality seriously
will grant that there could have been one
more pigeon than there actually is. What so
eases this conviction seems to be the ease
with which we imagine an extra pigeon
flapping about. With an eye to this example,
we might then follow Hume in saying
'. .. nothing we imagine is absolutely
impossible', or that what you can imagine is
possible.
If possibility is to remain objective, then
Hume's principle should be taken not meta-
physically but epistemologically. That is, it
is not that possibilities are, for example, the
images we form when we visualize. The
idea is rather that imagination is to know-
ledge of (mere) possibility as perception is to
knowledge of actuality. An empiricist can
accept that perception is basic to knowledge
of actuality without accepting Berkeley's
266
subjective idealist doctrine that to be is to be
perceived; and Hume's principle can be
taken as a way empiricism might try to
accommodate knowledge of possibility
without going idealist. (But doing so makes
the problem of naturalizing knowledge of
possibility more acute.) Note the analogy
claims only that imagination is sufficient for
knowledge of possibility, not that it is neces-
sary. For some of the arcana of, say, con-
temporary physics may be too arcane to
have been seen or even imagined, and yet
for all that may be actual and so possible.
Many contemporary analytical philo-
sophers deny Hume's principle, perhaps
because of some of the possibilities to which
it would commit them. But a principled
objection to Hume's principle, a serious
counterexample, should be a sentence to be
substituted for 'p' such that one can
imagine that p (and tell less-imaginative folk
a story that enables them to imagine that p)
plus a good argument that it is impossible
that p. No such counterexamples have been
forthcoming. which is evidence for Hume's
principle. Not that such a counterexample
would require for its second component an
epistemology for possibility alternative to
Hume's. It is that those who object to his
principle refuse to provide, bluster about
meaning or logic notwithstanding.
Much more could be said here. but our
space is limited, so how could one get from
Hume's principle to dualism. The major
premise, Hume's principle, says that what
you can imagine is possible. The conclUSion.
dualism, says that you could be dis-
embodied. So the obvious minor premise in
an argument for dualism says that you can
imagine being disembodied. This argument
is preferable to the one from dreaming and
the cogito because it is valid and it is for the
required conclusion.
To show how to imagine being dis-
embodied, begin by dividing mental func-
tions into three sorts: input, inside and
output. Input from the world includes the
five senses: sight, hearing, taste, touch and
smell (see PERCEPTIONS). Output to the
world includes ACTION. like moving oneself
about the world. Inside the mind lies the
rest of the mind, like THO UGH T and
EMOTION. Neither input nor output are
essential to the mind; disembodiment could
blind and paralyse. But the interest of
dualism is proportional to the extent of the
mind that could be disembodied, so since
input and output can be preserved, they
should. (On the other hand, Descartes, who
was after the essence of mind, may have
shrunk it to its inside, thus exposing himself
to charges of, say, intellectualism.)
Starting then with sight, can one imagine
seeing while disembodied? Suppose that, still
embodied, you awaken one morning but,
before raising your eyelids, you stumble
over to your mirror. Then, facing your
mirror, you raise your eyelids, and see in
the mirror that your eye sockets are empty;
you can imagine how your face would look
in the mirror if you had no eyes. Interested,
you probe the sockets with a finger; you can
imagine how it would look in the mirror
and feel to probe the empty socket and
channel where your optic nerve used to be.
Even more curious, you take your surgical
saw and cut off the top of your skull. In the
mirror you see that your brain pan is
empty; you can imagine how your empty
brain pan would look in the mirror and feel.
You've now imagined seeing without the
two body-parts, eyes and a brain, most
people think crucial to seeing. You don't
need your legs to see, so imagine them
away, and similarly for your arms, trunk
and the rest of your head. Now of course
what you see in the mirror is no longer
your face, for that is gone; what you see
now is the wall behind you opposite the
mirror.
Have we really imagined seeing without a
body? (Imagination is no more incorrigible
than sight.) This is partly a matter of what
sight is (and part of the interest of arguing
for the minor premise is the light it casts on
our faculties). Sight requires visual experi-
ence, and we have imagined having visual
experience while disembodied. But visual
experience is common to sight, visualizing
and dreaming, and thus does not suffice for
sight. At this point the tradition adds veridi-
cality. Veridicality is to experience as truth
DUALISM
is to thought; a thought is true when the
world is as the thinker thinks it to be, and a
visual experience is veridical when the
world is as it looks to the one with the
experience. The tradition distinguishes sight
from visualizing and dreaming by veridi-
cality. But veridicality is a bare conjunctive
property; all it requires is that it look to one
as if p, and p. It is easy to add veridicality to
our story about visual experience while dis-
embodied.
But Grice showed that veridical visual
experience is not sufficient for sight, since
the world could be as one visualizes it to
be, but only by accident. So, Grice inferred,
veridical visual experience is sight only if it
is caused by what makes it veridical.
(Grice did not think that veridical visual
experience caused by what makes it ver-
idical suffices for sight, but that is another
story.) So if our disembodied person with
veridical visual experience is to see what
makes his visual experience veridical, that
experience should be caused by what
makes it veridical. But how could there be
causal commerce between immaterial dis-
embodied people and the physical objects
they see?
Here we come up against what has
seemed since Descartes the most difficult
problem confronting dualism, the interac-
tion problem, the problem of how, if imma-
terial mind is as different from material
body as Descartes claims, mind and body
can interact causally (see REASONS AND
CAUSES). (It is plain that they do so inter-
act, since a kick in the shins, which is phy-
sical, hurts, which is mental, and fear,
which is mental, makes one's heart beat
faster, which is physical.) Not that the
interaction problem is as much a problem
about the nature of causation as it is a
problem about the natures of mind and
matter. Hume, for example, explains causa-
tion in terms of temporal priority, spatial
contiguity and constant conjunction. Con-
tiguity requires location. Suppose we could
locate a disembodied person's visual experi-
ence. (As a subjective preliminary, note
that visual experience is always along lines
of sight. In visualizing fantasies, these sub-
267
DUALISM
jective lines of sight need not coincide with
real lines in space, but when visual experi-
ence is veridical, they will, and the point,
or small region of their convergence is a
subjectively described preliminary place
where the disembodied person's visual
experience lies.) The crux of causation for
Hume is constant conjunction, by which he
intended bare conjunction: this and then
that. But bare constant conjunction
between physical and mental (or vice
versa) event types is no more problematic
than such conjunction between physical
and physical event types. Here judgment
seems called for: either Hume is right about
causation and there is no interaction
problem, or else there is an interaction
problem and we need a better view of cau-
sation to make sense of it. The second
seems the better option.
Quine argues that causation is the flow
of energy. Could any quantity pass
between things as different as mind and
matter are according to dualism? This for-
mulation seems to do better justice to the
interaction problem. But it also suggests a
treatment of that problem. Energy (or
mass-energy) is conserved, and conserva-
tion is a quantitative principle. So we need
intrinsically psychological quantities. To
cut a long story very short, we can give a
general account of quantity (satisfied by
temperature, mass, length and so on) that
is pretty well satisfied by desire, belief, how
much it looks to one as if such and such,
and other psychological phenomena. Once
we have such psychological quantities, we
may imagine that as light from objects
seen reaches the region of convergence
along the disembodied person's lines of
sight, it passes straight through but loses
some electromagnetic energy and, at a
fixed rate of conversion, that person
acquires or is sustained in visual experi-
ence of those objects seen. Here is a way to
imagine that the disembodied person's ver-
idical visual experience be caused by that
in virtue of which it is veridical, and thus
far, that he sees. So we have solved the
interaction problem. To be sure, we have
not imagined exactly how light energy con-
268
verts into the psychic energy implicit in
visual experience. But then neither do phy-
sicists tell us how mass turns into energy
when an atom bomb goes off, and if their
lacuna does not embarrass them, neither
need ours embarrass us. (Now we can
better locate visual experience where the
conversion from electromagnetic energy to
psychic energy occurs.)
A good deal more could, and should, be
said, especially about action, but our space
is limited, and it is already clear that our
version of dualism differs from that of Des-
cartes. His interaction problem was about
how an embodied person's mind is, for
example, acted on by objects he sees. It is
difficult to take Descartes seriously about
the pineal gland. For that would at best
tell us only where interaction occurs, not
what it is like. More deeply, Descartes
takes extension to be the essence of matter.
Under extension, Descartes seems to put all
geometrical properties including location.
So to keep mind immaterial, he must deny
it location; his view of the mind is not so
much other-wordly as a-wordly. We post-
Newtonians accept geometrical objects in
regions of space innocent of matter, so for
us location does not require physicality,
and thus disembodied minds can be some-
where in space without being thereby
embodied. Note too that when disembodied
people see on our model, light loses energy
where their visual experience occurs, and
if this loss were large enough, we could
see where a disembodied person is. From
one angle, it might look like a dimming
where a part of the disembodied person is.
So disembodied people could look like the
hazy holograms in the haunted house at
Disneyland. In other words, when we
think and imagine disembodiment through,
we rediscover the possibility of the folklore
of ghosts. Therefore, scientistic propaganda
from materialists notwithstanding, dualism
is the commonsense solution to the mind-
body problem.
See also An Essay on Mind section 3.5;
HISTORY; IDENTITY THEORY; INTROSPEC-
TION; PHYSICALISM; SUBJECTIVITY.
BIBLIOGRAPHY
Descartes, Meditations. In Philosophical Works
of Descartes, trans. S. Haldene and G. R. T.
Ross, 2 vols. New York: Dover, 1955.
Hart, W.D. 1988. The Engines of the Soul. Cam-
Kripke, S. 1971. Identity and necessity. In
DUALISM
Identity and Individuation, ed. M. K. Munitz.
New York University Press.
Penelhum, T. 1970. Survival and Disembodied
Existence. Atlantic Highlands, N.J: Human-
ities Press.
Williams, B. 1978. Descartes: The Project of
Pure Enquiry. Penguin.
WILLIAM D. HART
269
E
eliminativism Eliminativists believe there
to be something fundamentally mistaken
about the common-sense (sometimes called
'folk psychological') conception of the mind,
and they suggest that the way forward is to
drop part or all of this conception in favour
of one which does not use notions such as
belief, experience, sensation and the like.
The rationale for this suggestion is, in the
main, because these notions are fraught
with conceptual difficulties as well as being
recalcitrant to any REDUCTION to natural
science. Since the conception with which
they propose to replace the common-sense
conception is invariably physicalist or ma-
terialist, one finds eliminativism also called
'eliminative materialism'. (See CONNECTION-
ISM; DENNETT; An Essay on Mind section
3.7; FOLK PSYCHOLOGY; LEWIS; SUPER-
VENIENCE.)
SAMUEL GUTTENPLAN
emotion In recent philosophy of mind,
emotion has been treated as a poor relation.
One reason for this is that emotions seem to
overstep a threshold of messiness beyond
which even the most masochistic of theor-
eticians tend to lose heart. The sheer variety
of phenomena covered by the word
'emotion' and its close neighbours seems to
discourage theory. Confusion about the
nature of emotions is reflected in the histor-
ical vicissitudes of the words used to speak
of it. Baier (1990) has noted, for example,
that while 'emotion' used to mean 'violent
passion', we now seem to use 'passion' to
mean 'violent emotion'. Much of these shifts
in emphasis through time are no doubt
merely the outcome of random linguistic
drift. We should expect that some of them,
however, reflect real differences between dif-
270
ferent implicit background models of the
architecture of the mind.
Such models differ in two different ways.
One concerns the place of emotion within
the general economy of mind. For Plato in
the Republic, there seemed to have been
three basic components of the human mind:
the reasoning, the desiring, and the emotive
parts. In subsequent theories, however,
emotions as a category are apt be sucked
into either of two other areas, of which
emotions are then treated as a mere
satrapy: a peculiar kind of BELIEF, or a
vague kind of DESIRE or WILL, or even mere
apprehension of bodily condition (James,
1884). For reasons that will be sketched
below, Plato's side of this debate now seems
the stronger.
A second set of alternative models relate
to the taxonomy of emotions themselves.
One model, advocated by Descartes as well
as by many contemporary psychologists,
posits a few basic emotions out of which all
others are compounded. An alternative
model views every emotion as consisting in,
or at least including, some irreducibly spe-
cific component not compounded of any-
thing simpler. Again, emotions might form,
like colour, an indefinitely broad continuum
comprising a small number of finite dimen-
sions, like the sensitivity ranges of retinal
cones corresponding roughly to primary
hues. We might then hope for relatively
simple biological explanations for the rich
variety of emotions, though rigid bound-
aries between them would be logically
arbitrary.
To date cognitive science does not seem
to have provided any crucial tests to decide
the second sort of question. Cross-cultural
research strongly suggests that a number of
emotions have intertranslatable names and
universally recognizable expressions.
According to Paul Ekman (1989) these are
happiness, sadness, fear, anger, surprise,
and disgust (which some researchers,
however, account too simple to be called an
emotion (Panksepp, 1982)). Other emotions
are not so easily recognizable cross-
culturally, and some expressions are almost
as local as dialects. But then this is an issue
on which cognitive science alone should
not, perhaps, be accorded the last word:
what to a neurologist might be classed as
two tokens of the same emotion type might
seem to have little in common under the
magnifying lens of a Proust. (See COGNITIVE
PSYCHOLOGY; PSYCHOLOGY AND PHILO-
SOPHY.)
These complications do not suffice to
explain philosophy's neglect of the emotions.
(Philosophers, after all, tend rather to be
fond of complications.) Indeed, this neglect is
both relatively recent and already out of
fashion. Most of the great classical philos-
ophers - Plato, Aristotle, Spinoza, Descartes,
Hobbes, Hume - have had recognizable
theories of emotion. Yet in twentieth-
century Anglo-American philosophy and
psychology, the increasing attention
recently devoted to emotion has had an air
of innovation. Under the influence of a
'tough-minded' ideology committed to
BEHA VIOURISM, theories of ACTION or THE
WILL, and theories belief or knowledge,
had seemed more readily achievable than
theories of emotion. The recently dominant
Bayesian-derived economic models of
rational decision and agency are essentially
assimilative models - two-factor theories,
which view emotion either as a species of
belief, or as species of desire.
That enviably resilient Bayesian model
has been cracked, in the eyes of many phil-
osophers, by such refractory phenomena as
akrasia or 'WEAKNESS OF WILL'. In cases of
akrasia, traditional descriptive rationality
seems to be violated, insofar as the 'stron-
gest' desire does not win, even when paired
with the appropriate belief (DAVIDSON,
1980). Emotion is ready to pick up the slack
- if only we had a theory of that.
It is one thing, however, to recognize the
EMOTION
need for a theory of mind that finds a place
for the unique role of emotions, and quite
another to construct one. Emotions vary so
much in various dimensions - transpar-
ency, intensity, object-directedness, or sus-
ceptibility to rational assessment - as to cast
doubt on the assumption that they have
anything in common.
What is distinctive about emotions is
perhaps precisely what made them a theor-
etical embarrassment: that they have a
number of apparently contradictory proper-
ties. In what follows, we shall sketch five
areas in which emotions pose specific philo-
sophical puzzles: emotion's relation to cog-
nition; emotions and self-knowledge; the
relation of emotions to their objects; the
nature of emotional intensity; and the rela-
tion of emotions to rationality.
EMOTION AND COGNITION
It is a commonplace (whether true or false)
that emotions are in some sense 'subjective'.
Some have taken this to mean that they
reflect nothing but the peculiar conscious-
ness of the subject. But that conclusion
follows only if one adopts a fallacious equa-
tion of point of view and SUBJECTIVITY. The
existence of perspective does not invalidate
cognition: that emotional states are per-
spectival, therefore, need not bar them from
being cognitive or playing a role in cogni-
tion.
There are at least three ways in which
emotions have been thought to relate to
cognition.
(1) As stimulants of cognition: philosophers
have been interested in learning from
psycho-physiologists that you won't
learn anything unless the limbic system
- the part of the brain most actively
implicated in emotional states - is
stimulated at the time of learning
(Scheffler, 1991).
(2) Many emotions are specified in terms of
propositions: one can't be angry with
someone unless one believes that
person guilty of some offence; one can't
be jealous unless one believes that one's
271
EMOTION
emotional property is being poached on
by another. From this. it has been in-
ferred that emotions are (always? some-
times?) cognitive in the sense that they
involve PROPOSITIONAL ATTITUDES.
This claim is relatively weak. however.
since the existence of a propositional
attitude is at best a necessary. but not a
sufficient condition of the existence of
an emotion.
(3) The most literal interpretation of cogni-
tivism about emotions would be com-
mitted to ascribing to emotions a 'mind-
to-world direction of fit'. The expression
'direction of fit'. which is due to SEARLE
(1983). distinguishes between an essen-
tially cognitive orientation of the mind.
in which success is defined in terms of
whether the mind fits the world (a
mind-to-world direction of fit) and an
essentially conative orientation. in
which success is defined in terms of the
opposite. world-to-mind. direction of fit.
We will what does not yet exist. and
deem ourselves successful if the world is
brought into line with the mind's plan.
A view ascribing to emotions a true
mind-to-world direction of fit would involve
a criterion of success that depended on cor-
rectness with respect to some objective
property. Such a view was first defended by
Scheler (1954). and has in general had
more currency as a variant of an objectivist
theory of aesthetics than as a theory of
emotions as a whole.
To take seriously cognitivism in this
sense. is to give a particular answer to the
question posed long ago in Plato's Euthy-
phro: Do we love X - mutatis mutandis for the
other emotions - because X is lovable. or do
we declare X to be lovable merely because we
love it? One way to defend a modest objec-
tivism. in the sense of the first alternative. is
to explore certain analogies between
emotion and PERCEPTION. It requires first
that we define clearly what is to count as
'objectivity' in the relevant sense. Second. it
requires that we show that there is a valid
analogy between some of the ways in which
we can speak of perception as aspiring to
272
objectivity and ways in which we can say
the same of emotion.
Emotions are sometimes said to be sub-
jective in this sense: that they merely reflect
something that belongs exclusively and
contingently to the mind of the subject of
experience. and therefore do not covary
with any property that could be indepen-
dently identified. This charge presupposes a
sense of 'objective' that contrasts with 'pro-
jective. in something like the psycho-
analytic sense (see PSYCHOANALYTIC
EXPLANATION). In terms of the analogy of
perception. to say that emotions are uni-
versally subjective in this sense would be to
claim that they resemble hallucinations
more than veridical perceptions. The per-
ceptual system is capable of the sort of
vacuous functioning that leads to percep-
tual mistakes. Similarly. emotions may
mislead us into 'hasty' or 'emotional' judg-
ments (Solomon. 1984). On the other hand.
the lack of perceptual capacities can be a
crippling handicap in one's attempt to
negotiate the world: similarly a lack of ade-
quate emotional responses can hinder our
attempts to view the world correctly and act
correctly in it (Nussbaum. 1990; Thomas.
1989). This explains why we are so often
tempted to take seriously ascription of rea-
sonableness or unreasonableness. fittingness
or inappropriateness. for common emotions.
The big drawback of this view is that it is
quite unclear how independently to identify
the alleged objective property.
Closely related to the question of the cog-
nitive aspect of emotion is the question of its
passivity. Passivity has an ambiguous rela-
tion to subjectivity. In one vein. impressed
by the bad reputation of the 'passions' as
taking over our consciousness against our
will. philosophers have been tempted to take
the passivity of emotions as evidence of
their subjectivity. In another vein. however.
represented especially in the last few years
by Robert Gordon (1987). philosophers
have noted that the passivity of emotions is
sometimes precisely analogous to the pas-
sivity of perception. How the world is. is not
in our power. So it is only to be expected
that our emotions. if they actually represent
something genuinely and objectively in the
world, should not be in our power either.
To this extent, the cognitive model holds
out rather well, while at the same time sug-
gesting that our common notion of what
cognition amounts to may be excessively
narrow.
One more angle needs to be canvassed on
the cognitive role of the emotions. This is
their role not as specific cognitions, but as
providing the framework for cognition. This
we shall consider in the final section, about
the rationality of emotion. First, however,
let us look at a very specific sort of know-
ledge which emotions can either enhance or
distort.
EMOTIONS AND SELF-KNOWLEDGE
We often make the 'Cartesian' assumption
that if anyone can know our emotions it is
ourselves. Descartes said it thus: 'it is impos-
sible for the soul to feel a passion without
that passion being truly as one feels it'. (See
FIRST-PERSON AUTHORITY; INTROSPECTION;
THE SELF.) Barely a page later, however,
he noted that 'those that are most agitated
by their passions are not those who know
them best' (Descartes, 1984, 26, 28). In
fact. emotions are one of our avenues to self-
knowledge, since few kinds of self-knowledge
could matter more than knowing one's own
repertoire of emotional responses. At the
same time, emotions are both the cause
and the subject of many failures of self-
knowledge. Their complexity entails several
sources for their potential to mislead or be
misled. Insofar as most emotions involve
belief, they inherit the susceptibility of the
latter to SELF-DECEPTION. Recent literature
on self-deception has dissolved the air of
paradox to which this once gave rise (Fin-
garette, 1969). But there are also three dis-
tinct problems that are specific to emotions.
The first arises from the connection of
emotion with bodily changes. There is
something right in William James's notor-
ious claim that the emotion follows on,
rather than causing the voluntary and
involuntary bodily changes which express it
(James, 1884). Because some of these
EMOTION
changes are either directly or indirectly
subject to our choices, we are able to
pretend or dissimulate emotion. That
implies that we can sometimes be caught in
our own pretence. Sometimes we identify
our emotions by what we feel: and if what
we feel has been distorted by a project of
deception, then we will misidentify our own
emotions.
A second source of self-deception arises
from the role of emotions in determining
salience among potential objects of attention
or concern. Poets have always known that
the main effect of love is to redirect atten-
tion: when I love, I notice nothing but my
beloved, and nothing of his faults. When
my love turns to anger I still focus on him,
but now attend to a very different set of
properties. This suggests one way of con-
trolling or dominating my emotion: think
about something else, or think differently
about this object. But this carries a risk. It is
easier to think of something than to avoid
thinking about it; and to many cases of
emotional distress only the latter could
bring adequate relief. Besides, one is not
always able to predict, and therefore to
control, the effect that redirected attention
might produce. The best explanation for this
familiar observation requires us to take seri-
ously the hypothesis of the UNCONSCIOUS: if
among the associations that are evoked by
a given scene are some that I can react to
without being aware of what they are, then
I will not always be able to predict my own
reactions, even if I have mastered the not
altogether trivial task of attending to what-
ever I choose. Where the unconscious is,
self-deception necessarily threatens. (See
PSYCHOANALYTIC EXPLANATION.)
This brings us to the third source of emo-
tional self-deception: the involvement of
social norms in the determination of our
emotions. This possibility arises in two
stages from the admission that there are
unconscious motivations for emotions. First,
if I am experiencing an emotion that seems
altogether inappropriate to its occasion, I
will naturally confabulate an explanation
for it. A neurotic who is unreasonably
angry with his wife because he uncon-
273
EMOTION
sciously identifies her with his mother will
not rest content with having no reason for
his anger. Instead, he will make one up.
Moreover, the reason he makes up will typi-
cally be one that is socially approved (Averill,
1982).
When we are self-deceived in our emo-
tional response, or when some emotional
state induces self-deception, there are
various aspects of the situation about which
self-deception can take place. These relate to
the different kinds of intentional objects of
emotions.
INTENTIONAL OBJECTS
What does a mood, such as free-floating
depression or euphoria, have in common
with a precisely articulable indignation? The
first seems to have as its object nothing and
everything, and often admits of no par-
ticular justification; the second has a long
story to tell typically involving other people
and what they have done or said. Not only
those people, but the relevant facts about
the situations involved, as well as some of
the special facts about those situations,
aspects of those facts, the causal role played
by these aspects, and even the typical aims
of the actions motivated by the emotions,
can all in some context or other be labelled
objects of emotion. Objects are what we
emote at, with, to, because of, in virtue of,
or that (see INTENTIONALITY). This variety
has led to a great deal of confusion and
debate. A big question, for example, has
been the extent to which the objects of
emotions are to be identified with their
causes. This identification seems plausible;
yet it is easy to construct examples in which
being the cause of an emotion is intuitively
neither a necessary nor a sufficient condi-
tion for being its object: if A gets annoyed at
B for some entirely trivial matter, drunken-
ness may have caused A's annoyance, yet it
is in no sense its object. Its object may be
some innocent remark of B's, which while it
occasioned the annoyance cannot rightly be
said to be its cause. More precisely, it may
be a certain insulting quality in B's remark
which is, as a matter of fact, entirely ima-
274
ginary and therefore could not possibly be
its true cause.
The right way to deal with these com-
plexities is to embrace them. We need a tax-
onomy of the different sorts of possible
emotional objects. One might then distin-
guish different types of emotions according
to the different complex structures of their
object relations. Many emotions, such as
love, necessarily involve a target, or actual
particular at which they are directed.
Others, such as sadness, do not. On the
other hand, although there may be a
number of motivating aspects of the loved
one on which our love is focused, love lacks
a propositional object. Sadness, on the con-
trary, cannot be fully described without
specifying such a propositional object.
Depression or elation can lack all three
kinds of object. And so on.
Special puzzles arise for those emotions
that take propositional objects, such as
being sad that p, pleased that p, or embar-
rassed that p. Does p actually have to be
true for the emotion to be correctly attrib-
uted? It seems in general that they need
not, since a mistake about the relevant facts
seems to be one of the ways in which one
can be mistaken while experiencing an
emotion. Robert Gordon (1987), however,
has defended the interesting thesis that all
emotions taking propositional objects are
either 'factive', or 'epistemic'. Factive emo-
tions require for their correct ascription not
only that the proposition in question be
believed or thought probable, but that it
actually be true. Epistemic emotions, on the
other hand, require that the propOSitional
object be neither definitely believed nor defi-
nitely disbelieved. Thus one cannot be sad
that p unless it is true that p (although if it
is false one can be sad because one mis-
takenly believe that p); and one can fear
that p only if one is unsure whether p.
Whatever its merits, this controversial thesis
has an illuminating corollary: namely that
in most attributions of emotion, speaker,
hearer and subject share a common world.
Unless the uncertainty of the situation is
specifically in question, as in the epistemic
emotions, or unless the assumption is
explicitly defeated by the rider that the
subject doesn't quite live in our shared
world, we assume that we can all take the
same facts for granted. This would give the
emotions a kind of intrinsically realist bias,
and would help to explain why arguments
about emotions so often break down. They
break down because each participant
assumes a shared world which, in fact, con-
sists only of the set of scenarios that define
his or her own private emotional idiolect, or
repertoire of intelligible emotions.
INTENSITY
It seems to be an irreducible differentia of
emotions that they can be measured along
a dimension of intensity. This corresponds
neither to the strength of desire nor to a
belief's degree of confidence. What does
mild distaste have in common with the
most murderous rage? Is it just a matter of
degree? Or does intensity necessarily bring
with it differences in kind? Two different
sorts of considerations favour endorsing the
latter view. The difference between them
illustrates a characteristic methodological
dilemma faced by emotions research. The
first approaches taxonomy through social
significance: mild distaste is one thing, rage
quite another, in the sense that the circum-
stances in which the first or the second is
generally appropriate and acceptable are
radically disjoint. From this pOint of view,
then, they must obviously be classed as
entirely different phenomena. But a similar
response might be derived from an entirely
different approach: one might look at the
brain's involvement in the two cases and
find (perhaps) the first to be an essentially
cortical response, while the second involves
activity of the limbic system or even the
brain stem - what Paul Maclean (1975) has
dubbed the 'mammalian' or the 'crocodile'
brain. In this case the classification of the
two as entirely separate phenomena might
have a strictly physiological basis. How are
the two related?
The very notion of intensity is problematic
exactly to the extent that the emotions call
for disparate principles of explanation. Might
EMOTION
a physiological criterion settle the question?
One could stipulate that the most intense
emotion is the one that involves the greatest
quantity of physiological 'disturbance'
(Lyons, 1980). But this approach must
implicitly posit a state of 'normal' quietude
hard to pin down among the myriad differ-
ent measures of physiological activity one
might devise. To select a measure that will
count as relevant, one will inevitably have
to resort to another level of more functional
criteria: what are the types and levels of
physiological activity that are relevant to
the social functions subserved by those emo-
tions? And what are the mental functions
that should be deemed most important in
the context of the relevant demands of social
life? At that point, while physiological expla-
nations may be of great interest, there is no
hope from their quarter of any interesting
criteria for emotional intensity.
RA TION ALITY
There is a common prejudice that 'feelings',
a word now sometimes vulgarly used inter-
changeably with 'emotions', neither owe
nor can give any rational account of them-
selves. Yet we equally commonly blame
others or ourselves for feeling 'not wisely,
but too well', or for targeting inappropriate
objects. As we have seen, the norms appro-
priate to both these types of judgment are
inseparable from social norms, whether or
not these are endorsed. Ultimately they are
inseparable from conceptions of normality
and human nature. Judgments of reason-
ableness therefore tend to be endorsed or
rejected in accordance with one's ideologi-
cal commitments to this or that conception
of human nature. It follows that whether
these judgments can be viewed as objective
or not will depend on whether there are
objective facts to be sought about human
nature. On this question we fortunately do
not need to pronounce. It is enough to note
that there is no logical reason why judg-
ments of reasonableness or irrationality in
relation to emotions need be regarded as
any more subjective than any other judg-
ments of RATIONALITY in human affairs.
275
EMOTION
There is a further contribution that the
study of emotions can make to our under-
standing of rationality. The clearest notions
associated with rationality are coherence
and consistency in the sphere of belief, and
maximizing expected utility in the sphere of
action. But these notions are purely critical
ones. By themselves, they would be quite
incapable of guiding an organism towards
any particular course of action. For the
number of goals that it is logically possible
to posit at any particular time is virtually
infinite, and the number of possible strat-
egies that might be employed in pursuit of
them is orders of magnitude larger. More-
over, in considering possible strategies, the
number of consequences of anyone strategy
is again infinite, so that unless some drastic
preselection can be effected among the
alternatives their evaluation could never be
completed. This gives rise to what is known
among cognitive scientists as the 'Frame
Problem': in deciding among any range of
possible actions, most of the consequences
of each must be eliminated from considera-
tion a priori, i.e. without any time being
wasted on their consideration. That this is not
as much of a problem for people as it is for
machines may well be due to our capacity
for emotions. Emotions frame our decisions
in two important ways. First, they define
the parameters taken into account in any
particular deliberation. Second, in the
process of rational deliberation itself, they
render salient only a tiny proportion of the
available alternatives and of the conceivably
relevant facts. In these ways, then, emo-
tions would be all-important to rationality
even if they could not themselves be deemed
rational or irrational. For they winnow
down to manageable size the number of
considerations relevant to rational delibera-
tion, and provide the indispensable frame-
work without which the question of
rationality could not even be raised.
BIBLIOGRAPHY
Averill, J. 1982. Anger and Aggression: An
Essay on Emotion. New York: Springer-
Verlag.
276
Baier, A. 1990. What emotions are about.
Philosophical Perspectives, 4. 1-29.
Davidson, D. 1980. How is weakness of the
will possible? In Essays on Actions and
Descartes, R. 1984. The Passions of the Soul.
Vol. 1 of The Philosophical Writings of Des-
cartes, Trans. J. Cottingham, R. Stoothoff
and D. Murdoch. Cambridge University
Press.
de Sousa, R. 1987. The Rationality of Emotion.
Cambridge, MA. MIT Press.
Ekman, P., and Friesen, W.V. 1989. The argu-
ment and evidence about universals in
facial expressions of emotion. In Handbook of
Social Psychophysiology. New York: John
Wiley & Sons Ltd.
Fingarette, H. 1969. Self Deception. London:
Gordon, R.M. 1987. The Structure of Emotions.
James, W. 1984. What is an emotion? Mind,
19, 188-204.
Lyons, W. 1980. Emotion. Cambridge Uni-
versity Press.
MacLean, P.D. 1975. Sensory and perceptive
factors in emotional functions of the triune
brain. In Emotions: Their Parameters and
Measurement, ed. L. Lev. New York: Raven
Press.
Nussbaum, M. 1990. Love's Knowledge. Oxford
University Press.
Panksepp, J. 1982. Toward a general psycho-
biological theory of emotions. Behavioral and
Brain Sciences, 5,407-76.
Scheffler, 1. 1991. In Praise of the Cognitive
Emotions. London: Routledge & Kegan Paul.
Scheler, M. 1954. The Nature of Sympathy.
Trans. P. Heath, introduction by W. Stark.
Hamden, CT: Archon.
Searle, J .R. 1983. Intentionality. Cambridge
University Press.
Solomon, R.C. 1984. The Passions: The Myth
and Nature of Human Emotions. New York:
Doubleday.
Thalberg, 1. 1977. Perception, Emotion and
Action. New Haven: Yale University Press.
Thomas, L. 1989. Living Morally: A Psychology
of Moral Character. Philadelphia: Temple
University Press.
RONALD DE SOUSA
emergence see SUPERVENIENCE.
epiphenomenalism Mental phenomena
bodily sensations, sense experiences, emo-
tions, memories, beliefs, desires, intentions,
and the like - seem to have causal effects.
For example, the sharp pain in my toe
might make me wince. The wall's looking
red to me might make me think that it is
red. As I gaze over the edge of the cliff, my
fear of heights might make my breath
quicken and my heart beat faster. Remem-
bering an embarrassing situation might
make me blush. My desire to ring the alarm
and belief that I can do so by pressing the
button might result in my pressing the
button. Despite my fatigue, I might keep
grading papers, my intention to finish the
grading before I go to sleep sustaining my
effort. And as I reason through a problem,
one thought leads to another. Or so it
seems.
However, according to the traditional
doctrine of epiphenomenalism, things are
not as they seem: in reality, mental phe-
nomena can have no causal effects; they are
causally inert, causally impotent. Only
physical phenomena are causally effica-
cious. Mental phenomena are caused by
physical phenomena, but they cannot cause
anything. In short, mental phenomena are
epiphenomena.
The epiphenomenalist claims that mental
phenomena seem to be causes only because
there are regularities that involve types (or
kinds) of mental phenomena, and regular-
ities that involve types of mental and phys-
ical phenomena. For example, instances of a
certain mental type M (e.g. trying to raise
one's arm) might tend to be followed by
instances of a physical type P (e.g. one's
arm's rising). To infer that instances of M
tend to cause instances of P would be,
however, to commit the fallacy of post hoc,
ergo propter hoc. Instances of M cannot
cause instances of P; such causal transac-
tions are causally impossible. P-type events
tend to be followed by M-type events
because instances of such events are dual-
effects of common physical causes, not
because such instances causally interact.
EPIPHENOMENALISM
Mental events and states can figure in the
web of causal relations only as effects, never
as causes.
SOME APPARENT CONSEQUENCES OF THE
DOCTRINE
Epiphenomenalism is a truly stunning doc-
trine. If it is true, then no pain could ever be
a cause of our wincing, nor could some-
thing's looking red to us ever be a cause of
our thinking that it is red. A nagging head-
ache could never be a cause of a bad mood.
Moreover, if the causal theory of memory is
correct, then, given epiphenomenalism, we
could never remember our prior thoughts,
or an emotion we once felt, or a toothache
we once had, or having heard someone say
something, or having seen something; for
such mental states and events could not be
causes of memories. Furthermore, epipheno-
menalism is arguably incompatible with the
possibility of intentional action. For if, as
the causal theory of action implies, inten-
tional action requires that a desire for some-
thing and a belief about how to obtain what
one desires play a causal role in producing
behaviour, then, if epiphenomenalism is
true, we cannot perform intentional actions.
(See ACTION; INTENTION; REASONS AND
CAUSES.) Notice that we cannot, then, assert
that epiphenomenalism is true, if it is, since
an assertion is an intentional speech act (cf.
Malcolm, 1968). Still further, if epiphenom-
enalism is true, then our sense that we are
agents who can act on our intentions and
carry out our purposes is illusory. We are
actually passive bystanders, never agents; in
no relevant sense is what happens up to us.
Our sense of partial causal control over our
limb movements is illusory. Indeed, we
exert no causal control over even the direc-
tion of our attention! Finally, suppose that
reasoning is a causal process. Then, if epi-
phenomenalism is true. we never reason;
for there are no mental causal processes.
While one thought may follow another, one
thought never leads to another. Indeed,
while thoughts may occur, we do not
engage in the activity of thinking. How,
then, could we make inferences that
277
EPIPHENOMEN ALISM
commit the fallacy of post hoc, ergo propter
hoc, or make any inferences at all for that
matter?
One might try to dispute whether epi-
phenomenalism has some of these con-
sequences by disputing, for example, the
causal theory of memory, or the causal
theory of action, or the claim that reasoning
is a causal process. But rather than pursu-
ing such issues, let us instead ask what
considerations led to the doctrine of epi-
phenomenalism.
THE NEUROPHYSIOLOGICAL NO-GAP
ARGUMENT
As neurophysiological research began to
develop in earnest during the latter half of
the nineteenth century, it seemed to find no
mental influence on what happens in the
brain. While it was recognized that neuro-
physiological events do not by themselves
causally determine other neurophysiological
events, there seemed to be no 'gaps' in neu-
rophysiological causal mechanisms that
could be filled by mental occurrences.
Neurophysiology appeared to have no need
of the hypothesis that there are mental
events. (Here and hereafter, unless indicated
otherwise, I follow the common philosophi-
cal practice of using 'events' in a broad
sense to include states as well as changes.)
This 'no gap' line of argument led some the-
orists to deny that mental events have any
causal effects (Huxley, 1874). They rea-
soned as follows: if mental events have any
effects, among their effects would be neuro-
physiological ones; mental events have no
neurophysiological effects; thus, mental
events have no effects at all. The relation-
ship between mental phenomena and
neurophysiological mechanisms was likened
to that between the steam-whistle which
accompanies the working of a locomotive
engine and the mechanisms of the engine:
just as the steam-whistle is an effect of the
operations of the mechanisms but has no
causal influence on those operations, so too
mental phenomena are effects of the work-
ings of neurophysiological mechanisms, but
have no causal influence on their opera-
278
tions. (The analogy quickly breaks down, of
course: steam-whistles have causal effects,
but the epiphenomenalist alleges that
mental phenomena have no causal effects at
all.)
An early response to this 'no gap' line of
argument was that mental events (and
states) are not changes in (and states of)
an immaterial Cartesian substance (see
DUALISM), they are, rather, changes in (and
states of) the brain. While mental properties
or kinds are not neurophysiological proper-
ties or kinds, nevertheless, particular mental
events are neurophysiological events (see
Lewes, 1875; Alexander, 1920). According
to the view in question, a given event can
be an instance of both a neurophysiological
type and a mental type, and thus be both a
mental event and a neurophysiological
event. (Compare the fact that an object
might be an instance of more than one kind
of object; for example, an object might be
both a stone and a paper-weight.) It was
held, moreover, that mental events have
causal effects because they are neurophysio-
logical events with causal effects. This
response presupposes that causation is an
extensional relation between particular
events, that if two events are causally
related, they are so related however they
are typed (or described) (see ANOMALOUS
MONISM; DAVIDSON). But that assumption
is today widely held. We will accept it
throughout this essay. Given that the causal
relation is extensional, if particular mental
events are indeed neurophysiological events
with causal effects, then mental events are
causes, and epiphenomenalism is thus false.
TYPE (PROPERTY) EPIPHENOMENALISM
This response to the 'no gap' argument,
however, prompts a concern about the rel-
evance of mental properties or kinds to
causal relations. And in 1925, C. D. Broad
tells us that the view that mental events are
epiphenomena is the view 'that mental
events either (a) do not function at all as
causal-factors; or that (b) if they do, they do
so in virtue of their physiological character-
istics and not in virtue of their mental
characteristics' (1925, p. 473). If particular
mental events are physiological events with
causal effects, then mental events function
as cause-factors: they are causes. However,
the question still remains whether mental
events are causes in virtue of their mental
characteristics. As we noted, neurophysiol-
ogy appears to be able to explain neuro-
physiological occurrences without postulat-
ing mental characteristics. This prompts the
concern that even if mental events are
causes, they may be causes in virtue of their
physiological characteristics, but not in
virtue of their mental characteristics.
This concern presupposes, of course, that
events are causes in virtue of certain of
their characteristics or PROPERTIES. But it is
today fairly widely held that when two
events are causally related, they are so
related in virtue of something about each.
Indeed, theories of causation assume that if
two events x and yare causally related, and
two other events a and b are not, then there
must be some difference between x and y
and a and b in virtue of which x and yare,
but a and b are not, causally related. And
they attempt to say what that difference is;
that is, they attempt to say what it is about
causally related events in virtue of which
they are so related. For example, according
to so-called nomic subsumption views of
causation, causally related events will be so
related in virtue of falling under types (or in
virtue of having properties) that figure in a
'causal law'. It should be noted that the
assumption that causally related events are
so related in virtue of something about each
is compatible with the assumption that the
causal relation is an extensional relationship
between particular events (cf. McLaughlin,
1993). The weighs-less-than relation is an
extensional relation between particular
objects: if 0 weighs less than 0*, then 0 and
0* are so related however they are typed (or
characterized, or described). Nevertheless, if
o weighs less than 0*, then that is so in
virtue of something about each, namely
their weights and the fact that the weight of
one is less than the weight of the other.
Examples are readily multiplied. Extensional
relations between particulars typically hold
EPIPHENOMEN ALISM
in virtue of something about the particulars.
In what follows, we will remain neutral
concerning the competing theories of causa-
tion. But we will grant that when two
events are causally related, they are so
related in virtue of something about each.
Invoking the distinction between event
types and event tokens, and using the term
'physical', rather than the more specific
term 'physiological', we can, following
Broad, distinguish two kinds of epiphenom-
enalism (cf. McLaughlin, 1989):
TOKEN EPIPHENOMENALISM Mental
events cannot cause anything.
TYPE EPIPHENOMENALISM No event
can cause anything in virtue of falling
under a mental type.
(Property epiphenomenalism is the thesis that
no event can cause anything in virtue of
having a mental property.) The conjunction
of token epiphenomenalism and the claim
that physical events cause mental events is,
of course, the traditional doctrine of epiphe-
nomenalism, as characterized earlier. Token
epiphenomenalism implies type epiphenom-
enalism: for if an event could cause some-
thing in virtue of falling under a mental
type, then an event could be both a mental
event and a cause, and so token epiphe-
nomenalism would be false. Thus if mental
events cannot be causes, then events
cannot be causes in virtue of falling under
mental types. The denial of token epiphe-
nomenalism does not, however, imply the
denial of type epiphenomenalism, if a
mental event can be a physical event that
has causal effects. For, if so, then token epi-
phenomenalism is false. But type epiphe-
nomenalism may still be true. For it may be
that events cannot be causes in virtue of
falling under mental types. Mental events
may be causes in virtue of falling under
physical types, but not in virtue of falling
under mental types. Thus, even if token epi-
phenomenalism is false, the question
remains whether type epiphenomenalism is.
(It should be noted, parenthetically, that
much of the current literature on epiphe-
nomenalism focuses on type (or property)
279
EPIPHENOMENALISM
epiphenomenalism, rather than on token
epiphenomenalism. Keith Cambell's 'new'
epiphenomenalism (1970, ch. 6) is a
version of type (property) epiphenomenal-
ism for 'phenomenal' mental types (proper-
ties). Moreover, many of the discussions of
type epiphenomenalism in the current lit-
erature have focused on views espoused by
Donald Davidson. Davidson (1970) holds
that (i) causally related events must fall
under a strict law, that (ii) all strict laws are
physical laws, and that (iii) no mental type
(or property) either is or 'reduces to' a phy-
sical type (or property). Commitment to (i)-
(iii), it has been claimed, commits him as
well to type (property) epiphenomenalism.
(The literature is vast, but see, especially,
Honderich (1982), Kim (1984), Sosa
(1984), LePore and Loewer (1987), and
Antony (1989).) Davidson (1993),
however, denies that he is so committed;
and McLaughlin (1989) argues that (i}-(iii),
at least, would not commit Davidson to type
epiphenomenalism. (For discussions of
Davidson (1993), see McLaughlin (1993),
Kim (1993), and Sosa (1993).)
SOME APPARENT CONSEQUENCES OF TYPE
EPIPHENOMENALISM
Suppose, for the sake of argument, that type
epiphenomenalism is true. Why would that
be a concern if mental events are physical
events with causal effects? On our assump-
tion that the causal relation is extensional.
it could be true, consistent with type epi-
phenomenalism, that pains cause winces,
that desires cause behaviour, that percep-
tual experiences cause beliefs, that mental
states cause memories, and that reasoning
processes are causal processes. Nevertheless,
while perhaps not as disturbing a doctrine
as token epiphenomenalism, type epiphe-
nomenalism can, upon reflection, seem dis-
turbing enough.
Notice to begin with that 'in virtue of'
expresses an explanatory relationship.
Indeed, 'in virtue of' is arguably a near
synonym of the more common locution
'because of' (cr. McLaughlin, 1989). But, in
any case, the following seems to be true: an
280
event causes a G-event in virtue of being an
F-event if and only if it causes a G-event
because of being an F-event. 'In virtue of'
implies 'because of', and in the case in ques-
tion at least the implication seems to go in
the other direction as well. Suffice it to note
that were type epiphenomenalism consistent
with its being the case that an event could
have a'certain effect because of falling under
a certain mental type, then we would indeed
be owed an explanation of why it should be
of any concern if type epiphenomenalism is
true. We will, however, assume that type
epiphenomenalism is inconsistent with that.
We will assume that type epiphenomenalism
could be reformulated as follows: no event
can cause anything because of falling under
a mental type. (And we will assume that
property epiphenomenalism can be reformu-
lated thus: no event can cause anything
because of having a mental property.) To
say that c causes e in virtue of being F is to
say that c causes e because of being F; that
is, it is to say that it is because c is F that it
causes e. So understood, type epiphenomen-
alism is a disturbing doctrine indeed.
If type epiphenomenalism is true, then it
could never be the case that circumstances
are such that it is because some event or
state is a sharp pain, or a desire to flee, or a
belief that danger is near, that it has a
certain sort of effect. It could never be the
case that it is because some state is a desire
to X (impress someone) and another is a
belief that one can X by doing Y (standing
on one's head) that the states jointly result
in one's doing Y (standing on one's head). If
type (property) epiphenomenalism is true,
then nothing has any causal powers what-
soever in virtue of (because of) being an
instance of a mental type (or having a
mental property). For it could, then, never
be the case that it is in virtue of being an
instance of a certain mental type that a
state has the causal power in certain cir-
cumstances to produce some effect. For
example, it could never be the case that it is
in virtue of being an urge to scratch (or a
belief that danger is near) that a state has
the causal power in certain circumstances
to produce scratching behaviour (or fleeing
behaviour). If type-epiphenomenalism is
true, then the mental qua mental. so to
speak, is causally impotent. That may very
well seem disturbing enough.
What reason is there, however, for
holding type epiphenomenalism? Even if
neurophysiology does not need to postulate
types of mental events, perhaps the science
of psychology does. Note that physics has
no need to postulate types of neurophysio-
logical events; but that may well not lead
one to doubt that an event can have effects
in virtue of being (say) a neuron firing.
Moreover, mental types figure in our every-
day causal explanations of behaviour,
intentional action, memory, and reasoning.
What reason is there, then, for holding that
events cannot have effects in virtue of being
instances of mental types? This question
naturally leads to the more general question
of which event types are such that events
have effects in virtue of falling under them.
This more general question is best addressed
after considering a 'no gap' line of argu-
ment that has emerged in recent years.
THE NO-GAP ARGUMENT FROM PHYSICS
Current physics includes quantum mechan-
ics, a theory which appears able, in prin-
ciple, to explain how chemical processes
unfold in terms of the mechanics of sub-
atomic particles. Molecular biology seems
able, in principle, to explain how the phy-
siological operations of systems in living
things occur in terms of biochemical path-
ways, long chains of chemical reactions. On
the evidence, biological organisms are
complex physical objects, made up of mol-
ecules, atoms, and more fundamental phys-
ical particles (there are no entelechies or
elan vital). Since we are biological organ-
isms, the movements of our bodies and of
their minute parts, including the chemicals
in our brains, are, like the movements of all
physical entities, causally determined, to
whatever extent they are so, by events
involving subatomic particles and fields.
Such considerations have inspired a line of
argument that only events within the
domain of physics are causes.
EPIPHENOMEN ALISM
Before presenting the argument, let us
make some terminological stipulations: let
us henceforth use 'physical event (state),
and 'physical property' in a strict and
narrow sense to mean, respectively, a type of
event (state) and a property postulated by
(current) physics (or by some improved
version of current physics). Event types and
properties are postulated by physics if they
figure in laws of physics. Finally, by 'a
physical event (state), we will mean an
event (state) that falls under a physical type.
Only events within the domain of (current)
physics (or some improved version of
current physics) count as physical in this
strict and narrow sense.
Consider, then
The Token-Exclusion Thesis Only physical
events can have causal effects (i.e. as a
matter of causal necessity, only physi-
cal events have causal effects).
The premises of the basic argument for the
token-exclusion thesis are:
Physical Causal Closure Only physical
events can cause physical events.
Causation via Physical Effects As a
matter of at least causal necessity, an
event is a cause of another event if and
only if it is a cause of some physical
event.
These principles jointly imply the exclusion
thesis. The principle of causation via physi-
cal effects is supported on the empirical
grounds that every event occurs within
space-time, and by the principle that an
event is a cause of an event that occurs
within a given region of space-time if and
only if it is a cause of some physical event
that occurs within that region of space-
time. The following claim is offered in
support of physical closure:
Physical Causal Determination For any
(caused) physical event, p, there is a
chain of entirely physical events
leading to P, each link of which cau-
sally determines its successor.
281
EPIPHENOMENALISM
(A qualification: if strict causal determinism
is not true, then each link will determine
the objective probability of its successor.)
Physics is such that there is compelling
empirical reason to believe that physical
causal determination holds. Every physical
event will have a sufficient physical cause.
More precisely, there will be a deterministic
causal chain of physical events leading to
any physical event, P. Each link of the
chain will be an occurrence involving all
the particles and fields within some cross-
section of the backward lightcone of p, but
such links there will be. Such physical
causal chains are entirely 'gapless'. Now, to
be sure, physical causal determination does
not imply physical causal closure; the
former, but not the latter, is consistent with
non-physical events causing physical
events. However, a standard epiphenomen-
alist response to this is that such non-
physical events would be, without excep-
tion, overdetermining causes of physical
events, and it is ad hoc to maintain that
non-physical events are overdetermining
causes of physical events (see An Essay on
Mind section 3.4)
Are mental events within the domain of
physics? Perhaps, like objects, events can
fall under many different types or kinds. We
noted earlier that a given object might, for
instance, be both a stone and a paper-
weight. However, we understand how a
stone could be a paperweight. But how, for
instance, could an event of subatomic par-
ticles and fields be a mental event? Suffice it
to note for the moment that if mental
events are not within the domain of physics,
then, if the token-exclusion thesis is true, no
mental event can ever cause anything:
token epiphenomenalism is true.
THREE RESPONSES TO THE NO-GAP
ARGUMENT FROM PHYSICS
One might reject the token-exclusion thesis,
however, on the grounds that, typical
events within the domains of the special
sciences - chemistry, the life sciences, etc. -
are not within the domain of physics, but
nevertheless have causal effects. One might
282
maintain that neuron firings, for instance,
cause other neuron firings, even though
neurophysiological events are not within
the domain of physics. Rejecting the token-
exclusion principle, however, requires
arguing either that physical causal closure
is false or that the principle of causation via
physical effects is.
But one response to the 'no gap' argu-
ment from physics is to reject physical
causal closure. Recall that physical causal
determination is consistent with non-phys-
ical events being overdetermining causes of
physical events. One might concede that it
would be ad hoc to maintain that a non-
physical event, N, is an overdetermining
cause of a physical event p, and that N
causes P in a way that is independent of the
causation of P by other physical events.
However, perhaps N can cause a physical
event P in a way that is dependent upon P's
being caused by physical events. One might
argue that physical events 'underlie' non-
physical events, and that a non-physical
event N can be a cause of another event, X
(physical or non-physical), in virtue of the
physical event that 'underlies' N being a
cause of X. (This sort of response is devel-
oped in Kim, 1984 and Sosa, 1984.)
Another response is to deny the principle
of causation via physical effects. Physical
causal closure is consistent with non-
physical events causing other non-physical
events. One might concede physical causal
closure but deny the principle of causation
via physical effects, and argue that non-
physical events cause other non-physical
events without causing physical events.
This would not require denying either that
(1) physical events invariably 'underlie'
non-physical events or that (2) whenever a
non-physical event causes another non-
physical event, some physical event that
underlies the first event causes a physical
event that underlies the second. Claims (1)
and (2) do not imply the principle of causa-
tion via physical effects. Moreover, from the
fact that a physical event, p, causes another
physical event, P*, it may not follow that P
causes every non-physical event that P*
underlies. That may not follow even if the
physical events that underlie non-physical
events causally suffice for those non-phys-
ical events. It would follow from that that
for every non-physical event, there is a cau-
sally sufficient physical event. But it may be
denied that causal sufficiency suffices for
causation; it may be argued that there are
further constraints on causation that can
fail to be met by an event that causally suf-
fices for another. Moreover, it may be
argued that given the further constraints,
non-physical events are the causes of non-
physical events. (This sort of response is
developed in Yablo, 1992a, 1992b.)
However, the most common response to
the 'no-gap' argument from physics is to
concede it. and thus to embrace its conclu-
sion, the token-exclusion thesis, but to
maintain the doctrine of token physicalism.
the doctrine that every event (state) is
within the domain of physics. If special
science events and mental events are within
the domain of physics, then they can be
causes consistent with the token-exclusion
thesis (see Davidson, 1970; Fodor, 1975).
Now whether special science events and
mental events are within the domain of
physics depends, in part, on the nature of
events; and that is a highly controversial
topic about which there is nothing
approaching a received view (see Bennett.
1988). The topic raises deep issues that are
beyond the scope of this essay, issues con-
cerning the 'essences' of events, and the
relationship between causation and causal
explanation. In any case, suffice it to note
here that I believe that the same funda-
mental issues concerning the causal efficacy
of the mental arise for all the leading the-
ories of events. indeed for all the leading
theories of the relata of the causal relation.
The issues just 'pop up' in different places.
However, that cannot be argued here, and
it won't be assumed.
Since the token physicalist response to
the no-gap argument from physics is the
most popular response, we shall focus on it.
Moreover, we will assume, just for the sake
of argument, that special science events,
and even mental events, are within the
domain of physics. Of course, if mental
EPIPHENOMENALISM
events are within the domain of physics,
then, token epiphenomenalism can be false
even if the token-exclusion thesis is true: for
mental events may be physical events
which have causal effects.
THE TYPE EXCLUSION THESIS
However. concerns about the causal relev-
ance of mental properties and event types
would remain. Indeed, token physicalism,
together with a fairly uncontroversial
assumption. naturally leads to the question
of whether events can be causes only in
virtue of falling under types postulated by
physics. The assumption is that physics
postulates a system of event types that has
the following features.
Physical Causal Comprehensiveness
Whenever two physical events are
causally related, they are so related in
virtue of falling under physical types.
This, together with token physicalism,
implies:
Universal Physical Causal Comprehensive-
ness Whenever two events are causally
related, they are so related in virtue of
falling under physical types.
That thesis naturally invites the question of
whether the following is true:
The Type-Exclusion Thesis An event can
cause something only in virtue of
falling under a physical type (i.e. a type
postulated by physics).
The type-exclusion thesis offers one
would-be answer to our earlier question of
which event types are such that events
have effects in virtue of falling under them.
If the answer is the correct one, however,
then the fact (if it is one) that special
science events and mental events are within
the domain of physics will be cold comfort.
For type physicalism, the thesis that every
event type is a physical type, seems false.
Mental types seem not to be physical types
283
EPIPHENOMENALISM
in our strict and narrow sense. No mental
type, it seems, is necessarily coextensive (Le.
coextensive in every 'POSSIBLE WORLD ')
with any type postulated by physics. Given
that, and given the type-exclusion thesis,
type epiphenomenalism is true. However,
typical special science types also fail to be
necessarily coextensive with any physical
types, and thus typical special science types
fail to be physical types. Indeed, we individ-
uate the sciences in part by the event (state)
types they postulate. Given that typical
special science types are not physical types
(in our strict sense), if the type-exclusion
thesis is true, then typical special science
types are not such that events can have
causal effects in virtue of falling under
them.
SOME APPARENT CONSEQUENCES OF THE
TYPE EXCLUSION THESIS
We see, then, that, given that type physical-
ism is false for the reasons discussed above,
the type-exclusion thesis has truly incredible
consequences, in addition to implying type
epiphenomenalism. Since a neuron firing is
not a type of event postulated by physics,
given the type exclusion thesis, no event
could ever have any causal effects in virtue
of being a firing of a neuron! The neuro-
physiological qua neurophysiological is cau-
sally impotent! Moreover, if things have
causal powers only in virtue of their physi-
cal properties, then an HIV virus, qua HIV
virus, does not have the causal power to
contribute to depressing the immune
system; for being an HIV virus is not a phy-
sical property (in our strict sense). Similarly,
for the same reason, the SALK vaccine, qua
SALK vaccine, would not have the causal
power to contribute to producing an im-
munity to polio. Furthermore, if, as it seems,
phenotypic properties are not physical prop-
erties, phenotypic properties do not endow
organisms with causal powers conducive to
survival! Having hands, for instance, could
never endow anything with causal powers
conducive to survival since it could never
endow anything with any causal powers
whatsoever. But how, then, could pheno-
284
typic properties be units of natural selec-
tion? And if, as it seems, genotypes are not
physical types, then, given the type exclu-
sion thesis, genes do not have the causal
power, qua genotypes, to transmit the
genetic bases for phenotypes! How, then,
could the role of genotypes as units of her-
edity be a causal role? There seem to be
ample grounds for scepticism that any
reason for holding the type-exclusion thesis
could outweigh our reasons for rejecting it.
We noted that the thesis of universal
physical causal comprehensiveness (here-
after, 'upc-comprehensiveness' for short)
invites the question of whether the type-
exclusion thesis is true. But does upc-
comprehensiveness imply the type-exclusion
thesis? Could one rationally accept upc-
comprehensiveness while rejecting the type-
exclusion thesis?
'IN VIRTUE OF' DOES NOT IMPLY 'ONLY
IN VIR TUE OF'
Notice that there is a crucial one-word dif-
ference between the two theses: the exclu-
sion thesis contains the word 'only' in front
of 'in virtue of', while thesis of upc-compre-
hensiveness does not. This difference is rele-
vant because 'in virtue of' does not imply
'only in virtue of'. I am a brother in virtue
of being a male with a sister, but I am also
a brother in virtue of being a male with a
brother; and, of course, one can be a male
with a sister without being a male with a
brother, and conversely. Likewise, I live in
the state of New Jersey in virtue of living in
the city of New Brunswick, but it is also
true that I live in New Jersey in virtue of
living in the County of Middlesex; and one
can live in the County of Middlesex without
living in New Brunswick. Moreover, in the
general case, if something, x, bears a rela-
tion, R, to something y in virtue of x's being
F and y's being G, it does not follow that x
bears R to y only in virtue of x's being F and
y's being G. Suppose that x weighs less than
y in virtue of x's weighing 10lbs and y's
weighing 15 lbs. Then, it is also true that x
weighs less than y in virtue of x's weighing
under 12 lbs and y's weighing over 12 lbs.
And something can, of course, weigh under
12 lbs without weighing 10 lbs. To repeat:
'in virtue of' does not imply 'only in virtue
of'.
Why, then, think that upc-comprehen-
siveness implies the type-exclusion thesis?
The fact that two events are causally related
in virtue of falling under physical types does
not seem to exclude the possibility that they
are also causally related in virtue of falling
under non-physical types, in virtue of the
one being (say) a firing of a certain neuron
and the other being a firing of a certain
other neuron, or in virtue of one being a
secretion of enzymes and the other being a
breakdown of amino acids. Notice that the
thesis of upc-comprehensiveness implies
that whenever an event is an effect of
another, it is so in virtue of falling under a
physical type. But the thesis does not seem
to imply that whenever an event is an effect
of another. it is so only in virtue of falling
under a physical type. Upc-comprehensive-
ness seems consistent with events being
effects in virtue of falling under non-physical
types. Similarly. the thesis seems consistent
with events being causes in virtue of falling
under non-physical types.
Nevertheless. an explanation is called for
of how events could be causes in virtue of
falling under non-physical types if upc-com-
prehensiveness is true. The most common
strategy for offering such an explanation
involves maintaining there is a dependence-
determination relationship between non-
physical types and physical types. Upc-com-
prehensiveness. together with the claim that
instances of non-physical event types are
causes or effects. implies that. as a matter of
causal necessity. whenever an event falls
under a non-physical event type, it falls
under some physical type or other. The
instantiation of non-physical types by an
event thus depends. as a matter of causal
necessity. on the instantiation of some or
other physical event type by the event. It is
held that non-physical types are. moreover.
'realized by' physical types in physical con-
texts; although. a given non-physical type
might be 'realizable' by more than one
physical type. The occurrence of a physical
EPIPHENOMENALISM
type in a physical context in some sense
determines the occurrence of any non-
physical type that it 'realizes.'
Recall the considerations that inspired the
'no gap' argument from physics: quantum
mechanics seems able, in principle. to
explain how chemical processes unfold in
terms of the mechanics of subatomic par-
ticles; molecular biology seems able, in prin-
ciple. to explain how the physiological
operations of systems in living things occur
in terms of biochemical pathways. long
chains of chemical reactions. Types of sub-
atomic causal processes 'implement' types of
chemical processes; and certain types of
chemical processes implement types of
physiological processes. Many in the cogni-
tive science community hold that computa-
tional processes implement mental
processes; and that computational processes
are implemented. in turn. by neurophysiolo-
gical processes (see COGNITIVE PSYCHOL-
OGY; COMPUTATIONAL MODELS OF THE
MIND). On this view, the sciences form a
hierarchy with physics at the base. and
which includes in ascending order chem-
istry. biology. and psychology. Types of
'lower-level' causal processes implement
types of 'higher-level' processes. The types
of lower-level causal processes that imple-
ment a higher-level causal transaction
between A-type events and B-type events
will contain a realization of A as their first
stage and a realization of B as their final
stage. Since A and B may be 'multiply rea-
lizable. that is. realizable by more than one
lower-level type. there may be many distinct
lower-level processes that could implement
a given higher-level process. Every type of
higher-level causal transaction is. however.
implemented by at least one type of physical
causal process.
It should be stressed. however. that there
is no received view about the exact nature
of the realization relation. and thus, no
received view about the exact nature of the
implementation relation. (But for discus-
sions of such matters, see Putnam. 197 5a;
Hooker. 1981. pp. 496-506; McLaughlin,
1983; Jackson and Pettit, 1988; LePore and
Loewer, 1989; Block. 1990; Yablo, 1992a,
285
EPIPHENOMEN ALISM
1992b; Van Gulick, 1993.) Let us set aside
the general issues of the nature of the rea-
lization relationship, and of how, in part, by
virtue of being realized by physical types in
physical contexts, non-physical types can be
such that events have effects in virtue of
falling under them. Let us assume, however,
that a causally necessary condition for a non-
physical type to be such that events can be
causes in virtue of falling under it is that
the non-physical type be realizable by phys-
ical types in physical contexts. Turn, then,
to the case of mental types, in particular.
THE CAUSAL RELEVANCE OF CONTENT
AND CONSCIOUSNESS
Some terminology will prove useful: let us
stipulate that an event type, T, is a causal
type if and only if there is at least one type,
T*, such that something can cause a T* in
virtue of being a T. And by saying that an
event type is realizable by physical event
types or physical properties, let us mean
that it is at least causally possible for the
event type to be realized by a physical
event type. Given that non-physical causal
types must be realizable by physical types,
and given that mental types are non-
physical types, there are two ways that
mental types might fail to be causal. First,
mental types may fail to be realizable by
physical types. Second, mental types might
be realizable by physical types but fail to
meet some further condition for being
causal types. Reasons of both sorts can be
found in the literature on mental causation
for denying that any mental types are
causal. However, there has been much
attention paid to reasons of the first sort in
the case of phenomenal mental types (pain
states, visual states, etc.). And there has
been much attention to reasons of the
second sort in the case of intentional
mental types (Le. beliefs that P, desires that
Q, intentions that R, etc.). In what remains,
we will very briefly consider these reasons
in reverse order.
But first one preliminary: notice that
intentional states figure in explanations of
intentional actions not only in virtue of
286
their intentional mode (whether they are
beliefs, or desires, etc.), but also in virtue of
their contents (Le. what is believed, or
desired, etc.). For example, what causally
explains someone's doing A (standing on
his head) is that the person wants to X
(impress someone) and believes that by
doing A he will X. The contents of the belief
and desire (what is believed and what is
desired) seem essential to the causal expla-
nation of the agent's doing A. Similarly, we
often causally explain why someone came
to believe that P by citing the fact that the
individual came to believe that Q and in-
ferred P from Q. In such cases, the contents
of the states in question are essential to the
explanation. This is not, of course, to say
that contents themselves are causally effica-
cious; contents are not among the relata of
causal relations. The point is, rather, that
we characterize states when giving such
explanations not only as being in certain
intentional modes, but also as having
certain contents; we type states for the pur-
poses of such explanations in terms of their
intentional modes and their contents. We
might call intentional state types that
include content properties 'contentful inten-
tional state types'; but, to avoid prolixity, let
us call them 'intentional state types' for
short; thus, for present purposes, by 'inten-
tional state types' we will mean types such
as the belief that P, the desire that Q, and so
on, and not types such as belief, desire, and
the like (see INTENTIONALITY).
Although it was no part of Putnam's or
Burge's purpose to raise concerns about
whether intentional state types are causal,
their well-known 'TWIN EARTH' thought-
experiments have prompted such concerns
(Putnam, 1975b; Burge, 1979). These
thought-experiments are fairly widely held
to show that two individuals who are
exactly alike in every intrinsic physical
respect can have intentional states with dif-
ferent contents. If they show that, then
intentional state types fail to supervene on
intrinsic physical state types. The reason is
that what contents an individual's beliefs,
desires, etc. have, depends, in part, on
extrinsic, contextual factors. Given that, the
concern has been raised that states cannot
have effects in virtue of falling under inten-
tional state types.
One concern seems to be that states
cannot have effects in virtue of falling under
intentional state types because individuals
who are in all and only the same intrinsic
states must have all and only the same
causal powers. In response to that concern,
it might be pointed out that causal powers
often depend on context. Consider weight.
The weights of objects do not supervene on
their intrinsic properties: two objects can be
exactly alike in every intrinsic respect (and
thus have the same mass) yet have different
weights. Weight depends, in part, on ex-
trinsic, contextual factors. Nonetheless, it
seems true that an object can make a scale
read 10lbs in virtue of weighing 10 lbs.
Thus, objects which are in exactly the same
types of intrinsic states may have different
causal powers due to differences in their cir-
cumstances.
It should be noted, however, that on
some leading EXTERNALIST theories of
content, content, unlike weight, depends on
a historical context (see CONTENT. Call such
theories of content 'historical-externalist
theories'. On one leading historical-extern-
alist theory, the content of a state depends
on the learning history of the individual
(DRETSKE, 1988, 1993); on another, it
depends on the selection history of the
species of which the individual is a member
(Millikan, 1993). Historical-externalist the-
ories prompt a concern that states cannot
have causal effects in virtue of falling under
intentional state types. Causal state types, it
might be claimed, are never such that their
tokens must have a certain causal ancestry.
But, if so, then, if the right account of
content is a historical-externalist account,
then intentional types are not causal types.
Some historical-externalists appear to
concede this line of argument, and thus to
concede that states do not have causal
effects in virtue of falling under intentional
state types (see Millikan, 1993). However,
other historical-externalists attempt to
explain how intentional types can be causal,
even though their tokens must have appro-
EPIPHENOMENALISM
priate causal ancestries (see Dretske, 1993).
This issue is hotly debated, and remains
unresolved.
I shall conclude by noting why it is con-
troversial whether phenomenal state types
can be realized by physical state types. Phe-
nomenal state types are such that it is like
something for a subject to be in them; it is,
for instance, like something to have a throb-
bing pain (see CONSCIOUSNESS; QUALIA). It
has been argued that phenomenal state
types are, for that reason, subjective (see
SUBJECTIVITY): to fully understand what it
is to be in them, one must be able to take up
a certain experiential point of view
(Jackson, 1982; Nagel, 1986). For, it is
claimed, an essential aspect of what it is to
be in a phenomenal state is what it is like to
be in the state; only by taking up a certain
experiential point of view can one under-
stand that aspect of a phenomenal state.
Physical state types (in our strict and
narrow sense) are paradigm objective states,
i.e. non-subjective states. The issue arises,
then, as to whether phenomenal state types
can be realized by physical state types. How
could an objective state realize a subjective
one? This issue too is hotly debated, and
remains unresolved. Suffice it to note that if
only physical types and types realizable by
physical types are causal, and if phenom-
enal types are neither, then nothing can
have any causal effects in virtue of falling
under a phenomenal type. Thus, it could
never be the case, for example, that a state
causally results in a bad mood in virtue of
being a throbbing pain.
See also An Essay on Mind section 3.6; IDEN-
TITY THEORY; PHYSICALISM; SUPER-
VENIENCE.
REFERENCES
Alexander, S. 1920. Space, Time, and Deity. 2
vols. London: Macmillan.
Antony, 1. 1989. Anomalous monism and the
problem of explanatory force. Philosophical
Review, 2, 153-87.
Bennett, J. 1988. Events and Their Names.
Indianapolis: Hackett.
Block, N. 1990. Can the mind change the
287
EXPLANANS/EXPLANANDUM
world? In Meaning and Method: Essays in
Honor of Hilary Putnam. ed. G. Boolos. Cam-
Broad, C.D. 1925. The Mind and Its Place in
Nature. London: Routledge & Kegan Paul.
Midwest Studies in Philosophy, Vol. 4, 73-
121.
Cambell, K. 1970. Body and Mind. New York:
Anchor Books.
Davidson, D. 1970. Mental events. In Experi-
ence and Theory. ed. L. Foster and J. W.
Swanson. The University of Massachusetts
Press and Duckworth.
Davidson, D. 1993. Thinking causes. In Heil
and Mele 1993.
Dretske, F. 1988. Explaining Behavior: Reasons
in a World of Causes. Cambridge MA: MIT
Press/Bradford Books.
Dretske, F. 1993. Mental events as structuring
causes of behavior. In Heil and Mele 1993
Fodor, J. 1975. The Language of Thought. New
York: Thomas Y. Crowell.
Heil, J. and Mele, A. eds. 1993. Mental Causa-
tion. Oxford University Press.
Honderich, T. 1983. The argument for anom-
alous monism. Analysis, 42, 59-64.
Hooker, C.A. 1981. Toward a general theory of
reduction. Part III. Dialogue, 20, 496-529.
Huxley, 1874. Of the hypothesis that animals
are automata. Reprinted in Science and
Culture, 1882.
Jackson, F. 1982. Epiphenomenal qualia.
Philosophical Quarterly, 32, 127-36.
Jackson, F., and Pettit, P. 1988. Broad con-
tents and functionalism. Mind, 47, 381-
400.
Kim, J. 1984. Epiphenomenal and super-
venient causation. Midwest Studies in Philo-
sophy, Vol. 9, 257-70.
Kim, J. 1993. Can supervenience and 'non-
strict laws' save anomalous monism? In Heil
and Mele 1993.
LePore, E., and Loewer, B. 1987. Mind
Matters. Journal of Philosophy, 84, 630-42.
LePore, E., and Loewer, B. 1989. More on
making mind matter. Philosophical Topics,
17,175-91.
Lewes, G.H. 1875. Problems of Life and Mind,
vol. 2. London: Kegan Paul. Trench,
Turbner, & Co.
Malcolm, N. 1968. The conceivability of mech-
anism. Philosophical Review, 77,45-72.
288
McLaughlin, B.P. 1983. Event supervenience
and supervenient causation. Southern
Journal of Philosophy, Supplementary Volume
on Supervenience, 22, 71-92.
McLaughlin, B.P. 1989. Type dualism, type
epiphenomenalism, and the causal priority
of the physical. Philosophical Perspectives, 3,
109-35.
McLaughlin, B.P. 1993. On Davidson's
response to the charge of epiphenomenal-
ism. In Heil and Mele 1993, 27-40.
Millikan, R., 1993. Explanation in biopsychol-
ogy. In Heil and Mele 1993, 211-32.
Nagel. T. 1986. The View From Nowhere.
Oxford: Clarendon.
Putnam, H. 1975a. Philosophy and our
mental life. In Philosophical Papers, vol. 2,
ed. H. Putnam. Cambridge University Press.
Putnam, H. 197 5b. The meaning of
'meaning'. In Language, Mind, and Know-
ledge, ed. K. Gunderson. Minnesota Studies
in Philosophy of Science, 7. Minneapolis:
University of Minnesota Press, pp. l31-93.
Sosa, E. 1984. Mind-body interaction and
supervenient causation. Midwest Studies in
Sosa, E. 1993. Davidson's thinking causes. In
Heil and Mele 1993,
Van Gulick, R. 1993, Who's in charge here?
and who's doing all the work? In Heil and
Mele 1993.
Yablo, S. 1992a. Mental causation. Philosoph-
ical Review, 101. 245-80.
Yablo, S. 1992b. Cause and essence. Synthese,
93,403-449.
BRIAN P. McLAUGHLIN
explanans/explanandum This pair of
terms has had a wide currency in philo-
sophical discussions because it allows a
certain succinctness which is unobtainable
in ordinary English. Whether in science,
philosophy or in everyday life, one often
offers explanations. The particular state-
ments, laws, theories or facts that are used
to explain something are collectively called
the explanans: and the target of the expla-
nans - the thing to be explained - is called
the explanandum. Thus, one might explain
why ice forms on the surface of lakes (the
explanandum) in terms of the special prop-
erty of water to expand as it approaches
freezing point together with the fact that
materials less dense than liquid water float
in it (the explanans). The terms come from
two different Latin grammatical forms:
'explanans' is the present participle of the
verb which means explain; and 'explan-
andum' is a direct object noun derived from
that same verb.
SAMUEL GUTTENPLAN
extension see INTENSION.
externalism/internalism In giving an
account of what someone believes, does
essential reference have to be made to how
things are in the environment of the
believer? And, if so, exactly what relation
does the environment have to the belief?
Answering these questions involves taking
sides in the externalism/internalism debate.
To a first approximation, the externalist
holds that one's propositional attitudes
cannot be characterized without reference
to the disposition of objects and properties
in the world - the environment - in which
one is situated. The internalist thinks that
propositional attitudes (especially belief)
must be characterizable without such refer-
ence. (The motivation for internalism will
be discussed below.) The reason that this is
only a first approximation of the contrast is
that there can be different sorts of external-
ism. Thus, one sort of externalist might
insist that you could not have, say, a belief
that grass is green unless it could be shown
that there was some relation between you,
the believer, and grass. Had you never come
across the plant which makes up lawns and
meadows, beliefs about grass would not be
available to you. This does not mean that
you have to be in the presence of grass in
order to entertain a belief about it, nor does
it even mean that there was necessarily a
time when you were in its presence. For
example, it might have been the case that,
though you have never seen grass, it has
been described to you. Or, at the extreme,
perhaps grass no longer exists anywhere in
EXTERN ALISMjINTERN ALISM
the environment, but your ancestors'
contact with it left some sort of genetic
trace in you, and that trace is sufficient to
give rise to a mental state that could be
characterized as about grass.
Clearly, these forms of externalism entail
only the weakest kind of commitment to the
existence of things in the environment.
However, some externalists hold that pro-
positional attitudes require something
stronger. Thus, it might be said that in
order to believe that grass is green, you
must have had some direct experience -
some causal contact - with it during your
lifetime. Or an even stronger version might
hold that there are beliefs that require that
you be in direct contact with the subject
matter of these beliefs in order to so much
as have them. Obviously, such a strong
form of externalism is implausible in con-
nection with a general belief about grass,
for example, that it is green. But when it
comes to what are called singular beliefs,
matters are not so clear. For example, on
seeing something bird-like outside the
window of my study, I may say: 'that bird
was a greenfinch', thereby expressing what
I believe. Suppose, however, that I never did
see a bird on that occasion - it was only a
movement of a leaf which I had mistaken
for one. In this case, one sort of externalist
would insist that, since nothing in my
environment answers to the expression
'that bird' that I used, then I simply do not
have the belief that that bird was a green-
finch. And this is true even if I myself am
convinced that I have the belief. On this
strong externalist stance, propositional atti-
tudes become opaque to their possessors.
We can think we believe and desire various
things - that our attitudes have certain
contents - though we might well just be
wrong.
In contrast, the internalist would insist
that the contents of our attitudes can be
described in ways that do not require the
existence of any particular objects or prop-
erties in the environment, and this is so
even in the case of singular beliefs. There
are several motivating factors here. First,
there is the intuition that we do know the
289
EXTERN ALISMjINTERN ALISM
contents of our own minds. I may be wrong
about there being a bird, but how can I be
wrong about my believing that there is one?
One way the internalist might try to embar-
rass the externalist into agreeing here is to
ask how to explain our intuition that we
have some sort of first-person authority
with respect to the contents of our
thoughts. For, on the strong form of extern-
alism, what we actually think is dependent
on the environment, and this is something
that is as accessible to others as it is to
oneself. The second motivation comes from
the demands of action explanation. Suppose
that I reach for my binoculars just after
insisting that I saw the bird in the tree. The
obvious explanation for my action would
seem to mention, among other things, my
belief that there is such a bird. However,
since if the externalist is right, then I just do
not have any such belief, it is unclear how
to explain my reaching for the binoculars.
Finally, internalism can seem the obvious
way to deal with the otherwise puzzling
consequences of versions of the TWIN
EARTH thought experiment. Briefly, suppose
this time that I really do see the bird, but
suppose that my twin - someone who is a
molecular duplicate of me on a duplicate
planet called 'twin earth' - does not. (We
can stipulate that the only difference
between earth and twin earth at that very
time is that there really is a bird in the tree
on earth, but there is none on twin earth.)
As would generally be agreed, my twin
would say 'that bird is a greenfinch', whilst
pointing in the direction of the tree. After
all, being a molecular duplicate of me, one
would expect his behaviour to resemble
mine as closely as can be imagined. More-
over, it is difficult to deny that his saying
this is good evidence that he actually
believes it. The fact that my twin and I are
molecule-for-molecule the same is often
reckoned to imply that my twin and I are
290
psychological. as well as physical. dupli-
cates. Yet the strong externalist position
would be committed to saying that my twin
has no such belief, whilst I do, and this
because of the way things are in our re-
spective environments. Yet, if I were sud-
denly to be in my twin's shoes - if I were
instantaneously transported to twin earth
without any knowledge of the move - there
could be no doubt that I would say 'that
bird is a greenfinch'. And what reason
could be given for saying that my mental
state had changed during the transporta-
tion? Why, if my saying something counted
as evidence for my belief in the one case,
doesn't it count in the other?
Given these factors, the internalist is apt
to insist that beliefs and other attitudes
must be characterizable 'from the inside', so
to speak. What I share with my twin is a
content, though we obviously do not share
an environment that answers to that
content in the same way. In not being
answerable to how things are in the envir-
onment, it has been suggested that what I
and my twin share is a narrow content. The
broad or wide content does take the environ-
ment into consideration, and it is therefore
true that my twin and I do not share broad
content in the case imagined. However,
what the internalist insists is that only the
notion of narrow content is up to the task of
explaining the intuitions we have about
twin earth cases, explanation of action and
first-person authority. To be sure, there are
rejoinders available to the externalist in
respect of each of these intuitions, and only
the beginnings of the debate have been
sketched here. (See BELIEF; CONTENT; An
Essay on Mind section 2.1.3; PROPOSI-
TIONAL ATTITUDES; THOUGHT; THOUGHT
AND LANGUAGE.)
SAMUEL GUTTENPLAN
F
first-person authority Descartes (1596-
1650) insisted that the mind was as a
special kind of substance, one which con-
trasts sharply with material substance (see
HISTORY). Hence, the label 'Cartesian' tends
to be applied to any view that is D U A LI S T in
thinking of the mind as fundamentally dif-
ferent from matter. Accompanying this Car-
tesian dualism of substances is a dualism of
ways of knowing about minds and about
matter. The Cartesian conception has it that
we have access to the contents of our own
minds in a way denied us in respect of
matter. That is, we can know what we
think, feel and want, and know this with a
special kind of certainty that contrasts with
our knowledge of the physical world. Indeed,
Descartes thought that we could be mis-
taken about even the existence of our own
bodies, whilst we could not be in error about
what passes in our minds.
The doctrine that there is something
special about our knowledge of our own
minds goes naturally with Cartesian
dualism, but it is by no means necessary to
be a dualist to think this. Nor need one be
committed to anything as strong as the Car-
tesian view that self-knowledge is certain.
For whatever one's conception of mind and
matter, there does seem to be an asymmetry
between our knowledge of our own mind
and our knowledge of the external world
(including here our knowledge of other per-
son's minds). One way to put it, without
any specific commitment to dualism, is to
say that a subject (or self) has first-person
authority with respect to the contents of his
or her mind, whereas others (third persons)
can only get at these contents indirectly.
You count as an authority about your own
mind because you can know about it
directly or immediately and this is very dif-
ferent from the type of knowledge others
can have.
The notion of 'authority' is vague, but it
can be sharpened in a number of different
ways. First, one may say that first-person
authority consists in the infallibility of your
judgments about your own mental states; if
you claim sincerely to be in pain, then you
cannot be wrong. This is certainly the view
held by Descartes himself. A little less rad-
ically, one might think that such authority
only confers incorrigibility: your judgments
about your pains might be wrong, but you
cannot be corrected by others, you count as
the highest authority. Finally, one might
think that what counted most for first-
person authority was that your mental
states were transparently available to you: if
. you are in pain, then you know it. This is
known as self-intimation. And it should be
noted that this does not by itself require
that we are either infallible or incorrigible
about our pains or other mental contents.
The Cartesian conception of first-person
authority encourages the idea that the mind
is like a theatre whose show can only be
witnessed by the subject whose theatre it is.
This view of the mental as something like a
private performance is a highly metaphor-
ical way of motivating the idea of first-
person authority, though it is a metaphor
that comes naturally to us. Nonetheless,
since it is possible to accept some form of
first-person authority without any commit-
ment to dualism, much effort has been
expended in trying to explain the asymmetry
between first- and third-person knowledge of
the mind in a way that does not depend on
the so-called 'Cartesian theatre'. (See CON-
scIOusNEss; An Essay on Mind section
2.2.3; INTROSPECTION, THE SELF).
SAMUEL GUTTENPLAN
291
FODOR, JERRY A.
Fodor, Jerry A, 'Hey, what do you think
you're doing?' Fair question. When I was a
boy in graduate school. the philosophy of
mind had two main divisions: the mind/
body problem and the problem of other
minds, The project was to solve these
problems by doing conceptual analysis.
Nobody knew what conceptual analysis
was, exactly; but it seemed clear that lots of
what had once been considered the philo-
sophy of mind (lots of Hume and Kant, for
example) didn't qualify and was not, in fact,
philosophy at all. Instead, it was a mis-
guided sort of armchair psychology, best
read by flashlight beneath the covers.
Philosophical fashions change. It's gotten
harder to believe that there is a special
problem about the knowledge of other
minds (as opposed to other anything elses);
and these days we're all materialists for
much the reason that Churchill gave for
being a democrat: the alternatives seem
even worse. Correspondingly, there's a new
research agenda: to reconcile our mater-
ialism with the psychological facts; to
explain how minds qua material objects
could have the properties they do.
It is, however, reasonable to wonder
whether this is a research agenda in philo-
sophy. If, after all, you wanted to know how
rocks, trees, and spiral nebulas, qua mater-
ial objects, could have the properties that
they do, you would be well advised to
consult not a philosopher but a geologist, a
botanist, and an astrophysicist respectively.
To explain how photosynthesis works is to
explain how a plant qua material object
could photosynthesize; so reconciling ma-
terialism with the facts about plants isn't a
project distinct from the one that botanists
are already engaged on. Parity of argument
suggests that reconciling the facts about
minds with materialism is not a project dis-
tinct from the one that psychologists are
already engaged on, hence that psychology
needs a philosophy of minds about as much
as botany needs a philosophy of plants. This
is a line of thought about which few philo-
sophers of mind are enthusiastic.
On the other hand, some of the most per-
vasive properties of minds seem so myster-
292
ious as to raise the Kantian-sounding
question how a materialistic psychology is
even possible. Lots of mental states are con-
scious, lots of mental states are intentional,
and lots of mental processes are rational,
and the question does rather suggest itself
how anything that is material could be any
of these. This question is without a botan-
ical counterpart since, though there is
plenty that's puzzling about plants, their
materiality seems not to be much in doubt.
'How is a materialistic psychology pos-
sible?' is, moreover, a question that may
belong to philosophers by default; it's not
one that psychologists have had a lot to say
about. While behaviourism was in style,
psychologists assumed, in effect, that psy-
chology isn't possible, Now their research
strategy is largely to assume, without much
argument, that the conceptual apparatus of
commonsense folk-psychological explana-
tion can be adapted to scientific require-
ments. De facto, the status of this
assumption has become the main topic in
the philosophy of mind.
My philosophical concerns are mostly
with this question, so what follows is mostly
a discussion of its current status. For these
purposes, 'POLK PSYCHOLOGY' is primarily
intentional explanation; it's the idea that
people's behaviour can be explained by
reference to the contents of their beliefs and
desires. Correspondingly, the methodo-
logical issue is whether intentional explana-
tions can be co-opted to make science out
of. Similar questions might be asked about
the scientific potential of other folk-psycho-
logical concepts (CONSCIOUSNESS, for
example), but I won't discuss them here for
want of space and lack of progress to report.
What makes intentional explanations
problematic is that they presuppose that
there are intentional states, What makes
intentional states problematic is that they
exhibit a pair of properties which (with one
exception; see below) are never found to-
gether elsewhere (see INTENTION ALITY).
(1) Intentional states have causal powers.
Thoughts (more precisely, havings of
thoughts) make things happen; typically,
thoughts make behaviour happen. Self-
pity can make one weep, as can onions.
(2) Intentional states are semantically evalu-
able. BELIEFS, for example, are about
how things are and are therefore true or
false depending on whether things are
the way that they're believed to be.
Consider, by contrast: tables, chairs,
onions, and the cat's being on the mat.
Though they all have causal powers,
they are not about anything and are
therefore not evaluable as true or false.
If there is to be an intentional science, there
must be semantically evaluable things that
have causal powers. Moreover, there must
be laws about such things, including, in
particular, laws that relate beliefs and
desires to one another and to actions. If
there are no intentional laws, then there is
no intentional science. This claim is tenden-
tious, but I think we had better assume that
it's true. Perhaps scientific explanation isn't
always explanation by law subsumption, but
surely it often is, and there's no obvious
reason why an intentional science should
be exceptional in this respect. Moreover, one
of the best reasons for supposing that
common sense is right about there being
intentional states is precisely that there
seem to be many reliable intentional gen-
eralizations for such states to fall under. I
assume that many of the truisms of folk
psychology either articulate intentional
laws or come pretty close to doing so.
So, for example, it's a truism of folk psy-
chology that rote repetition facilitates recall.
(More generally, repetition improves
performance; 'How do you get to Carnegie
Hall?' 'Practice, practice. ') This general-
ization relates the content of what you learn
to the content of what you say to yourself
while you're learning it; so what it expres-
ses is, prima facie, a lawful causal relation
between types of intentional states. Real
psychology has lots more to say on this
topic, but it is, from the present point of
view, mostly much more of the same. To a
first approximation, repetition does causally
facilitate recall, and that it does is lawful.
There are, to put it mildly, many other
FODOR, JERRY A.
cases of such reliable intentional causal
generalizations. There are also many, many
kinds of folk psychological generalizations
about correlations among intentional states,
and these too are plausible candidates for
fleshing out as intentional laws. For
example, that anyone who knows what
7 + 5 is also knows what 7 + 6 is; that
anyone who knows what 'John loves Mary'
means also knows what 'Mary loves John'
means; and so forth.
Philosophical opinion about folk psycho-
logical intentional generalizations runs the
gamut from 'there aren't any that are really
reliable' to 'they are all platitudinously true,
hence not empirical at all'. I won't argue
about this here. Suffice it, on the one hand,
that the necessity of 'if 7 + 5 = 12 then
7 + 6 = 13' is quite compatible with the
contingency of 'if someone knows that 7 + 5
= 12, then he knows that 7 + 6 = 13';
and, on the other hand, that I'd be rich if I
could only find people prepared to bet
against the reliability of such folk psycholo-
gical truths.
So, then, part of the question 'how can
there be an intentional science' is 'how can
there be intentional laws?' I'll return to this
presently. First, a bit about laws per se.
LAWS PER SE
I assume that laws are true generalizations
that support counterfactuals and are con-
firmed by their instances. I suppose that
every law is either basic or not. Basic laws
are either exceptionless or intractably
statistical. The only basic laws are laws of
basic physics.
All non-basic laws, including the laws of
all the non-basic sciences, including, in par-
ticular, the intentional laws of psychology,
are 'c[eteris] p[aribus] laws;' they hold only
'all else being equal'. There is - anyhow
there ought to be - a whole department of
the philosophy of science devoted to the
construal of cp laws; to making clear, for
instance, how they can be explanatory,
how they can support counterfactuals, how
they can subsume the singularly causal
truths that instance them ... , etc. I omit
293
FODOR. JERRY A.
these issues in what follows because they
don't belong to philosophical psychology as
such. If the laws of intentional psychology
are cp laws, that is because intentional psy-
chology is a special (Le. non-basic) science,
not because it is an intentional science.
IMPLEMENT A TION
There is a further quite general property
that distinguishes cp laws from basic ones:
non-basic laws want mechanisms for their
implementation. Suppose, for a working
example, that some special science says that
being F causes xs to be G. (Being irradiated
by sunlight causes plants to photo-
synthesize; being freely suspended near the
earth's surface causes bodies to fall with
uniform acceleration; and so forth.) Then it
is a constraint on this generalization's being
lawful that How does being F cause xs to be
G? must have an answer. This is, I suppose
one of the ways special science laws are dif-
ferent from basic laws. A basic law says that
Fs cause (or are) Gs, punkt. If there were
any explaining how, or why, or by what
means Fs cause Gs, the law would ipso facto
be not basic but derived.
Typically - though not invariably - the
mechanism that implements a special
science law is defined over the micro-
structure of the things that satisfy the law.
The answer to 'how does sunlight make
plants photosynthesize?' implicates the
chemical structure of plants; the answer to
'how does freezing make water solid?' im-
plicates the molecular structure of water;
and so forth. In consequence, theories about
how a law is implemented usually draw
upon the vocabularies of two (or more)
levels of explanation.
If you are specially interested in the pecu-
liarities of aggregates of matter at the Lth
level (in plants, or minds, or mountains, as
it might be) then you are likely to be spe-
cially interested in implementing mechan-
isms at the L-lth level (the 'immediately'
implementing mechanisms); this is because
the characteristics of L-level laws can often
be explained by the characteristics of their
L-lth level implementations. You can
294
learn quite a lot about plants qua plants by
studying their chemical constitution. You
learn correspondingly less by studying their
subatomic constitution though, no doubt,
laws about plants are implemented, even-
tually, subatomic ally. The question thus
arises what mechanisms might immediately
implement the intentional laws of psychol-
ogy, thereby accounting for their character-
istic features. (We are, appearances to the
contrary notwithstanding, rapidly arriving
at a question that may actually have an
answer.)
Intentional laws subsume causal interac-
tions among mental processes; that much is
truistic. But here's something substantive,
something that a theory of the implementa-
tion of intentional laws ought to account
for: The causal processes that intentional
states enter into tend to preserve their
semantic properties. For example, thinking
true thoughts tends to cause one to think
more thoughts that are also true. This is no
small matter; the very rationality of thought
depends on such facts as that true thoughts
that ((P ..... Q) and (P)) tend to cause true
thoughts that Q.
A lot of what has happened in psychol-
ogy - notably since Freud - has consisted of
finding new and surprising cases where
mental processes are semantically coherent
under intentional characterization. Freud
made his reputation by showing that this
was true even fol' much of the detritus of
behaviour: dreams, verbal slips and the like.
But, as it turns out, the psychology of
normal mental processes is largely grist for
the same mill. For example, it turns out to
be theoretically revealing to construe per-
ceptual processes as inferences that take
specifications of proximal stimulations as
premises and yield specifications of their
distal causes as conclusions, and that are
reliably truth preserving in ecologically
normal circumstances. The psychology of
learning cries out for analogous treatment
(e.g. for treatment as a process of hypothesis
formation and confirmation).
Here's how things have gone so far in
this discussion. We started with the ques-
tion: 'How is intentional psychology
possible?' and this led us to 'how can there
be intentional laws?' which in turn raised
'how could intentional laws have imple-
menting mechanisms?' A mark of a good
answer to this last question is that it should
explain why mental processes are generally
coherent under semantic representation
(roughly, why they tend to preserve seman-
tic properties like truth). This is where cog-
nitive science starts.
THOUGHTS AS SYMBOLS
I remarked that intentional states, as
common sense understands them, have
both causal and semantic properties and
that the combination appears to be un-
precedented: propositions are semantically
evaluable, but they are abstract objects and
have no causal powers. Onions are concrete
particulars and have causal powers, but
they aren't semantically evaluable. Inten-
tional states seem to be unique in combin-
ing the two; that indeed, is what so many
philosophers have against them.
Well, almost unique. Suppose I write 'the
cat is on the mat'. On the one hand, the
thing I've written is a concrete particular in
good standing and it has, qua material
object, an open-ended galaxy of causal
powers. (It reflects light in ways that are
essential to its legibility; it exerts a small but
in principle detectable gravitational effect
upon the moon; etc.) And, on the other
hand, what I've written is about something
and is therefore semantically evaluable; it's
true if and only if there's a cat where it says
that there is. So, my inscription of 'the cat is
on the mat' has both content and causal
powers; as does my thought that the cat is
so mat. (I'm playing fast and loose with the
TYPE/TOKEN distinction for both inscrip-
tions and thoughts. Though it may seem to,
the point I'm making doesn't depend on
doing so; and it simplifies the exposition.)
What are we to make of this analogy
between thoughts and symbols? The history
of philosophical and psychological theoriz-
ing about the mind consists largely of
attempts to exploit it by deriving the causal/
semantic properties of thoughts from the
FODOR. JERRY A.
causal/semantic properties of symbols; in
effect, by supposing that thinking is a kind
of talking to oneself. (Or of making pictures
to oneself; for our purposes the difference
doesn't matter since pictures are a kind of
symbol too.) Correspondingly, according to
this tradition, the laws that govern inten-
tional processes are supposed to be imple-
mented by causal relations among these
mental symbols. Cognitive science, with its
'mental representations' and 'languages of
thought' is merely the most recent avatar of
this ancient idea. I'm inclined to take it very
seriously if only for lack of alternative can-
didates.
(NB From the present lofty perspective,
one can barely distinguish behaviourists
from mentalists. Both assimilate thinking to
talking to yourself, the major disagreement
being over whether what you talk to your-
self in when you think is English or Menta-
lese. It is pretty close to a literal truth that
the tradition of theorizing about the mind
offers only two options: either thinking is
talking to yourself, or there is no such thing
as thinking.)
Inclinations aside, however, it would tend
to prove this traditional pudding if the
hypothesis that intentional mental processes
are causal interactions among mental
symbols could somehow be made to explain
why intentional mental processes are gen-
erally semantically coherent.
(NB 'If mental processes weren't generally
semantically coherent, they wouldn't be
intentional. This is a conceptual truth.'
(Compare, among many others, Clark,
1989.) But assuming that the rationality of
intentional processes is conceptually neces-
sary makes no progress with the problem in
hand; it just puts the pea under a different
shell. The operative question now becomes:
'how can merely material processes meet
the rationality conditions on intentionality?'
Generally speaking, if a philosopher offers to
'dissolve' the problem that you're working
on, tell him to go climb a tree.}
The hypothesis that mental processes are
causal interactions was, in fact, the con-
straint that Empiricist psychology foundered
on. Empiricists assumed that trains of
295
FODOR. JERRY A.
thought consist of causally connected
sequences of 'Ideas' (the Empiricist's term of
art for mental representations). The seman-
tic coherence of intentional processes was to
be explicated by laws of association which,
in effect, adjust the causal relations among
Ideas to reflect corresponding relations
among the things that they're ideas of. (Salt
comes with pepper; so pepper-Ideas become
associated to salt-Ideas; so thinking of salt
makes you think of pepper.) According to
the more advanced ('brain-like') Empiricist
models, the associative laws are themselves
supposed to be implemented by neurological
mechanisms. For example, variations in the
strength of associative relations between
ideas might be achieved by varying the cor-
responding synaptic resistances. You can
find this doctrine already pretty mature in
books like Donald Hebb's The Organization of
Behavior.
Not a word of it was true, of course. As
everyone has known since Kant - but
somehow forgets every thirty years or so -
semantically coherent processes are not, in
general. associative; and associative pro-
cesses are not, in general. semantically
coherent. Associationism died of this (CON-
NECTIONISTS have recently been trying to
revive it. with predictably dismal un success)
and, when it did, nobody then had the slight-
est idea how merely material processes could
implement the intentional laws that govern a
semantically coherent mind. Or how merely
immaterial ones could either. There the
problem stood until Alan Turing had what
I suppose is the best thought about how the
mind works that anyone has had so far.
What TURING did was to take the tradi-
tional analogy between minds and symbols
absolutely seriously. Symbols have both
semantic and material properties. (See above:
inscriptions of 'the cat is on the mat' both
mean that the cat is on the mat and grav-
itationally attract the Moon.) 'Very well,
then,' Turing (more or less) said, 'perhaps
one could build a symbol manipulating
machine whose changes of state are driven
by the material properties of the symbols on
which they operate (for example, by their
weight, or their shape. or their electrical
296
conductivity). And perhaps one could so
arrange things that these state changes are
semantically coherent in, for example. the
following sense: Given a true symbol to play
with, the machine will reliably convert it
into other symbols that are also true.'
TURING was right about its being possible
to build such machines ('computers' as one
now says). Its a fundamental idea of
current theorizing that minds are machines
of this kind. and that it is because they are
that mental processes are, by and large.
reliably rational. If intentional states are
relations to mental symbols and mental pro-
cesses are implemented by causal relations
among these symbols, then we can begin to
understand how the laws that subsume
mental states could preserve their semantic
coherence. So we see how an implementa-
tion theory for intentional laws might be
possible.
Summary so far: An implementation
theory for the law Fs cause Gs answers the
question: 'How do Fs cause Gs?' It does so
by specifying a mechanism that the instan-
tiation of F is sufficient to set in motion, the
operation of which reliably produces some
state of affairs that is sufficient for the
instantiation of G. Postulating a computa-
tional level of psychological explanation
provides for an implementation theory for
intentional laws; one which promises to
account for (some? much? all?) of their
semantic coherence. Since computational
mechanisms really are mechanisms, we are
some way towards understanding how
something rational could be material
through and through, hence how an inten-
tional psychology might be possibly com-
patible with materialist ontological scruples.
So far, so good. In fact, so far very good
indeed.
If you'll grant me this general framework,
I will soon be able to tell you what it is that
I think I'm doing.
PROPERTY THEORIES FOR AN
INTENTIONAL PSYCHOLOGY
Remember that, in the typical case, the
vocabulary that specifies an implementing
FODOR. JERRY A .
L: F .... ... G ....
L1
M2
F
-
L2
Basic level:
Figure 1. A typical relation between implemented laws (Fs cause Gs) and implementing mechan-
isms. Laws at any level are implemented by mechanisms whose behaviour is controlled by laws at
the next level down. For example, the law that Fs cause Gs is implemented by mechanisms that
obey the law that (MIpS cause MIGs). This law is itself implemented by mechanisms at the next
lower level, and so on down to basic mechanisms. The present case assumes that both property F
and property G are multiply realized at level L1.
mechanism is different from the vocabulary
that specifies the law that it implements. In
consequence, implementing state Mp's
bringing about implementing state MG
explains how Fs cause Gs only modulo a
theory that explains how something's being F
could be sufficient for something's being Mp
and how something's being MG could be
sufficient for something's being G. Figure 1
gives the general idea. I'll borrow a term
from Robert Cummins (1983) and call any
explanation of this kind a property theory.
(Apparently Cummins assumes, as I should
not wish to do, that property theories are
ipso facto theories of property identity. I
think that solving some of the problems to
be discussed below may turn on rejecting
this assumption (see Fodor, 1994).) Where I
think we now are in the philosophical dis-
cussion of cognitive science is this: thanks
to Turing's efforts (and despite those of con-
nectionists) we can now see how there
might be implementation theories for inten-
tional laws on the assumption that mental
processes are computational. But we are not
at all clear how, on that assumption, there
can be property theories for intentional
laws. If we knew how there can be property
theories for intentional laws, we would
know how an intentional psychology is pos-
sible, and this part of the philosophy of
mind could justifiably opt for early retire-
ment. (See COGNITIVE PSYCHOLOGY, COMPU-
TA TION AL MODELS.)
The problem is this: two kinds of relations
between laws and mechanisms are recog-
nized by property theories in cases outside
psychology: reduction and multiple realiza-
tion; and the relation between intentional
laws and computational processes seems to
be of neither of these kinds. So somebody
needs to worry about what kind of relation
it is.
You get reduction whenever some L-Ievel
property F is identifiable with some L - 1 level
property Mp. So, for example, being water is
identical to being H
2
0 according to the
usual understanding. You get multiple reali-
zation whenever there is a disjunction of
L-1 level properties, such that the instan-
tiation of F is sufficient for the instantiation
of the disjunction but not for the instantia-
tion of any of its disjuncts (see figure 1). In
the classic case of multiple realization, a
higher-level property is 'functionally de-
fined', and the realizing disjunction includes
all and only the mechanisms that can
perform the defining function. (So, for
example, there is presumably some disjunc-
tion of mechanisms any of which might
serve as a carburettor, and such that every
carburettor is an instance of one of the
297
FODOR, JERR Y A.
disjuncts or other.} These days, most philo-
sophers of mind suppose that most psycho-
logical properties are multiply realized.
Now, it is built into both the notions of
reduction and of multiple realization that
the instantiation of an L - 1 level property
can be a sufficient condition for the instan-
tiation of the corresponding L-level property
and vice versa. If, for example, F in figure 1
is the same property as M1p, then, of
course, every instantiation of the latter is
ipso facto an instantiation of the former.
Similarly, if (M1p vs. M2p ... ) is the dis-
junction that realizes F, then the tokening
of any of the disjuncts is sufficient for F. It is
no accident that both of the concepts typ-
ically deployed by property theories should
have this feature. As we've seen, the exist-
ence of a mechanism that eventuates in MG
will not explain how Fs bring about Gs
unless the instantiation of MG is sufficient
for the instantiation of G and the instanti-
ation of F is sufficient to set the mechanism
in motion.
The upshot is that if the immediately
implementing mechanisms for intentional
laws are computational. then we need a
property theory that provides computation-
ally sufficient conditions for the instanti-
ation of intentional properties. This,
however, implies a nasty dilemma. On the
one hand, we've seen that the computa-
tional account of the implementation of
intentional laws is extremely well moti-
vated; it's the only explanation of the
semantic coherence of mental processes that
looks like having a chance to work. But, on
the other hand, there are well-known and
persuasive reasons to doubt that there could
be computationally sufficient conditions for
the instantiation of intentional properties. If
this dilemma can't be broken, it looks as
though the usual constraints on property
theories aren't satisfiable in the case of
intentional laws. So maybe the answer to
how is an intentional psychology even possible?
is that it's not.
The force of this dilemma depends, of
course, on the status of the claim that there
can't be computationally sufficient condi-
tions for instantiating intentional properties,
298
and this claim is not universally granted,
either in cognitive science or in the philoso-
phy of mind. For example, lots of people in
cognitive science hold computational ver-
sions of the theory that the intentional
content of a thought is determined by its
'inferential role' (roughly, by its causal role
in mental processes). Readers familiar with
the AR TIFICIAL INTELLIGENCE literature will
recognize 'procedural' semantics as a
version of this idea. Structuralism in lin-
guistiCS and functionalism in the philosophy
of language are among its other incarna-
tions. It would be nice to be able to believe
that inferential role views of content are
right since, if the intentional properties of
thoughts are determined by their inferential
roles, it gets a lot easier to see how inten-
tional laws could be computationally im-
plemented. Inference is precisely what
computers do best.
But inferential accounts of content don't
work; there can't be computationally suffi-
cient conditions for the instantiation of
intentional properties. (More precisely, there
can't be metaphysically sufficient computa-
tional conditions for the instantiation of
intentional properties. This does not rule it
out that there might be nomologically suffi-
cient computational conditions for instan-
tiation of intentional properties; indeed, I
think that this possibility deserves very
serious consideration. Here too, see Fodor,
1994.) That there can't be computationally
sufficient conditions is because, as 'extern-
alists' in the philosophy of language are
forever reminding us, the content of one's
thoughts depends, at least in part, on their
relations to things in the external world. It is,
for example, something about the way that
they are related to dogs that makes some of
one's thoughts dog-thoughts (rather than,
say, cat-thoughts); and it is something
about the way that they are related to cats
that makes others of one's thoughts cat-
thoughts (rather than, say, dog-thoughts).
Inferential role theories of content were
offered, in the first instance, as a corrective
to the '''Fido''-Fido Fallacy'; which,
however, increasingly appears not to have
been fallacious. (A further problem about
inferential role semantics is that it is, almost
invariably, suicidally holistic. For recent dis-
cussion, see Fodor, 1991; and Fodor and
Lepore, 1992.)
If externalism is right, then (some of) the
intentional properties of thoughts are essen-
tially extrinsic; they essentially involve
mind-to-world relations. But we are still fol-
lowing Turing in assuming that the compu-
tational role of a mental representation is
determined entirely by its intrinsic properties
(its weight, shape, or electrical conductivity,
as it might be). The puzzle is that it is, to
put it mildly, hard to see how the extrinsic
properties of thoughts could supervene on
their intrinsic properties; which is to say
that it's hard to see how there could be
computationally sufficient conditions for
being in an intentional state; which is to
say that it's hard to see how the immediate
implementation of intentional laws could be
computational.
So semantics and methodology join forces
to make trouble for cognitive science: meth-
odology requires property theories for
special science laws, and semantics suggests
that intentional properties can't be compu-
tational. But the idea that the implementa-
tion of intentional laws is computational is
the only serious cognitive science we've got;
without it, the semantic coherence of the
intentional is completely a mystery. It
would be nice to find a way out of this, and
much of the discussion that is now taking
place between philosophers of science,
semanticists and philosophers of mind is
concerned with how to do so.
AT LAST: WHAT I THINK I'M DOING
Folk psychology supposes that we act out of
our beliefs and our desires; it is both relent-
lessly causal and relentlessly intentional.
The analogy between thoughts and symbols
- discovered, I suppose, by Plato, and
refined by the likes of Descartes, Hume, Mill
and Freud - helps us to see how thoughts
could have both sorts of properties: we have
only to assume that there are mental
symbols. Turing's idea that mental pro-
cesses are computational helps us see how
FODOR. JERR Y A.
thinking could take one from truths to
truths: we have only to assume that mental
processes are driven by the intrinsic proper-
ties of mental symbols. Put the two ideas
together and one begins to have a glimmer
how something that is merely material can
nonetheless be rational. Thus far has the
representational theory of the mind effected
the transmutation of folk psychology into
science.
But both Plato's idea and Turing's depend
on there being something that fixes the rela-
tion between a mental symbol and its content
and keeps this relation fixed as the symbol's
causal role in computation unfolds. It's no
use having a mechanism in virtue of which
'All men are bald and Socrates is a man'
representations cause 'Socrates is bald'
representations if 'bald' has stopped
meaning bald in the course of the transac-
tion. What could this content-fixing some-
thing be? If contents just were causal roles
in computation, this property identity would
be the glue. But they aren't, and it isn't,
and some other story needs to be told.
I'm working on telling some other story.
Doing so is part of a general project to
which, in my view, most sensible psycholo-
gists and philosophers of mind have always
been more or less wittingly devoted: the
construction of a representational theory of
mind; the reduction of minds to symbols. I
sort of like the work. The pay is no good,
and the progress is very slow. But you do
get to meet interesting people.
See also CONCEPTUAL ROLE SEMANTICS;
CONTENT; FUNCTIONALISM; LANGUAGE OF
THOUGHT; MODULARITY; PSYCHOLOGY AND
PHILOSOPHY; THOUGHT; THOUGHT AND
LANGUAGE.
BIBLIOGRAPHY
Clark, A. 1989. Microcognition. Cambridge,
MA: MIT Press.
Cummins, R. 1983. The Nature of Psychological
Explanation. Cambridge, MA: MIT Press.
Fodor, J. 1991. A Theory of Content and Other
Essays. Cambridge, MA: MIT Press.
Fodor, J., and Lepore, E. 1992. Holism, A Shop-
per's Guide. Oxford: Basil Blackwell.
299
FOLK PSYCHOLOGY (1)
Fodor, J. 1994. The Elm and the Expert. Cam-
bridge, MA: MIT Press ..
Hebb, D. 1949. The Organization Of Behavior: A
Neuropsychological Theory. New York: Wiley.
JERRY. A. FODOR
folk psychology (1) What is folk psychol-
ogy? Until recently, many philosophers
would have answered something like this:
folk psychology (henceforth 'FP') is a 'con-
ceptual framework' and/or 'network of
principles' (perhaps, largely implicit) used
by ordinary people to understand, explain,
and predict their own and other people's
behaviour and mental states. Since these
ways of characterizing FP are often asso-
ciated with the claim that FP is a theory and
that claim is now controversial, a more
neutral formulation is called for.
Human beings are social creatures. And
they are reflective creatures. As such, they
continually engage in a host of cognitive
practices that help them get along in their
social world. In particular, they attempt to
understand, explain, and predict their own
and others' psychological states and overt
behaviour; and they do so by making use of
an array of ordinary psychological notions
concerning various internal mental states,
both occurrent and dispositional. Let us
then consider FP to consist, at a minimum,
of (a) a set of attributive, explanatory, and
predictive practices, and (b) a set of notions
or concepts used in those practices.
Whether it also consists of a set of laws,
generalizations, principles, or rules that are
implicated in (a) or that help to define (b) I
shall, for the moment, leave an open ques-
tion.
OVERVIEW OF THE LITERATURE
Broadly speaking, contemporary philosoph-
ical discussion of FP has been primarily
concerned with the question of FP's status
vis-a.-vis a future scientific theory of mind-
brain. There have been two relatively dis-
tinct strands to the discussion. The first has
identified the relevant scientific theory as a
theory deriving from neuroscience and has
300
fairly single-mindedly focused on the claim
that FP will eventually be eliminated in
favour of this theory. Churchland (1981)
has been the major proponent of this elim-
inativist claim. There is also a sizeable lit-
erature which takes on the eliminativist
challenge and tries to respond to it in
various ways. (See e.g. Kitcher, 1984;
Horgan and Woodward, 1985; Baker,
1987, unpublished; Boghossian, 1990.)
In contrast, the second strand takes the
relevant mind-brain theory to be a theory
derived from scientific psychology and the
major question to be whether FP will be
vindicated by scientific psychology. Sug-
gested answers have ranged from 'definitely
yes' (Fodor, 1975, 1987) to 'sort of, partly'
(Dennett, 1987) to 'possibly not' (Stich,
1983).
The question of the status of FP vis-a.-vis
the ultimate future scientific theory of the
mind-brain has been conducted under the
assumption that FP is also something like a
theory, consisting of a variety of general-
izations or laws connecting mental states
with other mental states and mental states
with behaviour. A number of papers
(Gordon, 1986, 1992; Goldman, 1989,
1992) have recently challenged this
assumption. Although the question of
whether FP is a theory (I shall call this the
question of FP's form) has taken on a life of
its own, it is also generally perceived as
being closely tied to the question of its
status. The reason is that it is generally
assumed that if FP is not a theory, then it
can't be a radically false theory (Stich and
Nichols, 1992); hence, there will be no
reason to want to eliminate it.
Although findings from and speculation
about scientific psychology have increas-
ingly been brought to bear on discussions
about FP, there is still a surprising amount
of dispute, especially of the eliminability
question, which operates as if scientific psy-
chology did not exist. I take this to be a
serious mistake and in what follows I will
attempt to redress the balance somewhat by
focusing on what scientific psychology can
tell us about FP. My view is that because a
scientific theory exists (namely, scientific
psychology) which is conceptually closer to
FP and 'higher level' than neuroscientific
theory, the question of whether FP will be
eliminated in favour of neuroscientific
theory can only be meaningfully asked as a
two-step question: will FP eventually be
eliminated in favour of scientific psychol-
ogy? And will scientific psychology even-
tually be eliminated in favour of
neuroscientific theory? (see COGNITIVE
PSYCHOLOGY). In what follows I will
address only the former question.
THE CONTENT OF FOLK PSYCHOLOGY
Given how we have defined FP, the
'content' of FP can be regarded as the parti-
cular concepts and practices (and if there
are generalizations or rules - the particular
generalizations or rules) employed by an
ordinary person in understanding, explain-
ing, and predicting human psychology,
whether his or her own or someone else's.
The dominant philosophical picture of the
content of FP is that it consists, first and
foremost, of concepts pertaining to our
various PROPOSITIONAL ATTITUDE states
(especially, the attitudes of BELIEF and
DESIRE) and, second, of practices (or prin-
ciples) connecting these mental attitudes to
each other, to perceptual stimuli, and to
ACTIONS. This content is sometimes de-
scribed by articulating a sample list of 'laws'
which either analytically or synthetically,
depending on your view, govern the con-
cepts or states in question. (See, for example,
Churchland, 1970, 1979, and FOLK PSY-
CHOLOGY (2), this volume.) There is also a
vast literature that attempts to describe spe-
cific aspects of mind, as we ordinarily con-
ceive it (e.g. belief, EMOTION, PRACTICAL
REASON, etc.) based primarily on various
'armchair' methodologies such as concept-
ual analysis. INTROSPECTION, insight, and
speculation. (For particularly good examples
of this sort of approach, see Audi, 1973a, b.)
Philosophers are not the only ones that
have concerned themselves with the
content of FP, however. Systematic descrip-
tions similar to the standard philosophical
conception have recently been attempted by
FOLK PSYCHOLOGY (1)
Wellman (1990, ch. 4), a developmental
psychologist, and d'Andrade (1987), an
anthropologist. A somewhat different
picture of FP emerges if we turn to social
psychology, a field that has been investigat-
ing FP for nearly 50 years under the labels
'self perception', 'person perception', and,
more recently, 'social cognition'. (For an
introduction to this body of research, see
Tedeschi, Lindskold, and Rosenfeld, 1985,
chs 2, 4, and 5.)
Two general points emerge from a study
of this literature. First, although there seem
to be aspects of adult folk psychology that
all humans beings share (at least at a
certain developmental age), there is also
considerable variation: historical, gender,
and individual. (Recent work by anthro-
pologists also provides evidence of sig-
nificant cultural variation. See d' Andrade,
1987; Lutz, 1985, 1987). Second, the
conceptual apparatus of folk psychology
encompasses not only concepts of mental
states such as the propositional attitudes
but also a vast array of concepts pertaining
to a person's personality traits and disposi-
tions.
A classical example of social psychology
research will make the latter point clear and
give a sense of how social psychologists
investigate FP. According to social psychol-
ogists, people not only make inferences
about the immediate psychological causes of
behaviour; they also, typically, go on to
make inferences about other people's per-
sonality characteristics. We consider each
other to be unhappy, vain, boring, unim-
aginative, humourless, moody, critical,
shrewd, reserved, artistic, cautious, prac-
tical, reliable, modest, tolerant, helpful,
sincere, sociable, and so on (Rosenberg,
Nelson, and Vivekananthan (1969) list 60
such characteristics). In particular, once a
person has assigned a few traits to another
individual, he or she has little hesitation in
assigning others. From the folk psycholog-
ical perspective, certain traits just seem to
'go together'. Such 'trait implications' have
been extensively studied by social psycholo-
gists beginning with a classic study by Asch
in 1946. Several findings have emerged.
301
FOLK PSYCHOLOGY (1)
One is that in making trait-trait infer-
ences, some traits matter more than others.
For example, when Asch gave subjects a
short list of traits describing a person and
asked them to generate a more elaborate
description of the person, he found that
traits like 'warm' versus 'cold' generated
bigger differences than traits like 'polite'
versus 'blunt'. He suggested that the first
group is central to impression formation
whereas the second group is peripheral.
A second finding has been dubbed the
'halo effect' (Wegener and Vallacher,
1977). Once an observer has decided that
another person has some positive character-
istics, he or she tends to attribute other
positive characteristics. Similarly, once a
negative attribution has been made, other
negative attributions will typically follow.
Warm people tend to be seen as generous,
good-natured, popular, wise, happy, and
imaginative, for example, whereas cold
people are seen as unsociable, humourless,
stern, and critical.
In sum, there is good reason to believe
that folk psychology is both more hetero-
geneous and richer than philosophers have
typically taken it to be.
THE FORM OF FOLK PSYCHOLOGY
A key ingredient in Churchland's (1981)
case for the eliminability of FP has been the
assumption that FP is a theory. This is also
an assumption embraced by many develop-
mental and social psychologists (Wellman,
1990; Astington, Harris, and Olson, 1988;
Bruner and Tagiuri, 1954; Heider, 1958;
Wegener and Vallacher, 1977; to name but
a few). In the past few years, the so-called
'theory-theory' (Morton, 1980) has been
challenged by Gordon (1986, 1992) and
Goldman (1989, 1992) who maintain that
we explain and predict our own and other's
behaviour not by adverting to a folk psy-
chological theory but by engaging in a form
of mental simulation. This view is called 'the
simulation view'.
The theory-theory attempts to explain
our FP practices by claiming that folk psy-
chological theory is a theory not merely in
302
some abstract, Platonic sense, but in the
sense that people engage in the FP practices
they do at least partly in virtue of having
such a theory. In cognitivist terms, the
clearest way to have a theory of some
domain X is to employ a set of representa-
tions or a complex representational struc-
ture whose CONTENT constitutes a theory of
X. Two features of such representational
structures bear on the theory vs. simulation
dispute. First, the content of a representa-
tional structure need not be consciously
accessible. Second, representational struc-
tures can have many different kinds of
representation 'bearers' (Von Eckardt,
1993). In particular, a mental representa-
tion can be 'borne' by a 'propositional' data
structure (on the assumption that the
mind-brain is a conventional computa-
tional device) or by a pattern of connections
or a node (see CONNECTIONISM) (on the
assumption that the mind-brain is a con-
nectionist device). (The latter is the sort of
view that Churchland (1989, and FOLK
PSYCHOLOGY (2), this volume) currently
has in mind.)
In contrast, the simulation view rejects
the idea that people have a folk psychologi-
cal theory in this sense. Instead, it claims
that our FP practices are largely based upon
a capacity to utilize our normal decision-
making capability in simulation mode. Fol-
lowing Stich and Nichols (1992), let us
assume that this capability rests, at least in
part, on a Practical Reasoning System
('PRS'). What Gordon and Goldman hypo-
thesize is that we predict and explain beha-
viour by running this system 'off-line', that
is, decoupled from our 'action control
system'. To predict another person's beha-
viour, we imagine ourselves in the other
person's situation and then decide what we
would do. Or in cognitivist terms, we input
a set of 'pretend' or 'simulated' beliefs and
desires (those we believe are held by the
other person) into PRS and see what deci-
sion it comes up with. To explain why
someone acted the way he or she did, we
use our PRS in an 'analysis-by-synthesis'
mode and ask: what beliefs and desires,
when input into PRS, could have resulted in
an INTENTION to perform the action in
question?
Although the theory vs. simulation
debate has produced a considerable amount
of both a priori and empirical argumenta-
tion, the question of what underlies our FP
practices is, at this point, still an open ques-
tion. None of the a priori arguments that
have been given are compelling, and there
is serious disagreement regarding the the-
oretical import of the various experimental
findings that have been brought to bear on
the issue. On my view, the principal reason
for the stalemate is that, to date, neither
theory has been articulated in sufficient
detail so that the opposing parties can agree
on what sort of findings would count for or
against each of the theories.
There are numerous points of insufficient
articulation. I will comment on only two.
According to the simulation theory, a
person simulates another person's situation
by running either his or her own PRS with
'pretend' inputs. But how do these systems
work? We are not given a clue. The omis-
sion is an important one because one of the
ways these systems might work is by
exploiting a theory or set of rules. That is,
as Goldman (1989, 1992) puts it: such
systems may be 'theory-driven' rather than
'process-driven' (Le. not involving a theory).
But, of course, if the simulation theory is to
constitute a competitor to the theory-theory,
it must be articulated in such a way that
the 'theory-driven' option is ruled out. The
difficulty is that it is by no means clear how
to do this. First, there is the conceptual
problem of saying when a system does or
does not involve a theory. Solving this
problem will, in part, involve coming to
grips with such ancillary questions as
whether it is possible to have simulation
without mental concepts (Gordon, 1992b,
says 'yes'; Churchland, 1989, says 'no'),
and if not, whether mental state concepts
can have content without being embedded
in a theory (again Churchland, 1989, says
'no'). And, assuming this first problem can
be solved, there is the theoretical problem of
devising a process-driven PRS that can
account for all the features our capacity for
FOLK PSYCHOLOGY (1)
reasoning actually has, including the notor-
iously troublesome feature of productivity
(see Von Eckardt, 1993, pp. 81-2).
If we distinguish between the 'theoretical
models' or systems being invoked in the
theory-theory and simulation view and
the claims ('theoretical hypotheses') vis-ii-vis
the real world being made for those models,
the lack of clarity we have dealt with thus
far concerns only the theoretical models.
Unclarity also exists with respect to the
respective theoretical hypotheses. In par-
ticular it is unclear what the scope of the
two theoretical hypotheses is supposed to
be. Is the claim that theory/simulation is
used every time we engage in FP practices or
only some of the time? And if the latter,
under what circumstances? This question
becomes particularly acute if one adopts the
eminently sensible view that when people
engage in their FP practices, they can and
do sometimes use simulation and they can
and do sometimes use theory.
THE STATUS OF FOLK PSYCHOLOGY
Philosophers have asked about the status of
FP in various ways. Some have asked
whether it is likely that FP will be eliminated
in favour of our ultimate future scientific
theory of the mind-brain (ST); some have
asked whether FP is eliminable; others,
whether FP should be eliminated. Still others
speak of vindication by ST or simply of the
truth or falsity of FP.
Under what conditions would we expect
that FP would actually be eliminated in
favour of ST? There are four: (1) FP would
have to be the sort of thing that is suitable
for elimination; (2) ST would have to be
able to do the explanatory, predictive, and
descriptive job FP now does; (3) FP would
have to be considered false to a significant
degree (otherwise we would have no moti-
vation for replacing it); and (4) various
practical conditions would have to be satis-
fied (e.g. a change in first-language learning
practices, government intervention, etc.) so
that replacement would actually occur. (See
An Essay on Mind section 3.7; ELIM-
IN ATIVISM.)
303
FOLK PSYCHOLOGY (1)
True?
~ ~
Yes (REALISM) No (ANTI-REALISM)
~ - - - - - - ~ ~
Largely Partly Largely
I I I
Due to scientific Due to scientific Due to scientific
vindication? vindication? vindication?
/ ~ / ~ / ~
Yes No Yes No Yes No
I I I
By what? By what? By what?
/ ~ / ~ / ~
Psychology Neuroscience Psychology Neuroscience Psychology Neuroscience
I
I
I
I
I
I
I I I I
I I I I
I I I I
!
Fodor
Horgan &
Woodward
X
I I I I
Baker
Graham
& Horgan
Von Eckardt X Dennett Stich Churchland
Figure 1. Taxonomy of positions on the truth of folk psychology.
For purposes of discussion I will follow
most philosophers in assuming that condi-
tion (1) is satisfied. In other words, I will
assume that either FP is a theory or, if it is
not, it is nevertheless, replaceable by ST. I
will also follow standard practice in ignor-
ing condition (4). Thus, the focus of atten-
tion will be on whether, using our best
current estimates about the nature of ST, ST
could take over FP's explanatory and pre-
dictive functions and whether there is
reason to believe that FP will eventually
prove to be false. My discussion will,
however, be somewhat idiosyncratic in that
I will take ST to be identical to the mind-
brain theory that eventually issues from sci-
entific psychology rather than the theory
that eventually issues from neuroscience.
Is there any reason to believe that ST
could not eventually serve the explanatory,
descriptive, and predictive functions of FP?
To my knowledge, this question has not
been directly addressed in the literature. My
own view is that the answer is 'no'. In fact,
there is every reason to believe that ST will
serve these cognitive functions and will
serve them better than FP can.
304
In the first place, the domain of scientific
psychology (that is, the phenomena it is
interested in explaining, predicting, and
describing) includes the domain of FP. (It
also encompasses a lot else besides, including
all but one of Churchland's (1981, this
volume) list of FP explanatory failures -
mental illness, creativity, memory, intelli-
gence differences, and the many forms of
learning. Sleep will probably turn out to be a
physiological rather than psychological phe-
nomenon.) Second, there is no reason to
believe that ST won't have the cognitive
resources to do the requisite describing,
explaining, and predicting. A negative case
of this sort can be made vis-ii-vis neuro-
scientific theory. That is, it can be argued
that neuroscientific theory will never be able
to explain our actions and capacities because
it won't 'capture the generalizations' (Fodor,
1981) or because it won't posit states or
entities with intentional properties (Von
Eckardt, 1993). But neither of these failures
is shared by scientific psychology which
(despite the claims of Stich (1983)) shows
every sign of retaining intentional modes of
description (see INTENTION ALITY).
Even if ST could replace FP, it is not likely
that we would actually seek to eliminate FP
in favour of ST unless we thought that FP
was largely false. Is there any reason to
believe that FP will eventually prove to be
false to a significant degree? This question
has attracted a lot of philosophical atten-
tion. In fact, philosophers have taken just
about every conceivable position on the
truth of FP. A tree taxonomizing the major
positions is depicted in figure 1. In sorting
out this debate, it is important to note that
there are a number of sources of disagree-
ment.
First, there is disagreement regarding the
role of vindication in establishing the truth
or falsity of FP. Roughly speaking, one
theory is said to be vindicated by another if
the ontological commitments and claims of
the first are somehow reflected in the
second. Precisely what this 'reflection'
comes to is a matter of controversy. See
Horgan and Woodward (1985). Many phi-
losophers, including leading proponents
both of realism (Fodor, 1975, 1987; Horgan
and Woodward, 1985) and anti-realism
(Stich, 1983; Churchland, 1981), share
the assumption that the ultimate arbiter of
the reality of FP entities and the truth
of FP generalizations is science. There are
others, however (Baker, 1987 unpublished;
Graham and Horgan, 1988; Horgan and
Graham, 1991) who strongly object to this
view, arguing that the probative and/or
epistemic situation in favour of FP is so
strong that scientific vindication or lack
thereof is irrelevant to assessing its onto-
logical and truth claims.
Second, philosophers who are committed
to vindication sometimes disagree about
what scientific psychology will be like.
Fodor (1975), for example, speculates that
cognitive psychology will continue to posit
a representational system which is sig-
nificantly language-like (in the sense of
being evaluable, having constituent struc-
ture, and being compositional). In contrast,
Ramsey, Stich, and Garon (1991) predicate
their discussion of vindication on the
serious possibility that the mind is a con-
nectionist device in which the encoding of
FOLK PSYCHOLOGY (1)
information is widely distributed and the
most appropriate interpretation is 'subsym-
bolic' rather than symbolic.
Finally, there is often disagreement about
what FP includes. There are those (like
myself) who adopt a very 'thick' conception
of FP maintaining that any concept or gen-
eralization ordinary people use in their FP
practices is fair game for inclusion in FP. In
contrast, others (like Baker, unpublished)
adopt an extremely 'thin' conception
according to which FP includes only the
assumption that there are persons with atti-
tudes identified by propositional content.
And then, of course, there is the typical
'middle-size' conception according to which
FP includes concepts of the propositional
attitudes and various qualitative mental
states but not concepts of traits or person-
ality characteristics.
The stand one takes on each of these
questions - on the role of vindication, on
the nature of our future scientific psychol-
ogy, and on the content of FP - makes a big
difference with respect to one's stance on
elimination. For example, philosophers who
downplay the importance of vindication or
who adopt a very 'thin' conception of FP
tend to regard FP as clearly true. In con-
trast, those who opt for vindication but
adopt a stringent criterion for vindication or
envisage ST as quite dissimilar to FP tend to
regard FP as most probably false.
My own view is this. If we take vindica-
tion to be relevant, ST to be conceptually
similar to present-day scientific psychology,
and FP to be 'thickly' conceived, then FP is
likely to be partly false and partly true. The
judgment that FP is likely to be partly false
need not rest on educated guesses regarding
future vindication. There are numerous
findings from scientific psychology (includ-
ing neuropsychology) which call various
aspects of FP into question. Both Dennett
(1987) and Stich (1983) describe phenom-
ena that, at a minimum, put stress on our
folk psychological modes of explanation.
These phenomena include blindsight
(Weiskrantz, 1983), the behaviour of split-
brain patients (Gazzaniga, 1985), blindness
denial and hemi-neglect (Churchland,
305
FOLK PSYCHOLOGY (1)
1986). and the fact that the explanations
people give of their own behaviour often do
not correspond with its true causes (Nisbett
and Wilson. 1977).
Even more telling are the many social
psychology studies indicating that the ways
in which we form impressions of other
people and the ways in which we typically
explain their behaviour are 'biased' in
various ways. Tedeschi. Lindskold. and
Rosenfeld (1985) summarize their chapter
'Perceiving Persons and Explaining Beha-
viors' as follows:
People make characteristic errors in
assigning causes for behavior. In
general. actors perceive the environ-
ment as the cause of their behavior.
while observers attribute the same
behavior to the actor. There is a uni-
versal self-serving bias associated with
the outcomes of behavior. Actors tend
to attribute success to something about
themselves. usually ability. while they
attribute failure to external circum-
stances. often task difficulty or bad
luck.
A number of factors are involved
when we form overall impressions of
strangers .... We are influenced by
both the type of information presented
and the order of presentation. Central
traits. such as warm and cold. have
been shown to highly influence people
in determining their impression of an
actor's behavior. A primacy effect
occurs when early information has
more impact on impressions than later
information (p. 123).
Clearly there is reason to believe that FP
is at least partly false. Why think that it is
partly true? Here the argument does rest on
a guess about what form a future theory of
the mind-brain will take. Fodor (1987. p.
10) proposes that we take scientific psychol-
ogy to vindicate propositional attitude FP
just in case 'it postulates states (entities.
events. whatever) satisfying the following
conditions: (1) They are semantically evalu-
able. (2) They have causal powers. (3) The
306
implicit generalizations of commonsense
belief/desire psychology are largely true of
them.' It is far too early to tell whether con-
dition (3) will be true of ST. However. there
is excellent reason to suppose that (1) and
(2) will be. The project of cognitive psychol-
ogy (and. more broadly. cognitive science)
is. in part. to explain in virtue of what we
have the cognitive capacities we do and in
virtue of what these capacities have the
basic properties they do. including such
properties as intentionality. pragmatic
evaluability. and productivity (Von Eckardt.
1993). Currently cognitive theorizing.
including connectionist theorizing. attempts
to explain our cognitive capacities and their
basic properties by positing mental states
that are both representational and causal.
Unless our conception of what we want
psychology to explain undergoes radical
revision. it is difficult to imagine any suc-
cessful theory getting by without such
posits. Thus. although it may turn out that
ST does not precisely vindicate the existence
of beliefs. desires. and so forth. it will surely
vindicate the existence of causally effica-
cious mental states with content. Hence. in
that respect at least. FP will turn out to be
true.
See also DENNETT; FODOR; LANGUAGE OF
THOUGHT; LEWIS; PSYCHOLOGY AND
PHILOSOPHY; REASONS AND CAUSES;
STALNAKER; THOUGHT; THOUGHT AND
LANGUAGE.
BIBLIOGRAPHY
Asch. S.E. 1946. Forming impressions of per-
sonality. Journal of Abnormal and Social Psy-
chology. 41. 258-90.
Astington. J.W .. Harris. P.L.. and Olson. D.R ..
eds. 1988. Developing Theories of Mind.
Audi. R. 1973a. The concept of wanting.
Philosophical Studies. 24. 1-21.
Audi. R. 1973b. Intending. Journal of Philo-
sophy. 70. 387-403.
Baker. L.R. 1970. Saving Belief. Princeton Uni-
versity Press.
Baker. L.R. 1987 unpublished. The cognitive
status of common sense.
Boghossian. P. 1990. The status of content.
Philosophical Review. 99. ] 57-84.
Bruner. J.S .. and Tagiuri. R. 1954. The
perception of people. In Handbook of Social
Psychology. ed. G. Lindzey Cambridge. MA.:
Addison -Wesley.
Churchland. P.M. 1970. The logical character
of action explanations. Philosophical Review.
79.214-36.
Churchland. P.M. 1979. SCientific Realism and
the Plasticity of Mind. Cambridge University
Press.
Churchland. P.M. 1981. Eliminative materi-
alism and the propositional attitudes. Journal
of Philosophy. 78. 67-90.
Churchland. P.M. 1989. A Neurocomputational
Perspective. Cambridge. MA: MIT Press.
Churchland. P.S. 1986. Neurophilosophy:
Toward a Unified Science of the Mindl Brain.
dAndrade. R. 1987. A folk model of the mind.
In Cultural Models in Language and Thought.
ed. D. Holland and N. Quinn. Cambridge
University Press.
Fodor. J. 1975. The [,anguage of Thought. Cam-
Fodor. J. 1981. Computation and reduction.
In Representations. ed. J. Fodor. Cambridge
MA.: MIT Press.
Fodor. J. 1987. Psychosemantics. Cambridge
MA.: MIT Press.
Gazzaniga. M. 1985. The Social Brain: Discover-
ing the Networks of the Mind. New York:
Basic Books.
Goldman. A. 1989. Interpretation psychol-
ogized. Mind and Language. 4. 161-85.
Goldman. A. 1992. In defense of the simula-
tion theory. Mind and Language. 7. 104-19.
Gordon. R. 1986. Folk psychology as simula-
tion. Mind and Language. 1, 158-71.
Gordon. R. 1992 The simulation theory:
Objections and misconceptions. Mind and
Language. 7. 11-34.
Graham. G.L .. and Horgan. T. 1988. How to
be realistic about folk psychology. Philoso-
phical Psychology. 1. 69-81.
Heider. F. 1958. The Psychology of Interpersonal
Relations. New York: Wiley.
Horgan. T . and Graham. G. 1991. In defense
of southern fundamentalism. Philosophical
Studies. 62. 107-34.
FOLK PSYCHOLOGY (1)
Horgan. T .. and Woodward. J. 1985. Folk psy-
chology is here to stay. Philosophical Review.
94.197-225.
Kitcher. P. 1984. In defense of intentional
psychology. Journal of Philosophy. 71. 89-
106.
Lutz. C. 1985. Ethnopsychology compared to
what?: Explaining behavior and con-
sciousness among the Ifaluk. In Person. Self,
and Experience: Exploring Pacific Ethnopsychol-
ogies. ed. G. White and J. Kirkpatrick.
Berkeley: University of California Press.
Lutz. C. 1987. Goals and understanding in
Ifaluk emotion theory. In Cultural Models in
Language and Thought. ed. D. Holland and N.
Quinn. Cambridge University Press.
Morton. A. 1980. Frames of Mind. Oxford Uni-
versity Press.
Nisbett. R .. and Wilson. T. 1977. Telling more
than we can know: Verbal reports on
mental processes. Psychological Review. 84.
231-59.
Ramsey. W .. Stich. S.P .. and Garon. J. 1991.
Connectionism. eliminativism. and the
future of folk psychology. In Philosophy and
Connectionist Theory. ed. W. Ramsey. S. P.
Stich and D. E. Rumelhart. Hillsdale. NJ:
Erlbaum.
Rosenberg. S .. Nelson. c.. and Vivekananthan.
P.S. 1969. A multidimensional approach to
the structure of personality impressions.
Journal of Personality and Social Psychology.
9.293-94.
Stich. S.P. 1983. From Folk Psychology to
Cognitive Science: The Case Against Belief.
Stich. S.P .. and Nichols. S. 1992. Folk psy-
chology: simulation or tacit theory? Mind
and Language. 7.35-71.
Tedeschi, J.T .. Lindskold. S .. and Rosenfeld. P.
1985. Introduction to Social Psychology. St
Paul. MN: West Publishing Co.
Von Eckardt. B. 1993. What is Cognitive
Science? Cambridge. MA.: MIT Press.
Wegener. D.M .. and Vallacher. R.R. 1977.
Implicit Psychology: An Introduction to Social
Cognition. Oxford University Press.
Weiskrantz. J. 1983. Evidence and scotomata.
Behavioral and Brain Sciences. 6. 464-7.
Wellman. H. 1990. The Child's Theory of Mind.
BARBARA VON ECKARDT
307
FOLK PSYCHOLOGY (2)
folk psychology (2) 'Folk psychology'
denotes the prescientific. common-sense
conceptual framework that all normally
socialized humans deploy in order to com-
prehend, predict, explain. and manipulate
the behaviour of humans and the higher
animals. This framework includes concepts
such as belief, desire. pain, pleasure. love, hate,
joy, fear. suspicion. memory. recognition,
anger, sympathy. intention, and so forth. It
embodies our baseline understanding of the
cognitive. affective, and purposive nature of
persons. Considered as a whole, it con-
stitutes our conception of what a person is.
The term 'folk psychology' is also in-
tended to portray a parallel with what
might be called 'folk physics'. 'folk chem-
istry'. 'folk biology'. and so forth. The term
involves the deliberate implication that
there is something theory-like about our
commonsense understanding in all of these
domains. The implication is that the rele-
vant framework is speculative. systematic.
and corrigible. that it embodies generalized
information. and that it permits explanation
and prediction in the fashion of any theore-
tical framework.
There is little disagreement about the
existence of this shared conceptual frame-
work. but there are important disagree-
ments about its nature. its functions. its
epistemology. and its future. In particular.
the claim that our common-sense concep-
tion of human nature is like an empirical
theory has been strongly contested by a
number of writers. as has the related claim
that it might be empirically false. These
issues are best addressed by rehearsing the
history of this notion.
ORIGINS OF THE IDEA
The first explicit portrayal of our collective
self-conception as importantly theory-like
appears in a landmark paper by Wilfrid
Sellars (1956). Sellars describes an imagin-
ary stage of human pre-history in which
people have acquired the use of language.
but have not yet developed the vocabulary
for. nor even any conception of. the
complex mental states and processes rou-
308
tinely recognized by modern humans. Their
explanatory resources for explaining human
behaviour are limited to a few purely dis-
positional terms. all of which can be oper-
ationally defined (like 'is soluble') in terms
of some observable circumstance (such as
being put in water) that is sufficient for an
observable behaviour (such as dissolving).
For this reason. Sellars refers to these
people. pejoratively. as 'our Rylean ances-
tors' (see R YLE). They can explain some
human behaviours. but only very few.
Being limited to a set of operationally
defined dispositional concepts. they have no
conception of the complex dance of occur-
rent internal states driving human beha-
viour. no conception of the internal
economy that is just waiting to be char-
acterized by a full-blown theory of human
nature.
As Sellars develops the story. this deficit is
repaired by a visionary theorist named
Jones. Taking as his model the overt
declarative utterances already current in his
society. he postulates the existence. within
all humans. of covert. utterance-like events
called 'thoughts'. These internal events are
postulated to have the same semantic and
logical properties as their overt counter-
parts. and to play an internal role com-
parable to the ongoing discursive and
argumentative role often performed by overt
speech. A suitable sequence of such internal
events - some rough chain of practical rea-
soning. for example - is thus fit to explain
certain human behaviours as the natural
outcome of hidden speech-like antecedents.
despite the absence of any overtly voiced
practical reasonings preceding the beha-
viour on that occasion.
A further postulation by Jones. this time
exploiting the model of external perceptual
objects. brings the range of qualitatively dis-
tinct SENSATIONS into the picture. These
are said to be internal perceivables. covert
objects that can provoke appropriate cogni-
tion and ACTION even in the absence of
their external public counterparts. Related
postulations bring INTENTIONS onto the
scene. and also BELIEFS and DESIRES as
relatively lasting states of any individual
(they are dispositions to have occurrent
thoughts and intentions. respectively). In
sum. Jones postulates the basic ONTOLOGY
of our current folk psychology. and assigns
to its elements their now familiar causal
roles. much to the explanatory and pre-
dictive advantage of everyone who gains a
command of its concepts.
Once learned by everyone. Jones's theory
gets a final boost when it turns out that
each adept can further learn to make spon-
taneous first-person ascriptions of the new
concepts. ascriptions which are strongly
consistent with the ascriptions made on
purely explanatory or third-person criteria.
Jones's society has now reached the same
conceptual level that we moderns enjoy.
And what has raised them to this position.
on Sellars' account. is their acquisition of a
novel explanatory framework with a novel
internal ontology. Sellars's lesson is that our
modern folk psychology has precisely the
same epistemological status. logical func-
tions. and modelling ancestry as the frame-
work postulated by Jones in the heroic
myth. It is. in short. an empirical theory.
DEVELOPMENT OF THE IDEA
Sellars made much of the fact that our
conception of the semantic properties of
thoughts is derivative upon an antecedent
conception of the semantic properties of
overt declarative utterances. Our accounts
of semantic properties in general. therefore.
should not take the semantic features of
thoughts as explanatory primitives fit for
illuminating the semantics of overt speech.
as is the common impulse (d. Grice. 1957;
Searle. 1983). Instead. Sellars proposed a
conceptual-inferential-role account that
would provide an independent but parallel
explanation for semantics in both domains.
A more salient development of Sellar's
account is the novel solution it provides to
an old sceptical problem: the Problem of
Other Minds. The Behaviourist attempt to
forge a 'logical' connection between inner
states and overt behaviour (see BEHA-
VIOURISM). and the Argument from Ana-
logy's attempt to forge an inductive
FOLK PSYCHOLOGY (2)
connection between them. can both be put
aside in favour of the quite different hypo-
thetico-deductive connection implied by
Sellars's account. Third-person ascriptions
of mental states are typically singular expla-
natory hypotheses from which we can draw.
in the context of folk psychology as a whole.
consequences concerning the subject's
observable behaviour. As with explanatory
hypotheses in general. these mentalistic
hypotheses are believable exactly to the
degree that they are successful in allowing
us to explain and to anticipate behaviour.
In the main. the ascription of mental states
to others is explanatorily and predictively
successful. So it is reasonable to believe that
other humans have mental states. Indeed.
on this account the same hypothesis would
be similarly reasonable as applied to any
creature. human or non-human. so long as
its behaviour yielded to the same explana-
tory strategy.
If folk psychology (hereafter: FP) is a
theory. then its concepts must be embedded
in a framework of laws. laws at least tacitly
appreciated by those adept in its use. In the
60s and 70s. this inference was imposed on
everyone by the unquestioned logical
empiricist assumption that any theory is a
set of sentences or propositions. typically
universal in their logical form. Accordingly.
some writers set about to 'recover' the laws
of FP from our common explanatory prac-
tices - from the factors ordinarily appealed
to in explanations of human behaviour. and
from the ways in which they are occasion-
ally subjected to criticism and defence
(Churchland. 1970. 1979).
Any search of this kind quickly turns up
hundreds of putative laws. all of which have
the familiar ring of the obvious. These range
from the very simple to the quite complex.
For example:
(1) People who suffer bodily damage gen-
erally feel pain.
(2) People who are angry are generally
impatient.
(3) People who fear that P generally hope
that not-Po
(4) People who desire that p. and believe
309
FOLK PSYCHOLOGY (2)
that Q is a means to p, and have no
overriding desires or preferred strat-
egies, will generally try to bring it about
that Q.
These 'laws', and thousands more like
them, were claimed to sustain common-
sense explanations and predictions in the
standard deductive-nomological or 'cover-
ing-law' fashion. And the specific content of
those laws was claimed to account for the
relevance of the explanatory factors stand-
ardly appealed to in our daily practice.
Examination of the fine logical structure of
the many FP laws involving the proposi-
tional attitudes also revealed deep parallels
with the logical structure of laws in the
various mathematical sciences (Churchland,
1979, 1981). Further, the portrait of FP as
a network of causal laws dovetailed neatly
with the emerging philosophy of mind
called FUNCTIONALISM (Putnam, 1960;
Fodor 1968; Lewis 1972).
Critics often objected that such 'causal
laws' are either strictly speaking false, or
else only vacuously true by reason of sheer
analyticity or implicit ceteris paribus clauses
(Wilkes, 1981, 1984; Haldane, 1988). But
defenders replied that folk-theoretic laws
should not be expected to be anything more
than rough-and-ready truths, and that the-
oretical laws in every science are to some
degree qualified by ceteris paribus clauses, or
restricted in their application to standard or
to idealized situations. All told, the develop-
ing case for FP's theoretical status was com-
pelling.
CONSEQUENCES OF THE IDEA
These proved alarming, at least to the idea's
many critics. First, Sellars' account yields a
modern version of the Kantian claim that
one can know oneself, in consciousness,
only as one represents oneself with one's
own concepts. On Sellars' view, one repre-
sents oneself with the concepts of FP, a
speculative empirical theory. Introspective
knowledge is thus denied any special epis-
temological status: one's spontaneous first-
person psychological judgments are no
310
better (and no worse) than one's spontane-
ous observation judgments generally (see
FIRST-PERSON AUTHORITY; INTROSPEC-
TION). They are all hostage to the quality of
the background conceptual scheme in
which they are framed. This contingency
did not trouble Sellars. He was entirely con-
fident that FP was empirically true. But it
does trouble some others who are less
willing to roll the dice against future experi-
ence.
A second consequence is that the tradi-
tional mind-body problem emerges as a
straightforward scientific question - as a
question of how the theoretical framework
of FP will turn out to be related to what-
ever neuropsychological theory might
emerge to replace it. If FP reduces smoothly
to a materialist successor theory, then the
IDENTITY THEORY will be vindicated. If it
proves disjunctively so 'reducible', then
Functionalism will be vindicated. If it
proves irreducible by reason of finding no
adequate materialist successor at all, then
some form of DUALISM will be vindicated.
And if it proves irreducible by reason of
failing utterly to map onto its successful
materialist successor theory, then a position
called Eliminative Materialism will be vindi-
cated (see An Essay on Mind section 3.7;
ELIMINATIVISM). The successor theory will
then displace Jones's antique theory in our
social and explanatory practices, and the
ontology of FP will go the way of phlogis-
ton, caloric fluid, and the crystal spheres of
ancient astronomy.
This eliminative possibility was urged as
real fairly early in the discussion (Feyer-
abend, 1963; Rorty, 1965). Later it was
defended as empirically the most likely
outcome (Churchland, 1981). The bare pos-
sibility of a wholesale rejection of FP is of
course a simple consequence of FP's specu-
lative theoretical status. The positive like-
lihood of its rejection requires more
substantial empirical premises. To this end,
Churchland cited three major empirical fail-
ings ofFP.
First, FP fails utterly to explain a con-
siderable variety of central psychological
phenomena: mental illness, sleep, creativity,
memory. intelligence differences. and the
many forms of learning. to cite just a few.
A true theory should not have such
yawning explanatory gaps. Second. FP has
not progressed significantly in at least 2500
years. The Greeks appear to have used
essentially the same framework that we
deploy. If anything, FP has been in steady
retreat during this period. as intentional
explanations have been withdrawn from yet
one domain after another - from the heav-
enly bodies, from the wind and the sea,
from a plethora of minor gods and spirits.
from the visitation of disease. and so forth.
FP has not shown the expansion and devel-
opmental fertility one expects from a true
theory. Last. FP shows no sign of being
smoothly integra table with the emerging
synthesis of the several physical, chemical,
biological. physiological, and neurocompu-
tational sciences (see COM PUT A T 10 N A L
MODELS). Since active coherence with the
rest of what we presume to know is a
central measure of credibility for any
theory. FP's emerging wallflower status
bodes ill for its future.
That FP is fated to be judged empirically
false is the most intriguing and alarming of
the three major consequences of Sellars's
original idea, but clearly it is not a direct
consequence of that idea alone. It requires
additional premises about the empirical fail-
ings of FP. These additional premises can
and have been hotly contested (Kitcher,
1984; Dennett. 1981; Horgan and Wood-
ward, 1985). Most of these authors are
quite prepared to accept or even to urge the
theoretical nature of FP; but they are unan-
imous in their defence of its empirical in-
tegrity and rough truth.
As these authors see it. FP is simply not
responsible for explaining most of the puz-
zling phenomena listed in the preceding
paragraph. Those problems are set aside as
the burden of some other theory. To the
second complaint. it is replied that folk
psychology has indeed changed somewhat
over the centuries. although its approx-
imate truth has never required of it more
than minor adjustments. And to the third
complaint there are voiced a number of
FOLK PSYCHOLOGY (2)
tertiam quids - proposed alternatives to
the stark choice. 'either reduce FP, or
eliminate it'. Here the varieties of non-
reductive materialism - FUNCTIONALISM
and ANOMALOUS MONISM, for example -
playa prominent role (see esp. Fodor. 1975;
Dennett, 1981; Davidson, 1970; Clark,
1989).
A more radical and purely a priOri
response to eliminative materialism dis-
misses it as simply incoherent, on the
grounds that in embracing or stating its
case it must presuppose the integrity of the
very framework it proposes to eliminate
(Baker, 1987; Boghossian. 1990). Consider.
for example, the evident conflict between
the eliminativist's apparent belief that FP is
false, and his simultaneous claim that there
are no beliefs.
A straightforward response concedes the
real existence of this and many other con-
flicts, but denies that they signal anything
wrong with the idea that FP might
someday be replaced. Such conflicts signal
only the depth and far-reaching nature of
the conceptual change being proposed.
Insofar, they are only to be expected, and
they do nothing to mark FP as unreplace-
able. Even if current FP were to permit no
coherent denial of itself within its own the-
oretical vocabulary, a new psychological
framework need have no such limitation
where the denial of FP is concerned. So
long as a coherent. comprehensive altern-
ative to FP can be articulated and explored,
then no argument a priori can rightly
single out FP as uniquely true of cognitive
creatures.
In this connection it is worth noting that
a similar 'incoherence argument' could be
deployed to permit the uncritically con-
servative defence of any framework for
understanding c o g n i ~ o n . no matter how
inadequate it migh( be. so long as it
happened to enjoy the irrelevant distinction
of being currently in use by the people
attempting to criticize it. In short. the
incoherence argument covertly begs the
question in favour of current FP, the very
framework being called into question
(Churchland. 1981; Devitt, 1990).
311
FOLK PSYCHOLOGY (2)
This response returns us to the empirical
issues raised two paragraphs ago. Whether
FP is false and whether it will fail to reduce
are empirical issues whose decisive settle-
ment must flow from experimental research
and theoretical development. not from any
arguments a priori. The empirical jury is
still out and there is ample room for reason-
able people to disagree.
However. it must be noted that. accord-
ing to the most fertile theoretical accounts
currently under exploration in computa-
tional neuroscience (Anderson and Rosen-
feld. 1988; Churchland and Sejnowski.
1992). the basic unit of occurrent cognition
is apparently not the sentence-like state. but
rather the high-dimensional neuronal acti-
vation vector (that is. a pattern of excitation
levels across a large population of neurons)
(see CONNECTIONISM). And the basic unit of
cognitive processing is apparently not the
inference from sentence to sentence. but
rather the synapse-induced transformation
of large activation vectors into other such
vectors. It is not certain that such accounts
of cognition are true. nor that even if they
are. FP will fail to find some reduction
thereto. But recent science already suggests
that Jones' lingua-formal theory - Folk Psy-
chology - fails utterly to capture the basic
kinematics and dynamics of human and
animal cognition.
CRITICISM AND DEFENCE OF THE IDEA
Many philosophers resist entanglement in
these empirical issues and reject the pos-
sibility of FP's demise by rejecting the idea
that FP is an empirical theory in the first
place. Some play down the predictive and
explanatory role of FP. and focus attention
instead on the many social activities
conducted with its vocabulary. such as
promising. greeting. joking. threatening.
congratulating. insulting. reassuring. invit-
ing. provoking. sympathizing. questioning.
demanding. cajoling. smpmg. offering.
advising. directing. confiding. and so forth
(Wilkes. 1981. 1984). On this view. FP is
less an empirical theory than an intricate
social practice. one in which all normal
312
humans learn to participate (see also
Putnam. 1988). A supporting consideration
is the clearly normative character of many of
the so-called 'laws' of FP. a feature at odds
with the presumably descriptive character
of any empirical theory. And if FP is not a
theory. then there is no danger that it
might be false and hence no question of its
being eliminated.
There is much to be said for the positive
half of this portrayal of FP. and it is surely
counter-productive to resist it. But its nega-
tive half betrays a shallow understanding of
what theories are and what the command
of one typically involves. Since Kuhn's
1962 book. The Structure of Scientific Revolu-
tions. it has been evident that learning the
theories peculiar to any discipline is not
solely or even primarily a matter of learning
a set of laws and principles: it is a matter of
learning a complex social practice. of enter-
ing a specialized community with shared
values and expectations. both of the world
and of each other. One slowly acquires the
right skills of recognition and categoriza-
tion. the right skills of instrumental and
symbolic manipulation. the right sorts of
expectations and the right standards of
communication and evaluation.
Moreover. during normal science the
exemplars of achieved understanding playa
strongly normative role. both in setting the
standards for further understanding. and
even in imposing a standard on nature
itself. Such peripheral phenomena as may
fail to conform to the current paradigm are
regularly counted as deviant. abnormal.
pathological. or at least non-ideal.
In sum. the claim that FP is an empirical
theory is entirely consistent with - indeed it
is explanatory of - the intricate practical life
enjoyed by its adepts. It is typical of theor-
etical adepts that their practical activities.
and their practical worlds. are transformed
by the relevant acquisition of knowledge. So
it is with children who master FP in the
normal course of socialization.
As regards immunity to elimination. we
should observe that practices can be dis-
placed just as well as theories. and for
closely related reasons. Becoming a medi-
eval alchemist, for example, was a matter of
learning an inseparable mix of theory and
practice. But when modern chemistry began
to flower, the medieval practice was dis-
placed almost in its entirety. Current chem-
ical practice would be unintelligible to an
alchemist. And given the spectacular power
of modern chemistry, no one defends or
mourns the passing of the alchemist's com-
paratively impotent practice.
The positive idea behind the projected
displacement of FP is the hope of a com-
parably superior social practice rooted in a
comparably superior account of human
cognition and mental activity. If better
chemical theory can sustain better chem-
ical practice, then better psychological
theory can sustain better social practice. A
deeper understanding of the springs of
human behaviour may permit a deeper
level of moral insight and mutual care.
Accordingly, a genuinely worthy scientific
replacement for FP need not be 'dehuma-
nizing', as so many fear. More likely it will
be just the reverse. Perversity of practice is
a chronic feature of our social history.
Think of trial by ordeal, purification by fire,
absolution by ritual, and rehabilitation by
exorcism or by long imprisonment with
other sociopaths. Against such dark and
impotent practices, any source of light
should be welcomed.
The 'criticism from practice' may be
easily turned aside, but a different line of
criticism cuts more deeply against the origi-
nal claim of theoretical status for FP. This
complaint focuses attention on the large
number of laws that presumably must be
stored in any FP adept, and on the psycho-
logical unreality of the idea that one's
running comprehension and anticipation of
one's ongoing social situation involves the
continual application of appropriate general
sentences somehow retrieved from memory,
and the repeated performance of complex
deductions from these and other premises in
order to achieve the desired comprehension
and anticipation (Gordon, 1986; Church-
land, 1988; Goldman, 1993). People are
generally unable to articulate the 'laws' on
which their running comprehension is
FOLK PSYCHOLOGY (2)
alleged to rest, and it is in any case myster-
ious how they could perform such prodi-
gious feats of retrieval and deductive
processing in the mere twinklings of time
typically involved in our ongoing social
commerce.
This wholly genuine difficulty moves
Gordon and Goldman to defend a refur-
bished version of the Argument from
Analogy, called the 'Simulation Theory', as
an account of our knowledge of other
minds. The problems with this venerable
approach are familiar. The capacity for
knowledge of one's own mind may already
presuppose the general knowledge that FP
embodies (Strawson, 1958), and a general-
ization from one's own case may be both
logically too feeble and explanatorily too
narrow in its scope to account for the full
range and robustness of one's general
knowledge of human nature (Churchland,
1984). But there is an alternative response
to our difficulty about knowing and deploy-
ing law-like sentences, one that strikes at
the legacy of logical empiricism itself.
The difficulties in claiming FP as an
explanatory theory stem not from that
claim itself, but rather from the logical
empiricist's crudely linguaformal conception
of theories as sets of sentences, and his
correlative conception of explanation as the
deduction of the explanandum from such
sentences. The psychological unreality of
this picture, noted above in connection
with FP explanations, is in fact a chronic
defect of the logical empiricist's account in
every theoretical domain in which cognitive
agents are adepts, including the established
sciences. This defect, and others, provide
compelling grounds for rejecting entirely
the classical picture of what theoretical
knowledge is and how it is deployed on spe-
cific occasions for recognition, explanation,
or prediction (Van Fraassen, 1980; Church-
land, 1989a, 1989b). This critical assess-
ment coheres with the already existing
positive epistemological traditions estab-
lished by Kuhn in America, by Heidegger
on the continent, and more recently by the
Connectionists in the field of artificial intel-
ligence and cognitive neurobiology. FP can
313
FOLK PSYCHOLOGY (2)
thus continue to be counted an explanatory
theory, but the claim that it has this status
needs to be reformulated within a new and
independently motivated story of what
theoretical knowledge, explanatory under-
standing, and pragmatic skills really are.
TRANSFORMATION OF THE IDEA
As sketched in the last section but one, the
emerging account of how brains embody
information has nothing to do with sen-
tences, or with states that are even remotely
sentence-like. A familiar analogy may help
introduce this alternative account.
A television screen embodies a sequence
of REPRESENTATIONS which are non-
sentential in nature, in their syntax as well
as their semantics. A specific TV representa-
tion has no logical structure: rather, it is a
specific pattern of activation or brightness
levels across a large population of tiny
screen pixels. A human retina embodies a
representation in much the same sense:
what matters is the pattern of activations
across the photoreceptors. These two
examples are overtly pictorial in their
'semantics', but this is an incidental feature
of the examples chosen. Tastes are also
coded as a pattern of activations across the
several types of gustatory neurons on the
tongue, sounds are coded as a pattern of
activations across the auditory neurons in
the cochlea, and smells are coded as a
pattern of activations across the olfactory
neurons, but none of these representations
is 'pictorial' in the familar two-dimensional
spatial sense. And yet such pattern coding -
or vector coding, as it is commonly called -
is extremely powerful. Since each vectorial
representation is one permutation of the
possible values of its elements, the number
of distinct things representable explodes as a
power function of the number of available
elements. Think how many distinct pictures
a TV screen can display, using only
200,000 pixels, and think how many more
the retina can embody, using fully 100
million pixels.
The suggestion of the preceding is that
the brain's basic mode of occurrent repre-
314
sentation is the activation vector across a
proprietary population of neurons - retinal
neurons, olfactory neurons, auditory
neurons, and so forth. Such activation
vectors have a virtue beyond their combina-
tori ally explosive powers of representation.
They are ideally suited to participate in a
powerful mode of computation, namely,
vector-to-vector transformation. An activa-
tion pattern across one neural population
(e.g. at the retina) can be transformed into
a distinct activation pattern (e.g. at the
visual cortex) by way of the axonal fibres
projecting from the first population to the
second, and by way of the millions of care-
fully tuned synaptic connections that those
fibres make with the neurons at the second
or target population.
That second population of neurons can
project to a third, and those to a fourth, and
so on. In this way, a sensory activation
pattern can undergo many principled
transformations before it finally finds itself,
profoundly transformed by the many inter-
vening synaptic encounters, reincarnated as
a vector of activations in a population of
motor neurons, neurons whose immediate
effect is to direct the symphony of muscles
that produce coherent bodily behaviour
appropriate to the original input vector at
the sensory periphery. The animal dodges a
seen snowball, freezes at the sound of a pre-
dator, or moves forward at the smell of food,
all as a result of its well-tuned synaptic con-
nections and their repeated transformation
of its representational vectors.
Those synaptic connections constitute a
second domain of stored information in the
brain, a domain beyond the occurrent
domain of fleeting neuronal activations.
The well-tuned synaptic connections
embody all of the creature's general know-
ledge and skills: of interpretation, of recog-
nition, of anticipation, and of coherent,
interactive behaviour. Here also is where
learning enters the picture. Learning is not a
matter of assembling a vast mass of sen-
tences, as on the classical account. Instead,
learning is a matter of configuring the tril-
lions of synaptic connections between
neurons so that incoming sensory vectors
are automatically and almost instanta-
neously transformed into appropriate 'pro-
totype' vectors at the higher populations of
cortical neurons, Such prototype vectors
constitute the brain's learned perceptual
and explanatory categories. These proto-
types typically involve more information
than is strictly present in the sensory input
on any given occasion, and they thus con-
stitute ampliative interpretations of that
input, interpretations that place the input
into an antecedently prepared context and
fund expectations of features so far unper-
ceived.
Further transformations produce further
activation vectors or vector sequences in
downstream neuronal populations, and
these lead quite quickly to appropriate
motor responses, since activation vectors
are also an ideal means to direct and co-
ordinate large populations of muscles. All of
this happens in milliseconds because the
relevant transformations are achieved by
massively parallel processing: the many ele-
ments in any input pattern go through the
matrix of synaptic connections simulta-
neously.
This brief sketch indicates how neuro-
computational ideas suggest a unified
account of perceptual recognition, explana-
tory understanding, prediction, and motor
control - an account untroubled by
problems of retrieval or speed of processing.
The motivation for this account derives
primarily from its apparent success in
accounting for the functional significance of
the brain's microstructure, and the striking
cognitive behaviours displayed by artificial
neural networks. A secondary motive
derives from the illumination it brings to
traditional issues in epistemology and the
philosophy of science. To learn a theoretical
framework is to configure one's synaptic
connections in such a fashion as to parti-
tion the space of possible neuronal activa-
tion patterns into a system or hierarchy of
prototypes. And to achieve explanatory
understanding of an event is to have
activated an appropriate prototype vector
from the waiting hierarchy (Churchland,
1989b).
FOLK PSYCHOLOGY (2)
Finally, a much smaller motive derives
from the relief this view provides to our
earlier difficulty with the theoretical status
of FP. No longer need FP labour under its
archaic portrayal as a set of universally
quantified sentences, and no longer need
its functions be falsely cast in terms of
laborious deductions. The claim that FP is
a corrigible theory need not be hobbled by
its initial logical positivist dress. Instead,
we can claim that FP, like any other
theory, is a family of learned vectorial pro-
totypes, prototypes that sustain recognition
of current reality, anticipation of future
reality, and manipulation of ongoing
reality.
As the heroic myth of Jones underscores,
FP does indeed portray human cognition in
terms of overtly sentential prototypes, viz. in
terms of the many propositional attitudes.
But there is no reason why it must be
correct in so representing our cognition, nor
in representing itself in particular. Perhaps
the internal kinematics and dynamics of
human and animal cognition is not at all
like the sentential dance portrayed in FP.
This recalls the position of eliminative
materialism discussed earlier. Perhaps we
harbour instead a kinematics of activation
patterns and a dynamics of vector-to-vector
transformations driven by learned config-
urations of synaptic connections. Evidently
it is not inconceivable that FP might
someday be challenged by a better account
of human nature. Evidently the process is
already underway. Jones would surely
approve.
See also CONCEPTS; CONTENT; COGNITIVE
PSYCHOLOGY; FODOR; INTENTIONALITY;
LANGUAGE OF THOUGHT; LEWIS; PROPOSI-
TIONAL ATTITUDES; PSYCHOLOGY AND
PHILOSOPHY; REASONS AND CAUSES; STAL-
NAKER; THOUGHT; THOUGHT AND LAN-
GUAGE.
BIBLIOGRAPHY
Anderson, J.A., and Rosenfeld, E., eds. 1988.
Neurocomputing: Foundations of Research.
315
FOLK PSYCHOLOGY (2)
Baker. L.R. 1987. Saving Belief. Princeton
University Press.
Boghossian. P. 1990. The status of content.
Philosophical Review. 99. 157-84.
Clark. A. 1989. Microcognition. Cambridge.
MA.: MIT Press.
Churchland. P.M. 1970. The logical character
of action explanations. Philosophical Review.
79: 2.
-- 1979. SCientific Realism and the Plasticity
of Mind. Cambridge University Press.
-- 1981. Eliminative materialism and the
propositional attitudes. Journal of Philosophy.
LXXVIII: 2.
-- 1984. Matter and Consciousness. Cam-
-- 1988. Folk psychology and the explana-
tion of human behaviour. Proceedings of the
Aristotelian Society. supp!. vo!' 62. 209-21.
Reprinted in Churchland. P.M. 1989.
-- 1989a. On the nature of theories: A neu-
rocomputational perspective. In Scientific
Theories: Minnesota Studies in the Philosophy
of Science. vol. XIV. ed. W. Savage. Minnea-
polis: University of Minnesota Press.
-- 1989b. On the nature of explanation: A
PDP approach. In Churchland. P.M. 1989.
-- 198 9c. A Neurocomputational Perspective.
Churchland. P.S .. and Sejnowski. T. 1992. The
Computational Brain. Cambridge. MA.: MIT
Press.
Davidson. D. 1970. Mental events. In Experi-
ence and Theory. ed. L. Foster and J.
Swanson. Amherst: University of Massachu-
setts Press.
Dennett. D.C. 1981. Three kinds of inten-
tional psychology. In Reduction. Time and
Reality. ed. R. Healey. Cambridge University
Press. Reprinted in Dennett. D.C .. The Inten-
tional Stance (Cambridge. MA.: MIT Press.
1987).
Devitt. M. 1990. Transcendentalism about
content. Pacific Philosophical Quarterly. 71:
4.247-63.
Feyerabend. P.K. 1963. Materialism and the
mind-body problem. Review of Metaphysics.
17.
Fodor. J.A. 1968. Psychological Explanation.
New York: Random House.
Fodor. J.A. 1975. The Language of Thought.
316
Goldman. A. 1992. The psychology of folk
psychology. Behavioral and Brain Sciences.
16.15-28.
Gordon. R. 1986. Folk psychology as simula-
tion. Mind and Language. 1: 2.
Grice. H.P. 1957. Meaning. Philosophical
Review. 66. 377-88.
Haldane. J. 1988. Understanding folk. Proceed-
ings of the Aristotelian Society. supp!. vo!' 62.
223-54.
Horgan. T .. and Woodward. J. 1985. Folk psy-
chology is here to stay. Philosophical Review
XCIV: 2.197-225.
Kitcher. P. 1984. In defense of intentional
psychology. Journal of Philosophy. LXXI: 2.
89-106.
Kuhn. T.S. 1962. The Structure of Scientific
Revolutions. University of Chicago Press.
Lewis. D. 1972. Psychophysical and Theor-
etical Identifications. Australasian Journal of
Philosophy L: 3. 249-58.
Putnam. H. 1960. Minds and Machines. In
Dimensions of Mind. ed. S. Hook. New York
University Press.
Rorty. R. 1965. Mind-body identity. privacy.
and categories. Review of Metaphysics. 1.
Searle. J.R. 1983. Intentionality: An Essay in
the Philosophy of Mind. Cambridge University
Press.
Sellars. W. 1956. Empiricism and the philo-
sophy of mind. In Minnesota Studies in the
Philosophy of Science. Vol 1. Minneapolis:
University of Minnesota Press. Reprinted in
Sellars. W. Science. Perception and Reality
(London: Routledge & Keegan Paul. 1963).
Strawson. P.F. 1958. Persons. In Concepts.
Theories. and the Mind-Body Problem: Minne-
sota Studies in the Philosophy of Science. Vol.
2. ed. Minneapolis: University of Minnesota
Press.
Wilkes. K. 1981. Functionalism. psychology.
and the philosophy of mind. Philosophical
Topics. 12: 1.
Wilkes. K. 1984. Pragmatics in science and
theory in common sense. Inquiry. 27: 4.
Van Fraassen. B. 1980. The Scientific Image.
PAUL M. CHURCHLAND
formal see SYNTAX/SEMANTICS.
functionalism (1) The functionalist thinks
of MENTAL STATES and events as causally
mediating between a subject's sensory
inputs and that subject's ensuing beha-
viour. Functionalism itself is the stronger
doctrine that what makes a mental state the
type of state it is - a pain, a smell of violets,
a belief that koalas are dangerous - is the
functional relations it bears to the subject's
perceptual stimuli, behavioural responses,
and other mental states.
RECENT HISTOR Y
(Most of the works to be mentioned without
citation are included in Block, 1980; Lycan,
1990; Rosenthal. 1991.)
Twentieth-century functionalism gained
its credibility in an indirect way, by being
perceived as affording the least objection-
able solution to the mind-body problem.
Disaffected from Cartesian D U A LIS M and
from the 'first-person' perspective of intro-
spective psychology, the behaviourists had
claimed that there is nothing to the mind
but the subject's behaviour and disposi-
tions to behave (see BEHAVIOURISM). For
example, for Rudolf to be in pain is for
Rudolf to be either behaving in a wincing-
groaning-and-favouring way or disposed to
do so (in that he would so behave were
something not keeping him from doing so);
it is nothing about Rudolf's putative inner
life or any episode taking place within
him.
Though behaviourism avoided a number
of nasty objections to dualism (notably
Descartes' admitted problem of mind-body
interaction), some theorists were uneasy;
they felt that in its total repudiation of the
inner, behaviourism was leaving out some-
thing real and important. U. T. Place spoke
of an 'intractable residue' of conscious
mental items that bear no clear relations to
behaviour of any particular sort. And it
seems perfectly possible for two people to
differ psychologically despite total similarity
of their actual and counterfactual beha-
FUNCTIONALISM (1)
viour, as in a Lockean case of 'inverted
spectrum'; for that matter, a creature might
exhibit all the appropriate stimulus-
response relations and lack mentation
entirely.
For such reasons, Place and Smart pro-
posed a middle way, the IDENTITY THEOR Y,
which allowed that at least some mental
states and events are genuinely inner and
genuinely episodic after all; they are not to
be identified with outward behaviour or
even with hypothetical dispositions to
behave. But, contrary to dualism, the episo-
dic mental items are not ghostly or non-
physical either. Rather, they are neurophy-
siological (see PAIN). They are identical with
states or events occurring in their owners'
central nervous systems. To be in pain is,
for example, to have one's c-fibres, or possi-
bly a-fibres, firing. A happy synthesis: The
dualists were wrong in thinking that mental
items are non-physical but right in thinking
them inner and episodic; the behaviourists
were right in their PHYSICALISM but wrong
to repudiate inner mental episodes.
However, PUTNAM (1960) and FODOR
(1968) pointed out a presumptuous impli-
cation of the identity theory understood as
a theory of types or kinds of mental items:
that a mental type such as pain has always
and eveywhere the neurophysiological char-
acterization initially assigned to it. For
example, if the identity theorist identified
pain itself with the firings of c-fibres, it
followed that a creature of any species
(earthly or science-fiction) could be in pain
only if that creature had c-fibres and they
were firing. But such a constraint on the
biology of any being capable of feeling pain
is both gratuitous and indefensible; why
should we suppose that any organism
must be made of the same chemical mate-
rials as us in order to have what can be
accurately recognized as pain? The identity
theorist had overreacted to the behaviour-
ists' difficulties and focused too narrowly
on the specifics of biological humans'
actual inner states, and in so doing they
had fallen into species chauvinism.
Fodor and Putnam advocated the
obvious correction: What was important
317
FUNCTION ALISM (1)
was not its being c-fibres (per se) that were
firing, but what the c-fibre firings were
doing, what their firing contributed to the
operation of the organism as a whole. The
role of the c-fibres could have been per-
formed by any mechanically suitable com-
ponent; so long as that role was performed,
the psychology of the containing organism
would have been unaffected. Thus, to be in
pain is not per se to have c-fibres that are
firing, but merely to be in some state or
other, of whatever biochemical description,
that plays the same functional role as did
the firings of c-fibres in the human beings
we have investigated. We may continue to
maintain that pain 'tokens', individual
instances of pain occurring in particular
subjects at particular times, are strictly
identical with particular neurophysiological
states of those subjects at those times, viz.
with the states that happen to be playing
the appropriate roles; this is the thesis of
'token identity' or 'token physicalism'. But
pain itself (the kind, universal or type) can
be identified only with something more
abstract: the causal or functional role that
c-fibre firings share with their potential
replacements or surrogates. Mental state-
types are identified not with neurophysio-
logical types but with more abstract func-
tional roles, as specified by state-tokens'
relations to the organism's inputs, outputs,
and other psychological states.
(Terminological note: In this article, the
label 'functionalism' is reserved for what
Block calls 'psychofunctionalism'; what
Block calls 'conceptual functionalism' is not
per se discussed here' (see FUNCTIONALISM
(2)). The term was originally applied only to
psychofunctionalism; Block has extended it
to cover the conceptual analyses of Lewis
and Armstrong.)
MACHINE FUNCTIONALISM
Putnam compared mental states to the
functional or 'logical' states of a computer:
Just as a computer program can be realized
or instantiated by any of a number of physi-
cally different hardware configurations, so
318
can a psychological 'program' be realized by
different organisms of various physiochemi-
cal composition, and that is why different
physiological states of organisms of different
species can realize one and the same mental
state-type. Where an identity theorist's type-
identification would take the form, 'To be in
mental state of type M is to be in the neuro-
physiological state of type N', Putnam's
machine functionalism (as we may call it)
has it that to be in M is to be merely in
some physiological state or other that plays
role R in the relevant computer program
(Le. the program that at a suitable level of
abstraction mediates the creature's total
outputs given total inputs and so serves as
the creature's global psychology). The phy-
siological state 'plays role R' in that it
stands in a set of relations to physical
inputs, outputs and other inner states that
matches one-to-one the abstract input/
output/logical-state relations codified in the
computer program. (For the beginnings of a
formalization of this idea, see FUNCTION-
ALISM (2).)
The functionalist, then, mobilizes three
distinct levels of description but applies
them all to the same fundamental reality.
A physical state-token in someone's brain
at a particular time has a neurophysiologi-
cal description, but may also have a func-
tional description relative to a machine
program that the brain happens to be real-
izing, and it may further have a mental
description if some everyday mental state is
correctly type-identified with the functional
category it exemplifies. And so there is after
all a sense in which 'the mental' is distinct
from 'the physical': Though presumably
there are no non-physical substances or
stuffs, and every mental token is itself
entirely physical. mental characterization is
not physical characterization, and the
property of being a pain is not simply the
property of being such-and-such a neural
firing. Moreover, unlike behaviourism and
the identity theory, functionalism does not
strictly entail that minds are physical; it
might be true of non-physical minds, so
long as those minds realized the relevant
programs.
COGNITIVE PSYCHOLOGY, ARTIFICIAL
INTELLIGENCE, AND THE COMPUTER
MODEL OF THE MIND
In a not accidentally similar vein, beha-
viourism in psychology has almost entirely
given way to 'cognitivism', Cognitivism is
roughly the view that (i) psychologists may
and must advert to inner states and episodes
in explaining behaviour, so long as the
states and episodes are construed through-
out as physical, and (ii) human beings and
other psychological organisms are best
viewed as in some sense informaticn-
processing systems, As COGNITIVE PSy-
CHOLOGY sets the agenda, its questions take
the form, 'How does this organism receive
information through its sense-organs,
process the information, store it, and then
mobilize it in such a way as to result in
intelligent behaviour?' The working lan-
guage of cognitive psychology is highly con-
genial to the functionalist, for cognitivism
thinks of human beings as systems of inter-
connected functional components, interact-
ing with each other in an efficient and
productive way,
Meanwhile, researchers in computer
science have pursued fruitful research pro-
grammes based on the idea of intelligent
behaviour as the output of skilful informa-
tion-processing given input. ARTIFICIAL
INTELLIGENCE (AI) is, roughly, the project
of getting computing machines to perform
tasks that would usually be taken to
demand human intelligence and judgment;
and computers have achieved some modest
successes, But a computer just is a machine
that receives, interprets, processes, stores,
manipulates and uses information, and AI
researchers think of it in just that way as
they try to program intelligent behaviour;
an AI problem takes the form, 'Given that
the machine sees this as input, what must it
already know and what must it accordingly
do with that input in order to be able to , , ,
[recognize, identify, sort, put together,
predict, tell us, etc,] , , , ? And how, then,
can we start it off knowing that and get it
to do those things?' So we may reasonably
attribute such success as AI has had to self-
FUNCTIONALISM (1)
conscious reliance on the information-pro-
cessing paradigm, And that in turn
mutually encourages the functionalist idea
that human intelligence and cognition
generally are matters of computational
information-processing (see COMPUT A-
TION AL MODELS),
TELEOLOGICAL FUNCTIONALISM
Machine functionalism supposed that
human brains may be described at each of
three levels, the first two scientific and the
third familiar to common sense: the bio-
logical. specifically neurophysiological; the
machine-program or computational; and
the everyday mental or folk-psychological
(see FOLK PSYCHOLOGY), Psychologists
would explain behaviour, characterized in
everyday terms, by reference to stimuli and
to intervening mental states such as beliefs
and desires, type-identifying the mental
states with functional or computational
states as they went. Such explanations
would themselves presuppose nothing about
neuroanatomy, since the relevant psycholo-
gical/computational generalizations would
hold regardless of what particular biochem-
istry might happen to be realizing the
abstract program in question,
Machine functionalism as described has
more recently been challenged on each of a
number of points, that together motivate a
specifically teleological notion of 'function'
(Sober speaks aptly of 'putting the function
back into Functionalism'),
(1) The machine functionalist still con-
ceived psychological explanation in the Pos-
itivists' terms of subsumption of data under
wider and wider universal laws, But FODOR,
DENNETT and Cummins (1983) have defen-
ded a competing picture of psychological
explanation, according to which beha-
vioural data are to be seen as manifesta-
tions of subjects' psychological capacities,
and those capacities are to be explained by
understanding the subjects as systems of
interconnected components, Each compon-
ent is a 'homunculus', in that it is identified
by reference to the function it performs,
and the various homuncular components
319
FUNCTIONALISM (1)
cooperate with each other in such a way as
to produce overall behavioural responses to
stimuli. The 'homunculi' are themselves
broken down into subcomponents whose
functions and interactions are similarly used
to explain the capacities of the subsystems
they compose, and so again and again until
the sub-sub- ... components are seen to be
neuroanatomical structures. (An auto-
mobile works - locomotes - by having a
fuel reservoir, a fuel line, a carburettor, a
combustion chamber, an ignition system, a
transmission, and wheels that turn. If one
wants to know how the carburettor works,
one will be told what its parts are and how
they work together to infuse oxygen into
fuel, and so on.) Nothing in this pattern
of explanation corresponds to the sub-
sumption of data under wider and wider
universal generalizations, or to the Posi-
tivists' Deductive-Nomological model of
explanation as formally valid derivation
from such generalizations.
(2) The machine functionalist treated
functional 'realization', the relation between
an individual physical organism and the
abstract program it was said to instantiate,
as a simple matter of one-to-one correspon-
dence between the organism's repertoire of
physical stimuli, structural states and beha-
viour, on the one hand, and the program's
defining input/state/output function on the
other. But this criterion of realization was
seen to be too liberal. since virtually any-
thing bears a one-to-one correlation of some
sort to Virtually anything else; 'realization'
in the sense of mere one-to-one correspon-
dence is far too easily come by (Block,
1978; Lycan, 1987, ch. 3); for example, the
profusion of microscopic events occurring in
a sunlit pond (convection currents, biotic
activity, or just molecular motion) undoubt-
edly yield some one-to-one correspondence
or other to any psychology you like, but this
should not establish that the pond is, or
has, a mind. Some theorists have proposed
to remedy this defect by imposing a
teleological requirement on realization: a
physical state of an organism will count
as realizing such-and-such a functional
description only if the organism has
320
genuine organic integrity and the state
plays its functional role properly for the
organism, in the teleological sense of 'for'
and in the teleological sense of 'function'
(see DRETSKE; TELEOLOGY). The state must
do what it does as a matter of, so to speak,
its biological purpose. This rules out our
pond, since the pond is not a single organ-
ism having convection currents or mole-
cular motions as organs. (Machine
functionalism took 'function' in its spare
mathematical sense rather than in a genu-
inely functional sense. NB, as used here, the
term 'machine functionalism' is tied to the
original libertine conception of 'realizing'; so
to impose a teleological restriction is to
abandon machine functionalism.)
(3) Of the machine functionalist's three
levels of description, one is commonsensical
and two are scientific, so we are offered a
two-levelled picture of human psychobiol-
ogy. But that picture is unbiological in the
extreme. Neither living things nor even
computers themselves are split into a purely
'structural' level of biological/physiochemi-
cal description and anyone 'abstract' com-
putational level of machine/psychological
description. Rather, they are all hier-
archically organized at many levels, each
level 'functional' with respect to those
beneath it but 'structural' or concrete as it
realizes those levels above it. This relativity
of the 'functional' !,structural' or 'software' /
'hardware' distinction to one's chosen level
of organization has repercussions for func-
tionalist solutions to problems in the philo-
sophy of mind (Lycan, 1987, ch. 5), and for
current controversies surrounding CON-
NEcTIONIsM and neural modelling.
(4) Millikan, Van Gulick, Fodor, Dretske
and others have argued powerfully that tel-
eology must enter into any adequate analy-
sis of the INTENTION ALITY or aboutness of
mental states such as beliefs and desires, by
reference to the states' psychobiological
functions. If teleology is needed to explicate
intentionality and machine functionalism
affords no teleology, then machine function-
alism is not adequate to explicate intention-
ality.
It would have been nice to stick with
machine functionalism, for the teleologizing
of functionalism comes at a price. Talk of
teleology and biological function seems to
presuppose that biological and other 'struc-
tural' /states of physical systems really have
functions in the teleological sense. The
latter claim is controversial, to say the least.
And if it is not literally true, then mental
states cannot be type-identified with tele-
ological states. But fortunately for the tele-
ological functionalist, there is now a small
but vigorous industry whose purpose is to
explicate biological teleology in naturalistic
terms, typically in terms of aetiology.
PROBLEMS WITH INTENTIONALITY
Functionalism, and cognitive psychology
considered as a complete theory of human
thought, inherit some of the same diffi-
culties that earlier beset behaviourism and
the identity theory. These remaining obsta-
cles fall into two main categories: intention-
ality problems and QUALIA problems.
PROPOSITION AL ATTITUDES such as
beliefs and desires are directed upon states
of affairs which mayor may not actually
obtain (e.g. that the Republican candidate
will win), and are about individuals who
mayor may not exist (e.g. King Arthur).
Franz Brentano raised the question of how
any purely physical entity or state could
have the property of being 'directed upon'
or about a non-existent state of affairs
or object; that is not the sort of feature
that ordinary, purely physical objects can
have.
The standard functionalist reply is that
propositional attitudes have Brentano's
feature because the internal physical states
and events that realize them represent actual
or possible states of affairs. (See REPRE-
SENTATION, LANGUAGE OF THOUGHT.)
What they represent (see CON TEN T) is deter-
mined at least in part by their functional
roles.
There are two difficulties. One is that of
saying exactly how a physical item's repre-
sentational content is determined; in virtue
of what does a neurophysiological state
represent precisely that the Republican candi-
FUNCTIONALISM (1)
date will win? An answer to that general
question is what Fodor has called a 'psy-
chosemantics', and several attempts have
been made.
The second difficulty is that ordinary pro-
positional attitude contents do not SUPER-
VENE on the states of their subjects' nervous
systems, but are underdetermined by even
the total state of that subject's head. Put-
nam's (1975) TWIN EARTH and indexical
examples show that, surprising as it may
seem, two human beings could be molecule-
for-molecule alike and still differ in their
beliefs and desires, depending on various
factors in their spatial and historical envir-
onments. Thus we can distinguish between
'narrow' properties, those that are deter-
mined by a subject's intrinsic physical com-
position, and 'wide' properties, those that
are not so determined, and representational
contents are wide (see EXTERNALISM/
INTERNALISM). Yet functional roles are,
ostensibly, narrow; how, then, can proposi-
tional attitudes be type-identified with func-
tional roles?
Functionalists have responded in either of
two ways: One is to understand 'function'
widely as well, specifying functional roles
historically and/or by reference to features
of the subject's actual environment. The
other is simply to abandon functionalism as
an account of content in particular, giving
some alternative psychosemantics for pro-
positional attitudes, but preserving func-
tionalism in regard to attitude types (thus
what makes a state a desire that P is its
functional role, even if something else
makes the state a desire that P). For more
on these issues, see FUNCTION ALISM (2).
PROBLEMS WITH QUALIA
The 'quale' of a mental state or event (par-
ticularly a SENSATION) is that state or
event's feel, its introspectible 'phenomenal
character'. Many philosophers have objec-
ted that neither functionalist metaphysics
nor any of the allied doctrines aforemen-
tioned can explain, illuminate or even toler-
ate the notion of what it feels like to be in a
mental state of such-and-such a sort. Yet,
321
FUNCTIONALISM (1)
say these philosophers, the feels are quintes-
sentially mental - it is the feels that make
the mental states the mental states they are.
Something, therefore, must be drastically
wrong with functionalism.
There is no single problem of qualia (d.
CONSCIOUSNESS); there are a number of
quite distinct objections that have been
brought against functionalism (some of
them apply to materialism generally). Space
permits mention of just three.
(1) Block (1978) and others have urged
various counterexample cases against func-
tionalism - examples in which some entity
seems to realize the right program but lacks
one of mentality's crucial qualitative
aspects. (Typically the 'entity' is a group,
such as the entire population of China
acting according to an elaborate set of
instructions, but our pond would also serve.
Neither, it seems, would be feeling anything
on its own.) Predictably, functionalists have
rejoined by arguing, for each example,
either that the proposed entity does not in
fact succeed in realizing the right program
(e.g. because the requisite teleology is
lacking) or that there is no good reason for
denying that the entity does have the rel-
evant qualitative states.
(2) Gunderson and Nagel have worried
over first-person/third-person asymmetries
and the perspectivalness or subjective point-
of-view-iness of consciousness. I can know
what it is like to have such-and-such a sensa-
tion only if I have had that sensation
myself; no amount of objective, third-person
scientific information would suffice. In reply,
functionalists have offered analyses of 'per-
spectivalness', complete with accounts of
'what it is like' to have a sensation, that
make those things compatible with func-
tionalism. Nagel and Jackson have argued,
further, for the existence of a special, intrin-
sically perspectival kind of fact, the fact of
'what it is like', which intractably and in
principle cannot be captured or explained
by physical science. Functionalists have
responded that the arguments commit a
LEIBNIZ'S-LAW fallacy (that of applying the
law in an intensional context); some have
added that in any case, to 'know what it is
322
like' is merely to have an ability, and
involves no fact of any sort.
(3) Levine and others have complained
that no functionalist theory can explain why
such-and-such a sensation feels to its
subject in just the way it does. The ques-
tion, 'But why do such-and-such functional
goings-on constitute or produce a sensation
like this?' seems always open. Some func-
tionalists contend that a functionalist psy-
chology could indeed explain that; others
try to show why it need not, consistently
with the truth of functionalism.
In sum, functionalism's besetting difficul-
ties are significant but not daunting. Func-
tionalism is not merely alive and well, but
still and rightly the reigning paradigm in
philosophy of mind.
See also CONCEPTUAL ROLE SEMANTICS; An
Essay on Mind section 3.6.3; LEWIS;
TURING.
BIBLIOGRAPHY
Block, N.J. 1978. Troubles with functionalism.
In Perception and Cognition: Minnesota Studies
in the Philosophy of Science, Vol. 9, ed. W.
Savage. Minneapolis: University of Minne-
sota Press. Excerpts reprinted in Lycan,
1990.
Block, N.J., ed. 1980. Readings in Philosophy
of Psychology, Vol. 1. Cambridge, MA.:
Explanation. Cambridge, MA.: MIT Press/
Bradford Books.
Fodor, J.A. 1968. Psychological Explanation.
New York: Random House.
Lycan, W. 1987. Consciousness. Cambridge,
MA.: MIT Press/Bradford Books.
Lycan, W.G., ed. 1990. Mind and Cognition: A
Reader. Oxford: Basil Blackwell.
Putnam, H. 1960. Minds and machines. In
Dimensions of Mind, ed. S. Hook. New York:
Collier Books.
Putnam, H. 1975. The meaning of'meaning'.
In Language, Mind and Knowledge: Minnesota
Studies in the Philosophy of Science, Vol. 7, ed.
Minnesota Press.
Rosenthal, D., ed. 1991. The Nature of Mind.
Shoemaker, S. 1981. Some varieties of func-
tionalism. Philosophical Topics, 12, 93-120.
WILLIAM G. LYCAN
functionalism (2) Functionalism is one of
the great 'isms' that have been offered as
solutions to the mind/body problem. The
cluster of questions that all of these 'isms'
promise to answer can be expressed as:
What is the ultimate nature of the mental?
At the most general level, what makes a
mental state mental? At the more specific
level that has been the focus in recent
years: What do thoughts have in common
in virtue of which they are thoughts? That
is, what makes a thought a thought? What
makes a pain a pain? Cartesian DUALISM
said the ultimate nature of the mental was
to be found in a special mental substance.
BEHAVIOURISM identified mental states with
behavioural dispositions; PHYSICALISM in
its most influential version identifies mental
states with brain states (see IDENTITY
THEOR Y). Functionalism says that mental
states are constituted by their causal
relations to one another and to sensory
inputs and behavioural outputs. Functional-
ism is one of the major theoretical develop-
ments of twentieth-century analytic
philosophy, and provides the conceptual
underpinnings of much work in cognitive
science.
A range of papers about functionalism
are to be found in five anthologies, Block
(1980), Lycan (1990), Rosenthal (1991),
Beakley and Ludlow (1992) and Goldman
(1993). (I will refer to these without dates
henceforth.) Expositions of one or another
version of functionalism are to be found in
the Introduction and the papers by Block,
Lewis and Putnam in Part III of Block, sec-
tions II and VII of Lycan, Section IIIB of
Rosenthal. Part I of Beakley and Ludlow
and Part III of Goldman. See also Chapter
12 of Shoemaker (1984) and Schiffer
(1987).
Functionalism has three distinct sources.
First, PUTNAM and FODOR saw mental
states in terms of an empirical computa-
tional theory of the mind. Second, Smart's
FUNCTIONALISM (2)
'topic neutral' analyses led Armstrong and
Lewis to a functionalist analysis of mental
concepts. Third, Wittgenstein's idea of
meaning as use led to a version of function-
alism as a theory of meaning, further devel-
oped by Sellars and later Harman. (See
FUNCTIONALISM (1) in this book and the
introduction to the functionalism section in
Block (1980)).
One motivation behind functionalism can
be appreciated by attention to artefact con-
cepts like carburettor and biological concepts
like kidney. What it is for something to be a
carburettor is for it to mix fuel and air in an
internal combustion engine - carburettor is
a functional concept. In the case of kidney,
the scientific concept is functional - defined
in terms of a role in filtering the blood and
maintaining certain chemical balances.
The kind of function relevant to the mind
can be introduced via the parity-detecting
automaton illustrated in the figure below,
which tells us whether it has seen an odd or
even number of 'l's (though it counts zero
as even). This automaton has two states, Sl
and S2; two inputs, '1' and blank (-); and
two outputs, it utters either the word 'Odd'
or 'Even'. The table describes two functions,
one from input and state to output, and
another from input and state to next state.
Each square encodes two conditionals speci-
fying the output and next state given both
the current state and input. For example,
the top-left box yields the following two
conditionals. (1) If the machine starts in Sl
Figure 1.
323
FUNCTIONALISM (2)
and sees a '1', it says 'Odd' (indicating that
it has seen an odd number of 'l's) and (2) if
the machine starts in Sl and sees a '1', it
goes to S2' The entire machine is specified
by eight such conditionals.
Now suppose we ask the question: 'What
is Sl?' The answer is that the nature of Sl is
entirely relational, and entirely captured by
the table. The nature of Sl is given by the 8
conditionals, including that when in Sl and
having seen a '1' the machine goes into
another state which is characterizable along
the same lines as Sl'
Suppose we wanted to give an explicit
characterization of 'Sl" We could do it as
follows.
Being in Sl = being in the first of two
states that are related to one another and to
inputs and outputs as follows: being in one
of the states and getting a '1' input results
in going into the second state and emitting
'Odd'; being in the second of the two states
and getting a '1' input results in going into
the first and emitting 'Even'; and so on, for
the remaining half of the table.
Here is a variant on this characterization
in which the quantification is more explicit:
Being in Sl = Being an x such that
:3 P:3 Q[If x is in P and gets a '1' input,
then it goes into Q and emits 'Odd'; if x
is in Q and gets a '1' input it goes into
P and emits 'Even'; and so on for the
remaining half of the table; and x is in
P] (Note: read ':3 P' as There is a prop-
erty P.)
This illustration can be used to make a
number of points. (1) According to func-
tionalism, the nature of a mental state is
just like the nature of an automaton state:
constituted by its relations to other states
and to inputs and outputs. All there is to Sl
is having some property which, together
with a '1' input results in a certain output
and a certain state transition, etc. Accord-
ing to functionalism, all there is to being in
pain is having some property that disposes
you to say 'ouch', wonder whether you are
ill, etc. (2) Because mental states are like
automaton states in this regard, the illus-
324
trated method for defining automaton states
is supposed to work for mental states as
well. Mental states can be totally character-
ized in terms that involve only logico-math-
ematical language and terms for input
signals and behavioural outputs. Thus func-
tionalism satisfies one of the desiderata of
behaviourism, characterizing the mental in
entirely non-mental language. (3) Sl is a
second-order state in that it consists in
having other properties, say mechanical or
hydraulic or electronic properties that have
certain relations to one another. These
other properties, the ones quantified over in
the definitions just given, are said to be the
realizations of the functional properties. So,
although functionalism characterizes the
mental in non-mental terms, it does so only
by quantifying over realizations of mental
states, which would not have delighted
behaviourists. (4) One functional state can
be realized in different ways. For example,
an actual metal and plastic machine satisfy-
ing the machine table might be made of
gears, wheels, pulleys and the like, in which
case the realization of Sl would be a
mechanical state; or the realization of Sl
might be an electronic state, and so forth.
(5) Just as one functional state can be rea-
lized in different ways, one physical state can
realize different functional states in different
machines. This could happen, for example, if
a single type of transistor were used to do
different things in different machines.
(6) Since Sl can be realized in many ways,
a claim that Sl is a mechanical state would
be false (at least arguably), as would a claim
that Sl is an electronic state. For this
reason, there is a strong case that function-
alism shows physicalism is false; if a crea-
ture without a brain can think, thinking
can't be a brain state. (But see the section
on functionalism and physicalism below.)
The notion of a realization deserves
further discussion. In the early days of func-
tionalism, a first-order property was often
said to realize a functional property in
virtue of a 1-1 correspondence between the
two realms of properties. But such a defini-
tion of realization produces far too many
realizations. Suppose, for example, that at t1
we shout 'one' at a bucket of water, and
then at t2 we shout 'one' again, We can
regard the bucket as a parity-detecting
automaton by pairing the physical config-
uration of the bucket at t1 with Sl and the
heat emitted or absorbed by the bucket at t1
with 'Odd'; by pairing the physical config-
uration of the bucket at t2 with S2 and the
heat exchanged with the environment at t2
with 'Even'; and so on, What is left out by
the post hoc correlation way of thinking of
realization is that a true realization must
satisfy the counterfactuals mentioned in the
table. To be a realization of Sl' it is not
enough to lead to a certain output and state
given that the input is a 'I'; it is also
required that had the input been a '0', the Sl
realization would have led to the other output
and state. Satisfaction of the relevant coun-
terfactuals is built into the notion of realiza-
tion mentioned in (3) above. (See Lycan,
1987.)
Suppose we have a theory of mental
states that specifies all the causal relations
among the states, sensory inputs and beha-
vioural outputs. Focusing on pain as a
sample mental state, it might say, among
other things, that sitting on a tack causes
pain and that pain causes anxiety and
saying 'ouch'. Agreeing for the sake of the
example, to go along with this moronic
theory, functionalism would then say that
we could define 'pain' as follows: being in
pain = being in the first of two states, the
first of which is caused by sitting on tacks,
and which in turn causes the other state
and emitting 'ouch'. More symbolicly
Being in pain = Being an x such that :I
P :I Q[sitting on a tack causes P and P
causes both Q and emitting 'ouch' and
x is in P]
More generally, if T is a psychological
theory with n mental terms of which the
seventeenth is 'pain', we can define 'pain'
relative to T as follows (the 'F1' ... 'Fn' are
variables that replace the n mental terms):
Being in pain = Being an x such that :I
F1 ... F
n
[T(F
1
... Fn) & x is in Fd
FUNCTIONALISM (2)
(The existentially quantified part of the
right-hand side before the '&' is the Ramsey
sentence of the theory T.) In this way, func-
tionalism characterizes the mental in non-
mental terms, in terms that involve quanti-
fication over realizations of mental states
but no explicit mention of them; thus func-
tionalism characterizes the mental in terms
of structures that are tacked down to reality
only at the inputs and outputs.
The psychological theory T just men-
tioned can be either an empirical psycholo-
gical theory or else a commonsense 'folk'
theory, and the resulting functionalisms are
very different. In the former case, which I
named psycho functionalism, the functional
definitions are supposed to fix the extensions
of mental terms. In the latter case, con-
ceptual functionalism, the functional defini-
tions are aimed at capturing our ordinary
mental concepts. (This distinction shows an
ambiguity in the original question of what
the ultimate nature of the mental is -
another ambiguity will be mentioned in the
next section.) The idea of psychofunctional-
ism is that the scientific nature of the
mental consists not in anything biological.
but in something 'organizational', ana-
logous to computational structure. Con-
ceptual functionalism, by contrast, can be
thought of as a development of logical
behaviourism. Logical behaviourists thought
that pain was a disposition to pain behaviour.
But as Geach and Chisholm pointed out,
what counts as pain behaviour depends on
the agent's beliefs and desires. Conceptual
functionalists avoid this problem by defining
each mental state in terms of its contribu-
tion to dispositions to behave - and have
other mental states. (See FUNCTIONALISM
(1); the introduction to the functionalism
section of Block, 1980. vol. 1.)
FUNCTIONALISM AND PHYSICALISM
Theories of the mind prior to functionalism
have been concerned both with (1) what
there is and (2) with what gives each type of
mental state its own identity, for example
what pains have in common in virtue of
which they are pains. We might say
325
FUNCTIONALISM (2)
(stretching these terms a bit) that (1) is a
matter of ONTOLOGY and (2) of metaphysics.
Here are the ontological claims: dualism told
us that there are both mental and physical
substances, whereas behaviourism and
physicalism are monistic, claiming that
there are only physical substances. Here are
the metaphysical claims: behaviourism told
us that what pains, for example have in
common in virtue of which they are pains is
something behavioural; dualism gave a non-
physical answer to this question, and phys-
icalism gives a physical answer to this ques-
tion.
Turning now to functionalism, it answers
the metaphysical question without answering
the ontological question. Functionalism tells
us that what pains have in common - what
makes them pains - is their function; but
functionalism does not tell us whether the
beings that have pains have any non-physi-
cal parts. This point can be seen in terms of
the automaton described above. In order to
be an automaton of the type described, an
actual concrete machine need only have
states related to one another and to inputs
and outputs in the way described. The
machine description does not tell us how
the machine works or what it is made of,
and in particular it does not rule out a
machine that is operated by an immaterial
soul, so long as the soul is willing to operate
in the deterministic manner specified in the
table. (See the papers by Putnam and by
Fodor in Block, 1980, vol. 1.).
In thinking about the relation between
functionalism and physicalism, it is useful to
distinguish two categories of physicalist
theses (see PHYSICALISM). One version of
physicalism competes with functionalism,
making a metaphysical claim about the
physical nature of mental state properties or
types (and is thus often called 'type' physic-
alism). As mentioned above, on one point of
view, functionalism shows that type physic-
alism is false. (Another point of view, that of
LEWIS will be mentioned later.)
However, there are more modest physic-
alisms whose thrusts are ontological rather
than metaphysical. Such physicalistic
claims are not at all incompatible with
326
functionalism. Consider, for example, a phy-
sicalism that says that every actual thing is
made up entirely of particles of the sort that
compose inorganic matter. In this sense of
physicalism, most functionalists have been
physicalists. Further, functionalism can be
modified in a physicalistic direction, for
example, by requiring that all properties
quantified over in a functional definition be
physical properties. Type physicalism is
often contrasted with token physicalism.
(The word 'teeth' in this sentence has five
letter tokens of three letter types.) Token
physicalism says that each pain, for
example, is a physical state, but token
physicalism allows that there may be
nothing physical that all pains share,
nothing physical that makes a pain a pain.
It is a peculiarity of the literature on func-
tionalism and physicalism that while some
functionalists say functionalism shows phys-
icalism is false (see the papers by Putnam,
Fodor, and Block and Fodor in Block, 1980),
others say functionalism shows physicalism
is true. (See LEWIS and the papers by Lewis
and by Armstrong in Block, 1980, and
Rosenthal, 1991.) In Lewis's case, the issue
is partly terminological. Lewis is a con-
ceptual functionalist about having pain.
'Having pain' on Lewis's regimentation,
could be said to be a rigid designator of a
functional property. (A rigid designator
names the same thing in each POSSIBLE
WORLD. 'The colour of the sky' is non-rigid,
since it names red in worlds in which the
sky is red. 'Blue' is rigid, since it names blue
even in worlds in which the sky is red.)
'Pain', by contrast, is a non-rigid designator
conceptually equivalent to a definite descrip-
tion of the form 'the state with such and
such a causal role'. The referent of this
phrase in us, Lewis holds, is a certain brain
state, though the referent of this phrase in a
robot might be a circuit state, and the refer-
ent in an angel would be a non-physical
state. Similarly, 'the winning number' picks
out '17' in one lottery and '596' in another.
So Lewis is a functionalist (indeed a con-
ceptual functionalist) about having pain. In
terms of the metaphysical issue described
above - what do pains have in common in
virtue of which they are pains - Lewis is a
functionalist. not a physicalist. What my
pains and the robot's pains share is a causal
role. not anything physical. Just as there is
no numerical similarity between 17 and
596 relevant to their being winning
numbers. there is no physical similarity
between human and Martian pain that
makes them pains. And there is no physical
similarity of any kind between human pains
and angel pains. However. on the issue of
the scientific nature of pain. Lewis is a phys-
icalist. What is in common to human and
Martian pain in his view is something con-
ceptual. not something scientific.
The claim that functionalism shows type
physicalism is false has received another
(related) type of challenge repeatedly over
the years. most effectively by Kim (1992).
Kim notes that psychofunctionalists have
taken the functional level to provide an
autonomous level of description that is the
right level for characterizing and explaining
the mental because it lumps together all the
different instantiations of the same mental
structure. those made of silicon with those
made of protoplasm. Kim argues that if
functionalism is true. then mental terms are
like 'jade' in not denoting natural kinds. A
natural kind is nomic (suitable for framing
laws) and projectible (see NATURAL KINDS).
Jadeite and nephrite are natural kinds. but
jade is not. If functionalism is true. then the
natural kinds in the realm of the mental are
physical rather than psychological - and
thus as far as the scientific nature of the
mental is concerned. physicalism is true.
functionalism being a theory of mental con-
cepts rather than mental kinds. So Kim ends
up as a functionalist. but. importantly. a
conceptual functionalist. not a psycho-
functionalist. The upshot. if Kim and Lewis
are right. is that the correct story about
mental concepts is functional. but the
correct story about the scientific nature of
the mental is physicalistic.
The psychofunctionalist reply should be
that the provinces of psychology and other
special sciences are physically disparate
entities that resemble one another function-
ally because they have been subjected to
FUNCTIONALISM (2)
forces like evolution and conscious design.
These forces have created a level of descrip-
tion - characterized by common properties
- genuine natural kinds. Consider econom-
ics - lawlike generalizations about capital
formation. inflation and the like are to a
large degree independent of the physical
nature of the economy. For example. one
set of economic laws can govern systems in
which the money is very different physi-
cally: gold. paper. beads. etc. Further. one
set of economic laws could be expected to
govern humans as well as silicon-based
robots and Martians. if they approximate to
rationality in roughly the way we do. But
the existence of a level of economic science
above the physical level presupposes aspects
of a similar level for psychology. As with
many areas of philosophy. the nub of the
conflict between physicalists and functional-
ists comes down to a question of how
sciences work.
White (1986) has further developed
Smart's famous argument for topic-neutral
analyses. the upshot of which is that
anyone who accepts any empirical identity
thesis. be it functionalist or physicalist.
should also be a conceptual functionalist.
Suppose that we accept an empirical iden-
tity thesis. say. that pain = state S17 where
S17 can be either a psycho functional state
or a brain state. Since this is not an a priori
truth (ex hypothesi). the terms flanking the
, =' sign must pick out this common refer-
ent via different modes of presentation of
the referent. in the manner of 'the evening
star = the morning star'. After all. if the
identity theorist believes both that he is in
pain and that he is in S17. these are distinct
beliefs. as is shown by the fact that he could
have believed that he was in pain but not in
S17' So the two terms must pick out the
same entity in virtue of different properties
of it. There is no mystery about how 'S17'
picks out the referent. but what is the mode
of presentation associated with 'pain'? Pre-
sumably. this will be some mental property.
say the phenomenal aspect of pain (see CON-
CEPTS; PROPER TY). But the identity theorist
will have to see this phenomenal aspect as
itself something functional or physical. so if
327
FUNCTIONALISM (2)
he resists conceptual functionalism, he will
suppose that there is another empirical iden-
tity, e.g. the phenomenal aspect = SI34' But
now we are back where we started, for this
identity raises the same issue as the original
one. The only plausible way to escape an
infinite regress, the argument concludes, is
to accept an a priori identity between some
mental property and a functional property,
for only in the case of an a priori identity
can the modes of presentation of terms
flanking the identity sign be the same, and
no physicalist or psychofunctionalist (or
dualist) identities are plausibly a priori.
Indeed, the only plausibly a priori identity is
a functionalist conceptual analysis, so any
identity theorist should be a conceptual
functionalist.
One way of resisting this conclusion
would be to adopt a holistic picture of how
words get their reference (see H 0 LIS M ). The
holistic psychofunctionalist can suggest that
we arrive at psycho functional identities in
choosing among theoretical perspectives on
the basis of evidence. If we want to know
what the meanings of mental terms are,
they should be given in terms of the entire
theory, and those 'definitions' will not be
happily classifiable as a priori or empirical.
Another way out (Loar, 1990) is to regard
the relation between 'pain' and its referent
as unmediated by any mode of presentation,
explaining how one can believe he is in
pain without being in SI7 by the different
functional roles of 'pain' and 'Sl7" The idea
is that to the extent that there is anything
associated with 'pain' in the category of
sense, it is a functional role, but this does
not require any descriptional element in the
background, taking 'pain' to its referent via
a mediating property, so the Smart-White
argument gets no purchase.
This sounds plausible as far as it goes, but
it does not go far enough, for one wants to
know more about the relation between
'pain' and its referent. Is 'pain' supposed to
be a name for, say, a brain state? Loar sup-
plies more detail, arguing that the concept
of a phenomenal property is a recognitional
disposition that is 'triggered' by its referent,
a physical or functional state. But couldn't
328
a physical or functional state trigger a
recognitional reaction without any phe-
nomenal feel, in the manner of blindsight?
To avoid such a possibility, it would seem
that Loar would have to bring in a mode of
presentation of the reference, a phenomenal
mode, and that would get him back in the
clutches of the Smart-White argument (see
QUALIA).
FUNCTIONALISM AND PROPOSITIONAL
ATTITUDES
The discussion of functional characteriza-
tion given above assumes a psychological
theory with a finite number of mental state
terms. In the case of monadic states like
pain, the sensation of red, etc., it does seem
a theoretical option to simply list the states
and their relations to other states, inputs
and outputs. But for a number of reasons,
this is not a sensible theoretical option for
belief-states, desire-states, and other propo-
sitional-attitude states. For one thing, the
list would be too long to be represented
without combinatorial methods. Indeed,
there is arguably no upper bound on the
number of propositions anyone of which
could in principle be an object of thought.
For another thing, there are systematic rela-
tions among beliefs: for example, the belief
that John loves Mary and the belief that
Mary loves John. These belief-states rep-
resent the same objects as related to each
other in converse ways. But a theory of the
nature of beliefs can hardly just leave out
such an important feature of them. We
cannot treat 'believes-that-grass-is-green',
'believes-that-grass-is-blue', etc. as unre-
lated primitive predicates. So we will need a
more sophisticated theory, one that involves
some sort of combinatorial apparatus. The
most promising candidates are those that
treat belief as a relation. But a relation to
what? There are two distinct issues here.
One issue is how to formulate the functional
theory. See Loar (1981) and Schiffer (1987)
for a suggestion in terms of a correspon-
dence between the logical relations among
sentences and the inferential relations
among mental states. A second issue is
what types of states could possibly realize
the relational propositional attitude states.
Field (1978) and Fodor (in Block, 1980, vol.
2, ch. 3) argue that to explain the pro-
ductivity of propositional attitude states,
there is no alternative to postulating a lan-
guage of thought, a system of syntactically
structured objects in the brain that express
the propositions in propositional attitudes.
(See Stalnaker (1984), chapters 1-3 for a
critique of Field's approach.) In later work,
Fodor (1987) has stressed the systematicity
of propositional attitudes mentioned above.
Fodor points out that the beliefs whose con-
tents are systematically related exhibit the
following sort of empirical relation: if one is
capable of believing that Mary loves John,
one is also capable of believing that John
loves Mary. Fodor argues that only a
language of thought in the brain could
explain this fact. (See FODOR; LANGUAGE OF
THOUGHTS; PROPOSITIONAL ATTITUDES.)
EXTERNALISM
The upshot of the famous 'TWIN EARTH'
arguments has been that meaning and
content are in part in the world and in the
language community. Functionalists have
responded in a variety of ways. One reaction
is to think of the inputs and outputs of a
functional theory as long-arm, as including
the objects that one sees and manipulates.
Another reaction is to stick with short-arm
inputs and outputs that stop at the surfaces
of the body, thinking of the intentional con-
tents thereby characterized as narrow -
supervening on the non-relational physical
properties of the body. There has been no
widely recognized account of what narrow
content is, nor is there any agreement as to
whether there is any burden of proof on the
advocates of narrow content to characterize
it (see CONTENT; EXTERNALISM/INTERNAL-
ISM; SUPERVENIENCE). (See the papers by
Burge, Loar and Stalnaker in Rosenthal,
1991; see also Goldman, 1993.)
MEANING
Functionalism says that understanding the
meaning of the word 'momentum' is a func-
FUNCTIONALISM (2)
tional state. On one version of the view, the
functional state can be seen in terms of the
role of the word 'momentum' itself in think-
ing, problem solving, planning, etc. But if
understanding the meaning of 'momentum'
is this word's having a certain function,
then there is a very close relation between
the meaning of a word and its function, and
a natural proposal is to regard the close
relation as simply identity, that is, the
meaning of the word just is that function
(see Peacocke, 1992).
Thus functionalism about content leads
to functionalism about meaning, a theory
that purports to tell us the metaphysical
nature of meaning. This theory is popular
in cognitive science, where in one version
it is often known as procedural semantics,
as well as in philosophy where it is often
known as CONCEPTUAL ROLE SEMANTICS.
The theory has been criticized (along with
other versions of functionalism) in
Putnam (1988) and Fodor and LePore
(1992).
HOLISM
Functionalism about content and meaning
appears to lead to holism. In general, transi-
tions among mental states and between
mental states and behaviour depend on the
contents of the mental states themselves. If I
believe that sharks are dangerous, I will
infer from sharks being in the water to the
conclusion that people shouldn't be swim-
ming. Suppose I first think that sharks are
dangerous, but then change my mind,
coming to think that sharks are not danger-
ous. However, the content that the first
belief affirms can't be the same as the
content that the second belief denies,
because the transition relations (e.g. the
inference from sharks being in the water to
what people should do) that constitute the
contents changed when I changed my mind
(see Fodor and LePore, 1992). A natural
functionalist reply is to say that some tran-
sitions are relevant to content individuation,
whereas others are not. But functionalists
have not told us how to do that. Appeal to a
traditional analytic/synthetic distinction
329
FUNCTIONALISM (2)
clearly won't do. For example, 'dog' and
'cat' would have the same content on such
a view. It could not be analytic that dogs
bark or that cats meow, since we can
imagine a non-barking breed of dog and a
non-meowing breed of cat. If 'Dogs are
animals' is analytic, so is 'Cats are animals'.
If 'Cats are adult kittens' is analytic, so is
'Dogs are adult puppies'. Dogs are not cats -
but then cats are not dogs. So a functional-
ist account will not find traditional analytic
inferential relations that will distinguish the
meaning of 'dog' from the meaning of 'cat'.
Other functionalists accept holism for
'narrow content', attempting to accom-
modate intuitions about the stability of
content by appealing to wide content.
(These issues are discussed in the 1993
volume of Mind and Language.)
QUALIA
Recall the parity-detecting automaton
described at the beginning of this article. It
could be instantiated by four people, each
one of whom is in charge of the function as
specified by a single box. Similarly, the
much more complex functional organiza-
tion of a human mind could 'in principle' be
instantiated by a vast army of people. We
would have to think of the army as con-
nected to a robot body, acting as the brain
of that body, and the body would be like a
person in its reactions to inputs. But would
such an army really instantiate a mind?
More pointedly, could such an army have
pain or the experience of red? If functional-
ism ascribes minds to things that don't have
them, it is liberal. Lycan (1987) suggests
that we include much of human physiology
in our theory to be functionalized to avoid
liberalism; that is, the theory T in the defi-
nition described earlier would be a psycho-
logical theory plus a physiological theory.
But that makes the opposite problem, chau-
vinism, worse. The resulting functional
description won't apply to intelligent Mar-
tians whose physiologies are different from
ours. Further, it seems easy to imagine a
simple pain-feeling organism that shares
little in the way of functional organization
330
with us. The functionalized physiological
theory of this organism will be hopelessly
different from the corresponding theory of
us. Indeed, even if one does not adopt
Lycan's tactic, it is not clear how pain could
be characterized functionally so as to be
common to us and the simple organism.
(See my 'Troubles with functionalism'
which appears in all the anthologies men-
tioned.)
Much of the force of the problems just
mentioned derives from attention to phe-
nomenal states like the look of red. Phe-
nomenal properties would seem to be
intrinsic to (non-relational properties of) the
states that have them, and thus phenom-
enal properties seem independent of the
relations among states, inputs and outputs
that define functional states. Consider, for
example, the fact that lobotomy patients
often say that they continue to have pains
that feel the same as before, but that the
pains don't bother them. If the concept of
pain is a functional concept, what these
patients say is contradictory or incoherent -
but it seems to many of us that it is intelli-
gible. (All the anthologies have papers on
this topic; see also Lycan, 1987; chapters 8,
9, 14, and 15 of Shoemaker, 1984; Hill,
1991.)
The chauvinism/liberalism problem
affects the characterization of inputs and
outputs. If we characterize inputs and
outputs in a way appropriate to our bodies,
we chauvinistically exclude creatures whose
interface with the world is very different
from ours, e.g. creatures whose limbs end in
wheels, or turning to a bigger difference,
gaseous creatures who can manipulate and
sense gases but for whom all solids and
liquids are alike. The obvious alternative of
characterizing inputs and outputs them-
selves functionally would appear to yield an
abstract structure that might be satisfied by,
e.g. the economy of Bolivia under manip-
ulation by a wealthy eccentric, and would
thus fall to the opposite problem of liberal-
ism.
It is tempting to respond to the chauvin-
ism problem by supposing that the same
functional theory that applies to me also
applies to the creatures with wheels. If they
thought they had feet, they would try to act
like us, and if we thought we had wheels,
we would try to act like them. But notice
that the functional definitions have to have
some specifications of output organs in
them. To be neutral among all the types of
bodies that sentient beings could have
would just be to adopt the liberal alternative
of specifying the inputs and outputs them-
selves functionally.
Lewis allows imperfect realizers, and it
might be thought that this device can avoid
the chauvinism problem. If a weird alien
believes that the sun is bigger than the
moon, he will be prepared to make certain
inferences and reject others, and these are
examples of a host of psychological general-
izations that have nothing to do with hands
and feet. Perhaps most of the psychological
generalizations about the aliens are the
same as for us, and so they can be con-
sidered to be imperfect realizers of our func-
tional organization. But this idea neglects
the crucial difference between a theory
(with all of its mental state terms) and its
Ramsey sentence (with all those terms
replaced by variables). If the alien's input
and output organs are mostly different from
ours, then the input/output constants in the
Ramsey sentence will be mostly different
from the constants in the Ramsey sentence
for us. And given that difference, the only
way the alien could count as a near-
realization of our Ramsey sentence is if near
isomorphism of the structure specified by
the Ramsey sentence is what is taken to
matter to realization - which makes inputs
and outputs not very relevant. But now we
are back to the economy of Bolivia problem
mentioned above.
TELEOLOGY
Many philosophers (see the papers by Lycan
and Sober in Lycan, 1990; FUNCTIONALISM
(2); Lycan, 1987) propose that we avoid lib-
eralism by characterizing functional roles
teleologically. We exclude the armies and
economies mentioned because their states
aren't Jor the right things. A major problem
FUNCTIONALISM (2)
for this point of view is the lack of an accep-
table TELEOLOGICAL account. Accounts
based on evolution smack up against the
swamp-grandparents problem. Suppose you
find out that your grandparents were
formed from particles from the swamp that
came together by chance. So, as it happens,
you don't have any evolutionary history to
speak of. If evolutionary accounts of the tel-
eological underpinnings of content are
right. your states don't have any content. A
theory with such a consequence should be
rejected.
CAUSATION
Functionalism dictates that mental proper-
ties are second-order properties that consist
in having other properties that have certain
relations to one another. But there is at
least a prima Jacie problem about how such
second-order properties could be causal and
explanatory in a way appropriate to the
mental. Consider, for example, provocative-
ness, the second-order property that consists
in having some first-order property (say
redness) that causes bulls to be angry. The
cape's redness provokes the bull, but does
the cape's provocativeness provoke the bull?
The cape's provocativeness might provoke
an animal protection society, but isn't the
bull too stupid to be provoked by it? (See
Block, 1990.)
Functionalism continues to be a lively
and fluid point of view. Positive develop-
ments in recent years include enhanced
prospects for conceptual functionalism and
the articulation of the teleological pOint of
view. Critical developments include prob-
lems with causality and holism, and con-
tinuing controversy over chauvinism and
liberalism.
BIBLIOGRAPHY
Beakley, B., and Ludlow, P., eds. 1992. Philo-
sophy oj Mind: Classical Problems/Contem-
porary Issues. Cambridge, MA: MIT Press.
Block, N., ed. 1980. Readings in Philosophy oj
Psychology, Vol. 1. Cambridge, MA: Harvard
University Press.
331
FUNCTIONALISM (2)
Block. N. 1990. Can the mind change the
world? In Meaning and method: Essays in
honor of Hilary Putnam. ed. G. Boolos. Cam-
Field. H. 1978. Mental representation. Erkent-
niss. 13. 9-61.
. Fodor. J. 1987. Psychosemantics. Cambridge.
MA.: MIT Press.
Fodor. J .. and Lepore. E. 1992. Holism. Oxford:
Basil Blackwell.
Goldman. A. 1993. Readings in Philosophy and
Cognitive Science. Cambridge. MA.: MIT
Press.
Hill. C.S. 1991. Sensations. Cambridge Uni-
versity Press.
Kim. J. 1992. Multiple Realization and the
metaphysics of reduction. Philosophy and
Phenomenological Research. LII: 1.
Loar. B. 1981. Mind and Meaning. Cambridge
University Press.
Loar. B. 1990. Phenomenal states. In Philo-
sophical perspectives 4: Action Theory and
332
Philosophy of Mind. ed. J. Tomberlin. Atasca-
dero: Ridgeview.
Lycan. W.G. 1987. Consciousness. Cambridge.
MA.: MIT Press.
Lycan. W.G.. ed. 1990. Mind and Cognition.
Rosenthal. D .. ed. 1991. The Nature of Mind.
Schiffer. S. 1987. Remnants of Meaning. Cam-
Shoemaker. S. 1984. Identity. Cause and Mind.
Ithaca: Cornell.
Stalnaker. R.C. 1984. Inquiry Cambridge. MA.:
MIT Press.
White. S.1. 1986. Curse of the qualia. Synth-
ese. 68. 333-68.
NED BLOCK
H
history: medieval and Renaissance philoso-
phy of mind Since the late 1960s there
has been a significant and fast-expanding
interest in medieval philosophy, and though
once largely confined to questions of logic
and general ontology, the range of this
interest has now extended to cover most
aspects of normative enquiry such as ethics,
politics and aesthetics (see Haldane, 1991,
1992). The philosophy of the Renaissance is
far less widely studied, though in recent
times there have been signs of a developing
interest and no doubt in the coming years
there will be an expansion and intensifica-
tion of this. One may doubt, however,
whether the renaissance is as likely as the
medieval period to catch and hold the inter-
ests of philosophers, as distinct from cul-
tural historians, for while the renaissance
produced striking innovations in the style of
speculative writing and saw the emergence
of secular humanism it had relatively little
to add to the philosophical systems devel-
oped in the middle ages. These systems were
themselves related to earlier ways of think-
ing, in particular to those of Plato and Aris-
totle, but the medievals added much to
what antiquity had produced. Here I shall
mention authors of the early middle ages
through whose work the ideas of antiquity
were communicated to later periods, and
writers of the early modern period who
paved the way for Descartes; but I will focus
upon the major figures of the high middle
ages.
The expression 'philosophy of mind' is a
modern one. In antiquity and in the medi-
eval and renaissance periods writers dis-
cussed questions that are of central interest
to present-day philosophers, such as the
relation of persons to their bodies, the struc-
ture of intentional ACTION and the nature
of IMAGINATION, MEMORY and mental
reference, but these were taken to be
aspects of psychology in an older sense of
the term, connoting the description (logos)
of the soul (psuche). The word 'psychology'
itself probably first appears in the sixteenth
century in writings by two German authors
- Johannes Freigius and Goclenius of
Marburg - published in 1575 and 1590,
respectively. Like their medieval pre-
decessors these authors considered the
study of the soul to be part of natural philo-
sophy and regarded the classic text on the
subject to be Aristotle's De Anima. Indeed, it
is barely an exaggeration to say that medi-
eval and renaissance philosophy of mind
consists of commentaries and reflections on
that work.
A HISTORICAL OVER VIEW
The writings of Aristotle were received into
the Latin West through two sources. First,
through the work of early translators and
commentators such as Boethius (480-524),
Cassiodorus (490-585) and Isidore of
Seville (570-636). However, these figures
knew very little of Aristotle's authentic
corpus and were themselves neo-Platonists
subscribing to the dualistic world-view
espoused by Plotinus (204-69) and, in
Christianized form, by St Augustine (354-
430). Isidore produced a work entitled De
Anima but it was Augustinian, not Aris-
totelian, in inspiration. In it he argued for
the immateriality and immortality of indi-
vidual souls on the basis of the non-empiri-
cal nature of the objects of thought, i.e.
transcendental essences and truths, and the
ancient principle that activities and their
objects are of like kind. From this period,
however, two important and long-lasting
333
HISTOR Y: MEDIEVAL AND RENAISSANCE PHILOSOPHY OF MIND
definitions emerged: Augustine's account of
the human soul as 'a rational substance
suited for ruling a body' (De Quantitate
Animae, 13) and Boethius' definition of a
person as 'an individual substance of
rational nature' (Contra Eutychen, 6). Both
are dualistic in intent. though while Boe-
thius' general account is worded so as to
allow for unembodied persons its formula-
tion permitted later non-dualists and even
materialists to claim that their views were
in accord with ancient tradition (see
DUALISM). Indeed, this definition would be
acceptable to a present-day philosopher
such as DAVIDSON who advocates 'ANOM-
ALOUS MONISM', a version of non-reductive
PHYSICALISM.
The second and more fruitful source of
Aristotelian writings was the Arab world. In
the century prior to Boethius, Syrian Chris-
tians translated ancient Greek texts into
Syriac, and several centuries later their suc-
cessors arrived in Baghdad where they
began to produce Arabic translations of
these works. In due course Moslem philoso-
phers became interested in these writings,
making commentaries on them and appro-
priating some of their ideas. The most
important of these Islamic figures were
Alfarabi (890-950), Avicenna (980-1037)
and Averroes (1126-98) - the last of
whom was to be a source of major disputes,
in both the medieval and renaissance
periods, about the human soul. During the
twelfth and thirteenth centuries, editions of
Aristotle in Greek and Arabic, and support-
ing commentaries, began to appear in the
West where they were rapidly translated
into Latin. A mark of the significance of
their reception is that St Thomas Aquinas
(1225-74), undisputedly the greatest medi-
eval philosopher, seems to have composed
his discussions of human nature and the
soul (Summa Theologiae, 1. 75-89; Quaes-
tiones Disputatae De Anima) and commentary
on Aristotle's text (Sententia super De Anima)
within a year or so of his Dominican associ-
ate William of Moerbeke (1215-86) having
translated Aristotle into Latin in about
1268.
Such developments in psychology as were
334
initiated during the renaissance were also
coincident with the production of new
translations of the De Anima. In the fifteenth
century philosophers were still using Moer-
beke's translation together with editions of
Greek, Arab and Latin commentaries; but as
the new humanism developed, inspired by
the antiquarian rediscovery of the classical
ages, writers became increasingly critical of
what they regarded as 'corrupt' medieval
Latin texts and they produced new Greek
and 'refined' Latin editions in Ciceronian
style. As important as the text of De Anima
itself were the early Greek commentaries,
for they provided secular authors with pre-
scholastic, pre-Arabic perspectives on the
Aristotelian idea of the soul. Furthermore,
the rediscovery of antiquity brought Pla-
tonic and neo-Platonist views back onto the
scene. During the fifteenth and early six-
teenth centuries psychology was no longer
pursued in a unitary scholastic fashion, and
by about 1550 the tradition of medieval
thought was more or less at an end. There
were scholastic authors writing after that
date - as there are even today. Francisco
Suarez (1548-1617) and John of St
Thomas ( 15 8 9 -1644) both took their
inspiration from Aristotle and Aquinas; but
by that point a new agenda was developing
in psychology as Hobbes (1588-1679) and
Descartes (1596-1650) tried in their very
different ways (materialist and dualist,
respectively) to relate the existence and
nature of the mind, conceived largely in
terms of CONSCIOllSNESS, to the human
body, by then thought of as a machine
composed out of material elements whose
intrinsic nature is geometrical - in effect,
atoms in the void.
MEDIEVAL SCHOLASTICISM
Descartes rejected many aspects of tradi-
tional Aristotelian psychology and intro-
duced new ways of thinking. Most
importantly, perhaps, he replaced the idea
of the soul as a single source of life and
reason with the notion that the latter is an
attribute of a distinct element, viz. the mind.
It is also worth observing, however, that
HISTORY: MEDIEVAL AND RENAISSANCE PHILOSOPHY OF MIND
much of what is regarded as definitive of
Cartesianism is to be found in earlier writ-
ings, in particular in the work of Augustine
and those scholastics whom he influenced
such as St Bonaventure (1217-74). It is
commonly assumed, for example, that Des-
cartes is the author of the proof that even if
one is deceived in believing one has a body,
nonetheless it is certain that one exists as a
thinking thing: 'Cogito ergo sum', 'I think,
therefore I am'. However, as contemporaries
of Descartes, such as Antoine Arnauld
(1612-94) and Andreas Colvius observed,
this argument (or a close relative of it) is
anticipated by Augustine in De Civitate Dei
XI, 26, where he insists 'Si JaIlor sum'
([Even] if I err, I am). Descartes' response to
these observations was either to set them
aside, as in his rejoinder to Arnauld (Reply
to the Fourth Set of Objections, in Haldane
and Ross, 1912), or else to claim, as in his
reply to Colvius, that Augustine was
making a different point (Letter to Colvius,
14 November 1640, in Kenny, 1970).
Similarly, it is common to credit Descartes
with the idea that a thinker is infallible with
respect to his own subjective psychology.
But again this is to be found in Augustine:
'[I]t could not possibly happen that the soul
should think about what is itself in the
same way that it thinks about what is not'
(De Trinitate, X, 10; see also De Vera Religione
39, 73). Equally, while finding dualism
compelling, Augustine is puzzled by the
relationship between the soul and the body
and the manner of their causal interaction
in the same ways as would trouble Des-
cartes. Augustine writes: 'The manner in
which spirits are united to bodies is alto-
gether wonderful and transcends the under-
standing of men (De Civitate Dei, XXI, 10);
and twelve hundred years later Descartes
comments: 'It does not seem to me that the
human mind is capable of conceiving quite
distinctly and at the same time both the dis-
tinction between mind and body, and their
union' (Kenny, 1970, p. 142).
However, the reception of Aristotelianism
in the thirteenth century brought with it a
more naturalistic perspective on the central
philosophical question of psychology; the
ontological nature of human persons (see
THE SELF). Clearly, in this context. N A T-
URALISM is not to be identified with physi-
calism - though in succeeding centuries
quasi-materialist views began to emerge.
For the Latin authors of the high middle
ages, however, materialism was unthink-
able. For one thing they were religious
believers, usually members of monastic reli-
gious orders (Albert the Great (1206-80)
and Aquinas were Dominicans; Bona-
venture, Duns Scotus (1266-1308) and
William of Ockham (1285-1349) were
Franciscans) and were committed to
upholding theological doctrines concerning
the human soul, such as that it is a spiritual
entity specially created by God, which sur-
vives bodily death and may exist for eter-
nity. In addition, many of them believed
that reason alone is capable of demonstrat-
ing the immateriality and immortality of the
soul. (The drift towards philosophical scepti-
cism is marked by Scotus' doubting that the
latter is provable (Opus Oxoniense, 4, 43)
and Ockham's denying that it is (Quodlibet,
r. 12).) There is not space here to consider
these proofs, but given the present-day
interest in the 'aboutness' of thought, it is
worth noting that several of them (such as
that offered by Aquinas in the Summa Theo-
logiae, I, 75) turn upon the nature of
INTENTION ALITY and its difference from
physical relations. (Indeed, the very term
and concept originate in the medieval
notion of esse intentionale, the 'intentional
being' of thoughts.)
While medieval philosophy remained
anti-materialist, the availability of the De
Anima together with sophisticated Arab
commentaries led to a new and important
phase in Western thinking about the nature
of the mind. The two central issues con-
cerned the ontological category oj the soul,
whether it is a particular thing in its own
right (hoc aliquid) or a supervenient quality
of something, namely a living human body
(see SUPERVENIENCE); and the identity oj the
soul, whether each person possesses one or
many souls, and indeed whether there is
but a single transcendent soul which ani-
mates the plurality of human beings. In
335
HISTORY: MEDIEVAL AND RENAISSANCE PHILOSOPHY OF MIND
order to see how these questions arise, it is
necessary to say something about the ideas
of the De Anima and how they differ from
the sorts of dualism espoused by Plato,
Augustine and Descartes,
As was noted, St Augustine thought of
persons as souls using bodies, That is to say
he conceived of human beings as compos-
ites of two distinct elements, the one ani-
mating ('anima' being the Latin equivalent
of 'psuche') and expressing itself through the
other. For Aristotle, by contrast, the soul is
thought of in terms of the general scheme of
form and matter. Matter is that out of which
something is made, form is the organizing
principle which makes it to be the thing it
is. If I cut out a paper circle what then lies
on the desk is, in Aristotelian vocabulary, a
'particular substance'. Although there is but
one thing present, nonetheless two aspects of
it can be identified: its matter (paper) and its
form (circularity). Similarly, Aristotle claims
that the organizing principle of a living
human being, its form, is the soul (De
Anima, II, 1). So here again we can say
there is one thing possessed of two aspects,
its matter and its form, the latter being in
this case a certain kind of animality.
This is an attractive view for several
reasons. One is that it offers the promise of
avoiding some of the problems of dualism
without lapsing into materialist reduction-
ism. Another is that it provides a way of
viewing reality as an ordered system of sub-
stances distinguished in point of various
'hylomorphic' (hyle = matter, morphe =
form) relationships. This is part of what
appealed to the medievals, but it was also to
prove a source of problems and disagree-
ments, and even contributed to the decline
of scholasticism in the renaissance. One set
of problems concerns the compatibility of
the Aristotelian view with the theological
doctrines listed above. If the soul stands to
the body as the shape stands to the paper,
then it seems to make no sense to say that
the soul is a something in itself which can
survive the body. Bonaventure's view is
interesting in this respect since he tries
unsuccessfully to combine the hylomorph-
ism of the De Anima with the dualism of
336
Augustine's De Quantitate Animae. He claims
that a living human being is indeed a com-
posite of form (soul) and matter (flesh): 'The
rational soul is the principle and form of the
human body' (Super Libros Sententiarum,
18); but he insists that the soul is itself a
hylomorphic union, in this case of spiritual
form and spiritual matter. The effort to
combine dualistic and hylomorphic per-
spectives is to some degree present in almost
all medieval writers of the thirteenth
century and is sometimes spoken of as the
'A ugustinian/ Aristotelian synthesis'.
Aquinas, who explicitly criticizes the
notion of spiritual matter, goes further in
the direction of Aristotle and away from
dualism; yet he also claims that the soul,
though the form of the body, is nonetheless
a subsistent, if incomplete, entity (a hoc
aliquid). Although it is difficult to make
sense of, his view merits greater considera-
tion than it has so far received from non-
scholastic philosophers. Such consideration
would do well to begin with the idea that as
a separable subject the soul is no more than
the locus of non-sensory mental acts; it is
not a subject in the post-Cartesian sense of
being a person. As Aquinas has it, 'Anima
mea non est ego' - 'my soul is not l' (Com-
mentarium Super Epistolas Pauli Apostoli (1
Cor.: 15)). On the one hand Aquinas is
appreciative of the difficulties with dualism
and the empirical and philosophical con-
siderations supporting a monistic view of
persons. On the other hand he believes that
persons are essentially psychological beings
and that thought is a non-organic activity.
Like his predecessors he associates the
nature of human beings as living thinking
individuals with their possession of rational
souls, but he takes from Aristotle the idea
that the soul is a form. Hence even if, unlike
other forms, it is a subsistent entity it is
incomplete in itself. In short, a human
person is a psychophysical substance, and a
separable soul is not a person but a minimal
subject.
In working out his views, Aquinas was
steering a course between a number of
opposing positions. Set against Augustinian
dualism was a form of radical Aristotelian-
HISTOR Y: MEDIEVAL AND RENAISSANCE PHILOSOPHY OF MIND
ism known, after the Arab commentator, as
Averroism. The main medieval proponent of
this was Siger of Brabant (1224-82), a
contemporary of Aquinas. The Averroists
were concerned with remarks in De Anima
where Aristotle appears to suggest that the
aspects of the soul that empower human
beings with thought (the active and passive
intellects) are a single universal principle,
that, in effect. there is but one mind dis-
tributed through many bodies. Aquinas
attacks this idea CMonopsychism') in his De
Unitate Intellectus Contra Averroistas, insist-
ing that since reason belongs to the soul
and the soul is the form of the human body
there are as many intellects as there are
human beings.
While the Averroists sought to posit a
single universal soul others argued that
each human being has several souls, i.e.
several organizing forms: the vegetative, the
sentient and the rational. Versions of this
view were held by the Jewish philosopher
Avicebron (1020-70) and later by Ockham
(Quodlibet 2). Again Aquinas argues against
it on the grounds that it is at odds with
what common sense and philosophy teach,
i.e. that each human being is a unity,
neither an instance of a universal intellect
nor a collection of individuals. Relatedly he
takes issue with those who view the
rational soul as something added to an
existing living being, and with those who
believe that it emerges naturally out of
matter having been transmitted through
sexual reproduction. The former position
might be termed 'dualist creationism', the
latter is known as 'traducianism'. What
Aquinas himself offers is a version of cre-
ationism - rational souls are not naturally
generated - but one in which of the act of
creation is the imposition by God of a single
soul possessed of vegetative, sentient and
rational powers upon embryonic matter
(Summa Contra Gentiles, II, 86).
RENAISSANCE HUMANISM
One may wonder about the philosophical
stability of Aquinas's moderate Aristotelian-
ism. Certainly in subsequent centuries
writers tended to move away from it in the
directions of either less, or more naturalistic
views. In the renaissance period the former
tendency was associated with the revival of
neo-Platonism and in particular with the
main figures of the Florentine Academy
such as Marsilius Ficinus (1433-99) who
translated Plato and Plotinus and whose
philosophical psychology involves a deliber-
ate return to the Christian dualism of St
Augustine (Theologia Platonica).
During the lifetime of Ficinus, Aristotle's
De Anima was re-translated at least twice
and this prompted renewed interest in the
views of pre-medieval commentators with
the result that there was a revival of A ver-
roism, particularly at the University of
Padua, and of Alexandrism. The latter view
takes its name from Alexander of Aphro-
disias (f 200) a commentator whose
account of De Anima is free of the dualistic
aspects associated with the Augustinian!
Aristotelian synthesis. The main renais-
sance Alexandrist was Pietro Pomponazzi
(1462-1525) who argued for a position not
unlike that of some present-day property
dualists. For while he accepted that thought
is a non-physical process he denied that this
implied that the subject of thought is any-
thing other than a living and mortal
human being (De Immortalitate Animae).
This view put him at odds with the declara-
tions of the Church which, during the Fifth
Lateran Council (1513), denounced Averro-
ism and Alexandrism and proclaimed that
individual immortality is philosophically
demonstrable - just the issue which in the
medieval period had divided Aquinas,
Scotus and Ockham.
The renaissance trend towards nat-
uralism led in due course to the kind of
empirical approach that laid the basis both
for materialism and for Descartes' separa-
tion of the philosophy of mind from the
natural science of living bodies. Equally,
though, as was seen earlier, one can view
Cartesianism as a revival of the Platonic-
Augustinean tradition. Either way it
emerges out of a rich and complex past, the
study of which promises to yield historical
and philosophical insights.
337
HISTORY: PHILOSOPHY OF MIND IN THE SEVENTEENTH AND EIGHTEENTH CENTURIES
BIBLIOGRAPHY
Adams, M. 1987. William of Ockham. Notre
Dame, IN: University of Notre Dame Press.
Cassirer, E., Kristeller, P., and Randall Jr., ).
eds. 1948. The Renaissance Philosophy of
Man. University of Chicago Press.
Copleston, F.C. 1972. A History of Medieval
Philosophy. London: Methuen.
Haldane, E., and Ross, G., eds. 1912. The Phi-
losophical Works of Descartes, Vol II. Cam-
Haldane, J. 1991. Medieval and renaissance
ethics. In A Companion to Ethics, ed. P.
Singer. Oxford: Basil Blackwell.
Haldane, J. 1992. Medieval and renaissance
aesthetics. In A Companion to Aesthetics, ed.
D. Cooper. Oxford: Basil Blackwell.
Hyman, A., and Walsh, J., eds. 1973. Philoso-
phy in the Middle Ages. Indianapolis: Hackett
Publishing Company.
Kenny, A., ed. 1970: Descartes: Philosophical
Letters. Oxford: Clarendon Press.
-- 1980. Aquinas. Oxford University Press.
-- 1993. Aquinas on Mind. London: Rout-
ledge.
Kirwan C. 1989. Augustine. London: Routledge
& Kegan Paul.
Kretzman, N., Kenny, A., and Pinborg, J., eds.
1982. The Cambridge History of Later Medi-
eval Philosophy, Cambridge University Press.
Schmitt, C., and Skinner, Q. eds. 1988: The
Cambridge History of Renaissance Philosophy.
JOHN HALDANE
history: philosophy of mind in the seven-
teenth and eighteenth centuries There was
no such field of philosophical enquiry as
the philosophy of mind in the seventeenth
and eighteenth centuries. Nevertheless, the
problems that are today regarded as the
central problems in the philosophy of mind
were first given something like their current
formulations in the works of the major
seventeenth- and eighteenth-century philo-
sophers, as were the basic positions to be
taken on these issues and some of the famil-
iar lines of argument. These problems and
the various leading responses to them came
up chiefly as side issues incidental to larger
controversies which have long since receded
338
from active philosophical interest. What
will try to do here, accordingly, in giving a
brief sketch of the treatment of some key
issues by philosophers in the period, is to
sketch also the larger context of the philoso-
phers concern with these issues; this often
changes our appreciation of the philoso-
pher's arguments and their purport. The
goal is not simply to point up anachronism
in the common pictures of these positions
and arguments, which are so frequently
invoked in contemporary discussions in
philosophy of mind, but rather to deepen
our appreciation of the issues themselves. It
is worth remarking, though, that there is
more than a little bit of anachronism to be
found in the understanding many con-
temporary writers have of positions and
arguments drawn from the early Modern
period, and that anachronism is not always
entirely benign.
As I've already noted, many of the issues
we take to be the central issues in philoso-
phy of mind were first formulated in the
seventeenth and eighteenth centuries.
Among these are: the so-called mind-body
problem, that is, the problem of the ontolo-
gical status of the mind and/or mental or
psychological states vis-a-vis the body and/
or the physical states of material systems;
the nature of personal identity; the nature
of mental representation (incarnated in our
period as the theory of ideas), and the
closely related question of whether there are
any innate ideas; the nature and content of
the emotions or passions; and others. In this
short essay I will give a brief overview of
the treatment of two large issues which are
a central part of the legacy left to con-
temporary philosophy of mind by the philo-
sophers of the seventeenth and eighteenth
centuries: the mind-body problem, and the
problem of personal identity.
THE MIND-BODY PROBLEM
Dualism
The mind-body problem, as we think of it
today, originated with Descartes. In his
extremely influential (to put it mildly) 1641
work, Meditations on First Philosophy,
HISTOR Y: PHILOSOPHY OF MIND IN THE SEVENTEENTH AND EIGHTEENTH CENTURIES
Descartes argued for a view that has been
known ever since as Cartesian dualism: the
view that the mind and the body are dis-
tinct substances, each of which could exist
apart from the other. (We will arrive at a
somewhat more precise formulation of the
view a little later on.) The argument he
gives for this thesis has several stages:
First, Descartes gives a battery of sceptical
arguments that call into doubt most if not
all of his preconceived beliefs and notions;
chief among these sceptical arguments are
the argument from dreaming, which calls
into question not only all of his particular
beliefs about external material objects, but
even the global belief that there are any
external material objects at all, and the evil
genius argument which calls into question
even those beliefs that are most certain,
such as that 2 + 3 = 5 or that a square has
four sides. (These arguments take up most
of the first of the six meditations.)
Next, at the beginning of the Second
Meditation Descartes delivers (almost) what
is one of the most famous philosophical
dicta of all time: Cogito ergo sum; 1 think,
therefore 1 am. (The classical formulation is
actually given in the Discourse on the
Method.) This is Descartes's Archimedean
point of certainty, which is to provide the
foundation for shoring up the whole edifice
of scientific knowledge, and it is crucially
important for understanding his later argu-
ment for dualism to understand how it
works. The basic idea is that even if we
accept the most radical sceptical thesis, the
evil genius hypothesis, which holds that we
are being deceived by an omnipotent being
who thus has the power to make us go
wrong even in those beliefs we hold most
firmly to be true, and think we know best,
we are committed by the acceptance of the
hypothesis itself to hold that it is certain
that we exist. For if the hypothesis is
correct, we are deceived by the evil genius;
but (a) it is impossible for there to be deceit
unless there is belief (false belief, but belief
nonetheless), and since belief is a mode of
thinking, there must be thought or think-
ing; and (b) since thought or thinking
cannot exist except in a thinking thing or
substance, there must be a thinking thing
or substance. Thus even the most radical
sceptical arguments possible cannot call
into question the belief that 1 exist as a
thinking thing; indeed, on the basis of these
arguments we can show that belief to be
certainly true.
The next stage, which takes up most of
the rest of the Second Meditation, has two
parts. First, on the basis of a lengthy analy-
sis of the Cogito argument Descartes estab-
lishes that the mind, i.e. what 'I' refers to as
it occurs in 'I think' or 'I am', must be (a) a
thinking thing, i.e. a thing which has,
among its states, states of thought or con-
sciousness; and (b) it must be essentially a
thinking thing, that is to say, it cannot have
any states which are not states of thought
or consciousness. The reason for (a) is that
the premiss of the Cogito argument, 'I
think' (or 'I am thinking') cannot be true
unless there currently exists a thought (and
as we have seen, thoughts cannot exist
unless they are thought by some thinker);
and the reason for (b) is that, since at this
point in the Meditations the dreaming argu-
ment is still in force, so that the meditator
must doubt that any external material
objects exist, to attribute any material or
physical states (which are the only kind of
states there are besides states of thought or
consciousness) to the thing referred to by 'I'
would be to consider it a material thing,
and thus call its existence into question,
thus losing the certainty of the knowledge
that 1 exist. It is on this basis that Descartes
concludes that he has a clear and distinct
conception of himself as res cogitans as a
thing that thinks and whose whole essence
it is to be a thinking thing. The second part
of this stage of the argument establishes the
complementary thesis about the conception
of body: the notorious wax example is
meant to show that body can be clearly and
distinctly conceived as a substance whose
whole essence is extension, i.e. having
spatial dimension or any of the qualities or
states (such as being in motion, or resis-
tance to being put into motion by contact
action, etc.) which presuppose the having of
spatial dimension. So by the end of the
339
HIS TOR Y: PHILOSOPHY OF MIND IN THE SEVENTEENTH AND EIGHTEENTH CENTURIES
Second Meditation Descartes has shown
that the mind (what T refers to) is clearly
and distinctly conceived as a substance
which has no states except states of
thought, and the body is clearly and dis-
tinctly conceived as a substance which has
no states except states of extension.
In the closing stage of the argument,
given in Meditation VI, Descartes goes from
conceivability to (real) possibility. Having
argued in Meditations III, IV and V that God
exists and is no deceiver, and thus having
now a rational basis for rejecting the evil
genius hypothesis, Descartes takes himself
to have established that whatever we
clearly and distinctly conceive to be true
must be as we conceive it. We are now in a
position to say not merely that we conceive
the mind to be a substance whose essence is
thought, and body to be a substance whose
essence is extension, but we can positively
assert that those are their natures. It follows
in turn from this that it is possible for the
mind to exist even if no bodies or states of
bodies exist, and for the body to exist even if
no minds or states of minds exist. Mind and
body are thus really distinct substances.
I've set out Descartes' argument at some
length not only because it is the argument
that initiates the modern philosophers' (and
our) preoccupation with the mind-body
problem but also because it exemplifies the
gulf between the seventeenth-century take
on these issues and our own. In the first
place, and on the grossest level, it should be
evident how inappropriate it is to character-
ize Cartesian dualism as consisting in the
claim that there is some mysterious or
incomprehensible 'soul-stuff' or 'mind-stuff'
that is the substance in which mental states
inhere. For Descartes, the substance of the
mind is eminently intelligible, since the
whole essence of the mind is captured in the
attribute of thought, and we have a clear
and distinct idea of this attribute. (There is,
of course, room to question how well Des-
cartes could do by way of providing a full
explication of this supposedly clear idea -
Locke, for one, argues that the notion of
thought is not at all clear - but right now
the issue has to do with Descartes's thesis as
340
he put it forward, not with how well he
could defend it.) If Cartesian dualism is in
this respect less silly a view than it is some-
times made out to be, in other respects it is
less well off than might be gathered from
current discussions. As the reconstruction
of the argument given above makes clear,
an ineliminable element of Descartes' a
priori argument for dualism is its invocation
of an omnipotent God to guarantee the
truth of our clear and distinct perceptions.
Without this, Descartes' argument is
exposed to the charge that at most it can
show that we must think of the mind as a
substance that is distinct from the body,
given our conception of the mind; but this
doesn't reach the question of what the
mind's nature actually is. Descartes' argu-
ment for the ontological conclusion thus
becomes an ignoratio elenchus. Even more
disturbing to the modern sensibility is the
absolutely ineliminable use Descartes makes
of the traditional Aristotelian-Scholastic
categories of substance and mode or acci-
dent. A crucial step in his argument is the
inference from the existence of a thought or
belief to the existence of a substance think-
ing that thought, or in the technical lingo,
in which that thought inheres; and equally
crucial is his analysis of the attributes of
thought and extension, which turn out to
be essential or defining principal attributes
sufficing for the existence and individuation
of an individual substance. Once these com-
mitments are given up, it is hard to see that
anything is left of the theory; yet modern
writers blithely invoke Cartesian dualism as
at least a possible position on the mind-
body problem, without troubling to clean
up these problems (if indeed they can be
cleaned up). (See DUALISM.)
I want now to connect up this discussion
of Descartes' arguments with the general
point with which I began, namely that
whatever it was Descartes took himself to be
doing when he argued for his dualism, he
did not take himself to be carving out a
position in the philosophy of mind. Why
then did he argue so prominently for
dualism? The most obvious suggestion is
that he might really be interested in laying
the basis for a proof of the immortality of
the soul. just as he says he is. But the basic
motivation for dualism in his case stems
from the basic goal of his philosophical
enterprise: to lay the foundations for the
new mechanistic philosophy of nature
demanded by the new science of the seven-
teenth century. According to the New Phi-
losophy, as it was called, all changes of state
of physical bodies are explicable by advert-
ing only to the mechanical affections of
material things: the sizes, shapes, and
motions of their parts. (This is the payoff of
the Cartesian claim that the essence of body
is extension.) Crucial to the enterprise of
establishing the New Philosophy is dis-
establishing the old one, viz. Aristotelian
Scholasticism, which was committed to
immaterial principles of organization and
causation in bodies-forms. Descartes'
radical break with traditional views of the
soul is thus really part of his attack on the
Scholastic doctrine of substantial forms. The
Scholastics of Descartes' time maintained
the Aristotelian - Thomist view of the soul.
according to which the rational soul is, like
the souls of plants and animals, the form of
the living creature. ('Form' here is used in
the technical sense it had in the context of
hylomorphic theory, where the form is the
essence or, as the lingo had it, the first
actuality of the individual substance, which
organizes the matter of the individual into a
living thing which has the essential char-
acteristics and functions typical of its
species; form, in short, 'makes a thing to be
what it is'.) Two crucial points about Aris-
totelian forms, are (1) that these forms are
themselves immaterial principles of unity
and causal action, and (2) they are inti-
mately bound up with formal causes, i.e.
with teleological causation in terms of final
principles. Descartes, who has claimed to
have established that the mind can have no
connection whatever with any states of
extension (or it would have to have some
such states itself, contrary to the conception
of mind yielded by the analysis of the Cogito
argument, must reject the traditional con-
ception of the rational soul as being both
the proper subject of states of THOUGHT or
CONSCIOUSNESS and the form of the living
human body; and reject it he does. (This
was one of the bases for the formal con-
demnation of Descartes' works.)
If the doctrine of the soul as the form of
the living body, which was the central case
for the doctrine of substantial forms, were to
be discarded, the hylomorphic theory as a
whole would go as well, paving the way for
the acceptance of the Mechanical Philoso-
phy. A further benefit in this regard is that
by getting rid of the soul in favour of the
Cartesian mind, a substance not only onto-
logically distinct from body but also causally
independent of it in its basic functions and
operations (those, namely, having to do
with pure intellect), Descartes delivers a
version of the Mechanical Philosophy free of
the suspicion of materialism and atheism
that attached to mechanistic atomism by
virtue of its origin in the works of ancient
atomists such as Democritus and Lucretius.
Pre-established Harmony
If we survey some of the leading positions on
the mind-body problem held by Descartes'
successors we'll find a similar degree of obli-
quity to the current problem-situation in
philosophy of mind as we found in Des-
cartes, although the obliquities lie in differ-
ent directions. Leibniz is a good case in
point. His official stand on the mind-body
problem stems from the doctrine he called
the 'pre-established harmony.' The common
view amongst philosophers of mind is that
this view is basically just Cartesian dualism
without the causal interaction Descartes
allowed between mind and body. The regu-
larities that Descartes put down to causal
interaction are produced instead by God's
having set up the events in the (separate,
and causally isolated) mental and physical
orders in synchronization; and the motiva-
tion for this view is widely supposed to have
been Leibniz's recognition of the problems
with conservation laws that attend dualistic
interaction. Some of this is right: Leibniz was
of course aware of the problems with con-
servation, and he did hold that the regu-
larities between the order of mental events
and the order of physical events was due to
341
God's having set up correspondences
between the orders. But there's at least as
much that is wrong in this picture of Leib-
niz's view as there is that is right. Leibniz is
not a dualist: for him there are no physical
substances, as such. The ultimate sub-
stances are monads, which are individual
substances that have as their states only
perceptions and appetitions; physical objects
are just appearances. The basis of Leibniz's
argument for this view, as well as for his
denial that there is causal interaction
between mind and body (the latter now con-
ceived in terms of a phenomenalist reduc-
tionism, which makes bodies onto logically
dependent upon perceptions or more exactly
on the contents of perceptions) lies in his
analysis of the notion of substance. Leibniz
does not, as Descartes did, accept virtually
whole and unrevised the Aristotelian-
Scholastic categories of substance and acci-
dent; he thinks the notion of substance to
be in need of explication, and he provides
this explication by way of a concept-
containment theory of truth and a related
theory of individual concepts. Leibniz
derives a number of metaphysical results
from this analysis, most notably his doc-
trines about the nature of necessity and
contingency. the causal isolation of sub-
stances. the unity and simplicity of sub-
stance. the principle of identity of
indiscernibles. and the doctrine of pre-estab-
lished harmony. Thus Leibniz arrives at his
position on the mind-body problem as a
consequence of claims he staked on these
more general issues. and not simply as a
response to the problem of conservation
laws; and in any case his version of pre-
established harmony is not straightfor-
wardly dualistic.
Dual Aspect Theory
Another position on the mind-body
problem tossed around by contemporary
philosophers of mind (the most notable
recent example is Brian O'Shaughnessy in
his work The Will: A Dual Aspect Theory) and
originating in our period is the dual aspect
theory, associated with Spinoza. On this
theory, neither mind nor body is a sub-
342
stance in its own right; instead, each is an
aspect (or mode - for Spinoza, each is an
infinite mode) of a single underlying sub-
stance. Some of the problems associated
with dual aspect theories, such as that the
underlying substance itself either has no
nature (Le. is a bare particular) or has a
mysterious nature, are not problems for
Spinoza, as he identifies the underlying sub-
stance as God. (Of course, to the modern
philosopher of mind this 'solution' to the
problems has all the advantages of theft
over honest toil (to swipe a phrase from
Bertrand Russell).) But here again the
theory is a consequence of more basic com-
mitments stemming from an analysis of the
notion of substance, although one quite dif-
ferent from Leibniz's. Spinoza faults Des-
cartes for not taking seriously (or perhaps
one should say literally) his own definition
of substance. If to be a substance is to be
able to exist on one's own. depending on no
other thing for existence, then Descartes's
res cogitans (mind) and res exstensa (body)
are not substances. since. as Descartes
acknowledged. mind and body depend upon
God for their existence. (Descartes' distinc-
tion between finite and infinite substances is
then treated as just a dodge.) The only
entity that can satisfy the traditional defini-
tion of substance. taken strictly. is God; so
God is the only substance, with mind and
body being just modes of this substance: dif-
ferent ways of expressing the essence of this
substance. From this analysis Spinoza draws
the conclusion that events in the mental
order must parallel events in the physical
order. as they express the same essence of
the same substance. While this conclusion
has some broad structural similarities to
Leibniz's pre-established harmony, there are
important differences: the regularities or
parallelism between the mental and physi-
cal orders are not the result of creating
intentions on God's part, and there is no
hint of the phenomenalist reductionism that
provides the context for Leibniz's doctrine.
Materialism
This view. which had its lease on life
renewed in the seventeenth century, is
perhaps the one position that is most con-
formable to its current incarnations (see
!DENTl TY THEORIES; PHYSICALISM).
Roughly, the doctrine is that all that exists is
matter in motion, and that mental states are
onto logically dependent on states of bodies.
Probably the most forceful and notorious
early expression of materialism in this period
is due to Hobbes. His motivation for em-
bracing materialism stems from his commit-
ment to mechanistic science, but he does not
base an argument for materialism on its sci-
entific credentials. Instead, he argues for
materialism on the basis of his logical
theory, wherein he replaces Aristotle's
central category of substance with the cat-
egory of body, on the grounds that the only
clear idea we can attach to the notion of
something that exists on its own is the idea
of body. Hobbes devotes much energy and
ingenuity to providing detailed accounts of
how sensation, appetition, and even higher-
order cognitive faculties ('reason', 'reason-
ing') can be reduced to material processes.
This stance was more of an embarrassment
to the champions of the new mechanistic
science than it was a help; Boyle, Charleton,
Locke, and other promoters of the New Phi-
losophy in England had to distance them-
selves from the out-and-out materialism of
Hobbes, which seemed to confirm the
common complaint that mechanistic atom-
ism was just the old materialistic, atheistic
philosophy of Democritus and Lucretius in
somewhat misleading new garb.
Locke is a particularly interesting case.
He was accused of being a materialist or at
least a fellow-traveller of materialism by
Leibniz and by many contemporary critics
in England; more recently, he has been seen
as a dualist. He was neither; as usual, he
preferred to remain agnostic in an area
where human reason could make no reli-
able pronouncements. Thus in a famous
passage in the Essay (Book IV, ch 3, 6) he
argued that both doctrinaire materialism
and doctrinaire dualism are misguided,
since our ideas of matter and thinking are
so impoverished. He did go on from this to
argue that we can find no inconsistency in
the suggestion that God superadded the
power of thought to merely material things,
so that we cannot rule it out that a purely
material substance might also be the subject
of mental states, and hence that a thinking
thing may be nothing but body. Once again,
Locke's interest in arguing for the possibility
of thinking systems of matter derived at
least as much from his enterprise of laying
the foundations for (Gassendi-Boyle style)
mechanistic atomism as from any other
motive; he took the superaddition of
thought to matter as a model for God-
ordained lawlike connections between
primary and secondary qualities, thus
securing the possibility of explanatory con-
nections between secondary qualities and
primary qualities even if these connections
are unintelligible to us.
PERSONAL IDENTITY
Locke
Locke's most important legacy to current
philosophers of mind lies in his formulation
and treatment of the problem of personal
identity. The pattern we've turned up in
connection with the mind-body problem
repeats itself here: Locke's treatment of the
issue looks primarily to larger framing ques-
tions which are no longer current, and his
actual position is somewhat different than
the common view has it. First, a brief
description of Locke's theory. Locke distin-
guishes between identity as determined by
unity of substance and identity as deter-
mined by the idea of the kind to which an
individual belongs. (Note that according to
his theory of real and nominal essences, the
idea of the kind, and consequently the kind
itself, is framed by us, or as Locke puts it, it
is 'the workmanship of the understanding'.)
So the identity of a mass or parcel of matter
is determined by unity of substance: as long
as the same atoms are united together, we
have the same parcel of matter. The identity
of an oak-tree or horse, on the other hand,
is determined by the idea of the kind:
numerically distinct (and differently sized
and shaped) parcels of matter may count as
the same oak-tree or the same horse (now a
sapling, later a mature tree; now a colt,
343
later an adult horse) as long as the change
in constituent matter is gradual and a
certain requisite organization of parts is pre-
served. Locke then suggests that the identity
of a person is of the latter sort: it is deter-
mined by our idea of a person, which is,
Locke says, that of 'a thinking intelligent
Being, that has reason and reflection, and
considers itself as itself, the same thinking
thing in different times and places'; and
since consciousness is thus the core defining
quality of persons, sameness of conscious-
ness constitutes personal identity. Locke is
often described as having put forward a
memory theory of personal identity, accord-
ing to which the criterion of personal iden-
tity is the ability to remember having done
a particular action or having had a parti-
cular experience, making one the same
person as the one who did that action or
had that experience. But this is not quite
right; the psychological continuity required
for memory is just an example of the sorts
of continuity which constitute sameness of
consciousness.
Perhaps Locke's most lasting contribution
to the topiC was the method he used in
arguing for his theory. He introduced the
method that is still the main vehicle of
argument on this topic, the method of
'puzzle-cases' (as J. 1. Mackie called it). To
show, for example, that the identity of a
person is not constituted by the identity of a
particular human body, he asks us to
imagine that a prince and a cobbler should
suddenly switch bodies, where the body of
the prince should now give voice to mem-
ories which only the cobbler could have
had, but can remember nothing of the
prince's life, and the same in reverse for the
body of the cobbler. If we can imagine this
as a genuine case of persons switching
bodies, we see that the conditions for being
the same person are distinct from those for
being the same human being. Similarly,
Locke argues against the immaterial sub-
stance theory of personal identity by asking
us to imagine that the soul, or immaterial
substance, that is the subject of our mental
states may in earlier times have been that of
Socrates' mental states; as long as we can
344
remember nothing of Socrates' experiences,
we are not the same person with Socrates.
On the other side, we can imagine that the
soul or immaterial substance that is the
subject of my states today might be a
numerically distinct soul from the one I had
yesterday (maybe the one I have today is
the one you had yesterday), but that the
memories had by the soul I had yesterday
were passed on to its successor; in spite of
the switch, I would still be the same person,
or so Locke thinks we would judge. Locke
concludes that the identity of a person does
not consist in the identity of a human being
(where that identity would be constituted
by sameness of a living human body), nor
in the identity of a soul or immaterial sub-
stance; as Locke's discussion of these and
many other examples is meant to show, our
intuitive judgments about when we have
the same person and when we don't suggest
an implicit commitment to the view that
identity of person is constituted by sameness
of consciousness and not by unity of sub-
stance, whether that substance is material
or immaterial.
As before, Locke's theory of personal
identity is motivated by larger issues. One
set of issues is theological; there was much
concern in seventeenth-century England
with soteriology, and especially with the
doctrine of the resurrection of the dead.
Locke's theory makes consciousness the
bearer of moral responsibility, so that one's
moral accountability extends to any action
to which one's consciousness extends,
whether or not one is in the same body or
has the same soul as when one did the
action. This theory thus cuts through all
the problems about which body it is that is
resurrected - the body as it was at death? as
it was in the prime of life? All particles of
matter that were ever a part of it? And the
toughest problem of all - what happens to
cannibals and their victims? Locke can say
that God simply attaches the consciousness,
with its awareness of past actions, to any
body, fashioned from any matter, and that
body thereby becomes the person's body (so
that the person can be said to live again).
Probably the major goal of Locke's dis-
cussion of personal identity is, however, to
establish that we have no need of souls,
considered as substantial forms of the living
human animal, to provide for the identity of
persons. Locke was at least as much inter-
ested as was Descartes in jettisoning the
Scholastic doctrine of forms and paving the
way for the mechanistic philosophy of
nature. Accordingly, he presents the discus-
sion of personal identity as an illustration of
his general thesis about identity: except in
those very limited cases in which identity is
determined by unity of substance, identity is
determined by the idea of the kind to which
the individual belongs. So plants, animals,
and persons, which were the Scholastics'
prime examples of things whose souls, or
substantial forms, determined their identity,
are now seen to have their identity deter-
mined by our ideas of the kinds. There is no
reason, on this account, to postulate sub-
stantial forms to account for the identity of
persons, or of any other natural object.
Hume
Hume's treatment of personal identity, like
Locke's, is presented as an illustration of a
general theory of identity. But where Locke
took himself to be explicating something
real - the identity of individual things,
persons - Hume instead is exposing the
operations of mind that produce the fiction
of a person as an individual thing. The
famous 'bundle theory' of personal identity
is thus not a theory setting out the condi-
tions under which persons have their iden-
tity as real individual things, but instead
provides an account of the mistakes of
thought that make it seem as if there were
such things. Hume calls on his general
principles of association of ideas to account
for the mind's propensity confusedly to tele-
scope resembling perceptions into a (fic-
tional) numerically identical perception, and
thus mistake a train of distinct but highly
resembling perceptions for a single enduring
one. It is this fictional identity that gives rise
to the conviction of real personal identity.
(It bears noting that Hume himself, in the
Appendix to the Treatise of Human Nature,
expressed deep reservations about his
account, and did not return to the topic in
his later reworking of the Treatise, the
Inquiry concerning Human Understanding.)
Kant
The last great philosopher of our period was
Kant; and in his work both of the large
issues we have traced, the mind-body
problem and the problem of personal iden-
tity, come together. Kant's most direct and
extended treatment of these issues is con-
tained in the charmingly titled section 'The
Paralogisms of Pure Reason' in the Critique
of Pure Reason. This section, like the other
sections in this part of the Critique (the
Transcendental Dialectic), is designed to
show that human reason will inevitably go
wrong if it tries to come to demonstrative
certainty regarding the metaphysical status
of the soul. The four paralogisms - concern-
ing the soul as substance, the simplicity of
the soul, the soul as personality (Le. as the
metaphysical basis of personal identity) and
the soul as related to external objects via
perception - all involve the same mistake,
that of taking a merely formal unity for a
property of empirical things. Kant's analysis
is quite complex, but it can be sketched as
follows. The basic principle that 'It must be
possible for the "I think" to accompany all
of my representations' (CPR B 131) cor-
rectly sets out the condition for being a
thinking thing, but it does so only in a
wholly general way: all thinking things as
such must satisfy the condition, so it cannot
provide for the individuation of anyone
thinking thing as opposed to any or all
others. For that, we need to consider the
particular causal relations that tie percep-
tions and other states of consciousness to
particular material objects in space and time
- in short, we need to consider the SELF
(what 'I' refers to) as an empirical self. But
then we find that the metaphysical attri-
butes of the self (substancehood, simplicity,
personality) that we derived from the pure
concept of a self (defined only by the formal
'I think') cannot with any confidence be
asserted of the empirical self, because the
empirical emptiness of the pure concept was
exploited in the derivations. Any attempt to
345
establish a priori metaphysical conclusions
about the self must therefore come to
naught. As a result of Kant's work, issues in
the philosophy of mind became almost inex-
tricably connected with general epistemolo-
gical issues, and it is only relatively recently
that they have been pulled apart.
This brief overview of the high points of
the seventeenth- and eighteenth-century
treatments of these issues has, I hope, sug-
gested two things: first, that there might be
some value to contemporary philosophers of
mind in looking more closely than is
perhaps usually done at the historical roots
of the problems that they deal with; and
second, that if one does look back in this
way, it is important to pay attention to the
context of the discussion; otherwise we may
just be projecting current preoccupations,
not to say prejudices, back in time.
BIBLIOGRAPHY
Primary sources
Rene Descartes, The Philosophical Writings of
Descartes, Vols. 1 and 2, trans. J. Cotting-
ham, R. Stoothoff and D. Murdoch. Cam-
bridge University Press, 1985.
David Hume, A Treatise of Human Nature, ed.
L. A. Selby-Bigge, rev. P. H. Nidditch.
Oxford. Clarendon Press, 1978.
Immanuel Kant, Critique of Pure Reason, trans.
Norman Kemp Smith. London: Macmillan &
Co, Ltd., 1929.
Gottfried Wilhelm Leibniz, Leibniz: Philosophical
Essays, trans. R. Ariew and D. Garber.
Indianapolis, Indiana, and Cambridge, MA:
Hackett Publishing Company, 1989.
John Locke, An Essay Concerning Human
Understanding, ed. P. H. Nidditch. Oxford
University Press, 1975.
Baruch Benedictus Spinoza, Ethics I and Des-
cartes's PrinCiples of Philosophy, in The Col-
lected Works of Spinoza, ed. and trans. E. M.
Curley. Princeton University Press. 1985.
Secondary sources
On Descartes
Curley, E. 1978. Descartes against the Skeptics.
346
Mathews, G. 1992. Thought's Ego: Augustine
and Descartes. Ithaca: Cornell University
Press.
Williams, B. 1978. Descartes: The Project of
Pure Enquiry. Atlantic Highlands, N.J.
Humanities Press.
Wilson, M. 1978. Descartes. London: Rout-
ledge & Kegan Paul.
On Leibniz
Adams, R.M. 1983. Phenomenalism and cor-
poreal substance in Leibniz. In Midwest
Studies in Philosophy, Vol. 8 ed. P. French, T.
Uehling and H. Wettstein. Minneapolis: Uni-
versity of Minnesota Press.
Mates, B. 1986. The Philosophy of Leibniz.
Sleigh, R.C. 1990. Leibniz and Arnauld: A Com-
mentary on their Correspondence. New Haven:
Yale University Press.
On Spinoza
Bennett J. 1984. A Study of Spinoza's Ethics.
Indianapolis: Hackett Publishing Co.
Curley, E. 1969. Spinoza's Metaphysics: An
Essay in Interpretation. Cambridge, MA.:
Curley, E. 1988. Behind the Geometrical
Method: A Reading of Spinoza's Ethics. Prin-
ceton University Press.
On Locke
Atherton, M. 1983. Locke's theory of personal
identity. In Midwest Studies in Philosophy,
Vol. 8, ed. P. French, T. Uehling, and H.
Wettstein. Minneapolis: University of Min-
nesota Press.
Ayers, M. 1991. Locke, 2 vols. London and
New York: Routledge & Kegan Paul.
Winkler K. 1991. Locke on personal identity.
Journal of the History of Philosophy, 29,
201-26.
On Hume
Flage, D.E. 1990. David Hume's Theory of Mind.
London and New York: Routledge & Kegan
Paul.
Stroud, B. 1977. Hume London and Boston:
On Kant
Allison, H. 1983. Kant's Transcendental Ideal-
ism. New Haven: Yale University Press.
Ameriks, K. 1982. Kant's Theory of Mind: An
Analysis of the Paralogisms of Pure Reason.
Oxford: Clarendon Press.
Kitcher, P. 1990. Kant's Transcendental Psy-
chology. New York: Oxford University Press.
General
Buchdahl, G. 1969. Metaphysics and the Philo-
sophy of Science. Cambridge, MA.: MIT Press.
Loeb, L. 1981. From Descartes to Hume: Con-
tinental Metaphysics and the Development of
Modern Philosophy. Ithaca: Cornell Uni-
versity Press.
Yolton, J. 1983. Thinking Matter: Materialism
in Eighteenth Century Britain. Minneapolis:
University of Minnesota Press.
Yolton, J. 1984. Perceptual Acquaintance from
Descartes to Reid. Minneapolis: University of
Minnesota Press.
EDWIN McCANN
holism Holism is the idea that the ele-
ments of a system have significance in
virtue of their interrelations with each
other. Of course, this definition is so general
that it could be relevant to virtually any
subject matter, and this threatens to rob the
notion of interest. For holism to be worth
discussing - for it to have any grip on our
view of some subject matter - we require a
more precise specification of the system and
type of significance that is at issue.
In the philosophy of mind, holism has
been most widely discussed in connection
with both the contents of propositional atti-
tudes and the meanings of linguistic con-
structions. For example, insofar as you
think that the meaning of a given sentence
in a language depends on its inferential or
evidential connections to other sentences in
that language, you are a holist about
meaning. Or, insofar as you think that, in a
given case, the content of someone's belief
depends on the inferential or evidential con-
nections to other beliefs that the person
holds, you are a holist about content. More-
HOLISM
over, even these formulations are too im-
precise. Sharpening them would require
that we go into detail about the notions of
'inferential' or 'evidential' connections, and
we would also have to say just how exten-
sive these connections must be. A very
moderate holism about language might
maintain that the meanings of certain
words depends on connections to a fairly
restricted range of other words. Thus, if one
argued that 'there are tigers' as used by a
given speaker could only mean that there
were tigers if that speaker was prepared to
infer from this sentence such things as
'there are animals' and 'there are
mammals', then this holism need not be
seen as having drastic consequences. But if
you insisted that the meaning of 'there are
tigers' depended on the epistemic attitudes
of the speaker (his tendency to voice dissent
or assent) to every other sentence of his
language, then the holism would seem to
have very drastic consequences indeed. For
it suggests that no two speakers could ever'
mean the same thing by some given sen-
tence, since it is almost certainly true that
any two speakers will differ somewhere in
their epistemic attitudes. My thinking that it
is very unlikely that tigers make good pets
and your thinking it less unlikely, would
make the sentence 'there are tigers' mean
something different in my mouth than in
yours.
Most recently, holism about linguistic
meaning has tended to trade under the label
'CONCEPTUAL ROLE SEMANTICS', though
this phrase can also be pressed into service
to describe a holism about propositional
attitude contents.
See also FODOR; FUNCTIONALISM.
SAMUEL GUTTENPLAN
347
identity theories What is the relation
between mind and physical reality? Well-
established schools of thought give starkly
opposing answers to this question. Descartes
insisted that mental phenomena are non-
physical in nature. This view seems inviting
because mental phenomena are indis-
putably different from everything else.
Moreover, it's safe to assume that all
phenomena that aren't mental have some
physical nature. So it may seem that the
best way to explain how the mental differs
from everything else is to hypothesize that
mind is non-physical in nature.
But that hypothesis is not the only way to
explain how mind differs from everything
else. It's also possible that mental phenom-
ena are instead just a special case of phys-
ical phenomena; they would then have
properties that no other physical phenom-
ena have, but would still themselves be
physical. This explanation requires that we
specify what is special about mental phe-
nomena which makes them different from
everything else. But we must specify that in
any case, just in order to understand the
nature of the mental. Characterizing mental
phenomena negatively, simply as not being
physical, does little to help us understand
what it is for something to be mental.
The claim that mental phenomena are a
special kind of physical phenomenon is the
root idea of mind-body materialism (also
called PHYSICALISM). And the version of
that thesis that's been most widely defended
over the last three decades is the identity
theory of mind.
In Descartes' time the issue between
materialists and their opponents was framed
in terms of substances. Materialists such as
Thomas Hobbes and Pierre Gassendi main-
tained that people are physical systems with
348
I
abilities that no other physical systems
have; people, therefore, are special kinds of
physical substance. Descartes' DUALISM, by
contrast, claimed that people consist of two
distinct substances that interact causally: a
physical body and a non-physical unex-
tended substance. The traditional concep-
tion of substance, however, introduces
extraneous issues, which have no bearing
on whether mental phenomena are physical
or non-physical. And in any case, even
those who agree with Descartes that the
mental is non-physical have today given up
the idea that there are non-physical sub-
stances. It's now widely accepted on all
sides that people are physical organisms
with two distinctive kinds of states: physical
states such as standing and walking, and
mental states such as thinking and feeling.
Accordingly, the issue of whether the
mental is physical or non-physical is no
longer cast in terms of whether people, and
other creatures that have the ability to
think and sense, are physical or non-phys-
ical substances. Rather, that question is put
in terms of whether the distinctively mental
states of thinking, sensing, and feeling are
physical states or non-physical states. The
identity theory is the materialist thesis that
every mental state is physical, that is, that
every mental state is identical with some
physical state.
PROPER TIES OF MENTAL STATES
If mental states are identical with physical
states, presumably the relevant physical
states are various sorts of neural states. Our
concepts of mental states such as thinking,
sensing, and feeling are of course different
from our concepts of neural states. of what-
ever sort. But that's no problem for the
identity theory. As J. J. C. Smart (1962),
who first argued for the identity theory,
emphasized, the requisite identities don't
depend on our concepts of mental states or
the meanings of mental terms. For a to be
identical with b, a and b must have exactly
the same properties, but the terms 'a' and
'b' need not mean the same (see LEIBNIZ'S
LAW).
But a problem does seem to arise about
the properties of mental states. Suppose pain
is identical with a certain firing of c-fibres.
Although a particular pain is the very same
state as a neural firing, we identify that
state in two different ways: as a pain and as
a neural firing. The state will therefore have
certain properties in virtue of which we
identify it as a pain and others in virtue of
which we identify it as a neural firing. The
properties in virtue of which we identify it
as a pain will be mental properties, whereas
those in virtue of which we identify it as a
neural firing will be physical properties.
This has seemed to many to lead to a kind
of dualism at the level of the properties of
mental states. Even if we reject a dualism of
substances and take people simply to be
physical organisms, those organisms still
have both mental and physical states. Simi-
larly, even if we identify those mental states
with certain physical states, those states will
nonetheless have both mental and physical
properties. So disallowing dualism with
respect to substances and their states simply
leads to its reappearance at the level of the
properties of those states.
There are two broad categories of mental
property. Mental states such as thoughts
and desires, often called PROPOSITIONAL
ATTITUDES, have CONTENT that can be
described by 'that' clauses. For example, one
can have a thought, or desire, that it will
rain. These states are said to have inten-
tional properties, or INTENTIONALITY. SEN-
SA TIONS, such as pains and sense
impressions, lack intentional content, and
have instead qualitative properties of
various sorts.
The problem just sketched about mental
properties is widely thought to be most
pressing for sensations, since the painful
IDENTITY THEORIES
quality of pains and the red quality of visual
sensations seem to be irretrievably non-
physical. So even if mental states are all
identical with physical states, these states
appear to have properties that aren't physi-
cal. And if mental states do actually have
non-physical properties, the identity of
mental with physical states won't sustain a
thoroughgoing mind-body materialism.
Smart's reply to this challenge is that,
despite initial appearances, the distinctive
properties of sensations are neutral as
between being mental or physical; in the
term Smart borrowed from Gilbert RYLE,
they are topic neutral. My having a sensa-
tion of red consists in my being in a state
that's similar, in respects that we need not
specify, to something that occurs in me
when I'm in the presence of certain stimuli.
Because the respect of similarity isn't speci-
fied, the property is neither distinctively
mental nor distinctively physical. But every-
thing is similar to everything else in some
respect or other. So leaving the respect of
similarity unspecified makes this account
too weak to capture the distinguishing
properties of sensations.
A more sophisticated reply to the diffi-
culty about mental properties is due inde-
pendently to D. M. Armstrong (1968) and
David LEWIS (1972), who argue that for a
state to be a particular sort of intentional
state or sensation is for that state to bear
characteristic causal relations to other par-
ticular occurrences. The properties in virtue
of which we identify states as thoughts or
sensations will still be neutral as between
being mental or physical, since anything
can bear a causal relation to anything else.
But causal connections have a better
chance than similarity in some unspecified
respect of capturing the distinguishing prop-
erties of sensations and thoughts.
This causal theory is appealing (see the
following section). Still, it's misguided to try
to construe the distinctive properties of
mental states as being neutral as between
being mental or physical. To be neutral as
regards being mental or physical is to be
neither distinctively mental nor distinctively
physical. But since thoughts and sensations
349
IDENTITY THEORIES
are distinctively mental states, for a state to
be a thought or a sensation is perforce for it
to have some characteristically mental
property. We inevitably lose the distinctively
mental if we construe these properties as
being neither mental nor physical.
Not only is the topic-neutral construal
misguided; the problem it was designed to
solve is equally so. That problem stemmed
from the idea that the mental must have
some non-physical aspect, if not at the level
of people or their mental states, then at the
level of the distinctively mental properties of
those states (see PROPER TY).
But the idea that the mental is in some
respect non-physical cannot be assumed
without argument. Plainly, the distinctively
mental properties of mental states are
unlike any other properties we know about.
Only mental states have properties that are
at all like the qualitative properties of sensa-
tions. And arguably nothing but mental
states have properties that are anything like
the intentional properties of THOUGHTS and
DESIRES. But this doesn't show that these
mental properties are not physical proper-
ties. Not all physical properties are like the
standard cases; so mental properties might
still be special kinds of physical properties.
Indeed, it's question begging to assume
otherwise. The doctrine that the mental
properties of mental states are non-physical
properties is simply an expression of the
CARTESIAN doctrine that the mental is
automatically non-physical.
To settle whether or not those mental
properties are non-physical, we would need
a positive account of what those properties
are. Proposals are available that would
account for intentional properties wholly in
physical terms (see DENNETT, DRETSKE, and
FODOR), and perhaps one of these will prove
correct. It's been more difficult to give a
positive account of the qualitative properties
of sensations, and that's led some to con-
clude that such properties will inevitably
turn out to be non-physical. But it's plainly
unfounded to infer from the difficulty in
explaining something to its being non-
physical.
It's sometimes held that properties should
350
count as physical properties only if they can
be defined using the terms of physics. This is
far too restrictive. Nobody would hold that
to reduce biology to physics, for example,
we must define all biological properties
using only terms that occur in physics. And
even putting REDUCTION aside, if certain
biological properties couldn't be so defined,
that wouldn't mean that those properties
were in any way non-physical. The sense of
'physical' that's relevant here must be
broad enough to include not only biological
properties, but also most commonsense,
macroscopic properties. Bodily states are
uncontroversially physical in the relevant
way. So we can recast the identity theory as
asserting that mental states are identical
with bodily states.
TYPES AND TOKENS
There are two ways to take the claim that
every mental state is identical with some
bodily state. It might mean identity at the
level of types, that is, that every mental-
state type is identical with some physical-
state type. Such type identity would hold if
all the instances of a particular type of
mental state are also instances of a parti-
cular type of bodily state. This is called the
type identity theory.
But the identity claim might instead
mean only that every instance of a mental
state is identical with an instance of a
bodily state, of some type or other. On this
construal, the various types of mental state
wouldn't have to correspond to types of
bodily state; instances of a single mental
type might be identical with tokens of dis-
tinct bodily types. This weaker claim is
known as the token identity theory.
There's reason to doubt that the type
identity theory is true. It's plausible that
organisms of different species may share at
least some types of mental state - say, pain
- even if their anatomical and physiological
differences are so great that they can't share
the relevant types of bodily state. No single
bodily-state type would then correspond to
these mental-state types. This possibility is
called the multiple realizability of mental
states. It's conceivable, of course, that
biology will someday type physiological
states in a way that corresponds tolerably
well with types of mental state, but we can
have no guarantee that this will happen.
Even if no physiological types correspond
to types of mental state, the causal theory of
Lewis and Armstrong would allow us to
identify types of mental state with types
described in other terms. On Lewis's version
of the theory, mental states are whatever
states occupy the causal roles specified by
all our commonsense psychological plati-
tudes, taken together. The various types of
mental state correspond to the various
causal roles thus specified; mental-state
tokens are of a particular mental type if
they occupy the causal role that defines that
type. These causal roles involve causal ties
to behaviour and stimuli and to other states
that occupy these causal roles. Such a
theory, which defines mental-state types in
terms of causal roles, is often called FUNC-
TIONALISM.
One could imagine that the individual
states that occupy the relevant causal roles
turn out not to be bodily states; for example
they might instead be states of an Cartesian
unextended substance. But it's over-
whelmingly likely that the states that do
occupy those causal roles are all tokens of
bodily-state types. So the causal theory,
together with this empirical likelihood, sus-
tains at least the token identity theory.
Moreover, this version of the causal theory
bypasses the problematic idea that the
mental properties of those states are neutral
as between being mental or physical, since
mental-state types are determined by our
psychological platitudes.
To defend the type identity theory as well,
however, would require showing that all
mental-state tokens that occupy a particular
causal role also fall under a single physio-
logical type. Lewis (1980) expects sub-
stantial uniformity of physiological type
across the tokens of each mental-state type,
at least within particular populations of
creatures. But if tokens of different physiolo-
gical types do occupy the same causal role,
that would undermine the type identity
IDENTITY THEORIES
theory, or at least make it relative to certain
populations.
Multiple realizability is the possibility that
mental-state types are instantiated by states
of distinct physiological types. It's an
empirical matter whether that's actually the
case. If it is, physical-state types don't corre-
spond to mental-state types, and the type
identity theory is false.
But one might, with Hilary PUTNAM
(1975), construe the type identity theory
more strongly, as claiming that the mental
properties that define the various types of
mental state are identical with physical
properties. And that's false even if the
tokens of each mental-state type fall under a
single physiological type; the property of
occupying a particular causal role is plainly
not identical with the property of belonging
to a particular physiological type. On this
construal, no empirical findings are needed
to refute the type identity theory. But it's
more reasonable to construe the type iden-
tity theory less strongly, as requiring the
claim only that all tokens of a particular
mental-state type fall under a single physio-
logical type.
Donald DAVIDSON (1970) has used differ-
ent considerations to argue that mental-
state types correspond to no physiological
types, but that the token identity theory is
nonetheless correct. Plainly, mental and
bodily events cause each other. Moreover,
as Davidson reasonably holds, one event
token can cause another only if that causal
connection instantiates some explanatory
law. But Davidson also insists that an event
token belongs to a particular mental type
only relative to certain background assump-
tions about meaning and RATIONALITY.
Tokens of physical events, by contrast,
belong to whatever physical type they do
independently of any such background
assumptions. Davidson infers that there can
be no strict laws connecting physical and
mental events. But if so, how can mental
and bodily events cause each other? (See
REASONS AND CAUSES.)
Davidson's solution relies on the fact that
explanatory laws describe events in parti-
cular ways and a different description of the
351
IDENTITY THEORIES
same events might not sustain the explana-
tory connection. So the impossibility of laws
connecting mental and physical events
means only that no laws can connect physi-
cal events, described as such, with mental
events, described as such. To interact caus-
ally, events must figure in explanatory laws.
So each mental-event token that interacts
causally with a bodily event can figure in a
law only if that mental-event token can also
be described in purely physical terms. The
considerations that preclude laws connect-
ing mental with physical events presumably
show also that no physical types correspond
to any mental-state types. But since we can
describe every mental-event token in physi-
cal terms, that token will be identical with
some physical-event token. This intriguing
argument is difficult to evaluate, mainly
because it's unclear exactly why back-
ground assumptions about meaning and
rationality should preclude laws connecting
events described in mental terms with those
described physically.
In order for causal interactions between
mental and bodily events to fall under laws
that describe events solely in physical terms,
physically indistinguishable events must be
mentally indistinguishable, though not
necessarily the other way around. That
relationship is known as SUPERVENIENCE;
in this case, mental properties would be said
to supervene on physical properties. Jaegwon
Kim (1984) has usefully explored such
supervenience as a way to capture the rela-
tion between mental and physical.
ELIMINATIVE MATERIALISM
The Cartesian doctrine that the mental is in
some way non-physical is so pervasive that
even advocates of the identity theory have
sometimes accepted it, at least tacitly. The
idea that the mental is non-physical under-
lies, for example, the insistence by some
identity theorists that mental properties are
really neutral as between being mental or
physical. To be neutral in this way, a prop-
erty would have to be neutral as to whether
it's mental at all. Only if one thought that
being mental meant being non-physical
352
would one hold that defending materialism
required showing that ostensibly mental
properties are neutral as regards whether or
not they're mental.
But holding that mental properties are
non-physical has a cost that is usually not
noticed. A phenomenon is mental only if it
has some distinctively mental property. So,
strictly speaking, a materialist who claims
that mental properties are non-physical
would have to conclude that no mental
phenomena exist. This is the ELIMINATIVE-
MATERIALIST position advanced by
Richard Rorty (1979). (See An Essay on
Mind section 3.7; ELIMINATIVISM.)
According to Rorty, 'mental' and 'phys-
ical' are incompatible terms. Nothing can be
both mental and physical; so mental states
cannot be identical with bodily states. Rorty
traces this incompatibly to our views about
incorrigibility; 'mental' and 'physical' are
incompatible terms because we regard as
incorrigible reports of one's own mental
states, but not reports of physical occur-
rences. But he also argues that we can
imagine a people who describe themselves
and each other using terms just like our
mental vocabulary, except that those people
don't take the reports made with that vo-
cabulary to be incorrigible. Since Rorty takes
a state to be a mental state only if one's
reports about it are taken to be incorrigible,
his imaginary people don't ascribe mental
states to themselves or each other. But the
only difference between their language and
ours is that we take as incorrigible certain
reports which they don't. So their language
has no less descriptive or explanatory power
than ours. Rorty concludes that our mental
vocabulary is idle, and that there are no dis-
tinctively mental phenomena.
This argument hinges on building incor-
rigibility into the meaning of the term
'mental'. If we don't, the way is open to
interpret Rorty's imaginary people as simply
having a different theory of mind from ours,
on which reports of one's own mental states
aren't incorrigible. Their reports would thus
be about mental states, as construed by
their theory. Rorty's thought experiment
would then provide reason to conclude not
that our mental terminology is idle, but
only that this alternative theory of mental
phenomena is correct. His thought experi-
ment would thus sustain the non-
eliminativist view that mental states are
bodily states. Whether Rorty's argument
supports his eliminativist conclusion or the
standard identity theory, therefore, depends
solely on whether or not one holds that the
mental is in some way non-physical.
Paul M. Churchland (1981) advances a
different argument for eliminative materi-
alism. According to Churchland, the com-
monsense conceptions of mental states
contained in our present FOLK PSYCHOLOGY
are, from a scientific point of view, radically
defective. But we can expect that eventually
a more sophisticated theoretical account
will replace those folk-psychological concep-
tions, showing that mental phenomena, as
described by current folk psychology, don't
exist. Since that account would be inte-
grated into the rest of science, we'd have a
thoroughgoing materialist treatment of all
phenomena. So this version of eliminativist
materialism, unlike Rorty's, does not rely on
assuming that the mental is non-physical.
But even if current folk psychology is
mistaken, that doesn't show that mental
phenomena don't exist, but only that they
aren't the way folk psychology describes
them as being. We could conclude they
don't exist only if the folk-psychological
claims that turn out to be mistaken actually
define what it is for a phenomenon to be
mental. Otherwise, the new theory would
still be about mental phenomena, and
indeed would help show that they're iden-
tical with physical phenomena. Church-
land's argument, like Rorty's, depends on a
special way of defining the mental, which
we needn't adopt. It's likely that any argu-
ment for eliminative materialism will
require some such definition, without which
the argument would instead support the
identity theory.
NECESSARY IDENTITY
Early identity theorists insisted that the
identity between mental and bodily events
IDENTITY THEORIES
was contingent. meaning simply that the
relevant identity statements were not con-
ceptual truths. That leaves open the ques-
tion of whether such identities would be
necessarily true on other construals of
necessity.
Saul A. Kripke (1980) has argued that
such identities would have to be necessarily
true if they were true at all. Some terms
refer to things contingently, in that those
terms would have referred to different
things had circumstances been relevantly
different. Kripke's example is 'The first Post-
master General of the U.S.', which, in a dif-
ferent situation, would have referred to
somebody other than Benjamin Franklin.
Kripke calls these terms non-rigid desig-
nators. Other terms refer to things neces-
sarily, since no circumstances are possible
in which they would refer to anything else;
these terms are r.igid designators.
If the terms 'a' and 'b' refer to the same
thing and both determine that thing neces-
sarily, the identity statement 'a = b' is
necessarily true. Kripke maintains that the
term 'pain' and the terms for the various
brain states all determine the states they
refer to necessarily; no circumstances are
possible in which these terms would refer to
different things. So if pain were identical
with some particular brain state, it would
be necessarily identical with that state. But
Kripke argues that pain can't be necessarily
identical with any brain state, since the tie
between PAINS and brain states plainly
seems contingent. He concludes that they
cannot be identical at all.
This argument applies equally to the
identity of types and tokens. Whenever the
term 'pain' refers to a state, it refers to that
state rigidly; similarly with the various
terms for brain states. So if an individual
occurrence of pain were identical with an
individual brain state, it would be neces-
sarily identical with it. Since they can't be
necessarily identical, they can't be identical
at all.
Kripke notes that our intuitions about
whether an identity is contingent can
mislead us. Heat is necessarily identical
with mean molecular kinetic energy; no cir-
353
IDENTITY THEORIES
cumstances are possible in which they
aren't identical. Still, it may at first sight
appear that heat could have been identical
with some other phenomenon. But it
appears this way, Kripke argues, only
because we pick out heat by our sensation
of heat, which bears only a contingent tie to
mean molecular kinetic energy. It's the sen-
sation of heat that actually seems to be con-
nected contingently with mean molecular
kinetic energy, not the physical heat itself.
Kripke insists, however, that such reason-
ing cannot disarm our intuitive sense that
pain is connected only contingently with
brain states. That's because for a state to be
pain is necessarily for it to be felt as pain.
Unlike heat, in the case of pain there's no
difference between the state itself and how
that state is felt, and intuitions about the
one are perforce intuitions about the other.
Kripke's assumption about the term 'pain'
is open to question. As Lewis notes, one
need not hold that 'pain' determines the
same state in all possible situations; indeed,
the causal theory explicitly allows that it
may not. And if it doesn't, it may be that
pains and brain states are contingently
identical. But there's also a problem about a
substantive assumption Kripke makes about
the nature of pains, namely, that pains are
necessarily felt as pains. First impressions
notwithstanding, there is reason to think
not. There are times when we are not
aware of our pains, for example when we're
suitably distracted. So the relationship
between pains and our being aware of them
may be contingent after all, just as the rela-
tionship between physical heat and our sen-
sation of heat is. And that would disarm the
intuition that pain is connected only con-
tingently with brain states.
SUBJECTIVE FEATURES
Kripke's argument focuses on pains and
other sensations, which, because they have
qualitative properties, are frequently held to
cause the greatest problems for the identity
theory. Thomas Nagel (1974) traces the
general difficulty for the identity theory to
the CONSCIOUSNESS of mental states. A
354
mental state's being conscious, he urges,
means that there's something it's like to be
in that state. And to understand that, we
must adopt the point of view of the kind of
creature that's in the state. But an account
of something is objective, he insists, only
insofar as it's independent of any particular
type of point of view. Since consciousness is
inextricably tied to points of view, no objec-
tive account of it is possible. And that
means conscious states cannot be identical
with bodily states.
The viewpoint of a creature is central to
what that creature's conscious states are
like because different kinds of creatures
have conscious states with different kinds of
qualitative property. But the qualitative
properties of a creature's conscious states
depend, in an objective way, on that crea-
ture's perceptual apparatus. We can't
always predict what another creature's con-
scious states are like, just as we can't
always extrapolate from microscopic to
macroscopic properties, at least without
having a suitable theory that covers those
properties. But what a creature's conscious
states are like depends in an objective way
on its bodily endowment. which is itself
objective. So these considerations give us no
reason to think that what those conscious
states are like is not also an objective
matter.
If a sensation isn't conscious, there's
nothing it's like to have it. So Nagel's idea
that what it's like to have sensations is
central to their nature suggests that sensa-
tions cannot occur without being conscious.
And that in turn seems to threaten their
objectivity. If sensations must be conscious,
perhaps they have no nature independently
of how we're aware of them, and thus no
objective nature. Indeed, it's only conscious
sensations that seem to cause problems for
the identity theory (see SUBJECTIVITY).
The assumption that mental states are
invariably conscious, like the supposition
that they're non-physical. is basic to the
Cartesian view. But sensations do occur
that aren't conscious. A mental state's
being conscious consists in one's being con-
scious of it in a way that's intuitively direct
and unmediated. But as already noted, dis-
tractions often make us wholly unaware of
our sensations. Sensations that aren't con-
scious also occur in both subliminal percep-
tion and peripheral vision, as well as in
more esoteric contexts. (See Weiskrantz,
1986.)
Sensations can, moreover, have qualita-
tive properties without being conscious.
Qualitative properties are sometimes called
QUALIA, with the implication that we must
be conscious of them; but we needn't be
bound by that term's implications. Qualita-
tive properties are simply those properties
by means of which we distinguish among
the various kinds of sensations when
they're conscious. But a sensation's being
conscious makes no difference to what its
distinguishing properties are; its being con-
scious consists simply in one's being con-
scious of those properties in a suitable way.
When a sensation isn't conscious its distin-
guishing properties seem to cause no diffi-
culty for the identity theory. And since
those properties are the .same whether or
not the sensation is conscious, there's
nothing about those properties that under-
mines the identity theory. We would
assume otherwise only if we held, with
Nagel and Kripke, that sensations must all
be conscious. (See Rosenthal, 1986.)
Perhaps multiple realizability refutes the
type identity theory; but there are ample
arguments that support the token identity
theory. Moreover, the arguments against
the token theory seem all to rely on
unfounded Cartesian assumptions about the
nature of mental states. The doctrine that
the mental is in some way non-physical is
straightforwardly question begging, and it's
simply not the case that all sensory states
are conscious. It is likely. therefore, that the
identity theory, at least in the token version,
is correct.
See also An Essay in Mind section 3.6.2.
BIBLIOGRAPHY
Armstrong, D.M. 1968. A Materialist Theory of
the Mind. New York: Humanities Press.
IMAGERY
Churchland, P.M. 1981. Eliminative material-
ism and the propositional attitudes. Journal
of Philosophy, LXXVIII: 2, 67-90.
ence and Theory, ed. L. Foster and J. W.
Swanson: Amherst, MA.: University of Mas-
sachusetts Press.
venient causation. In Midwest Studies in Phi-
losophy, Vol 9: Minneapolis: University of
Minnesota Press.
Kripke, S.A. 1980. Naming and Necessity. Cam-
bridge, MA.: Harvard University Press. Ori-
ginally published in Semantics of Natural
Language, ed. D. Davidson and G. Harman.
Dordrecht: D. Reidel Publishing Company,
1972.
Lewis, D. 1980. Mad pain and Martian pain.
In Readings in Philosophy of Psychology, Vol.
I, ed. N. Block. Cambridge, MA.: Harvard
University Press.
Lewis, D. 1972. Psychophysical and theo-
retical identification. Australasian Journal of
Philosophy, L: 3,247-58.
Nagel, T. 1974. What is it like to be a bat? The
Philosophical Review, LXXXIII: 4, 435-50.
Putnam, H. 1975. The nature of mental
states. Mind Language and Reality: Philosophi-
cal Papers, Vol. 2. Cambridge University
Press.
Rorty, R. 1979. Philosophy and the Mirror of
Nature. Princeton University Press.
Rosenthal, D.M., ed. 1991. The Nature of Mind.
New York: Oxford University Press.
Rosenthal, D.M. 1986. Two concepts of con-
sciousness. Philosophical Studies, XLIX: 3,
329-59.
Smart, J.J.C. 1962. Sensations and brain pro-
cesses. In The Philosophy of Mind, ed. V. C.
Chappell. Englewood Cliffs, N.J.: Prentice-
Hall, Inc.
Weiskrantz, 1. 1986. Blindsight. Oxford Uni-
versity Press.
DAVID M. ROSENTHAL
imagery Imagery has played an enor-
mously important role in philosophical
conceptions of the mind. The most popular
view of images prior to this century has
been what we might call 'the picture
theory'. According to this view. held by
such diverse philosophers as Aristotle.
355
IMAGERY
Descartes, and Locke, mental images -
specifically, visual images - are significantly
picture-like in the way they represent
objects in the world. Despite its widespread
acceptance, the picture theory of mental
images was left largely unexplained in the
traditional philosophical literature. Admit-
tedly most of those who accepted the theory
held that mental images copy or resemble
what they represent. But little more was
said.
This century the pictorial view of images
has come under heavy philosophical attack.
Three basic sorts of objections have been
raised. First, there have been challenges to
the sense of the view: mental images are
not seen with real eyes; they cannot be
hung on real walls; they have no objective
weight or colour. What, then, can it mean
to say that images are pictorial? Secondly,
there have been arguments that purport to
show that the view is false. Perhaps the best
known of these is founded on the charge
that the picture theory cannot satisfactorily
explain the indeterminacy of many mental
images. Finally, there have been attacks on
the evidential underpinnings of the theory.
Historically, the philosophical claim that
images are picture-like rested primarily on
an appeal to INTROSPECTION. And today
introspection is taken to reveal a good deal
less about the mind than was traditionally
supposed. This attitude towards introspec-
tion has manifested itself in the case of
imagery in the view that what introspection
really shows about visual images is not that
they are pictorial but only that what goes
on in imagery is experientially much like
what goes on in seeing.
So, the picture theory has been variously
dismissed by many philosophers this cen-
tury as nonsensical, or false, or without
support. Perhaps the most influential altern-
ative has been a view that is now com-
monly known as descriptionalism. This view
has some similarity with the claim made by
the behaviourist, J. B. Watson, that imaging
is talking to one's self beneath one's breath;
for the basic thesis of descriptionalism is
that mental images represent in the manner
of linguistic descriptions. This thesis,
356
however, should not be taken to imply that
during imagery there must be present inner
tokens of the imager's spoken language
either in any movements of the imager's
larynx or in the imager's brain. Rather the
thought is that mental images represent
objects in some neural code which is, in
important respects, language-like.
Descriptionalism remains popular in phi-
losophy today (see e.g. Dennett, 1981), and
it also has significant support in con-
temporary psychology. Nevertheless the tide
has begun to turn again back towards the
pictorial view, and I think it is fair to say
that the picture theory is now quite widely
regarded as intelligible, perhaps true, and
certainly not a theory in search of facts to
explain. What has been responsible for this
change in attitude more than anything else
is the work of some cognitive psychologists,
notably Stephen Kosslyn. In response to a
large body of experimental data on imagery,
Kosslyn and his co-workers have developed
an empirical version of the pictorial view
that seems much more promising than any
of its philosophical predecessors.
In the following section of this entry, I
present a brief sketch of Kosslyn's theory. In
the section after, I turn to recent descrip-
tional theories of imagery, and I make some
observations of my own about imagistic
representation. In the final section I layout
three further issues concerning imagery
that are of interest to philosophers.
KOSSLYN'S THEORY
According to Kosslyn, mental images are to
be conceived of on the model of displays on
a cathode-ray tube screen attached to a
computer. Such displays are generated on
the screen by the computer from informa-
tion that is stored in the computer's
memory. Since there are obvious differences
between mental images and screen displays,
a question arises as to just what the
respects are in which the former are sup-
posed to be like the latter.
Kosslyn suggests that before we answer
this question, we reflect upon how a picture
is formed on a monitor screen and what
makes it pictorial. We may think of the
screen itself as being covered by a matrix in
which there are a large number of tiny
squares or cells. The pattern formed by
placing dots in these cells is pictorial.
Kosslyn asserts, at least in part because it
has spatial features which correspond to
spatial features of the represented object. In
particular, dots in the matrix represent
points on the surface of the object and rela-
tive distance and geometrical relations
among dots match the same relations
among object points. Thus if dots A, B, and
C in the matrix stand respectively for points
PI, P
2
, and P
3
on the object surface, then if
PI is below P
2
and to P3'S left (as the object
is seen from a particular point of view) then
likewise A is below B and to C's left (as the
screen is seen from a corresponding point of
view). Similarly, if PI is further from P
2
than from P 3 then A is further from B than
from C.
Kosslyn's reasoning now becomes more
opaque. The main strand of thought which
is to be found in Kosslyn's writings seems to
be that although mental images lack the
above spatial characteristics, they none-
theless function as if they had those char-
acteristics. Thus, in Kosslyn's view, it is not
literally true that mental images are pic-
tures. Rather the truth in the picture theory
is that mental images are functional pictures.
This claim itself is in dire need of clari-
fication, of course. But before I try to clarify
its meaning, I want briefly to sketch certain
other aspects of Kosslyn's position. Consider
again a cathode-ray tube screen on which a
picture is displayed. The screen may be
thought of as the medium in which the
picture is presented. This medium is spatial
and it is made up of a large number of basic
units or cells some of which are illuminated
to form a picture. Analogously, according to
Kosslyn, there is a functional spatial
medium for imagery made up of a number
of basic units or cells. Mental images, on
Kosslyn's view, are functional pictures in
this medium,
Kosslyn hypothesizes that the imagery
medium, which he calls 'the visual buffer',
is shared with visual perception. In veridical
IMAGERY
perception, any given unit in the medium,
by being active, represents the presence of a
just-noticeable object part at a particular
two-dimensional spatial location within the
field of view. In imagery, the same unit, by
being active, represents the very same
thing. Thus, imaged object parts are rep-
resented within an image as having certain
viewpoint-relative locations they do not in
fact occupy, namely those locations they
would have occupied in the field of view
had the same object parts produced the
same active units during normal vision.
We are now in a position to see what
Kosslyn means by the thesis that mental
images are functional pictures. The basic
idea is that a mental image of an object 0
(and nothing else) functions like a picture of
o in two respects: (1) every part of the
image that represents anything (Le. any
active cell in the buffer) represents a part of
0; (2) two-dimensional relative distance
relations among parts of 0 are represented
in the image via distance relations among
corresponding image parts (active cells).
Distance in the imagery medium is not a
matter of actual physical distance, however.
Instead it is to be explicated in terms of
number of intervening cells - the greater
the distance represented, the greater the
number of active cells representing adjacent
portions of the object. Thus, Kosslyn's pro-
posal does not require that cells represent-
ing adjacent object parts themselves be
physically adjacent (as in a real picture).
Rather, like the cells in an array in a com-
puter, they may be widely scattered. What
matters to the representation of adjacency is
that the processes operating on such cells
treat them as if they were adjacent.
The remaining elements of Kosslyn's
theory concern the processes that operate
on images in the visual buffer. Kosslyn pos-
tulates that there are three sets of processes,
namely those that 'generate', 'inspect', and
'transform' the images. The generation
process is decomposable, in Kosslyn's view,
into further processes that combine together
information about object parts and rela-
tions. The inspection process is also really a
number of different processes that examine
357
IMAGERY
patterns of activated cells in the buffer
thereby enabling us to recognize shapes,
spatial configurations and other character-
istics of the imaged objects. For example, if I
form an image of a racehorse, it is the
inspection process that allows me to decide
if the tip of its tail extends below its rear
knees. Similarly, if I image two equilateral
triangles of the same size, one upright and
one inverted with its tip touching the
middle of the base of the upright one, the
inspection process is what enables me to
recognize the diamond-shaped parallelo-
gram in the middle. Finally, there are trans-
formation processes. These processes
'rotate', 'scale in size', or 'translate' the pat-
terns of activated cells in the buffer.
That, then, in crude outline is Kosslyn's
theory (for further discussion, see Block
1983; Tye, 1988, 1991). Why should we
believe such a view? I cannot possibly go
into all the relevant experimental evidence
here. Instead I shall simply mention one
important experiment. In this experiment,
Kosslyn required subjects to study the map
shown in figure 1.
When the subjects had become familiar
enough with the map to be able to draw it,
Figure 1.
358
they were asked to form a mental image of it
and then to focus in on one particular object
in the image, and this request was repeated
for different objects. It was found that the
farther away an object was from the place
on the image presently being focused on, the
longer it took to focus on that new object.
So, for example, shifting attention from the
'tree part' of the image to the 'hut part' took
longer than shifting attention from the 'tree
part' to the 'lake part'.
Kosslyn claims that this experiment
shows that mental images can be scanned
at fixed speeds. How exactly does this
hypothesis explain the results? After all,
images, on the pictorial view, are not so
constituted that their parts bear the same
relative distance relations to one another as
the object parts they represent. The answer
goes as follows. Scanning across a mental
image involves accessing the appropriate
image parts serially (either by shifting the
locus of attention across a stationary image
or by translating the imaged pattern across
the visual buffer so that different aspects of
the pattern fall under a fixed central focus
of attention). More specifically, in the map
case, scanning across the image involves
accessing one after another the members of
a sequence of representationally simple
image parts, each of which represents a dif-
ferent, just-noticeable location on the map
situated on a line connecting the figures
represented at the beginning and end of
scanning. Thus, if the image has parts A, B,
and C which represent respectively map
parts X, Y, and Z, and X is nearer to Y than
to Z, then scanning across the image from
A to B will involve accessing one after
another fewer image parts than scanning
from A to C. Thus, assuming a fixed scan-
ning speed, Kosslyn's view predicts that the
time it takes to scan from A to B will be
shorter than the time it takes to scan from
A to C. And this is indeed the result we get.
I want now to mention one serious diffi-
culty for Kosslyn's view. It concerns the
representation of the third dimension. How
is this to be accomplished, if mental images
are functional pictures? On Kosslyn's view,
the visual buffer is arranged like a two-
dimensional monitor screen in that the cells
in the buffer themselves only represent
surface patches in two dimensions. The
third dimension gets represented via pro-
cesses that inspect the contents of the buffer
and interpret depth cues much as we do
when viewing the display on a (real)
monitor screen. The major problem with
this view is that it conflicts with the results
of an experiment conducted by Steven
Pinker. In this experiment, subjects exam-
ined an open box, in which five toys were
suspended at different heights, until they
were able to form an accurate image of the
display with their eyes closed. The subjects
then scanned across their images by ima-
gining a dot moving in a straight line
between the imaged objects. It was found
that scanning times increased linearly with
increasing three-dimensional distance
between the objects. What this result
strongly suggests is that mental images
cannot simply represent in the manner of
two-dimensional screen displays.
Let us now take a quick look at the view
of some influential cognitive scientists that
images represent in the manner of struc-
tural descriptions.
DESCRIPTIONAL THEORIES
A structural description of an object is
simply a complex linguistic representation
whose basic non-logical semantic parts
represent object parts, properties, and spatial
relationships. The explicit representation of
properties and spatial relations is one key
difference between structural descriptions
and functional pictures. Consider, for
example, the representation of relative dis-
tance relations in functional pictures. We
saw earlier that this is achieved indirectly
via the number of image parts: more parts,
more distance. No explicit representation is
possible here, since every part of a func-
tional picture that represents anything
represents an object part. In a structural
description this is not the case, however.
The fact that A is further from B than from
C can be represented by some such proposi-
tion as 'F(abc)', where 'F' is a symbol for the
IMAGER Y
relation, and 'a', 'b', and 'c' are symbols for
the object parts. Since 'F' is as much a part
of 'F(abc)' as 'a', 'b', and 'c', there is in this
proposition a representational part that
doesn't represent an object part.
Another key difference between structural
descriptions and functional pictures arises
with respect to syntax. Structural descrip-
tions have syntactic parts, the contents of
which, together with their syntactic combi-
nations, determine the overall representa-
tional content. In 'F( abc)', for example, 'F'
belongs to a different syntactic category
than 'a', 'b', and 'c'. Moreover the content
of 'F(abc)' is different from the content of
'F(bac)' even though the parts are the same.
For a functional picture there are no syn-
tactically distinguishable parts; and there is
no syntactic order (so that two functional
pictures that differ in what they represent
must differ in what some of their parts
represent).
Zenon Pylyshyn maintains that mental
images are structural descriptions no differ-
ent in kind from the representations
involved in other areas of cognition. In
Pylyshyn's view there is an inner language
within which mental representation what-
ever its stripe is confined. This inner lan-
guage is largely unconscious and is not
itself a natural public language, although it
is translated into such a language when we
talk. According to Pylyshyn, Kosslyn's
experiments on imagery can be explained
by reference to the task demands placed on
subjects by the experimenter's instructions
together with facts the subjects already
know. Consider, for example, the map scan-
ning experiment. When subjects are told to
scan across their images to an object they
are not presently focusing on, they interpret
the instructions as requiring them to con-
struct inner representations something like
the representations they would undergo
were they actually scanning across the map
with their eyes. The subjects know that it
takes longer for their eyes to scan greater
distances. So they set a mental clock ticking
for a length of time that permits them to
mimic the response they would give in the
real scanning case.
359
IMAGERY
It is important to realize that although
this account is compatible with Kosslyn's
claim that subjects are accessing serially
image parts that represent adjacent just-
noticeable map parts situated on a line con-
necting the appropriate figures, it does not
presuppose that Kosslyn's claim is true. This
is because the representations the subjects
construct in response to the instructions
may well not represent any just-noticeable
map parts, and even if some such parts are
represented, not all of those lying on the
appropriate line need be. Perhaps, for
example, there is a sequence of representa-
tions like the following: 'The hut is in the
centre of the field of view', 'The hut is a
little to the left and below the centre of the
field of view', 'The hut is as far below and to
the left of the centre of the field of view as
the lake is above and to the right', 'The lake
is in the centre of the field of view'. And
perhaps there is also a further symbolic
representation of the distance on the map
between the lake and the hut. On this con-
ception of what is going on, for different dis-
tances scanned there may be no
corresponding difference in the number of
representations or parts thereof. So, Pyly-
shyn claims, the map-scanning experiment
does not demonstrate that subjects are
transforming a special picture-like image.
And what is true in this one case is true,
Pylyshyn thinks, for Kosslyn's other experi-
ments on imagery. Subjects, in responding
to the demands placed on them by the tasks
they are set (demands built into the instruc-
tions), draw upon their (frequently tacit)
knowledge of the world and their own
visual systems. So, the data Kosslyn has col-
lected give us no information about images
as they really are.
It is far from clear that the appeal to tacit
knowledge can explain all the imagery data,
contrary to Pylyshyn's claim (see Kosslyn,
1981). There are, however, other descrip-
tional theories of imagery that have no need
of the doctrine of tacit knowledge (e.g.
Hinton, 1979). So, descriptionalism cannot
be refuted simply by attacking the doctrine
of tacit knowledge.
Who, then, is on the right track about
360
image representation? My own view is that,
given the plentiful evidence of shared mech-
anisms and representations in imagery and
vision, images are best modelled on cer-
tain representations called '2%-D sketches',
which are posited in the most promising
theory of vision we have, namely the
theory of David Marr. 2%-D sketches are
arrays very like Kosslyn's functional pic-
tures except that their cells contain symbols
for the features of any tiny patch of surface
lying on its associated line of sight (e.g.
depth, orientation, presence of edge). Mental
images, I believe, represent in something
like the manner of 2%-D sketches to which
global sentential interpretations have been
affixed, for example 'This represents a horse'
(see Tye, 1991). So, mental images are, in
some respects, like pictures but in others
they are like descriptions. The truth about
images, I maintain, is that they are a mixed
breed.
I want now to mention some additional
matters of dispute.
FURTHER PHILOSOPHICAL ISSUES
One issue discussed historically by philoso-
phers and still of interest today concerns the
nature and origins of the vagueness that
infects many mental images. As I noted at
the beginning of this entry, image inde-
terminacy has often been considered a
problem for the picture theory. Is this really
the case? Moreover, just how much inde-
terminacy is possible in images?
Another topic much discussed of late is
the subjective character of mental images
(and visual percepts). It is widely held in
philosophy that mental images have, over
and above their representational contents,
intrinsic, introspectively accessible qualities
partly in virtue of which they have those
contents (see INTROSPECTION; SUB-
JECTIVITY). On this view, a mental image of
a zebra, for example, has certain intrinsic
qualities accessible to consciousness partly
in virtue of which it represents a zebra. This
is supposedly why imaging a zebra 'feels'
different to one from merely thinking of a
zebra without forming any image: in the
latter case. one's MENTAL REPRESENTATION
lacks the phenomenal qualities or visual
QUALIA. as they are sometimes called. that
are present in the former. Several philo-
sophers have recently argued that this
whole conception of the phenomenal or
subjective character of mental images is
mistaken.
A third issue of interest is the causal role
of image content. This issue may be illus-
trated as follows. Suppose you form an
image of a snake and suppose you have a
snake phobia. Your image may cause you
to perspire. In the event that this happens.
it is the fact that your image has the
content that it does (the fact that it is of a
snake and not of a bird. say) that is respon-
sible for its causing you to perspire. If. as
seems likely. a complete neurophysiological
accont may be given of the origins of your
perspiring. how is it possible for the content
of your image (apparently not a neurophy-
siological property) to make any difference
to your behaviour? And what reason is
there to suppose that the content of your
image is causally efficacious in this
instance. given that the contents of things
elsewhere are sometimes causally inert. as.
for example. when the pitch. but not the
content. of the sounds produced by a
soprano causes a glass to shatter?
There is no general agreement among
philosophers about how to settle the above
issues any more than there is among cogni-
tive scientists about how to understand the
nature of imagistic representation. Never-
theless. it is hard to come away from a
reading of recent work in cognitive science
or philosophy on imagery without the
feeling that real progress is being made.
See also COGNITIVE PSYCHOLOGY; COMPUTA-
TIONAL MODELS; CONSCIOUNESS; FUNCTION-
ALISM; IMAGINATION; LANGUAGE OF
THOUGHT; PERCEPTUAL CONTENT; REASONS
AND CAUSES; REPRESENTATION.
BIBLIOGRAPHY
Block. N .. ed. 1981. Imagery. Cambridge. MA.:
MIT Press.
IMAG IN A TION
Block. N. 1983. Mental pictures and cognitive
science. Philosophical Review. 92. 499-542.
Dennett. D. 1981. The nature of images and
the introspective trap. In Block. 1981.
Hinton. G. 1979. Some demonstrations of the
effects of structural descriptions in mental
imagery. Behavioral and Brain Sciences. 3.
231-50.
Kosslyn. S. 1980. Image and Mind. Cambridge.
MA.: Harvard University Press.
Kosslyn. S. 1981. The medium and the
message in mental imagery. In Block. 1981.
Pinker. S. 1980. Mental imagery and the third
dimension. Journal of Experimental Psychol-
ogy: General. 109. 354-71.
Pylyshyn. Z. 1981. The imagery debate:
analog media versus tacit knowledge. Psy-
chological Review. 88. 16-45. Reprinted in
Block. 1981.
Tye. M. 1988. The picture theory of mental
images. Philosophical Review. 97. 497-520.
Tye. M. 1991. The Imagery Debate. Cambridge.
MA.: MIT Press.
Watson. }.B. 1928. The Ways of Behaviorism.
New York: Harper.
MICHAEL TYE
imagination It is probably true that philo-
sophers have shown much less interest in
the subject of the imagination during the
last fifteen years or so than in the period
just before that. It is certainly true that
more books about the imagination have
been written by those concerned with lit-
erature and the arts than have been written
by philosophers in general and by those
concerned with the philosophy of mind in
particular. This is understandable in that
the imagination and imaginativeness figure
prominently in artistic processes. especially
in romantic art. Indeed. those two high
priests of romanticism. Wordsworth and
Coleridge. made large claims for the role
played by the imagination in views of
reality. although Coleridge's thinking on
this was influenced by his reading of the
German philosophy of the late eighteenth
and early nineteenth centuries. particularly
Kant and Schelling. Coleridge distinguished
between primary and secondary imagina-
tion. both of them in some sense productive.
361
IMAGIN A TION
as opposed to merely reproductive. Primary
imagination is involved in all perception of
the world in accordance with a theory
which, as we shall see, Coleridge derived
from Kant, while secondary imagination,
the poetic imagination, is creative from the
materials that perception provides. It is this
poetic imagination which exemplifies imagi-
nativeness in the most obvious way.
Being imaginative is a function of
THOUGHT, but to use one's imagination in
this way is not just a matter of thinking in
novel ways. Someone who, like Einstein for
example, presents a new way of thinking
about the world need not be by reason of
this supremely imaginative (though of
course he may be). The use of new CON-
CEPTS or a new way of using already exist-
ing concepts is not in itself an
exemplification of the imagination. What
seems crucial to the imagination is that it
involves a series of perspectives, new ways
of seeing things, in a sense of 'seeing' that
need not be literal. It thus involves, whether
directly or indirectly, some connection with
perception, but in different ways, some of
which will become evident later. The aim of
subsequent discussion here will indeed be to
make clear the similarities and differences
between seeing proper and seeing with the
mind's eye, as it is sometimes put. This will
involve some consideration of the nature
and role of images.
IMAGINATION AND PERCEPTION
Connections between the imagination and
PERCEPTION are evident in the ways that
many classical philosophers have dealt with
the imagination. One of the earliest ex-
amples of this, the treatment of phantasia
(usually translated as 'imagination') in Aris-
totle's De Anima III. 3 , seems to regard the
imagination as a sort of half-way house
between perception and thought, but in a
way which makes it cover appearances in
general. so that the chapter in question has
as much to do with perceptual appearances,
including illusions, as it has to do with, say,
imagery. Yet Aristotle also emphasizes that
imagining is in some sense voluntary, and
362
that when we imagine a terrifying scene we
are not necessarily terrified, any more than
we need be when we see terrible things in a
picture. How that fits in with the idea that
an illusion is or can be a function of the
imagination is less than clear. Yet some
subsequent philosophers, Kant in particular,
followed in recent times by P. F. Strawson,
have maintained that all perception
involves the imagination, in some sense of
that term, in that some bridge is required
between abstract thoughts and their percep-
tual instances. This comes out in Kant's
treatment of what he calls the 'schematism',
where he rightly argues that someone
might have an abstract understanding of
the concept of a dog without being able to
recognize or identify any dogs. It is also
clear that someone might be able to classify
all dogs together without any under-
standing of what a dog is. The bridge that
needs to be provided to link these two abil-
ities Kant attributes to the imagination.
In so arguing Kant goes, as he so often
does, beyond Hume who thought of the
imagination in two connected ways. First,
there is the fact that there exist, Hume
thinks, ideas which are either copies of
impressions provided by the senses or are
derived from these. Ideas of imagination are
distinguished from those of memory, and
both of these from impressions of sense, by
their lesser vivacity. Second, the imagina-
tion is involved in the processes, mainly
association of ideas, which take one from
one idea to another, and which Hume uses
to explain, for example, our tendency to
think of objects as having a continuing exis-
tence, even when we have no impressions
of them. Ideas, one might suggest, are for
Hume more or less images, and imagination
in the second, wider, sense is the mental
process which takes one from one idea to
another and thereby explains our tendency
to believe things which go beyond what the
senses immediately justify. The role which
Kant gives to the imagination in relation to
perception in general is obviously a wider
and fundamental role than Hume allows.
Indeed one might take Kant to be saying
that were there not the role that he, Kant.
insists on there would be no place for the
role which Hume gives it. Kant also allows
for a free use of the imagination in connec-
tion with the arts and the perception of
beauty, and this is a more specific role than
that involved in perception in general.
SEEING-AS AND IMAGES
In seeing things we normally see them as
such and suches (see PERCEPTION). But
there are also special. imaginative, ways of
seeing things, which WITTGENSTEIN
emphasized in his treatment of 'seeing-as' in
his Philosophical Investigations II. xi and else-
where. To see a simple triangle drawn on a
piece of paper as standing up, lying down,
hanging from its apex and so on is a form of
seeing-as which is both more special and
more sophisticated than simply seeing it as
a triangle. Both involve the application of
concepts to the objects of perception, but
the way in which this is done in the two
cases is quite different. One might say that
in the second case one has to adopt a
certain perspective, a certain point of view,
and if that is right it links up with what I
said earlier about the relation and difference
between thinking imaginatively and think-
ing in novel ways.
Wittgenstein (1953, p. 212e) used the
phrase 'an echo of a thought in sight' in
relation to these special ways of seeing
things, which he called 'seeing aspects'.
Roger Scruton has spoken of the part played
in it all by 'unasserted thought' (Scruton,
1974, chs 7 and 8), but the phrase used by
Wittgenstein brings out more clearly one
connection between thought and a form of
sense-perception which is characteristic of
the imagination. Wittgenstein (1953, p.
213e) also compares the concept of an
aspect and that of seeing-as with the
concept of an image, and this brings out a
point about the imagination that has not
been much evident in what has been said so
far - that imagining something is typically a
matter of picturing it in the mind and that
this involves images in some way (see
IMAGER Y). This aspect of the imagination is
crucial for the philosophy of mind, since it
IMAGINATION
raises the question of the status of images,
and in particular whether they constitute
private objects or states in some way.
Sartre, in his early work on the imagination
(Sartre, 1940), emphasized, following
Husserl, that images are forms of conscious-
ness of an object, but in such a way that
they 'present' the object as not being;
hence, he said, the image 'posits its object
as nothingness'. Such a characterization
brings out something about the role of the
form of consciousness of which the having
of imagery may be a part; in picturing
something the images are not themselves
the object of consciousness. The account
does less, however, to bring out clearly just
what images are or how they function.
PICTURING AND SEEING WITH THE
MIND'S EYE
As part of an attempt to grapple with this
question, RYLE too has argued that in pic-
turing, say, Helvellyn (the mountain in the
English Lake District), in having it before
the mind's eye, we are not confronted with
a mental picture of Helvellyn; images are
not seen. We nevertheless can 'see' Helvel-
lyn, and the question is what this 'seeing' is,
if it is not seeing in any direct sense. One of
the things that may make this question dif-
ficult to answer is the fact that people's
images and their capacity for imagery vary,
and this variation is not directly related to
their capacity for imaginativeness. While an
image may function in some way as a
REPRESENT A TION in a train of imaginative
thought, such thought does not always
depend on that; on the other hand, images
may occur in thought which are not really
representational at all, are not, strictly
speaking, 'of' anything. If the images are
representational. can one discover things
from one's images that one would not know
otherwise? Many people would answer 'No',
especially if their images are generally frag-
mentary, but it is not clear that this is true
for everyone. On the other hand, and this
affects the second point, fragmentary
imagery which is at best ancillary to the
process of thought in which it occurs may
363
IMAGIN ATION
not be in any obvious sense representa-
tional. even if the thought itself is 'of' some-
thing.
Another problem with the question what
it is to 'see' Helvellyn with the mind's eye is
that the 'seeing' in question mayor may
not be a direct function of MEMOR Y. For one
who has seen Helvellyn, imagining it may
be simply a matter of reproducing in some
form the original vision, and the vision may
be reproduced unintentionally and without
any recollection of what it is a 'vision' of.
For one who has never seen it the task of
imagining it depends most obviously on the
knowledge of what sort of thing Helvellyn is
and perhaps on experiences which are rele-
vant to that knowledge. It would be surpris-
ing, to say the least, if imaginative power
could produce a 'seeing' that was not con-
structed from any previous seeings. But that
the 'seeing' is not itself a seeing in the
straightforward sense is clear, and on this
negative point what Ryle says, and others
have said, seems clearly right. As to what
'seeing' is in a positive way, Ryle answers
that it involves fancying something and
that this can be assimilated to pretending.
Fancying that one is seeing Helvellyn is
thus at least like pretending that one is
doing that thing. But is it?
IMAGINING AND PRETENDING
There is in fact a great difference between,
say, imagining that one is a tree and pre-
tending to be a tree. Pretending normally
involves doing something, and even when
there is no explicit action on the part of the
pretender, as when he or she pretends that
something or other is the case, there is at all
events an implication of possible action.
Pretending to be a tree may involve little
more than standing stock-still with one's
arms spread out like branches. To imagine
being a tree (something that I have found
that some people deny to be possible, which
is to my mind a failure of imagination) need
imply no action whatever. (Imagining being
a tree is different in this respect from ima-
gining that one is a tree, where this means
believing falsely that one is a tree; one can
364
imagine being a tree without this commit-
ting one to any beliefs on that score.) On
the other hand, to imagine being a tree does
seem to involve adopting the hypothetical
perspective of a tree, contemplating perhaps
what it is like to be a fixture in the ground
with roots growing downwards and with
branches (somewhat like arms) blown by
the wind and with birds perching on them.
Imagining something seems in general to
involve contemplating some change or
partial change of identity on the part of
something or other, and in imagining being
something else, such as a tree, the partial
change of identity contemplated is in
oneself. The fact that the change of identity
contemplated cannot be complete does not
gainsay the point that it is a change of iden-
tity which is being contemplated. One might
raise the question whether something about
the SELF is involved in all imagining (see e.g.
Bernard Williams, 1973, ch. 3). Berkeley
even suggested that imagining a solitary
unperceived tree involves a contradiction, in
that to imagine that is to imagine oneself
perceiving it. (There is a similar argument in
Schopenhauer's The World as Will and Repre-
sentation, II, 1.) In fact there is a difference
between imagining an object, solitary or
not, and imagining oneself seeing that
object. The latter certainly involves putting
oneself imaginatively in the situation pic-
tured; the former involves contemplating
the object from a point of view, indeed from
that point of view which one would oneself
have if one were viewing it. Picturing some-
thing thus involves that point of view to
which reference has already been made, in a
way that clearly distinguishes picturing
something from merely thinking of it.
MUST IMAGES BE PICTORIAL?
Imagining a scene need not involve having
mental images which are for the person
concerned anything like pictures of the
scene. Certainly, to revert to a point already
made in connection with Ryle, to picture a
scene is not to regard inwardly a picture of
the scene. This does not rule out the possi-
bility that an image might come into one's
mind which one recognizes as some kind of
depiction of a scene. But when actually pic-
turing a scene, it would not be right to say
that one imagines the scene via a con-
templation of an image which plays the part
of a picture of it. Moreover, it is possible to
imagine a scene without any images occur-
ring, the natural interpretation of which
would be that they are pictures of that
scene. It is not impossible for one imagining,
say, Helvellyn to report on request the
occurrence of images which are not in any
sense pictures of Helvellyn - not of that par-
ticular mountain and perhaps not even of a
mountain at all. That would not entail that
he or she was not imagining Helvellyn; a
report of what was imagined might still be
relevant to or associated with Helvellyn,
even if the images reported would not be
thought by others to be of Helvellyn.
This raises a question which is asked by
Wittgenstein (1953, p. 177e) - 'What
makes my image of him into an image of
him?' To which Wittgenstein replies 'Not its
looking like him', and he goes on to suggest
that a person's account of what his imagery
represents is decisive. Certainly it is so when
the process of imagination which involves
the imagery is one that the person engages
in intentionally. The same is not true, as
Wittgenstein implicitly acknowledges in the
same context, if the imagery simply comes
to mind without there being any intention;
in that case one might not even know what
the image is an image of.
THE ROLE AND NATURE OF IMAGES
All this complicates the question what the
status of mental images is. It might seem
that they stand in relation to imagining as
SENSATIONS stand to perception, except
that the occurrence of sensations is a
passive business, while the occurrence of an
image can be intentional, and in the
context of an active flight of imagination is
likely to be so. Sensations give perceptions a
certain phenomenal character, providing
their sensuous, as opposed to conceptual,
content (see CONTENT; PERCEPTUAL
CONTENT). In thinking, where operation
IMAG IN A TION
with concepts looms largest and has
perhaps the overriding role, it still seems
necessary for our thought to be given a
focus in thought-occurrences such as
images. These have sometimes been char-
acterized as symbols which are the material
of thought, but the reference to symbols is
not really illuminating. Nevertheless, while
a period of thought in which nothing of this
kind occurs is possible, the general direction
of thought seems to depend on such things
occurring from time to time. (See Hamlyn,
1976.) In the case of the imagination
images seem even more crucial, in that
without them it would be difficult, to say
the least, for the point of view or perspec-
tive which is important for the imagination
to be given a focus.
I say that it would be difficult for this to
be so, rather than impossible, since it is
clear that entertaining a description of a
scene, without there being anything like a
vision of it, could sometimes give that per-
spective. The question still arises whether a
description could always do quite what an
image can do in this respect. This point is
connected with an issue over which there
has been some argument among psycholo-
gists, such as S. M. Kosslyn and Z. W. Pyly-
shyn, concerning what are termed
'analogue' versus 'propositional' theories of
representation. This is an argument con-
cerning whether the process of imagery is
what Pylyshyn (1986) calls 'cognitively
penetrable', i.e. such that its function is
affected by beliefs or other intellectual pro-
cesses expressible in propositions, or
whether, on the other hand, it can be inde-
pendent of cognitive processes although
capable itself of affecting the mental life
because of the pictorial nature of images
(what Pylyshyn calls their 'analogue
medium'). One example, which has loomed
large in that argument, is that in which
people are asked whether two asymme-
trically presented figures can be made to
coincide, the decision on which may entail
some kind of mental rotation of one or more
of the figures. Those defending the 'analo-
gue' theory point to the fact that there is
some relation between the time taken and
365
INNATENESS
the degree of the rotation required, this sug-
gesting that some process involving chan-
ging images is entailed. For one who has
little or no imagery this suggestion may
seem unintelligible. Is it not enough for one
to go through an intellectual working out of
the possibilities, based on features of the
figures that are judged relevant? This could
not be said to be unimaginative as long as
the intellectual process involved reference to
perspectives or points of view in relation to
the figures, the possibility of which the
thinker might be able to appreciate. Such
an account of the process of imagination
cannot be ruled out, although there are
conceivable situations in which the 'ana-
logue' process of using images might be
easier. Or at least it might be easier for
those who have imagery most like the
actual perception of a scene; for others the
situation might be different.
The extreme of the former position is
probably provided by those who have so-
called 'eidetic' imagery, where having an
image of a scene is just like seeing it, and
where, if it is a function of memory as it
most likely is, it is clearly possible to find
out details of the scene imagined by intro-
spection of the image. The opposite extreme
is typified by those for whom imagery, to
the extent it occurs at all, is at best ancillary
to propositionally styled thought. But, to
repeat the point made repeatedly before,
that thought, even when unasserted, will
not count as imagination unless it provides
a series of perspectives on its object. Because
images are or can be perceptual analogues
and have a phenomenal character ana-
logous to what sensations provide in per-
ception they are most obviously suited, in
the workings of the mind, to the provision
of those perspectives. But in a wider sense,
imagination enters the picture wherever
some link between thought and perception
is required, as well as making possible ima-
ginative forms of seeing-as. It may thus jus-
tifiably be regarded as a bridge between
perception and thought.
See also IMAGERY; PERCEPTION; REPRE-
SENTATION; THOUGHT.
366
BIBLIOGRAPHY
Hamlyn, D.W. 1976. Thinking. In Con-
temporary British Philosophy, Fourth Series,
ed. H. D. Lewis. London: Allen & Unwin.
Ishiguro, H. ] 966. Imagination. In British
Analytical Philosophy, ed. B. A. O. Williams
and A. Montefiore. London: Routledge &
Kegan Paul.
Ishiguro, H. ] 967. 'Imagination', PASS, XLI.
37-56.
Pylyshyn, Z. 1986. Computation and Cognition.
Ryle, G. 1949. The Concept of Mind. London:
Hutchinson.
Sartre, J.-P. 1972. The Psychology of Imagina-
tion. London: Methuen. First published as
L'imaginaire, 1940.
Scruton, R. 1974. Art and Imagination.
London: Methuen.
Strawson, P.F. 1974. Freedom and Resentment.
London: Methuen.
Warnock, M. 1976. Imagination. London:
Faber & Faber.
Williams, B.A.O. ] 973. Problems of the Self.
tions. Oxford: Basil Blackwell.
Wittgenstein, L. 1967. Zettel, esp. 621 ff.
D. W. HAMLYN
incorrigibility see FIRST-PERSON AUTHOR-
ITY.
infallibility see FIRST-PERSON AUTHORITY.
innateness The debate about whether
there are innate ideas is very old. Plato in
the Meno famously argues that all of our
knowledge is innate. Descartes and Leibniz
defended the view that the mind contains
innate ideas; Berkeley, Hume and Locke
attacked it. In fact, as we now conceive
the great debate between European
Rationalism and British Empiricism in the
seventeenth and eighteenth centuries, the
doctrine of innate ideas is a central bone
of contention: rationalists typically claim
that knowledge is impossible without a
significant stock of general innate concepts
or judgments; empiricists argued that all
ideas are acquired from experience. This
debate is replayed with more empirical
content and with considerably greater
conceptual complexity in contemporary
cognitive science. most particularly within
the domains of psycho linguistic theory
and cognitive developmental theory (see
COGNITIVE PSYCHOLOGY; DEVELOPMENTAL
PSYCHOLOGY).
One must be careful not to caricature the
debate. It is too easy to see the debate as
one pitting innatists. who argue that all
concepts. or all of linguistic knowledge is
innate (and certain remarks of Fodor (1975.
1987) and of Chomsky (1966. 1975) lend
themselves to this interpretation) against
empiricists who argue that there is no
innate cognitive structure to which one
need appeal in explaining the acquisition of
language or the facts of cognitive develop-
ment (an extreme reading of Putnam
(1975. 1992). Harman (1969. 1975). or
Nelson (1987) might give this impression).
But this debate would be a silly and a sterile
debate indeed. For obviously. something is
innate. Brains are innate. And the structure
of the brain must constrain the nature of
cognitive and linguistic development to
some degree. Equally obviously. something is
learned. and is learned as opposed to merely
grown as limbs or hair grow. For not all of
the world's citizens end up speaking
English. or knowing the Special Theory of
Relativity. The interesting questions then all
concern exactly what is innate. to what
degree it counts as knowledge. and what is
learned. and to what degree its content and
structure are determined by innately speci-
fied cognitive structures. And that is plenty
to debate about.
The arena in which the innateness debate
has been prosecuted with the greatest
vigour is that of language acquisition. and
so it is appropriate to begin there. But it will
be important to see how this debate has
more recently been extended to the domain
of general knowledge and reasoning abilities
through the investigation of the develop-
ment of object constancy - the disposition to
conceive of physical objects as persistent
when unobserved. and to reason about
INN ATENESS
their properties and locations when they are
not perceptible. To this we will turn after an
initial discussion of the innateness con-
troversy in psycho linguistics.
The most prominent exponent of the
innateness hypothesis in the domain of lan-
guage acquisition is CHOMSKY (1966.
1975). His research and that of his col-
leagues and students is responsible for
developing the influential and powerful
framework of transformational grammar
that dominates current linguistic and psy-
cholinguistic theory. This body of research
has amply demonstrated that the grammar
of any human language is a highly sys-
tematic. abstract structure and that there
are certain basic structural features shared
by the grammars of all human languages.
collectively called universal grammar. Varia-
tions among the specific grammars of the
world's languages can be seen as reflecting
different settings of a small number of para-
meters that can. within the constraints of
universal grammar. take any of several dif-
ferent values. All of the principal arguments
for the innateness hypothesis in linguistic
theory rely on this central insight about
grammars. The principal arguments are
these: (1) the argument from the existence
of linguistic universals; (2) the argument
from patterns of grammatical errors in early
language learners; (3) the poverty of the
stimulus argument; (4) the argument from
the ease of first language learning; (5) the
argument from the relative independence of
language learning and general intelligence;
(6) the argument from the modularity of
linguistic processing. We will consider each
of these arguments and the plausible replies
in turn.
Innatists argue (Chomsky 1966. 1975)
that the very presence of linguistic uni-
versals argues for the innateness of linguis-
tic knowledge. but more importantly and
more compellingly that the fact that these
universals are. from the standpoint of com-
municative efficiency. or from the stand-
point of any plausible simplicity criterion.
adventitious. There are many conceivable
grammars. and those determined by Uni-
versal Grammar are not ipso facto the most
367
INNATENESS
efficient or the simplest. Nonetheless all
human languages satisfy the constraints of
Universal Grammar. Since neither the com-
municative environment nor the com-
municative task can explain this
phenomenon, it is reasonable to suppose
that it is explained by the structure of the
mind - and therefore by the fact that the
principles of Universal Grammar lie innate
in the mind and constrain the language
that a human can acquire.
Linguistic empiricists, on the other hand,
reply that there are alternative possible
explanations of the existence of such adven-
titious universal properties of human lan-
guages. For one thing, such universals
could be explained, Putnam (1975, 1992)
argues, by appeal to a common ancestral
language, and the inheritance of features of
that language by its descendants. Or it
might turn out (Harman, 1969, 1975) that
despite the lack of direct evidence at present
the features of Universal Grammar in fact
do serve either the goals of communicative
efficacy or simplicity according to a metric
of psychological importance. Finally, empiri-
cists point out, the very existence of Uni-
versal Grammar might be a trivial logical
artefact (Quine, 1968): for one thing, any
finite set of structures will have some fea-
tures in common. Since there are a finite
number of languages, it follows trivially
that there are features they all share. More-
over, it is argued, many features of Uni-
versal Grammar are interdependent. So in
fact the set of fundamental principles shared
by the world's languages may be rather
small. Hence even if these are innately
determined, the amount of innate know-
ledge thereby required may be quite small
as compared with the total corpus of
general linguistic knowledge acquired by
the first language learner.
These replies are rendered less plausible,
innatists argue, when one considers the fact
that the errors language learners make in
acquiring their first language seem to be
driven far more by abstract features of
grammar than by any available input data
(Marcus et aI., 1992). So, despite receiving
correct examples of irregular plurals or past
368
tense forms for verbs, and despite having
correctly formed the irregular forms for
those words, children will often incorrectly
regularize irregular verbs once acquiring
mastery of the rule governing regulars in
their language. And in general, not only the
correct inductions of linguistic rules by
young language learners, but more import-
antly, given the absence of confirmatory
data and the presence of refuting data, chil-
dren's erroneous inductions are always con-
sistent with Universal Grammar, often
simply representing the incorrect setting of
a parameter in the grammar. More gen-
erally, innatists argue (Chomsky 1966,
1975; Crain, 1991) all grammatical rules
that have ever been observed satisfy the
structure-dependence constraint. That is,
many linguists and psycho linguists argue
that all known grammatical rules of all of
the world's languages, including the frag-
mentary languages of young children must
be stated as rules governing hierarchical
sentence structures, and not governing, say,
sequences of words. Many of these, such as
the constituent-command constraint gov-
erning anaphor, are highly abstract indeed,
and appear to be respected by even very
young children (Solan, 1983; Crain, 1991).
Such constraints may, innatists argue, be
necessary conditions of learning natural
language in the absence of specific instruc-
tion, modelling and correction, conditions in
which all first language learners acquire
their native languages.
An important empiricist reply to these
observations derives from recent studies of
CONNECTIONIST models of first language
acquisition (Rummel hart and McClelland,
1986, 1987). Connectionist systems, not
previously trained to represent any subset of
Universal Grammar that induce grammars
which include a large set of regular forms
and a few irregulars also tend to over-
regularize, exhibiting the same U-shaped
learning curve seen in human langlJage
acquirers. It is also noteworthy that "on-
nectionist learning systems that induce
grammatical systems acquire 'accidentally'
rules on which they are not explicitly
trained, but which are consistent with those
upon which they are trained, suggesting
that as children acquire portions of their
grammar, they may accidentally 'learn'
other consistent rules, which may be
correct in other human languages, but
which then must be 'unlearned' in their
home language. On the other hand, such
'empiricist' language acquisition systems
have yet to demonstrate their ability to
induce a sufficiently wide range of the rules
hypothesized to be comprised by Universal
Grammar to constitute a definitive empirical
argument for the possibility of natural lan-
guage acquisition in the absence of a pow-
erful set of innate constraints.
The poverty of the stimulus argument has
been of enormous influence in innateness
debates, though its soundness is hotly con-
tested. Chomsky (1966, 1975) notes that
(1) the examples of the target language to
which the language learner is exposed are
always jointly compatible with an infinite
number of alternative grammars, and so
vastly underdetermine the grammar of the
language; and (2) the corpus always con-
tains many examples of ungrammatical
sentences, which should in fact serve as fal-
sifiers of any empirically induced correct
grammar of the language; and (3) there is,
in general, no explicit reinforcement of
correct utterances or correction of incorrect
utterances, either by the learner or by those
in the immediate training environment.
Therefore, he argues, since it is impossible
to explain the learning of the correct
grammar - a task accomplished by all
normal children within a very few years -
on the basis of any available data or known
learning algorithms, it must be that the
grammar is innately specified, and is merely
'triggered' by relevant environmental cues.
Opponents of the linguistic innateness
hypothesis, however, point out that the cir-
cumstance that Chomsky notes in this
argument is hardly specific to language. As
is well known from arguments due to Hume
(1978), Wittgenstein (1953), Goodman
(1972), and Kripke (1982), in all cases of
empirical abduction, and of training in the
use of a word, data underdetermine the-
ories. This moral is emphasized by Quine
INN ATENESS
(1954, 1960) as the principle of the under-
determination of theory by data. But we
nonetheless do abduce adequate theories in
science, and we do learn the meanings of
words. And it would be bizarre to suggest
that all correct scientific theories or the facts
of lexical semantics are innate.
But, innatists reply, when the empiricist
relies on the underdetermination of theory
by data as a counterexample, a significant
dis analogy with language acquisition is
ignored: the abduction of scientific theories
is a difficult, laborious process, taking a
sophisticated theorist a great deal of time
and deliberate effort. First language acquisi-
tion, by contrast, is accomplished effortlessly
and very quickly by a small child. The enor-
mous relative ease with which such a
complex and abstract domain is mastered
by such a naIve 'theorist' is evidence for the
innateness of the knowledge achieved.
Empiricists such as PUTNAM (1975,
1992) have rejoined that innatists under-
estimate the amount of time that language
learning actually takes, focusing only on
the number of years from the apparent
onset of acquisition to the achievement of
relative mastery over the grammar. Instead
of noting how short this interval, they
argue, one should count the total number of
hours spent listening to language and
speaking during this time. That number is
in fact quite large, and is comparable to the
number of hours of study and practice
required in the acquisition of skills that are
not argued to derive from innate structures,
such as chess playing or musical composi-
tion (see Simon, 1981). Hence, they argue,
once the correct temporal parameters are
taken into consideration, language learning
looks like one more case of human skill
acquisition than like a special unfolding of
innate knowledge.
Innatists, however, note that while the
ease with which most such skills are
acquired depends on general intelligence,
language is learned with roughly equal
speed, and to roughly the same level of
general syntactic mastery regardless of
general intelligence. In fact even sig-
nificantly retarded indiViduals, assuming no
369
INNATENESS
special language deficit, acquire their native
language on a time-scale and to a degree
comparable to that of normally intelligent
children, The language acquisition faculty
hence appears to allow access to a sophisti-
cated body of knowledge independent of the
sophistication of the general knowledge of
the language learner.
Empiricists reply that this argument
ignores the centrality of language to a wide
range of human activities, and consequently
the enormous attention paid to language
acquisition by retarded youngsters and their
parents or caretakers. They argue as well
that innatists overstate the parity in linguis-
tic competence between retarded children
and children of normal intelligence.
Innatists point out that the MODULARITY
of language processing is a powerful argu-
ment for the innateness of the language
faculty. There is a large body of evidence,
innatists argue, for the claim that the pro-
cesses that subserve the acquisition, under-
standing and production of language are
quite distinct and independent of those that
subserve general cognition and learning.
( See Fodor, 1975; Garfield, 1989; Chomsky,
1966, 1975 for more on modularity
theory.) That is, language learning and lan-
guage processing mechanisms and the
knowledge they embody are domain specific
- grammars and grammatical learning and
utilizations mechanisms are not used
outside of language processing. They are
information ally encapsulated - only linguis-
tic information is relevant to language
acquisition and processing. They are man-
datory - language learning and language
processing are automatic. Moreover, lan-
guage is subserved by specific dedicated
neural structures, damage to which pre-
dictably and systematically impairs linguis-
tic functioning, and not general cognitive
functioning. All of this suggests a specific
'mental organ', to use Chomsky's phrase,
that has evolved in the human cognitive
system specifically in order to make lan-
guage possible. The specific structure of this
organ simultaneously constrains the range
of possible human languages and guides the
learning of the child's target language, later
370
making rapid on-line language processing
possible. The principles represented in this
organ constitute the innate linguistic
knowledge of the human being. Additional
evidence for the early operation of such an
innate language acquisition module is
derived from the many infant studies that
show that infants selectively attend to
soundstreams that are prosodically appro-
priate, that have pauses at clausal bound-
aries, and that contain linguistically
permissible phonological sequences. (See the
literature reviewed in Karmiloff-Smith,
1991.)
This argument, of course, depends on the
modularity thesis, which is highly con-
tentious in contemporary cognitive science.
Critics of the innateness hypothesis argue
that the processes involved in language
acquisition and processing are merely
instances of general learning strategy
(Putnam, op. cit.; Rummelhart and McClel-
land, op. cit., that language learning and
processing requires semantic information,
access to which violates modularity
(Marslen-Wilson and Tyler, 1987; Jacken-
doff, 1983; Putnam, op. cit.) and that much
highly skilled behaviour that is not innate is
nonetheless automated.
A particularly strong form of the innate-
ness hypothesis in the psycho linguistic
domain is FODOR'S (1975, 1987) LAN-
GUAGE OF THOUGHT hypothesis. Fodor
argues not only that the language learning
and processing faculty is innate, but that
the human representational system exploits
an innate language of thought which has
all of the expressive power of any learnable
human language. Hence, he argues, all
concepts are in fact innate, in virtue of the
representational power of the language of
thought. This remarkable doctrine is hence
even stronger than classical rationalist doc-
trines of innate ideas: whereas Chomsky
echoes Descartes in arguing that the most
general concepts required for language
learning are innate, while allowing that
more specific concepts are acquired, Fodor
echoes Plato in arguing that every concept
we ever 'learn' is in fact innate.
Fodor defends this view by arguing that
the process of language learning is a process
of hypothesis formation and testing, where
among the hypotheses that must be for-
mulated are meaning postulates for each
term in the language being acquired. But in
order to formulate and test a hypothesis of
the form 'X' means Y, where 'X' denotes a
term in the target language, prior to the
acquisition of that language, the language
learner, Fodor argues, must have the
resources necessary to express 'y'. Therefore
there must be, in the language of thought, a
predicate available co-extensive with each
predicate in any language that a human
can learn. Fodor also argues for the lan-
guage of thought thesis by noting that the
language in which the human information
processing system represents information
cannot be a human spoken language, since
that would, contrary to fact, privilege one of
the world's languages as the most easily
acquired. Moreover, it cannot be, he argues,
that each of us thinks in our own native
language since that would (a) predict that
we could not think prior to acquiring a lan-
guage, contrary to the original argument,
and (b) would mean that psychology would
be radically different for speakers of different
languages, which d o ~ n o t appear to be the
case. Hence, Fodor argues, there must be a
non-conventional language of thought, and
the facts that the mind is 'wired' in mastery
of its predicates together with its expressive
completeness entail that all concepts are
innate.
This argument has been criticized (Gar-
field, 1989a; Putnam, 1975, 1992) for
relying on a misleading and unsub-
stantiated model of language learning - the
hypothesis formation and testing model.
Rather, it is argued by these critics, follow-
ing Ryle (1949), Sellars (1963), and Witt-
genstein (1953), that language acquisition
- in particular the acquisition of the mean-
ings of words, which is what is at issue in
this argument - is first and foremost the
acquisition of a set of skills, a species of
'knowing-how'. So, when we learn the
meaning of the word 'green' in English, on
this view, we learn how to use that word
correctly, and not a new fact of the form
INNATENESS
'green' means the same as 'x' where 'X' is
an expression in the language of thought.
And, it is argued, the abilities that permit
the acquisition of these skills are quite pos-
sibly of the same kind as those that enable
the acquisition of a wide range of cognitive
skills. On this view, no innate language of
thought is needed to explain first language
acquisition, and hence no corresponding set
of innate ideas is demanded to explain the
acquisition of the concepts we employ when
we use natural language.
It is important to emphasize once again
that in reviewing these arguments and
counter-arguments, what is at stake is not
whether anything at all is innate. Even the
most extreme empiricist with regard to lan-
guage learning must agree that there are
innately specified cognitive mechanisms
that attune human infants to language and
that predispose humans to learning lan-
guages. Even the most extreme nativist
must agree that without an environment
sufficient to support language development,
including prominently adults speaking the
target language and interacting in a sup-
portive way with the child, no child learns
its first language. The debate concerns just
how much of what is eventually learned is
constrained by what is innate, and how
much is determined by the environment.
And it must be emphasized that though
some of the arguments are primarily con-
ceptual, the most Significant arguments are
empirical in character, and the resolution of
the innateness debate in the domain of lan-
guage acquisition is a thoroughly empirical
matter.
It is also important to ask to what degree
extra-linguistic knowledge about the world
is innate. A fertile field for clarifying this
debate has been that body of research
devoted to examining the development of
object constancy in infants. Piaget (1930,
1952, 1954, 1969, 1974) argued that
object constancy - the recognition that
objects remain existent when they pass out
of sight and the ability to take into account
in reasoning and acting the existence of
unseen objects - does not emerge until
about 9 months of age, and emerges as a
371
INNATENESS
consequence of cognitive development
involving both simple neurobiological
maturation and cognitive development con-
ditioned by learning. The results of Baillar-
geon (1987a, 1987b), Baillargeon and
Graber (1988), Baillargeon, Spelke and
Wasserman (1985), Baillargeon and
Hanko-Summes (1990) and Spelke (1982,
1988, 1990) cast doubt on this timetable,
however. They demonstrate that infants as
young as three months of age respond dif-
ferentially to 'possible' and 'impossible' tra-
jectories for objects behind screens, even
when what renders those trajectories possi-
ble or impossible are the positions of other
hidden objects that the infants have been
shown prior to the fixing of the screen.
Infants seem at very early ages to be aware
of constraints on objects' support as well.
These results are noteworthy because they
are achieved at ages at which infants could
be expected to have very little if any empiri-
cal familiarity with the constraints on object
motion and permanence to which they are
evidently responding.
One plausible interpretation of these
results is that infants come to the world
with innate knowledge - or at least the
innately determined propensity to acquire
such knowledge as a consequence of
maturation - of certain physical or meta-
physical properties of physical objects. How
else could they respond differentially to
these situations at such an early age? These
data certainly make a strong case for this
interpretation. On the other hand one must
move cautiously in inferring innateness
from precocity. For one thing there may be
learning by observation between birth and
the age at which object constancy is
observed that accounts for the phenom-
enon. For another, the differential response
that is observed at these early ages (differ-
ential looking time is the dependent
measure in these studies) does not indicate
with any certainty the presence of a full-
fledged concept of an object, let alone that
of constancy, and may be reckoned an
innate but non-epistemic precursor of that
concept, even if we grant the innate basis
for the observed phenomena.
372
Again, the issues at stake in the debate
concerning the innateness of such general
concepts pertaining to the physical world
cannot be as stark as a dispute between a
position according to which nothing is
innate and one according to which all
empirical knowledge is innate. Rather the
important - and again, always empirical
questions concern just what is innate, and
just what is acquired, and how innate equip-
ment interacts with the world to produce
experience. 'There can be no doubt that all
our knowledge begins with experience ....
But though all our knowledge begins with
experience it does not follow that it all arises
out of experience' (Kant, 1929, p. 1).
See also PSYCHOLOGICAL EXPLANATION.
BIBLIOGRAPHY
Baillargeon. R. 1987a. Object permanence in
3.5- and 4.5-month-old-infants. Develop-
mental Psychology. 23. 655-64.
Baillargeon, R. 1987b. Young infants' reason-
ing about the physical and spatial char-
acteristics of hidden objects. Cognitive
Development, 2, 178-200.
Baillargeon. R. In press. Reasoning about the
height and location of a hidden object in
4.5- and 6.5-month-old-infants. Cognition.
Baillargeon, R .. and Graber. M. 1988. Evid-
ence of location memory in 8-month-old-
infants in a non-search AB task. Develop-
mental Psychology, 24,502-11.
Baillargeon. R., and Hanko-Summes. S. 1990.
Is the top object adequately supported by
the bottom object? Young infants' under-
standing of support relations. Cognitive
Development, 5, 29-54.
Baillargeon, R .. Spelke, E.S., and Wasserman,
S. 1985. Object permanence in five-month-
old infants. Cognition, 20, 19l-208.
Chomsky, Noam. 1966. Cartesian linguistics.
In Acquisition and Use of Language. New
York: Harper and Row. In Stich. 1975.
Chomsky. Noam. 1975. Reflection of Language.
New York: Pantheon.
Crain, S. 1991. Language Learning. Behavioral
and Brain Sciences. 14: 4. 597-649.
Fodor. J. 1975. The Language of Thought. New
York: Thomas Y. Crowell, Co. Reprinted by
Harvard University Press. 1979.
Fodor, J. 1987. Psychosemantics: The Problem of
Meaning in the Philosophy of Mind. Cam-
bridge, MA.: MIT' Press.
Garfield, J.L. 1989a. Mentalese and mental se:
keeping language and thought distinct.
Conference on Mind, Meaning and Nature,
Wesleyan University.
Garfield, J.L., ed. 1989b. Modularity in Know-
ledge Representation and Natural Language
Understanding. Cambridge, MA.: MIT Press.
Goodman, N. 1972. The new riddle of induc-
tion. In Problems and Projects. Indianapolis:
Hackett.
Harman, G. 1969. Linguistic competence and
empiricism. In Language and Philosophy, ed.
S. Hook. New York: NYU Press.
Harman, G. 1975. Psychological aspects of the
theory of syntax. In Stich. 1975.
Hume, D. 1978. A Treatise of Human Nature.
Ed. Selby-Bigge and Niditch. Oxford Uni-
versity Press.
J ackendoff, R. 1983. Semantics and Cognition.
Kant, I. 1929. Critique of Pure Reason. Trans.
N. Kemp Smith. London: Macmillan. Origi-
nal work published in 1 787.
Karmiloff-Smith, A. 1991. Beyond modularity:
Innate constraints and developmental
change. In The Epigenesis of Mind: Essays on
Biology and Cognition, ed. S. Carey and R.
Gelman. Hillsdale, NJ: Lawrence Erlbaum
Associates, Publishers.
Kripke, S. 1982. Wittgenstein on Rules and
Private Language. Cambridge: Harvard Uni-
versity Press.
Marcus, G., S. Pinker, M. Ullman, M. Hol-
lander, T.J. Rosen, and F. Xu, 1992. Over-
regularization in Language Acquisition.
Chicago. Monographs of the Society for
Research in Child Development.
Marslen-Wilson, W.D., and L.K. Tyler. 1987.
Against modularity. In The Modularity of
Mind ed. J. L. Garfield. Cambridge, MA: MIT
Press.
Nelson, K. 1987. Nativist and functionalist
views of cognitive development; reflection
on Keil's review of making sense: The
acquisition of shared meaning. In Cognitive
Development. Norwood, New Jersey: Ablex
Publishing Corporation.
Piaget, J. 1929. The Child's Conception of the
World. London: Routledge & Kegan Paul.
Piaget, J. 1930. The Child's Conception of Physi-
INNATENESS
cal Causality. London: Routledge & Kegan
Paul.
Piaget, J. 1952. The Origins of Intelligence in
Childhood. New York: International Uni-
versities Press.
Piaget, J. 1954. The Construction of Reality in
the World. New York: Basic Books.
Piaget, J. 1969. The Child's Conception of
Movement and Speed. New York: Basic
Books.
Piaget, J. 1974. Understanding Causality. New
York: Norton.
Plunkett, K., and Marchman, V. 1990.
Regular and irregular morphology and the
psychological status of rules of grammar. In
Proceedings of the 17th Annual Meeting of the
Berkeley Linguistics Society. Berkeley, CA:
Berkeley Linguistics Society.
Plunkett, K., and Marchman, V. 1991. U-
multilayered perception: implications for
child language acquisition. Cognition, 38,
43-102.
Putnam, H. 1975. The 'innateness hypothesis'
and explanatory models in linguistics. In
Stich. 1975.
Putnam. Hilary. 1992. What is innate and
why: comments on the debate. In The Philo-
sophy of Mind. ed. B. Beakley and P. Ludlow.
Cambridge, MA.: MIT' Press.
Quine. W.V. 1954. The scope and language of
science. Bicentennial Conference at Colum-
bia University. Reprinted in The Ways of
Paradox and Other Essays, 1975. Cambridge,
Quine. W.V. 1960. Word and Object. Cam-
Quine. W.V. 1968. Linguistics and philosophy.
Lecture. Reprinted in Stich. 1975.
Rummelhart, D.E.. and McClelland. J.L. 1986.
On learning the past tenses of English verbs.
In Parallel Distributed Processing: Explorations
in the Microstructure of Cognition: Vol. 2.
Psychological and Biological Models, J.L.
McClelland. D.E. Rummelhart, and the PDP
Research Group. Cambridge. MA.: MIT'
Rummelhart. D.E., and McClelland. J.L.
1987. Learning the past tenses of English
verbs: implicit rules or parallel distributed
processing? In Mechanisms of Language
Acquisition. ed. B. MacWhinney. Hillsdale.
N.J. Erlbaum.
373
INTENSION AL
Ryle, G. 1949. The Concept of Mind. New York:
Barnes and Noble.
Sellars, W. 1963. Empiricism and the philoso-
phy of mind. In Science, Perception and
Reality. New York: Routledge & Kegan Paul.
Simon, H.A. 1981. The Sciences of the Artificial,
2nd edition. Cambridge MA.: MIT Press.
Solan, L. 1983. Pronominal Reference: Child
Language and the Theory of Grammar. Reidel.
Spelke, E.S. 1982. Perceptual knowledge of
objects in infancy. In Perspectives on Mental
Representation, ed. J. Mehler, M. Garrett, and
E. Walker. Hillsdale, N.J. Lawrence Erlbaum
Associates.
Speike, E.S. 1988. Where perceiving ends and
thinking begins: The apprehension of objects
in infancy. In Perceptual Development in
Infancy, ed. A Yonas. Minnesota Symposium
on Child Psychology, vol. 20, 191-234.
Hillsdale, N.J. Lawrence Erlbaum Associates.
Speike, E.S. 1990. Principles of object percep-
tion. Cognitive Science 14, 29-56.
Stich, P. 1975. Innate Ideas. University of Cali-
fornia Press.
tion. Oxford: Basil Blackwell.
JAY L. GARFIELD
intensional The word 'intensional' was
coined as the converse of 'extensional' and
both notions figure prominently in seman-
tics and logic. When logicians speak of the
extension of a predicate such as, for
example, 'is a horse', they have in mind the
set of creatures that this predicate picks out.
But in addition to its extension, a predicate
can be understood to have an intension,
some feature - grasped by anyone who has
linguistic mastery of the predicate - that
determines the extension. There are various
proposals in the literature for more specific
accounts of intensions in respect of pred-
icates. For example, some think of the
intension as something like the meaning of
the predicate, whilst others prefer to think
of it in more formal terms as the principle
(,function') that would determine the set of
things which are horses in any POSSIBLE
WORLD.
Predicates are not the only linguistic con-
structions which have extensions as well as
374
intensions. Most importantly, and due
mainly to the work of the logician and philo-
sopher Frege, it is usual to describe sentences
as having them. Thus, the extension of a
sentence is taken to be a truth value (either
true or false, or, in multi-valued logics, some
further value). And the intension of a sent-
ence could be variously understood as the
thought expressed by it, its meaning, or the
function from the sentence to a truth value
in any possible world. Additionally, one can
think of proper names as having both inten-
sion and extension. The extension of the
name 'Mark Twain' is the human being who
wrote various well-known books. Its inten-
sion is less easy to describe in everyday
terms: some think of it as the individual
concept that determines the author; others
prefer to stick with the possible-world defini-
tion and say that the intension of - 'Mark
Twain' is the function that picks out that
author in any possible world. In logic and
the philosophy of language, the whole ques-
tion of how to understand intensions is
fraught with difficulties, but the outlines of
the notion are clear enough.
Of special importance in the philosophy of
mind is the use of this pair of terms in con-
nection with sentences attributing proposi-
tional attitudes. Consider the sentence:
'John believes that Mark Twain wrote
Huckleberry Finn.' The sentence contained
within this attitude report, 'Mark Twain
wrote Huckleberry Finn' happens to be true.
In the terminology introduced above, one
would say that its extension was the truth
value true. Moreover, since Mark Twain was
Samuel Clemens, one could replace the
author's name in this sentence without
changing its extension. For 'Samuel
Clemens wrote Huckleberry Finn' is also
true. When such replacement or substitu-
tion does not change the truth value, the
sentence within which the substitutions are
made is said to constitute an 'extensional
context'. Returning now to the attitude
report that contains the sentence about
Mark Twain, one can see problems for sub-
stitution. The believer, John, may not know
that 'Mark Twain' was the name adopted
by Samuel Clemens and many feel that, in
this event, it is just false to say: 'John
believes that Samuel Clemens wrote Huckle-
berry Finn.' The fact that substitution fails
in this way to preserve extension (truth
value) is captured by saying that the belief
report constitutes an intensional context.
That is, it is a context where one can pre-
serve truth value only by substituting
expressions with the same intension. (It
should be noted here that it is controversial
whether one ought to appeal to intensions
in saying what is going on in this case.)
Though the word 'intension' has a tech-
nical sense in semantics deriving from its
opposite number 'extension', there is what
can be an initially confusing similarity
between 'intensional' and 'intentional'. The
second of this pair - the one spelled with a
't' in place of's' - has been widely used to
describe the feature of propositional atti-
tudes by which they are said to be about or
directed to something. Thus, with the
earlier example in mind, one would say that
John's belief is about or directed to Mark
Twain and his authorship of Huckleberry
Finn. Propositional attitudes have what phi-
losophers call 'INTENTIONALITY'. And con-
fusion can arise because, as was noted
above, propositional attitude reports, con-
stitute intensional (with an's') contexts.
See also BELIEF; CONCEPTS; PROPOSITIONAL
ATTITUDES.
SAMUEL GUTTENPLAN
intention Our concept of intention is a
central part of a web of concepts at the
heart of our commonsense conceptions of
intelligent agency. Included also in this web
are concepts of various attitudes, like belief,
knowledge, and desire; general notions of
meaning; and concepts that classify actions
- for example, concepts of intentional action
and of free action. (See FOLK PSYCHOLOGY.)
We use our concept of intention both to
classify ACTIONS and to characterize minds.
We do the former when we describe an
action as intentional or done with a certain
intention; we do the latter when we say
whether someone intends to do something,
INTENTION
now or later (Anscombe, 1963). A common
approach is to begin with action. We ask:
what is it to act intentionally, or with a
certain intention? A natural answer is that
for me to act intentionally is for my action
to be explainable, in an appropriate way, by
what I want (desire) and what I believe
(Anscombe, 1963; Davidson, 1980, essay
1). (See BELIEF, DESIRE.) My walking
towards the telephone is intentional. for
example, because it is explainable by my
desire to contact my friend and my belief
that I can do so by using the phone. Given
my desire and belief I have a reason to walk
towards the phone and I act because of that
reason; that is why my walking towards the
phone is intentional; and that is why I so
walk with the intention of contacting my
friend. To classify an action as intentional.
or as done with a certain intention, is, then,
to say it has the appropriate relation to the
agent's desires and beliefs. One acts inten-
tionally, or with a certain intention, just in
case one's action stands in the appropriate
relation to one's desires and beliefs; no dis-
tinct state or event of intending is involved
(DAVIDSON, 1980, essay 1).
We not only act intentionally and with
certain intentions. We also form intentions
concerning the future. What can such a
relational conception say about intending to
do something later? As Davidson himself
later noted (1980, essay 5), future-directed
intending cannot in general be simply a
relation between belief, desire, and action;
for though I now intend (for example) to go
swimming tomorrow there may never be a
relevant action. Nevertheless, many theor-
ists in this tradition try to extend to the case
of intending to do something later, the
machinery already introduced in the
account of intention in action. This leads
naturally to a reductive conception, one that
tries to analyse future-directed intending in
terms of the agent's relevant d e s ~ r ~ s and
beliefs. -
We can call this collection of views - the
methodological decision to begin with
intentional action, the relational conception
of intentional action and of action done
with an intention, and the reduction of
375
INTENTION
future-directed intending to desire and belief
- the desire-belief model of intention. Along
with this view has gone a certain (roughly,
Humean) conception of PRACTICAL REA-
SONING - reasoning about what to do. In
practical reasoning we weigh our relevant
desire-belief reasons for and against our
different alternative options. Such reasoning
typically issues in action. There is no clear
place in the model for a decision or
intention which can then function as a dis-
tinctive input into further practical reason-
ing; for decision and intention about the
future are understood solely in terms of
desire and belief, and intention in action is
simply implicit in the relation between
desire, belief, and action.
The desire-belief model of intention has
been, at least until recently, the standard
view in a wide range of philosophical
discussions of action. It is, for example, typi-
cally assumed in debates about whether
reasons are causes, including the large,
recent literature on the role of CONTENT in
psychological explanation (see REASONS
AND CAUSES). But recent years have
seen a re-examination of the desire-belief
model of intention. This has lead to the
development both of some more elaborate
versions of the desire-belief model and of
competing models according to which
intending is a distinctive kind of psychologi-
cal attitude.
A main theme in this re-examination of
the desire-belief model has been the idea of
commitment to action. Consider my inten-
tion now to go to Boston next week. In
intending so to act later I seem to be in
some sense committed to or settled on so
acting. In some important sense, the issue of
whether so to act is no longer a live issue for
me: the issue is settled. In being settled on
so acting I seem to go beyond merely
having desire-belief reasons for and/or
against this option and its alternatives. Yet
there is as yet no action.
What is this commitment to action that
seems characteristic of intention? My com-
mitment now to going to Boston next week
does not reach its ghostly hand over time
and control my action next week; that
376
would be action at a distance. Nor should
such commitment mean that I am irrevoc-
ably settled on so acting. Such irrevocability
would clearly be irrational; after all, things
change and we do not always correctly
anticipate the future. But if my commitment
now does not itself control my action, and is
subject to review and change later, why
bother? Why not just cross my bridges
when I come to them? So, on the one hand,
it may seem that an account of intention
really must provide for a distinctive kind of
commitment to action. Yet, on the other
hand, the very idea of this commitment
leads to puzzles of its own.
One might try to come to terms with
commitment while staying within the
desire-belief framework by emphasizing the
relation between an intention to A and a
belief that one will. In particular, one might
try to see intention as itself a special kind of
belief about one's own conduct. In intend-
ing now to A later I see the question of
whether so to act as settled in the sense that
I now am confident that I will so act. My
intention to A is not simply some desire-
belief reason in favour of my A-ing, or even
the fact that such a reason is in some sense
the strongest of my relevant, competing
reasons. In intending to A I go beyond such
reasons and believe I will in fact A. But
though I go beyond my desire-belief reasons
for A-ing, my intention is not a fundament-
ally new kind of attitude. In this way one
can try to accommodate the special com-
mitment involved in intention without
going beyond the resources of the desire-
belief framework.
Of course, many beliefs about one's future
conduct are not intentions. I might believe I
will miss a shot in a basketball game and
yet not intend to miss, and in a race I might
believe I will wear down my sneakers but
not intend to wear them down. (In the ter-
minology of Bentham (1789) I might be
said 'obliquely' to intend to wear down my
sneakers, whereas I 'directly' intend to run.
But our concern here is with so-called
'direct' intention.) A defender of a belief-
conception of intention will need to specify
further the type of belief about my own
conduct in which my intention consists.
Velleman (1989), for example, argues that
my intention to A is my belief, adopted out
of my desire for its truth, that I will A as a
result of this very belief. This introduces two
further ideas. The idea that intentions
involve a kind of self-referentiality is devel-
oped in a number of recent studies, many of
which reject the identification of intention
and belief. (See Donagan, 1987; Harman
1986; SEARLE, 1983.) The idea that inten-
tions are beliefs that are adopted out of rel-
evant desires raises issues that had been
discussed earlier by Grice (1971).
Grice had argued that my intention to A
does indeed require a belief on my part that
I will A. In this respect Velleman's account
follows Grice's. But Grice went on to argue
that my intention cannot just be identified
with some such belief. Suppose I intend to
go to London tomorrow. If this is a belief of
mine that I will go, we may ask what my
justification is for that belief. Grice worried,
roughly, that if my justification for my belief
were my desire to go, that would make my
case one of wishful thinking. But if my jus-
tification were instead some form of induc-
tive evidence (as when I expect I will sneeze
when I am exposed to your cat) then my
belief would be an ordinary prediction and
not an intention. Grice concluded that my
intention to go, while it involved my belief
that I will go, could not be identified simply
with my belief. Instead, my intention also
involves a special attitude of willing that I
go. Willing that I go does not itself require
that I believe I will; but it does involve
something like wishing wholeheartedly that
I go - so willing is not merely desiring. To
intend to go is to will that I go and to
believe that, as a result of such willing, I
will go. My intention involves the belief that
I will go; but this belief is based on evidence
that includes my knowledge that that is
what I will. (See THE WILL.)
Grice and Velleman agree that an inten-
tion to A at least involves a belief that one
will A. But this may be challenged. In a
later paper (1980, essay 5) Davidson gave
up on his earlier claim that talk of intention
was never talk about some state or event of
INTENTION
intending. But he also argued that an inten-
tion to A does not require a belief that one
will A (though it does require that one not
think it impossible that one will A). I might
intend to achieve some goal and yet be
doubtful about my ultimate success. Rather
than understand intention as consisting, at
least in part, in a belief that one will so act,
Davidson argued that we could see inten-
tion as a special kind of evaluation of
conduct. Davidson treats a desire to act in a
certain way as a prima facie judgment that
so acting would be, in some respect, desir-
able. An intention to act is also an evalu-
ative judgment; but it is not merely a prima
facie judgment. It is what Davidson calls an
'all-out judgment'. My intention to go to
London, for example, is my 'all-out judg-
ment' that my so acting would be best. In
reaching such an all-out evaluative judg-
ment I am settled on so acting. And David-
son took great pains to argue that, contrary
to initial appearances, this view is com-
patible with the possibility of weak-willed
intentions - intentions that are contrary to
one's best judgment about what to do (see
WEAKNESS OF WILL).
Return now to practical reasoning. As
noted, it is common to see such reasoning
as consisting in the weighing of desire-belief
reasons for and against conflicting alter-
natives. But we also sometimes decide to act
in a certain way and then reason about
how. I can decide - and so form an inten-
tion - to go to London next week. Given
this new intention I will need to reason
about how I am going to get there. In such
reasoning my intention to go to London
provides an end, and my reasoning aims at
settling on appropriate means. A general
conception of practical reasoning should
make room for, and explain the inter-
relations between, both kinds of practical
reasoning - the weighing of desire-belief
reasons, and reasoning about how to do
what one already intends to do.
These observations suggest yet another
way of approaching intention. We try to say
how intentions function in practical reason-
ing, not only as outputs of such reasoning,
but also as characteristic inputs to further
377
INTENTION
reasoning. Castaneda (1975) has developed
a version of this approach. He sees intention
as the acceptance of a distinctive kind of
content - what he calls a 'first-person prac-
tition'. A practition is, roughly, the content
normally involved in commands and pre-
scriptions. If, for example, I order you to
open the window my command has the
content: You to open the window. Typical
commands are addressed to others. But
intentions are analogous to self-addressed
commands - to intend is to accept a jirst-
person practition. To intend to open the
window is to endorse the practition: I to
open the window. This is not to say that for
me so to intend I must literally perform a
speech act of commanding myself so to act.
It is only to say that the contents of intend-
ings have a structure and logic analogous
to that of the contents of ordinary com-
mands and prescriptions.
Castaneda tries to develop an account of
the special logical relations between such
practitions and ordinary propositions, the
contents of beliefs. At the heart of this
special logic is the idea that a practition, S
to A, neither implies nor is implied by its
corresponding proposition that S A's. The
underlying motivation for trying to develop
such a special logic includes two ideas. (a)
We understand intention in large part in
terms of its roles in practical reasoning; (b)
We understand those roles by articulating a
special logic of the contents that are dis-
tinctive of intention.
But we can accept (a) without following
Castaneda in also accepting (b). We can try
to layout the roles of intention in practical
reasoning - including its roles as a dis-
tinctive input to such reasoning - without
supposing that we need some special logic
of the contents of intendings. For we can
distinguish a theory of reasoning from a
theory of logical implication (Harman,
1986).
The idea is to say what intentions are in
part by saying how they function in prac-
tical reasoning, without assuming that to
do this we must articulate a special logic of
the distinctive contents of intentions. In
particular, we try to articulate the roles of
378
future-directed intentions in practical rea-
soning, and we try to say how these roles
differ from those played by desires and
beliefs (Harman, 1986; Bratman, 1987).
An account of the roles of intentions in
practical reasoning should support, and be
supported by, answers to our earlier ques-
tion: why bother with intentions for the
future anyway? Here one may argue
(Bratman, 1987) that there are two main
answers. First, we are not frictionless delib-
erators. Deliberation is a process that takes
time and uses other resources, so there are
limits to the extent of deliberation at the
time of action. By settling on future-directed
intentions we allow present deliberation to
shape later conduct, thereby extending the
influence of Reason on our lives. Second, we
have pressing needs for coordination, both
intra-personal and social; and future-
directed intentions playa central role in our
efforts at achieving such coordination.
Future-directed intentions typically play
these roles as elements of larger, partial
plans. My intention to go to London this
Sunday helps coordinate my various activ-
ities for this weekend, and my activities
with the activities of others, by entering
into a larger plan of action. Such plans will
typically be partial and will need to be filled
in as time goes by with appropriate specifi-
cations of means, preliminary steps, and the
like. In filling in such partial plans in stages
we engage in a kind of practical reasoning
that is distinctive of planning agents like us
and that is, on the view being considered,
central to our understanding of intention.
Such approaches seek to articulate the basic
principles involved in such planning. In this
way they try to shed light on the nature of
intention and on its relations to other psy-
chological attitudes, like belief and desire
(Harman, 1986; Bratman, 1987).
Such planning conceptions of intention
focus on intention as a state of mind, and
pay special attention to future-directed
intentions and plans. But it can be asked
whether the structures that are thereby
revealed are peculiar to the special case of
sophisticated agents engaged in planning
for the future, and not generally applicable
to all cases of intentional action (see Wilson.
1989). Are there agents capable of inten-
tional action but not capable of such future-
directed planning? Even for planning
agents. are there some intentional activities
(e.g. relatively spontaneous and yet inten-
tional conduct) that do not plausibly
involve the kinds of psychological structures
highlighted by a focus on future-directed
intention?
These queries return us to Anscombe's
(1963) general question of how to under-
stand the relation between intention as a
state of mind and the characterization of
action as intentional or done with a certain
intention. According to what has been
called 'the Simple View' intentional A-ing
always involves intending to A. But some
argue that there are cases in which one
intentionally As and yet. though one does
intend something one does not. strictly
speaking. intend to A (Harman. 1986;
Bratman. 1987). The concerns raised in the
previous paragraph suggest that planning
conceptions of intention may drive an even
greater wedge between intention as a state
of mind and the characterization of action
as intentional. If so. we need to ask whether
that is an acceptable conclusion. Anscombe
(1963. p. 1) had urged that it is implausible
that 'intention' is equivocal in these differ-
ent contexts. In contrast. Velleman (1989.
pp. 112-13) argues that we should give up
on the search for a unitary meaning of
'intention' .
These debates about intention interact
with a wide range of issues. For example. it
is common to suppose that an agent's
desires and beliefs give her reasons for
action. If we see intention as a distinctive
attitude. over and above an agent's desires
and beliefs. we will want to know how this
attitude affects what reasons the agent has
to act in various ways and what it is
rational of the agent to do (see RATION-
ALITy). Again. we frequently talk about the
intentions of groups of agents. Speaking for
you and me. I might say that we intend to
sing a duet together. We are social crea-
tures: shared intentions and shared inten-
tional activities pervade our lives together.
INTENTION ALITY (1)
A theoretical approach to intention should
help us understand these important phe-
nomena.
BIBLIOGRAPHY
Anscombe. G.E.M. 1963. Intention. 2nd edn.
Ithaca: Cornell University Press.
Bentham. J. 1789. An Introduction to the Prin-
ciples of Morals and Legislation. London:
Methuen. 1982.
Bratman. M.E. 1987. Intention. Plans. and
Practical Reason. Cambridge MA.: Harvard
University Press.
Castaneda. H.N. 1975. Thinking and Doing.
Dordrecht: Reidel.
Davidson. D. 1980. Essays on Actions and
Donagan. A. 1987. Choice: The Essential
Element in Human Action. London: Routledge
& Kegan Paul.
Grice. H.P. 1971. Intention and uncertainty.
Proceedings of the British Academy. 57.263-
79.
Harman. G. 1986. Change in View. Cambridge
MA.: MIT Press.
Searle. J. 1983. Intentionality. Cambridge Uni-
versity Press.
Velleman. J.D. 1989. Practical Reflection.
Princeton University Press.
Wilson. G.M. 1989. The Intentionality of
Human Action: Revised and Enlarged Edition.
Stanford University Press.
MICHAEL E. BRATMAN
intentionality (1) The concept of inten-
tionality was originally used by medieval
scholastic philosophers. It was reintroduced
into European philosophy by Franz Bren-
tano in the nineteenth century. In its
current usage the expression 'intentionality'
refers to that property of the mind by which
it is directed at. about. or of objects and states
of affairs in the world. Intentionality. so
defined. includes such mental phenomena
as BELIEF. DESIRE. INTENTION. hope. fear.
love. hate. lust. disgust (see EMOTION). and
MEMOR Y as well as PERCEPTION and inten-
tional ACTION.
The concept of intentionality is a source
of at least two sorts of confusion. First there
379
INTENTIONALITY (1)
is a temptation to confuse intentionality-
with-a-t, the capacity of the mind to rep-
resent objects and states of affairs in the
world, with intensionality-with-an-s, the
property of certain sentences by which they
fail certain sorts of tests for extensionality
(more of this latter distinction later). A
second sort of confusion for English speak-
ers is to suppose mistakenly that 'intention-
ality' as a technical notion in philosophy
has some special connection with 'intend-
ing' in the ordinary sense, in which, for
example, one intends to go to the movies
tonight. Intending in the ordinary sense is
just one form of intentionality along with
belief, desire, hope, fear, etc.
INTENTIONALITY AND CONSCIOUSNESS
Some intentional states are conscious, some
not. The belief that George Washington was
the first president can be consciously enter-
tained, but a person can have that belief
while he or she is sound asleep. In such a
case, the intentional state is unconscious
(see CONSCIOUSNESS; THE UNCONSCIOUS).
Furthermore, many forms of consciousness
are intentional. but many are not. Thus, a
conscious desire to drink a cold glass of beer
or a conscious fear of snakes are both inten-
tional. but a feeling of pain or a sudden
sense of anxiety, where there is no object of
the anxiety, are not intentional. In short,
there is an overlap between consciousness
and intentionality but they are by no means
coextensive.
THE IRREDUCIBILITY OF
INTENTION ALITY
Many attempts have been made to eliminate
intentionality or analyse it away in terms of
some simpler notions. Thus, in the behaviour-
ist period of the philosophy of mind, many
philosophers felt that having a state of belief
or desire was simply a matter of being dis-
posed to behave in certain ways, given
certain stimuli (see BEHAVIOURISM). Sub-
sequent functional theories try to analyse
intentionality in terms of causal relations (see
FUNCTIONALISM). For example, on function-
380
alist accounts, a belief is analysed as a
certain functional state which is caused by
external stimuli, and which, in conjuction
with other states such as desires, causes
certain sorts of external behaviour. More
recent versions of functionalism try to iden-
tify intentional states with computational
states. The idea is that being in a mental
state is just being in a certain state of a com-
puter program, and the mind is construed as
a computer program running in the wetware
of the brain. On this view, called 'strong arti-
ficial intelligence' or 'computer functional-
ism' the mind is to the brain as the program
is to the hardware (see ARTIFICIAL INTEL-
LIGENCE; COMPUTATIONAL MODELS).
All of these attempts to analyse intention-
ality fail for the reason that they try to
reduce intentionality to something else (see
REDUCTION). As soon as one recognizes the
existence of intentionality as a genuine phe-
nomenon, one is committed to rejecting any
reductive or eliminative account of inten-
tionality; for such accounts implicitly deny
that the phenomenon exists as a genuine
feature of the world.
This discussion will not attempt to estab-
lish the existence of intentional states, such
as beliefs and desires, but will simply take
their existence for granted and explore their
logical properties.
THE STRUCTURE OF INTENTIONAL
STATES
I have characterized intentionality as that
feature of the mind by which it is directed at
or about or of objects or states of affairs in
the world, but that only raises the further
question, what is meant by 'directed at',
'about', or 'of'? This question becomes more
preSSing when one realizes that an inten-
tional state can be about something even
though the thing it is about does not exist.
Thus a child can have a belief that Santa
Claus will come on Christmas Eve even
though Santa Claus does not exist. The sim-
plest way to explain the structure of inten-
tional states is to compare them with speech
acts. It is not surprising that intentional
states and speech acts have parallel struc-
tures since every speech act is the expres-
sion of a corresponding intentional state.
For example. every statement is an expres-
sion of a belief. Every promise is the expres-
sion of an intention. etc. Notice that this
point holds even when the speaker is insin-
cere. even when he does not have the belief.
intention. etc .. which he expresses.
There are several structural similarities
between speech acts and intentional states.
among them the following three:
(1) The distinction familiar in the theory of
speech acts between the force or type of the
speech act and the propositional content of
the speech act carries over exactly into the
distinction in intentional states between
the type of the intentional state and its
propositional content.
In the theory of speech acts. there is a
familiar distinction between illocutionary
force and propositional content. A speaker
can perform three different types of speech
act in uttering the three different sentences:
Leave the room!
You will leave the room.
Will you leave the room?
even though there is something common to
all three utterances. In each utterance. the
propositional content. that you will leave
the room. is expressed. but it is expressed in
speech acts with different illocutionary
forces. The first has the force of an order or
request; the second has the force of the pre-
diction; and the third has the force of a
question. The general structure of the
speech act as exemplified by these three
cases is F(p). where the 'F' marks the
illocutionary force of an order. question.
statement. etc.; and the 'po marks the pro-
positional content. that you will leave the
room. These distinctions carryover exactly
to intentional states. Just as one can assert.
query. or order that you leave the room. so
one can hope that you will leave the room.
fear you will leave the room. believe you
INTENTIONALITY (1)
will leave the room. wish you would leave
the room. etc. In each case. the same pro-
positional content - that you will leave the
room - is presented in a different psycho-
logical mode - the psychological mode of
hope. fear. belief. etc. The structure of such
intentional states is S(p).
(2) The distinction between different directions
of fit. also familiar from the theory of speech
acts. carries over to intentional states.
Statements. for example. are supposed to
represent an independently existing reality.
and insofar as they succeed or fail in repre-
senting it accurately. they are said to be
true or false. Thus. for example. if one
makes the statement that John has left the
room. one's statement will be true or false
depending on whether or not John has left
the room. But orders. commands. promises
are not like statements in that they are not
supposed to match an independently exist-
ing reality. but rather. they are supposed to
bring about changes in reality so that the
world comes to match the propositional
content of the order. command. or promise.
Thus. if an order is given to John. 'John.
please leave the room!'. then the order is
not said to be true or false. but is said to be
obeyed or disobeyed. depending on whether
or not John's subsequent behaviour comes
to match the propositional content of the
order.
In the first sort of case - statements. asser-
tions. descriptions. etc. - the utterance has
the word-to-world direction of fit; it is true or
false depending on whether or not the words
match the world. But in the second sort of
case - orders. commands. promises. etc. -
the utterance has the world-to-word direc-
tion of fit; and the utterance is said to be
obeyed. fulfilled. kept. etc.. depending on
whether or not the world comes to match
the propositional content of the utterance.
'Truth' and 'falsity' are names for success
and failure in the word-to-world direction of
fit.
Some utterances with a propositional
content do not have a direction of fit of
either word-to-world or world-to-word.
381
INTENTION ALITY (1)
These are cases where the propositional
content is simply taken for granted as a pre-
supposition. Thus, for example, if I apologize
for stepping on your foot, or congratulate
you on winning the race, in each case there
is a propositional content - that I stepped
on your foot, that you won the race - but
the aim of the utterance is neither to assert
that propositional content nor to try to
change the world by getting the world to
match the propositional content. In all three
cases, the propositional content is simply
taken for granted, and we can say therefore
that such speech acts have the null direc-
tion of fit. Similarly there are intentional
states with a propositional content where
the truth of the propositional content is
taken for granted. Feeling glad that you
won the race or feeling sorry that I stepped
on your foot have propositional contents,
but like the corresponding speech acts, they
have the null direction of fit.
In summary, there are certain exact par-
allels in the way that the notion of direction
of fit applies to both speech acts and inten-
tional states. Beliefs, like statements, have
the word (or mind)-to-world direction of fit,
and like atatements, they can be said to be
true or false. Desires and intentions have
the world-to-word (or mind) direction of fit,
and like orders and promises, they can be
said to be carried out or fulfilled, but cannot
be said to be true or false. Joy and sadness,
sorrow and gladness like apologies and
thanks have no direction of fit.
(3) The notion of conditions of satisfaction
applies generally to both speech acts and to
intentional states.
Statements are said to be true or false;
orders are said to be obeyed or disobeyed;
promises are said to be kept or broken.
What stands to the statement being true is
what stands to the order being obeyed is
what stands to the promise being kept. In
each case we can say quite generally that
the speech act will be satisfied or not satis-
fied depending on whether or not the pro-
positional content comes to match the
world with the appropriate direction of fit.
382
For all such cases we can say that the
speech act represents its conditions of satis-
faction, and the illocutionary force deter-
mines the direction of fit with which it
represents its conditions of satisfaction.
Exactly analogously in the structure of
intentional states, what stands to the
belief's being true is what stands to the
desire's being fulfilled is what stands to the
intention's being carried out. In each case
the intentional state with a direction of fit
has conditions of satisfaction, and we can
say that the intentional state is a repre-
sentation of its conditions of satisfaction,
and the psychological mode determines the
direction of fit with which the intentional
state represents its conditions of satisfaction.
On this account, then, the key to under-
standing intentionality is representation in
a special sense of that word that we can
explain from our theory of speech acts.
Every intentional state with a direction of fit
is a representation of its conditions of satis-
faction. Intentional states in general have
both a propositional content and a psycho-
logical mode, and the psychological mode
will determine the direction of fit with
which the intentional state represents its
conditions of satisfaction.
This account of intentionality is quite
general, but it still does not account for two
sorts of phenomena. First, what about those
intentional states that have the null direc-
tion of fit. In what sense do they have con-
ditions of satisfaction? And secondly, what
about those intentional states that do not
have an entire propositional content, such
as admiring Einstein or hating Hitler?
If we consider such intentional states
which do not have a direction of fit - as
being glad that the Republicans won the
election, or feeling sorry that the value of
the dollar has declined - we find that each
of these contains both beliefs and desires,
and these component beliefs and desires
have a direction of fit. In general one can
say, for example, that in order to be glad
that p, one must believe it to be the case
that p and to want it to be the case that p.
In order to be sorry that p, one must believe
it to be the case that p and want it to be the
case that not p. This phenomenon is char-
acteristic of all of those intentional states
with a propositional content which do not
have a mind-to-world or world-to-mind
direction of fit: all of these contain beliefs
and desires. and the component beliefs and
desires do have a direction of fit.
This suggests a pattern of analysis for the
EMOTIONS which in the space of this article
I can only sketch. Emotions such as love
and hate. jealousy. anger. and lust all have
certain features which are somewhat
unusual among intentional states. They all
matter to us in ways that many of our
beliefs and desires are regarded with relative
indifference. For example. I believe it is
raining somewhere in the world right now.
but I don't care much about it. When it
comes to our emotions. however. of love
and hatred. jealousy. and lust etc .. we do
care desperately about these. Why? In
general. emotions are very strong agitated
forms of desire. It is because the emotion is
itself a form of strong desire. typically a
desire caused by a belief which is also a
component of the emotion. that one cares
so strongly about one's emotions. A typical
strong emotion will contain a whole
package of desires. Lovers. for example.
notoriously have varied and complex desires
toward the beloved. For most of what philo-
sophers have called 'the emotions' - love.
hate. anger. jealousy. envy. etc. - an essen-
tial component of the emotion is an agitated
form of desire.
How do we analyse those intentional
states which only make reference to a single
object and do not contain an entire proposi-
tional content. such as loving Sally or
admiring Mother Theresa? Though none of
these can be completely analysed into pro-
positional contents that have directions of
fit. in general. they require the presence of
beliefs and desires. and the beliefs and
desires do have a direction of fit.
THE INTENTIONALITY OF PERCEPTION
AND ACTION
It is characteristic of discussions of inten-
tionality that the paradigm cases discussed
INTENTION ALITY (1)
are usually beliefs or sometimes beliefs and
desires. However. the biologically most basic
forms of intentionality are in perception and
in intentional action. These also have
certain formal features which are not
common to beliefs and desires. Consider a
case of perception. Suppose I see my hand
in front of my face. What are the conditions
of satisfaction? First. the perceptual experi-
ence of the hand in front of my face has as
its condition of satisfaction that there be a
hand in front of my face. Thus far the con-
dition of satisfaction is the same as the belief
that there is a hand in front of my face. But
with perceptual experience there is this dif-
ference: in order that the intentional
content be satisfied. the fact that there is a
hand in front of my face must cause the
very experience whose intentional content
is that there is a hand in front of my face.
This has the consequence that perception
has a special kind of condition of satisfac-
tion that we might describe as 'causally self-
referential'. The full conditions of satisfac-
tion of the perceptual experience are. first
that there be a hand in front of my face.
and second. that there is a hand in front of
my face caused the very experience of
whose conditions of satisfaction it forms a
part. We can represent this in our canonical
form. S(p). as follows:
Visual experience (that there is a hand
in front of my face and the fact that
there is a hand in front of my face is
causing this very experience.)
Furthermore. visual experiences have a kind
of conscious immediacy not characteristic of
beliefs and desires. A person can literally be
said to have beliefs and desires while sound
asleep. But one can only have visual experi-
ences of a non-pathological kind when one
is fully awake and conscious because the
visual experiences are themselves forms of
consciousness.
Event memory is a kind of halfway house
between the perceptual experience and the
belief. Memory. like perceptual experience.
has the causally self-referential feature.
Unless the memory is caused by the event of
383
INTENTION ALITY (1)
which it is the memory. it is not a case of
a satisfied memory. But unlike the visual
experience. it need not be conscious. One
can literally be said to remember something
while sound asleep. Belief. memory and per-
ception all have the mind-to-world direction
of fit. and memory and perception have the
world-to-mind direction of causation. (See
MEMOR Y; PERCEPTION; PERCEPTU AL
CONTENT.)
Intentional action has interesting sym-
metries and asymmetries to perception. Like
perceptual experiences. the experiential
component of intentional action is causally
self-referential. If. for example. I am now
walking to my car. then the condition of
satisfaction of the present experience is that
there be certain bodily movements. and that
this very experience of acting cause those
bodily movements. Furthermore. like per-
ceptual experience. the experience of acting
is typically a conscious mental event.
However. unlike the case of perception and
memory. the direction of fit of the experi-
ence of acting is world-to-mind. and the
direction of causation is mind-to-world. My
intention will only be fully carried out if the
world changes so as to match the content of
the intention (hence world-to-mind direc-
tion of fit) and the intention will only be
fully satisfied if the intention itself causes
the rest of the conditions of satisfaction
(hence. mind-to-world direction of causa-
tion).
Furthermore. just as in the case of the
cognitive faculties. we needed to distinguish
between perception and memory. so in the
case of the volitional faculties. we need to
distinguish between the actual experience of
acting which we have while carrying out
an intentional action and a prior intention
that we sometimes form prior to the carry-
ing out of the action. We may dub these
two forms of intentionality as the intention-
in-action and the prior intention. Both
intentions-in-action and prior intentions are
causally self-referential. but the prior inten-
tion. unlike the intention-in-action. need
not be a conscious experience. As with
memory. a man who is sound asleep can be
said to have prior intentions.
384
THE NETWORK AND THE BACKGROUND
All intentional states only function in rela-
tion to other intentional states. and thus. in
any case of intentionality. the intentional
state only functions within a network of
intentionality. For example. if I form the
intention to go to Europe. I can only have
that intention within a network of other
beliefs. desires. and intentions. I must
believe that Europe is a certain distance
away. that it can be reached by airplane.
that the cities I wish to visit are located
within Europe. etc. I must desire that the
plane I am on will go to Europe and not to
Asia. that it will be able to take off. etc. I
must intend to buy a plane ticket. to pack
my bags. to go to the airport. to get on the
plane. etc. One intentional state only func-
tions in connection with an indefinite
number of other intentional states. This
phenomenon is called the 'Network' of
intentionality.
If we follow out the threads in the
Network. we soon reach a series of mental
capacities that are not themselves further
intentional states. There is a large bedrock
of capacities. abilities. tendencies. etc .. that
are simply taken for granted or presupposed.
but do not themselves form the structure of
the intentional network. Thus. for example.
in my dealings with the world. I presuppose
the solidity of objects and the traversability
of three-dimensional space. I presuppose the
persistence of continents such as Europe
and America. and that most of human life
goes on at or near the surface of the earth.
All of this forms a pre-intentional Back-
ground to my forming the intention to go to
Europe. but it is not itself a matter of further
intentional contents. This Background of
intentionality consists in the various abil-
ities. skills. and competences that I have for
engaging in various physical and social
activities. I know how to walk across the
room. to buy a plane ticket. to get on an
airplane. to sit in a seat. etc. In general.
where intentional action is concerned. one
may say that an agent's abilities rise to the
level of the Background skill. but for that
very reason. they reach to the bottom of the
physical exercise of that skill. Let us con-
sider each of these points in turn.
The man who has the ability to walk from
his home to his office does not have to form
a separate intention for each leg movement.
He simply forms the intention to walk to his
office, and then he simply does it. His inten-
tionality rises to the level of his Background
abilities. Nonetheless, each of the subsidiary
voluntary movements within the execution
of his intention is performed intentionally.
Thus, each leg movement is intentional,
though there is no separate intention deter-
mining the leg movements. How can this
be? The answer is simply that the top-level
intention, by invoking the Background
ability, by deriving its conditions of satisfac-
tion from the Background ability, governs
each of the voluntary movements within the
execution of the skill determined by the
Background ability. The only intention is
the intention to walk to his office. But the
foot and leg movements are not thereby
rendered unintentional. They are intention-
ally performed as part of the higher-level
intentional act. (See ACTION; THE WILL.)
INTENTIONALITY-WITH-A-T AND
INTENSION ALITY - WITH-AN-S
A standard confusion in the philosophical
literature is to suppose that there is some
special connection between intentionality-
with-a-t and intensionality-with-an-s. Some
authors even allege that these are identical.
But in fact, the two notions are quite dis-
tinct. Intentionality-with-a-t is that property
of the mind by which it is directed at, or is
about objects and states of affairs in the
world. Intensionality-with-an-s is that phe-
nonemon by which sentences fail to satisfy
certain tests for extensionality. Let us now
consider the relations between them.
There are many standard test for exten-
sionality, but the two most common in the
literature are substitutability of identicals
(LEIBNIZ'S LAW) and existential inference.
The principle of substitutability states that
co-referring expressions can be substituted
for each other without changing the truth
value of the statement in which the sub-
INTENTION ALITY (1)
stitution is made. The principle of existential
inference states that any statement which
contains a referring expression implies the
existence of the object referred to by that
expression. But there are statements that do
not satisfy these principles and such state-
ments are said to be intensional with
respect to these tests for extensionality. An
example of each is as follows.
From the statement that
(1) The sheriff believes that Mr Howard is
an honest man
and
(2) Mr Howard is identical with the notor-
ious outlaw, Jesse James
it does not follow that
(3) The sheriff believes that the notorious
outlaw, Jesse James, is an honest man.
This is a failure of the substitutability of
identicals.
From the fact that
(4) Billy believes that Santa Claus will
come on Christmas Eve
it does not follow that
(5) There is some x such that Billy believes
x will come on Christmas Eve.
This is a failure of existential inference.
Thus, statements (1) and (4) fail tests for
extensionality and hence are said to be
intensional with respect to these tests.
What, then, exactly is the relation
between intentionality-with-a-t and inten-
sionality-with-an-s? Notice that the sen-
tences that are intensional-with-an-s are
about states that are intentional-with-a-t.
The truth conditions of these intensional-
with-an-s sentences do not require that the
world be as represented by the original
intentional states, but only that the content
385
INTENTION ALITY (2)
of the intentional state be as represented in
the sentences about those intentional states.
Since intentional-with-a-t states are repre-
sentations, and since the content of the
representation can be reported indepen-
dently of whether or not it is satisfied, or
even independently of whether or not the
objects purportedly referred to by the repre-
sentation even exist, the report of the inten-
tional state does not commit the person
making the report to the existence of the
objects referred to by the original repre-
sentation (existential generalization); nor
does the report necessarily remain true
under substitution of co-referring expres-
sions in the report (substitutability).
The explanation for the failure of the tests
is that the ground floor intentional states in
the minds of the sheriff and Billy are repre-
sentations, but their reports in sentences
such as (1) and (4) are representations of
representations. The truth of the representa-
tion of the representation depends not on
how things are in the real world represented
by the original intentional representation
(the original intentional state), but rather,
how they are in the mental world of that
intentional representation. And that mental
representation can be reported accurately
even though the objects purportedly referred
to by that representation do not exist. This
accounts for the failure of existential gen-
eralization. And the expressions occurring
in the report, since they are not used to
refer to any such objects, but only express
the content of a representation, are not
subject to the law of the substitutability of
co-referring expressions. The substitution of
any such expression may fail to preserve the
mental content of the original intentional
state being reported, and thus such sub-
stitutions cannot guarantee sameness of
truth value. This accounts for the failure of
the substitutability test.
On this account, intentionality-with-a-t
and intensionality-with-an-s are quite dis-
tinct phenomena. The only connection is
that characteristically reports of intentional-
with-a-t states are intensional-with-
an-s reports, for the reasons I have just
given. Other sorts of sentences such as
386
modals, for example, also are intensional-
with-an-s.
THE INTENTIONALITY OF MEANING
SO far I have defined intentionality in such
a way that it applies only to mental phe-
nomena. Such states as beliefs and desires
are intrinsically intentional. But just as
beliefs and desires can represent states of
affairs in the world, so can sentences, pic-
tures, symbols and a host of other non-
mental phenomena. In all such cases the
intentionality of the mind is imposed on
some non-mental phenomena. Mental states
have intrinsic intentionality, material objects
in the world that are used to represent
something have derived intentionality. The
most important form of derived intention-
ality is in language and there is a special
name in English for this form of intention-
ality. It is called 'meaning' in one of the
many senses of that word.
See also An Essay on Mind section 2.1.
CONTENT; FOLK PSYCHOLOGY; PRO-
POSITION AL ATTITUDES ; REPRESENTATION;
THOUGHT AND LANGUAGE
JOHN R. SEARLE
intentionality (2) Many mental states and
activities exhibit the feature of intention-
ality: being directed at objects. Two related
things are meant by this. First, when one
desires or believes or hopes, one always
desires or believes or hopes something. Let's
assume that belief report (l) is true:
(1) Dan Quayle believes that George Bush
is a Republican.
(1) tells us that a subject, Quayle, has a
certain attitude, belief, to something, desig-
nated by the nominal phrase that George
Bush is a Republican and identified by its
content-sentence,
(2) George Bush is a Republican.
Following Russell and contemporary
usage I'll call the object referred to by the
that-clause in (1) and expressed by (2) a
proposition. Notice that this sentence might
also serve as Quayle's belief-text, a sentence
he could utter to express the belief that (1)
reports him to have. Such an utterance of
(2) by itself would assert the truth of the
proposition it expresses, but as a part of (1)
its role is not to assert anything, but to
identify what the subject believes. This same
proposition can be the object of other atti-
tudes and of attitudes of other people. Dole
may regret that Bush is a Republican,
Reagan may remember that he is, Bucha-
nan may doubt that he is.
The second way in which mental states
and activities are directed at objects has to
do with more familiar sorts of objects:
spatiotemporal objects like persons and
things, abstract particulars like numbers,
and universals like the property of being a
Republican or the relation of standing next
to. The truth of a proposition requires that
certain objects will have or come to have
certain properties or stand in certain rela-
tions. The attitude is about these objects,
properties, and relations. Quayle's belief, for
example, is about George Bush and the prop-
erty of being a Republican. While we can
have attitudes about ourselves and even our
own mental states, we have an enormous
number of attitudes about other things (see
PROPOSITION AL ATTITUDES). These things
may be quite remote from us. I have never
seen Bill Clinton in person, never talked to
him. But I have many beliefs, hopes, desires,
doubts, and fears about him. We all have
attitudes about people who are long dead,
like Aristotle and Caesar, and things that are
millions of miles away, like the planet Pluto.
This can seem rather puzzling. What exactly
is the relation between a subject, and the
objects about which they have attitudes?
The concept of intentionality was intro-
duced into modern philosophy by Franz
Brentano, who took what he called 'inten-
tional inexistence' to be a feature that dis-
tinguished the mental from the physical
(Brentano, 1960). In this article, we focus
on two puzzles about the structure of inten-
tional states and activities, an area in which
the philosophy of mind meets the philoso-
INTENTION ALITY (2)
phy of language, logic, and ontology. We
need to note that the term intentionality
should not be confused with the terms
intention and intension. Intentions, such as
Bush's intention to run for re-election, are
one kind of intentional state. Intensions are
properties or concepts, as opposed to exten-
sions: objects and sets of objects. To use
Russell's imperfect but memorable example,
featherless biped that is not a plucked chicken
has the same extension as human being; but
a different intension. There is an important
connection between intensions and inten-
tionality, for semantical systems, like exten-
sional model theory, that are limited to
extensions, cannot provide plausible
accounts of the language of intentionality
(see INTENSIONAL).
TWO PUZZLES
The attitudes are philosophically puzzling
because it is not easy to see how the inten-
tionality of the attitudes fits with another
conception of them, as local mental phenom-
ena.
Beliefs, desires, hopes, and fears seem to
be located in the heads or minds of the
people that have them. Our attitudes are
accessible to us through INTROSPECTION.
Quayle can tell that he believes Bush to be a
Republican just by examining the 'contents
of his own mind'; he doesn't need to investi-
gate the world around him. We think of
attitudes as being caused at certain times by
events that impinge on the subject's body,
specifically by perceptual events, such as
reading a newspaper or seeing a picture of
an ice-cream cone (see REASONS AND
CA USES). These attitudes can in turn cause
changes in other mental phenomena, and
eventually in the observable behaviour of
the subject. Seeing the picture of an ice-
cream cone leads to a desire for one, which
leads me to forget the meeting I am sup-
posed to attend and walk to the ice-cream
shop instead. All of this seems to require
that attitudes be states and activities that
are localized in the subject.
But the phenomenon of intentionality
suggests that the attitudes are essentially
387
INTENTION ALITY (2)
relational in nature; they involve relations
to the propositions at which they are dir-
ected and at the objects they are about.
These objects may be quite remote from the
minds of subjects. An attitude seems to be
individuated by the agent, the type of atti-
tude (belief, desire, etc.), and the proposition
at which it is directed. It seems essential to
the attitude reported by (1), for example,
that it is directed towards the proposition
that Bush is a Republican. And it seems
essential to this proposition that it is about
Bush. But how can a mental state or activ-
ity of a person essentially involve some
other individual? The difficulty is brought
out by two classical problems, which I will
call no-reference and co-reference.
Consider,
(3) Elwood believes the King of France is
bald.
It seems that if France were a monarchy,
and had a king, that king would be a con-
stituent of the proposition that Elwood
believed; his belief would be about him.
Since there is no king, there must be no
such proposition; what then does Elwood
believe? This is the no-reference problem.
Compare (1) and (4),
(4) Quayle believes that the person who
will come in second in the election is a
Republican.
In September 1992, (1) was surely true and
(4) was probably false. And yet Bush was
the person who was going to come in
second; that is, Bush and the person who will
come in second in the election co-referred. But
then it seems that the propositions Quayle is
said to believe by (1) and (4) are the same.
But then how can (1) be true and (4) be
false? This is the co-reference problem.
THE CLASSICAL SOLUTION
The classical solution to these problems is to
suppose that intentional states are only
indirectly related to concrete particulars,
like George Bush, whose existence is con-
388
tingent, and that can be thought about in a
variety of ways. The attitudes directly
involve abstract objects of some sort, whose
existence is necessary, and whose nature
the mind can directly grasp. These abstract
objects provide concepts or ways of thinking
of concrete particulars. On this view the
propositions:
that George Bush is a Republican,
that the person who will come in
second in the election is a Republican
are quite different, involving different con-
cepts. These concepts correspond to different
inferential/practical roles in that different
PERCEPTIONS and MEMORIES give rise to
these BEL I E F s, and they serve as reasons for
different ACTIONS. If we individuate propo-
sitions by concepts rather than individuals,
the co-reference problem disappears.
This proposal has the bonus of also
taking care of the no-reference problem.
Some propositions will contain concepts
that are not, in fact, of anything. These pro-
positions can still be believed, desired, and
the like.
This basic idea has been worked out in
different ways by a number of authors. The
Austrian philosopher Ernst Mally thought
that propositions involved abstract par-
ticulars that encoded properties, like being
the loser of the 1992 election, rather than
concrete particulars, like Bush, who exem-
plify them. There are abstract particulars
that encode clusters of properties that
nothing exemplifies, and two abstract
objects can encode different clusters of prop-
erties that are exemplified by a single thing
(see Zalta, 1988). The German philosopher
Gottlob Frege distinguished between the
sense and the reference of expressions. The
senses of George Bush and the person who
will come in second in the election are differ-
ent, even though the references are the
same. Senses are grasped by the mind, are
directly involved in propositions, and incor-
porate modes of presentation of objects (see
An Essay on Mind section 2.1.3; CONCEPTS).
For most of the twentieth century, the
most influential approach was that of the
British philosopher Bertrand Russell. Russell
(1905, 1929) in effect recognized two kinds
of propositions. Singular propositions consist
of particulars and properties or relations.
An example is a proposition consisting of
Bush and the property of being a Repub-
lican. General propositions involve only
universals. The general proposition corre-
sponding to someone is a Republican would
be a complex consisting of the property of
being a Republican and the higher-order
property of being instantiated. (The terms
singular proposition and general proposition
are from Kaplan (1989).)
Russell's theory of descriptions gives us
general propositions where we might have
thought we were getting singular ones.
Consider (5),
(5) The person who will come in second in
the election is a Republican.
Since Bush will lose the election, one might
suppose that this expresses the singular pro-
position we mentioned above. But, in fact, it
expresses the same general proposition
expressed by (6),
(6) There is a unique person that will come
in second in the election, and he or she
is a Republican.
Even (2) turns out not to express a singular
proposition on Russell's theory. Ordinary
proper names like Bush are hidden descrip-
tions. Where", is some crucial set of Bush's
properties, (1) reports that Quayle believes
that the'" is a Republican.
Similarly, (3) tells us that Elwood has a
belief about a number of universals, such as
being a King of France and being bald. All
of these universals can exist, even though
there is no King of France, so the fact that
Elwood has the belief he has does not imply
that there is such a person.
DIRECT REFERENCE
Over the past twenty-five years, the hidden
descriptions treatment of proper names has
INTENTIONALITY (2)
come in for a lot of criticism (see Donnellan,
1970; Kripke, 1972). According to this cri-
tique, the classical solution provides at best
a partial solution to the no-reference and
co-reference problems, and at worst rests on
a mistaken conception of intentionality.
To make the hidden descriptions treat-
ment of proper names work, we need a
description that (i) denotes the bearer of the
proper name, (ii) provides the correct
content for the proposition expressed by the
sentence in which the proper name occurs.
There seem to be two places to look for such
a description, in the mind of the subject,
and in the rules of language.
Someone like Quayle, who knows Bush
well, would associate a rich set of descrip-
tions or conditions with his name. Let's
assume that our description, the", incorpor-
ates all the facts that Quayle believes most
firmly about Bush. Then, on the hidden
descriptions view, it seems that (2) ex-
presses the same proposition as (7):
(7) The", is a Republican.
But is this right? Suppose, as seems likely,
that one of the properties Quayle is most
sure about, with respect to Bush, is that he
is a Republican. Then this property will be
incorporated into '" and (7) will be a trivial
proposition. But (2) is not trivial. Even
Quayle could probably imagine circum-
stances in which George Bush might not
have become a Republican.
Or consider Clinton, who also believes
that George Bush is a Republican. Intui-
tively, Clinton and Quayle believe the same
thing, that George Bush is a Republican.
But, on the hidden descriptions view, they
really do not, since the complex of things
Clinton associates most firmly with Bush
will not be exactly the same as "', the
complex that Quayle associates with him. If
Clinton uses sentence (1) to describe
Quayle, which proposition is at issue? The
one based on Clinton's conception of Bush
or the one based on Quayle's?
Finally, suppose that Elwood has heard of
Bush, like virtually everyone in the world.
But Elwood's beliefs about Bush, like his
389
INTENTION ALITY (2)
beliefs about many things, are confused and
fragmentary. He is not sure what party
Bush belongs to, or whether he is King or
President. All the true things he believes
about Bush don't amount to enough to pick
Bush out uniquely. Still, it seems that as
long as Elwood has heard about Bush -
careless as he may be reading the news-
paper articles, inattentive as he may be lis-
tening to the radio, inept as he may be in
remembering the little he manages to
understand - he can have beliefs about
Bush, even if they are mostly wrong.
The second place to look for the appro-
priate description is the linguistic conditions
of reference. It seems there must be some
relation 1t that obtains between a name and
the object to which it refers. It has been
suggested, for example, that the relation is
basically causal: a is the bearer of N iff a
stands at the beginning of a certain sort of
causal chain that leads to the use of N. This
may be more or less independent of the
speaker's belief, since many speakers do not
have the foggiest idea what 1t might be.
It does not seem, however, that the asso-
ciated description, the x such that 1t(N,x)
meets our second condition. For, whatever 1t
might turn out to be, it does not seem that
(2) expresses the same proposition as (8),
(8) The x such that 1t(George Bush, x) is a
Republican.
Intuitively, (2) doesn't tell us anything
about the name George Bush, and (8)
doesn't tell us anything about the person
George Bush. What (2) says could be true,
even if no one were named George Bush
(although we couldn't express it this way),
and what (8) says could be true even if
George Bush were a Democrat, so long as
someone else was named George Bush, and
that person were a Republican.
The conclusion to which these considera-
tions seem to point is, in Russellian terms,
that a statement like (2) expresses a sin-
gular proposition after all, and (1) attributes
to Quayle a belief in a singular proposition.
In David Kaplan's terminology, names are
directly referential (Kaplan, 1989). This does
390
not mean (as it might seem to) that the link
between a name and the object to which it
refers is unmediated. It means that the
object a name refers to is directly involved
in the propositions expressed by sentences
in which the name occurs.
But then the classical solutions to the co-
reference and no-reference problems are at
least very incomplete. We'll focus on the
first. Recall that Bill Clinton's original name
was BilI Blythe - he took Clinton, his step-
father's name, while a teenager. Suppose
that at one point Elwood is willing to assert
(9) but not (10):
(9) Bill Clinton is a Democrat.
( 10) Bill Blythe is a Democrat.
Then he learns about Clinton having two
names, and becomes willing to assert (10).
It seems that there was an important
change in Elwood's beliefs. And yet it seems
that the belief he expressed earlier with (9)
and the one he expressed later with (10)
express exactly the same singular proposi-
tion, with Clinton and the property of being
a Democrat as constituents.
Things are still more complicated, for
there are strong arguments that indexicals
(I, you, now) and demonstratives (this <1>,
that <1 are also directly referential. Let u be
a use by Clinton of:
(11) I am a Democrat.
What proposition does this express? Since
uses of the word I refer to the speaker, one
can associate the following condition with
the I in use u of (11),
(12) Being the speaker of u.
One might then be tempted to suppose that
(11) expresses the same proposition as
(13) The speaker of u is a Democrat.
But this isn't plausible. The proposition
expressed by u, Clinton's utterance of (11),
will be true in circumstances in which
Clinton is a Democrat, but doesn't utter u.
In uttering u, he says that he is a Democrat.
This was no doubt true before he uttered u,
and would have been true even if he had
not uttered u. The utterance u says nothing
about itself. In uttering (lI), Clinton would
be confirming Elwood's belief that he is a
Democrat, a belief Elwood would express
with (9). It seems that u expresses a sin-
gular proposition.
Now suppose that Clinton himself was
ignorant of, or has forgotten, his original
name. Imagine that someone trustworthy
tells him that Bill Blythe is a Democrat but
doesn't reveal that Bill Blythe is Clinton
himself. Clinton is willing to assert both (9)
and (lO). Given that names and indexicals
are directly referential, by doing this he
expresses his beliefs in exactly the same sin-
gular proposition. But Clinton clearly has
two beliefs in the same proposition: yet
another version of the co-reference problem.
MENTAL REPRESENTATIONS
Russell considered his theory to be a version
of realism, in the sense that the objects of
the attitudes were taken to be objective
entities, external to the mind of the subject.
Realists in this sense suppose that the mind
can directly grasp such objects as Frege's
senses, Mally's abstract objects, or Russell's
universals. Cognition of concrete particulars
is indirect, mediated by cognition of abstract
objects.
An equally natural response to the
puzzles is to reject realism in this sense, in
favour of the alternative that intentionality
basically involves having mental repre-
sentations: ideas, thoughts, or mental terms
and sentences in the subject's mind. Cogni-
tion of external objects, both abstract and
concrete, is mediated by concrete particu-
lars in the mind. There are as many vari-
eties of this approach as there are theories
of mental representations. According to a
very straightforward theory patterned after
the enthusiastic FODOR (1981) and the
sceptical Stich (1983), the representations
are best thought of as terms, predicates, and
sentences of a 'LANGUAGE OF THOUGHT'.
INTENTIONALITY (2)
This approach inherits both the traditional
empiricist distinction between simple and
complex ideas, and the theme of compos i-
tionality from the philosophy of language.
One supposes that the basic expressions of
the language of thought gain their meaning
from their role in perception, cognition, and
action, and the meaning of complex expres-
sions is determined by the meanings of their
parts.
An advocate of this view can adapt Rus-
sell's theory of descriptions to the language
of thought, and adopt his partial solutions
to the no-reference and co-reference prob-
lems. A mental description may be denota-
tionless, and different mental descriptions
may denote the same object. But there is no
need to adopt the hidden descriptions view
of basic expressions like names and index-
icals. One can suppose that the reference of
these basic terms is determined by their
cognitive role and their links with percep-
tion and action, rather than by any hidden
descriptive content.
Consider the mental equivalents of (9),
(lO), and (1I). We can imagine Elwood to
have formed two mental names (opened
two mental 'files') for Clinton that are not
internally connected, based on different
causal interactions with Clinton, as child-
hood friend and as candidate. When Elwood
reads about Clinton in the newspapers, new
predicates become associated with the
second name but not with the first. We can
suppose that the mental analogue of I is
that term in the language of thought that a
person uses to keep track of information
gained in the special ways one can gain
information about oneself. The mental sen-
tences corresponding to (9), (10), and (1I)
will have different causal and cognitive
roles, even though it is the political affili-
ation of the same person that makes them
each true.
So far, so good, but the story is very
incomplete. To flesh out this sort of account,
one needs to understand the relation of the
internal mental representations involved in
the attitude to the texts that express them
and the content sentences of reports that
describe them. With respect to texts, the
391
INTENTIONALITY (2)
most natural view is that the mental repre-
sentation involved in an attitude is synony-
mous with the text that expresses it.
The question of content sentences is more
complicated. Let's return to (1). What
exactly does this tell us about Quayle's
mental representations on this theory?
The natural hypothesis is that Quayle's
mental representation should have the same
meaning as (2), which is both the content
sentence of the belief report and Quayle's
belief text. However, things are not so
simple. Again, it is proper names and index-
icals that provide problems. Suppose Clinton
has a mental sentence with the same
meaning as (10) in the belief structure of his
mind. How would a knowledgeable person,
Quayle, say, report this belief of Clinton's?
He would not use the English translation of
the sentence in Clinton's head and say (14),
(14) Clinton believes that I am a Democrat.
This would be to say that Clinton believes
the false proposition that Quayle is a Demo-
crat, not the true one that Clinton is. He
would say, instead,
(15) Clinton believes that he is a Democrat.
The connection between the content sen-
tences in accurate attitude reports and
mental sentences in the mind then is looser
than having the same meaning. Clinton's
mental version of I and Quayle's use of the
pronoun he have the same reference, but
not the same meaning. It seems that Quayle
is not characterizing Clinton's belief in
terms of the meaning of the mental words
in Clinton's head, but rather in terms of the
objects to which those words refer.
TWO-TIERED VIEWS
With these last considerations, propositions
sneak back into the mental representations
account, albeit with a somewhat diminished
status. Propositions are not directly grasped
by the mind, but are tools we use to
describe something important that different
attitudes, involving different subjects and
392
different ways of thinking about objects,
may have in common.
On the family of views I shall call two-
tiered, our original notion of the object of
the attitudes assimilates two different levels
of comparison among attitudes (see Barwise
and Perry, 1983; Crimmins, 1992; Crim-
mins and Perry, 1989; Kaplan, 1989;
Perry, 1993; Richard, 1990; Salmon, 1986;
Salmon and Soames, 1988; Schiffer, 1978,
1990). Consider the hopes of Clinton, Bush,
and Perot as the election draws near. They
each might say,
(16) I hope that I win the election.
Their hopes are similar at the level of
mental representations, but different, indeed
incompatible, at the level of the proposition
hoped for. On the other hand, Hillary
Clinton would not say (16), although she
hopes for the same thing that Clinton does:
(17) Bill Clinton hopes that he wins the
election, and so does Hillary Clinton.
On a typical view of this sort, for an attitude
report to be true. the subject must have a
mental sentence <!>(Cl) in the appropriate
structure, which expresses the same proposi-
tion as the content sentence of the report.
For (1) to be true, Quayle must have some
sentence that expresses the singular propo-
sition that Bush is a Republican. This sen-
tence need not be a translation of the
content sentence; the terms in it need not
have the same meaning as those in the
content sentence, only the same reference.
Indeed, as our reflections above about
Quayle, Clinton. and (14) and (15) showed,
sometimes the content sentences cannot be
translations of the mental sentences. The
job of the attitude reporter is to express from
his or her perspective the same proposition
that the subject's mental sentence expresses
from the subject's perspective. Propositions
are not directly grasped by minds, but are
artefacts of our method of keeping track of
truth conditions across differences in subject
and mode of identification.
On this view, the language we use to
report the attitudes is basically incomplete
and bound to be misleading at times. An
attitude involves an agent having a certain
sort of mental representation, which, given
the agent's identity and circumstances,
determines a certain proposition (which is
usually thought of as a structured proposi-
tion of the sort Russell provides, although
other approaches are possible). It is the
subject and mental representation that are
crucial to the occurrence of the attitude; if
circumstances are wrong, no proposition
may be determined, or the same proposition
may be determined by quite different atti-
tudes. But the attitude report focuses on the
agent and proposition, only providing in-
direct information about the mental rep-
resentation. It is this incompleteness that
accounts for the no-reference and co-
reference problems.
Consider our example involving three
ways of referring to Clinton. Bill Clinton and
Bill Blythe correspond to two ways anyone
can think of Clinton, and I corresponds to
another, the 'self-thinking' way of thinking
about Clinton, which is a way of thinking
we can each use to think about ourselves.
Now if Quayle utters (15), we will naturally
suppose that Clinton's mental sentence is I
am a Democrat. But the content sentence of
(15) identifies only the proposition Clinton
believes, not how he believes it. For all (15)
tells us, Clinton's mental sentence might be
Bill Blythe is a Democrat or (looking at
himself in a mirror, noticing the moderately
liberal demeanor, but not recognizing
himself), that man is a Democrat. Basically,
attitude reports explicitly identify only two
parameters of the three that are involved in
the attitudes.
Among philosophers who offer two-tiered
accounts, there is agreement that our atti-
tude reports are looser than is envisaged on
the other accounts, and rely more on prag-
matic factors to communicate facts about
the mental representations involved in the
attitudes. For example, unless we are told
otherwise, we naturally expect that people
believe things about themselves in the first-
person way, and not only by an abandoned
name from their youth. Hence we would
INTENTIONALITY (2)
normally infer from Quayle's utterance of
(15) that Clinton has a belief he would
express with I am a Democrat and not just
one he would express with Bill Blythe is a
Democrat.
When we get to details, however, two-
tiered theorists disagree not only about the
mechanisms involved, but even about the
basic facts about which attitude reports are
true and false in problematic situations.
Recall Elwood, who a few pages back was
willing to assert (9) but not (10):
(9) Bill Clinton is a Democrat.
(10) Bill Blythe is a Democrat.
Now consider (18), (19) and (20):
(18) Elwood believes that Bill Clinton is a
Democrat.
(19) Elwood believes that Bill Blythe is a
Democrat.
(20) Elwood doesn't believe that Bill Blythe
is a Democrat.
On one approach, each report has an impli-
cit quantifier that ranges over the hidden
parameter (see Barwise and Perry, 1983;
Salmon, 1986). So (18) and (19) are true
and (20) is false because Elwood does
believe the proposition that Clinton is a
Democrat in some way or another - namely,
thinking of him as Bill Clinton. (20) suggests
that he believes the proposition when he
thinks of Clinton as Bill Clinton, while (19)
suggests that he believes it when he thinks
of Clinton as Bill Blythe. On the present
approach, this is conceived as merely a
matter of pragmatics. In the case of Elwood,
we would be reluctant to assert (19), since
it suggests that Elwood is in a state that
would lead him to assert (10). This reluc-
tance to assert (19) may be mistaken for an
intuition of the falsity of (19). But on the
present approach, it is literally true. Simi-
larly, we might be willing to assert (20),
since it suggests, correctly, that Elwood
would not assert (10). And this willingness
393
INTENTIONALITY (2)
might be mistaken for an intuition of the
truth of (20). But, on the present approach,
(20) is literally false.
Another approach takes it that the ways
of thinking about Clinton are inexplicit or
un articulated parts of what (18), (19), and
(20) are about (see Crimmins, 1992; Crim-
mins and Perry, 1989; Schiffer, 1978). For
a belief report to be true, the subject has to
believe the proposition identified by the
content sentence in virtue of thinking about
the objects in the way provided by the context.
The report is about certain ways, however,
and doesn't merely quantify over them. The
words Bill Clinton in (18) suggest that it is
thinking of Bill Clinton as Bill Clinton that is
at issue; given this, (18) says that Elwood
believes the proposition in this way. For the
truth of (19) and (20), however, the other
way of thinking about Clinton, as Bill
Blythe, is relevant. On this view, (18) and
(19) are true, and (20) is false, given the
facts about Elwood.
On the two-tiered view, the attitudes are
local mental phenomena, involving subjects
and mental representations. But the cogni-
tive role of these mental representations has
to be understood in terms of the interactions
of the agent with external objects, abstract
and concrete, for we use these objects to
classify and describe the attitudes.
Earlier we noted a tension between two
conceptions of the attitudes, the intentional
or object-directed conception, and the
natural conception of them as local mental
phenomena. Each tier of the two-tiered view
corresponds to one of these conceptions.
Mental representations are local phenom-
ena, located in the heads or minds of the
people that have them. When we describe
these representations in terms of the objects
they are representations of, however, our
descriptions involve not only what goes on
in the mind, but also the circumstances that
link representations with various parts of an
external reality.
Quayle's belief that Bush was a Repub-
lican was no doubt acquired on the basis of
perception, and involved a change in the
system of representations in his mind. This
change is a local mental phenomenon,
394
which can be cited in explanations of
Quayle's actions, since it no doubt altered
the ways in which he was disposed to
behave in various circumstances.
This change is linked, by the circum-
stances of the perception that caused it. to
George Bush. In certain circumstances, it is
useful and convenient to describe the
mental representation in terms of the object
that gave rise to it. We think that Quayle
will be good at recognizing Bush on sub-
sequent occasions, so that the knowledge of
his party affiliation will flow along these
identifications and guide his actions towards
Bush, however described or encountered.
Our two-tiered system works best in such
circumstances. When knowledge of a single
object is not integrated in this way - as in
the case of Elwood, who did not recognize
Clinton as the boy he once knew as Bill
Blythe - the two-tiered approach can be
confusing and misleading. This is even more
true in the case in which an internal repres-
entation has come into play with no proper
external referent. The co-reference and no-
reference problems call our attention to
these limitations of the two-tiered system. In
practice, however, we seldom have a
problem in making the details of the case
clear. Where singular propositions do not
readily make clear what is believed and not
believed, general propositions come quickly
to the rescue. Elwood believed that there
was a kid he once knew named 'Bill Blythe'.
and there is a man running for President
named 'Bill Clinton' and they are not the
same and the latter is a Democrat.
See also An Essay on Mind section 2.1 ;
CONTENT; INTENTIONALITY (1); THOUGHTS;
BIBLIOGRAPHY
Brentano. F. 1960. The distinction between
mental and physical phenomena. In Realism
and the Background of Phenomenology, ed.
R. M. Chisholm. Atascadero. CA: Ridgeview
Publishing.
Barwise, J., and J. Perry. 1983. Situations and
Attitudes. Cambridge. MA.: MIT Press. Brad-
ford Books.
Castaneda. H. 1990. Thinking and the Structure
of the World. Berlin: W. de Gruyter.
Crimmins. M. 1992. Talk About Beliefs. Cam-
bridge. MA: MIT Press/Bradford Books.
Crimmins. M .. and Perry. J. 1989. The prince
and the phone booth: reporting puzzling
beliefs. The Journal of Philosophy. LXXXVI.
685-711.
Donnellan. K. 1970. Proper names and identi-
fying descriptions. Synthese. 21 3-31.
Evans. G. 1982. Varieties of Reference. Oxford
University Press.
Fodor. J. 1981. Propositional Attitudes. In
Representations. Brighton: Harvester Press.
Frege. G. 1970. On sense and reference. In
Translations from the Philosophical Writings of
Gottlob Frege. ed. and trans. P. Geach and
M. Black. Oxford: Basil Blackwell.
Hintikka. J. 1975. The Intentions of Intention-
ality. Dordrecht: D. Reidel.
Kaplan. D. 1989. Demonstratives. In Themes
From Kaplan. ed. 1. Almog. J. Perry and H.
Wettstein. Oxford University Press.
Kripke. S. 1972. Naming and Necessity. In
Semantics of Natural Language. Dordrecht: D.
Reidel. 253-354.
Peacoke. C. 1986. Thoughts: an Essay on
Content. Oxford: Basil Blackwell.
Perry. J. 1993. The Problem of the Essential
Indexical. Oxford University Press.
Richard. M. 1990. Propositional Attitudes.
Russell. B. 1905. On denoting. Mind 14. 479-
93.
Russell. B. 1929. Knowledge by acquaintance
and knowledge by description. In Mysticism
and Logic. New York: W. W. Norton.
Salmon. N. 1986. Frege's Puzzle. Cambridge.
Salmon. N.. and S. Soames, eds. 1988.
Propositional Attitudes. Oxford University
Press.
Searle, J. 1983. Intentionality. Cambridge Uni-
versity Press.
Schiffer. S. 1978. The basis of reference.
Erkenntnis. 13. 171-206.
Schiffer. S. 1990. The mode-of-presentation
problem. In Propositions and Attitudes, ed.
C. A. Anderson and J. Owens. Stanford. CA.:
CSU.
Stalnaker. R. 1985. Inquiry. Cambridge. MA.:
INTROSPECTION
nitive Science. Cambridge, MA.: MIT Press/
Bradford Books.
Zalta, E. 1988. Intensional Logic and the Meta-
physics of Intentionality. Cambridge, MA.:
JOHN PERRY
introspection The term 'introspection'
derives from the Latin words spicere ('to
look') and intra (within), and is believed to
have first made its appearance in the second
half of the seventeenth century. It has been
used by philosophers and psychologists in a
number of different though related senses.
In its broadest sense it refers to the non-
inferential access each person has to a
variety of current mental states and events
- SENSATIONS. feelings, THOUGHTS. etc. -
occurring in that person. When the term is
used in this broad sense it is an open ques-
tion whether introspection is appropriately
thought of as a kind of PERCEPTION or
observation. involving an 'inner sense'. But
sometimes the use of the term is tied to the
view (suggested by its etymology) that our
introspective access to our mental states is
perceptual or quasi-perceptual. Some
writers distinguish introspection from a pre-
introspective awareness of mental phenom-
ena. saying that one is not properly speak-
ing introspecting unless one is not only
aware of some mental phenomenon but
aware that one is aware of it. And some
have suggested that introspection should be
thought of as a kind of low-level theorizing
about what is going on in one's mind, the
data for this theorizing presumably being, at
least in part, the knowledge gained from
'pre-introspective' awareness. It is perhaps
introspection in this last sense that 'intro-
spective' people are given to.
In this essay 'introspection' will be used
in its broadest sense, i.e to include what
some would classify as pre-introspective
awareness. We are concerned, then, with
something allowed to exist by all except
behaviourist philosophers - a special. non-
inferential. access which each person has to
(some of) his or her own mental states, and,
going with this, a special authority that
395
INTROSPECTION
attaches to (some) first-person mental state
ascriptions (see FIRST-PERSON AUTHORITY).
Whether this special access and authority
deserve the name 'privileged access' is a
further question.
An influential tradition in modern philo-
sophy, often associated with Descartes,
claims that CONSCIOUSNESS is the essence of
mentality, that each mind is 'transparent' to
itself, and that the awareness each mind has
of its current states and processes yields
knowledge having the highest possible
degree of certainty. Such a view has often
gone with a version of foundationalist epis-
temology which takes all empirical know-
ledge to be grounded, ultimately, in the
knowledge (here called 'introspective') each
mind has of its own states, including its
sensations and perceptual experiences. To
single out two important components of this
view, it holds first that the judgments a
person makes about his or her current
mental states are 'incorrigible', or 'infallible',
and it holds that the mental states are, by
their very nature, 'self-intimating'. To say
that a judgment is incorrigible is to say that
it is impossible that it could be shown to be
mistaken, while to say that a judgment is
infallible is to say that it is impossible that it
could be mistaken. Incorrigibility doesn't
straightforwardly entail infallibility, but it is
not easy to see how a kind of judgment
could be incorrigible without being infal-
lible. To say that a mental state is neces-
sarily self-intimating means that it follows
from someone's having the state that the
person is aware of having it, or, on a weaker
version of the notion, that the person would
be aware of having the state if he or she
considered the matter.
One source of the infallibility doctrine is
Descartes' Cogito (see HISTOR Y). The propo-
sitional content 'I am thinking' (and like-
wise such contents as 'I am conscious' and
'I have some beliefs') is such that, necessa-
rily, if it is judged to be true, or even con-
sidered, it must be true. The same is true
more generally of contents of the form 'I am
having a thought that P'. Judgments with
such contents are necessarily self-verifying.
But where a judgment is self-verifying, its
396
infallibility can be explained on purely
logical grounds and provides no evidence of
the existence of a perceptual or quasi-
perceptual access to mental states which is
immune to error. And most judgments
thought to be infallible are not self-
verifying, as is shown by the fact that their
negations, e.g. 'I am not in pain', are not
self-falsifying. If they are infallible, it must
be for another reason.
For a variety of reasons, 'CARTESIAN'
views about our access to our own minds,
including claims about infallibility and self-
intimation, have fallen out of favour. Freud
persuaded many that beliefs, wishes, and
feelings (e.g. of hostility, or jealousy) are
sometimes unconscious, and even sceptics
about Freudian theory acknowledge that
there is a such a thing as SELF-DECEPTION
about one's motives and attitudes. Recent
work in psychology has revealed that in
certain circumstances people are regularly
wrong in the claims they make about their
reasons for action, and that such claims are
sometimes the result of 'confabulation'. Psy-
chologists have also discovered such phe-
nomena as 'blindsight', in which a patient
cortically blind in some portion of the visual
field sincerely denies seeing anything there
but, when forced to make 'guesses' about
what is there, guesses correctly well above
chance. Cartesian views about introspective
access have also been challenged on more
purely philosophical grounds. The alleged
impossibility of imagining cases in which
someone is mistaken about whether he or
she is in PAIN has been challenged on the
basis of a variety of examples. For example,
there is the 'fraternity initiation' example, in
which the blindfolded subject is told that a
knife is about to be applied to his throat,
and takes himself to feel pain when a piece
of ice is applied.
Lying behind many of these challenges is
what has been called the 'distinct exist-
ences' argument, advanced by (among
others) David Armstrong. This takes it as
given that introspective awareness of a
mental state involves having a belief that
one has that mental state. And it takes it as
obvious that where one has a mental state
and the belief that one has it, the state and
the belief are 'distinct existences', i.e. such
that it is at least logically possible that
either of them should exist without the
other. If one's toothache could, logically,
exist without one's believing that one has a
toothache (even if one considers the
matter), that shows that toothaches are not
self-intimating. If the belief that one has a
toothache could, logically, exist without
being accompanied by a toothache, that
shows that such beliefs are not infallible.
Sometimes the claim that the first-order
mental state (e.g. the toothache) and the
introspective belief are distinct existences is
supported by the claim that if the belief con-
stitutes knowledge of the first-order state, this
must be because it is caused by it via a reli-
able mechanism, and that, as Hume taught
us, causes and effects are always distinct
existences.
Implicit here is a simple yet compelling
account of the nature of introspective self-
knowledge. On this view, our special access
to our mental states consists simply in the
fact that each of us is so constituted that,
under certain conditions, being in a mental
state produces, via some mechanism in the
mind or brain, a belief that one is in a
mental state of that sort. The belief counts
as knowledge because the mechanism is by
and large reliable, i.e. by and large produces
true beliefs.
Many who take this view present it as a
version of the perceptual, or 'inner sense',
model of introspective self-knowledge. Many
others have denied that introspective self-
knowledge should be thought of on the
model of sense-perception. But it is not alto-
gether clear what is at stake when it is
affirmed or denied that introspection is a
kind of perception. One probably should not
think of the issue as that of whether intro-
spection shares a 'real essence' of perception
which unites the various 'outer' senses.
There may be no such thing. It is better to
think of the question as whether introspec-
tion fits one or another stereotype of percep-
tion. And different stereotypes need to be
distinguished.
John Locke said that we perceive our
INTROSPECTION
ideas and the operations of our minds, and
so might be held to conceive of introspec-
tion (what he called 'reflection') as inner
sense. But Locke clearly did not think of
introspection on the model of sense-
perception as he thought it actually is - for he
thought that our perceptual access to exter-
nal things is mediated by their production
in us of 'ideas' of them, and he certainly did
not think of our introspective access to our
ideas as mediated in this way. If he thought
of introspection on the model of perception,
the operative stereotype of perception was
not sense-perception as it actually is but
rather sense-perception as we naIvely take it
to be. This is not the stereotype modern
advocates of the inner sense model have in
mind, for, unlike Locke, they think of the
deliverances of introspection as being
subject to error in the same way those of
sense-perception are.
One stereotype of perception, of which
Locke's was perhaps a special case, is what I
shall call the object-perception stereotype.
This takes visual perception as its paradigm
of perception, and takes it that perception is
in the first instance perception of (non-
factual) objects, and only derivatively per-
ception of facts. The central idea is that we
perceive facts by perceiving the objects
involved in them and identifying those
objects as being of certain kinds, or as being
certain particular things, by perceiving their
intrinsic features and their relations to other
objects.
Introspection differs in obvious ways from
visual perception. There is no organ of
introspection whose disposition is under our
voluntary control in the way the orienta-
tion of our eyes is. And while visual percep-
tion involves our having visual experiences
of the object, which constitute its appearing
to us in certain ways (ways in which it may
or may not be), no one supposes that intro-
specting a pain, say, involves one's having
an experience of the pain, distinct from the
pain itself, which constitutes its appearing
to one in some way. Introspective aware-
ness of a sensation does not involve having
yet another sensation that is 'of' the first
one. For some this is enough to show that
397
INTROSPECTION
introspection is not inner sense. But there
are other parts of the object-perception
stereotype that are independent of this, to
which some kinds of introspective aware-
ness have been thought to conform.
It is primarily the introspective awareness
of sensations and sensory states that philo-
sophers have found it plausible to think of
in terms of the object-perception model. For
it is here that it is natural to think of there
being non-factual objects that we are aware
of, such that it is by being aware of these,
and of their intrinsic features, that we are
aware of whatever facts we are aware of. It
is not natural to think of our awareness of
beliefs, for example, in this way. Nothing
seems to answer to the description: becom-
ing aware of something in oneself, and
identifying it as the belief that the cold war
ended in 1991. While we speak of being
aware of beliefs, it would seem inappropri-
ate to say that one is aware that one
believes that so-and-so by being aware of
the belief that so-and-so. Being aware of the
belief just is being aware of the fact that one
believes a certain thing - so here we do not
have object-awareness in the appropriate
sense. By contrast, it is very natural to say
that one is aware that one has a pain or an
itch by being aware of the pain or itch. And
it is natural to say that one is aware that
one is seeing a yellow after-image by being
aware of the after-image and seeing that it
is yellow.
But here the question of whether a per-
ceptual model of introspection is appropriate
becomes inextricably bound out with ques-
tions about the nature of the mental phe-
nomena to which we have introspective
access. What makes the inner sense model
irresistible in the case of sensory phenom-
ena is the acceptance of what has been
called the 'act-object' conception of sensa-
tion. Applied to the case of after-imaging,
this says that the phenomenon we describe
by saying 'r see a red after-image' involves
the existence somewhere in the mind of an
entity that is actually red. It is questionable
whether one can accept this view about
after-imaging without being committed to
full-scale acceptance of the sense-datum
398
theory of perceptual experience - the view
that every case of perceiving or seeming to
perceive involves 'sensing' or 'immediately
perceiving' a phenomenal object, a sense-
datum, distinct from the external object (if
any) that is perceived, and having whatever
sensory features one's experience represents
the external object as having. This theory is
almost universally rejected. And in the
opinion of many, its rejection requires the
rejection of the act-object conception across
the board. Some who reject the act-object
conception favour an 'adverbial' account of
sensory experience, according to which
what the act-object theorist describes as
seeing a red image is more appropriately
described as 'being appeared to redly'.
Others hold that the mistake of the act-
object conception is to misconstrue the
intentional object of a sensory state as an
actually existing object (see INTENTION-
ALITY). Thus, just as Ponce de Leon could
think of and look for the fountain of youth,
even though there is no such thing, so the
person experiencing a red after-image can
be seeing something red even though in fact
there is nothing red, either outside his mind
or inside it, that he is seeing. If any such
view is adopted, it is not clear that any of
the entities we are left with in the sensory
realm qualify as 'objects' in the sense
required for the applicability of the object-
perception model of introspection. What we
are left with are states of being appeared
to redly, or experiencings with certain rep-
resentational contents. And just as being
aware of a belief does not seem something
distinct from being aware that one has such
and such a belief, it is not clear that being
aware of such an experiencing, or state, is
distinct from being aware that one is under-
going such an experiencing.
When philosophers attack the idea that
introspection is perception, what they are
often attacking is what r am calling the
object-perception model of perception. But it
is far from clear that recent philosophers
who have endorsed a perceptual model of
introspection have meant to be endorsing
the object-perception model. Some of them
explicitly reject the act-object conception of
sensation. And some of them clearly want
their account of introspection as perception
to apply as much to the case of awareness
of beliefs and desires as to the case of
awareness of sensations, and give no indica-
tion that they think that the former con-
forms to the object-perception model. Nor
does it seem that someone who speaks of
introspection as perception is thereby com-
mitted to endorsing the object-perception
model - for there is a good deal of paradig-
matic sense perception, e.g. perception by
smell, that does not conform to this model
(if I say that I 'smell a skunk', I do not
mean that there is a particular skunk I
smell). It seems plausible that these philoso-
phers are operating with a broader concep-
tion of perception than this. The key
elements of this broad conception are, first,
that perception involves the production by
the object or state of affairs perceived, via a
reliable belief-producing mechanism, of a
belief about that object or state of affairs,
and, second, that the existence of the object
or state of affairs perceived is logically inde-
pendent of its being perceived and of there
being the belief-producing mechanisms that
would be required for its perception. Call
these, respectively, the causal condition and
the independence condition. It seems plaus-
ible that these conditions are satisfied in
ordinary cases of sense perception. Is it also
plausible to suppose that they are satisfied
in cases of introspective awareness of one's
own mental states?
This brings us back to the issue of
whether mental phenomena or states of
affairs are, in any good sense, 'self-intimat-
ing'. If they are, our awareness of them fails
the independence condition, and is not
perceptual even in the broad sense. The
Cartesian idea that the mind is completely
transparent to itself has been thoroughly
discredited. But one needn't accept that idea
in order to question the view that all mental
phenomena and states of affairs satisfy the
independence condition. The latter view
implies that for every sort of mental state,
there could be creatures that have states of
that sort but are totally devoid of intro-
spective access to those states - as we might
INTROSPECTION
put it, creatures who are introspectively
self-blind with respect to that sort of state.
Since such self-blindness is supposed to be a
perceptual disability, not a cognitive one, it
would seem that if it is possible at all it
should be possible for it to occur in crea-
tures who are cognitively and conceptually
on a par with us. Could such creatures be
self-blind with respect to their own beliefs?
Such a creature would have to have an
appreciable degree of RATIONALITY. And
such rationality would require a sensitivity
on the part of the creature to what the con-
tents of its belief-desire system are, since
only so could it make the revisions of its
belief-desire system, in the light of new
experience, that rationality requires. It is
arguable that such sensitivity would require
some introspective access to its beliefs and
desires. Also, it is questionable whether it is
coherent to suppose that there could be a
creature, possessed of the concepts of belief
and desire, who has the ability to express
beliefs and desires linguistically but not the
ability to report beliefs. Or consider the case
of sensations. Could there be a creature
with normal human intelligence and con-
ceptual capacity who has pains but is
totally without introspective access to its
pains? Or who has a perSistent ringing in its
ears, but is incapable of becoming intro-
spectively aware of it? If not, it would seem
that even the broad perceptual model for
introspection should be rejected.
If one takes the view that mental states
are defined or constituted by their causal or
functional roles, then one possible view is
that it is essential to the functional roles of
some kinds of mental states that under
certain conditions these mental states
produce in their subjects' introspective
awarenesses of them, these awarenesses
consisting of 'higher-order beliefs', or
'higher-order thoughts', about the states
that produce them. This would be a func-
tionalist version of the view that certain
kinds of mental states are self-intimating
(see FUNCTIONALISM). Among the problems
faced by such a view is that of explaining
how, if at all, we can allow for the apparent
fact that we share many kinds of mental
399
INTROSPECTION
states with creatures who lack the capacity
for introspective awareness.
EMOTION; IMAGERY; IMAGINATION; PER-
CEPTUAL CONTENT; QUALIA; THE SELF;
SUBJECTIVITY.
BIBLIOGRAPHY
Alston, W.P. 1971. Varieties of privileged
access. American Philosophical Quarterly, 8,
223-41.
the Mind. London: Routledge & Kegan Paul.
Descartes, R. 1985. Meditations on first philo-
sophy. In The Philosophical Writings of Des-
cartes, ed. J. Cottingham, R. Stoothhoff, and
D. Murdoch. Cambridge University Press.
Hill, C.S. 1991. Sensations: A Defense of Type
Materialism. Cambridge University Press.
400
Locke, J. 1975. An Essay Concerning Human
Understanding, ed. P. H. Nidditch. Oxford:
Clarendon Press.
Lyons, W.E. 1986. The Disappearance of Intro-
spection. Cambridge, MA: MIT Press/Brad-
ford Books.
Myers, G.E. 1986. Introspection and self-
knowledge. American Philosophical Quarterly,
23, 199-207.
Nisbett, R., and Wilson, T. de C. 1977. Telling
more than we know: verbal reports on
mental processes. Psychological Review, 84,
231-59.
Shoemaker, S.S. 1988. On Knowing One's
Own Mind. Philosophical Perspectives, 2,
183-209.
SYDNEY SHOEMAKER
inverted spectrum see FUNCTIONALISM;
QUALIA.
Language of Thought (1) The Representa-
tional Theory of the Mind arises with the
recognition that thoughts have contents
carried by mental representations. For
Abelard to think, for example, that Pegasus
is winged is for Abelard to be related to a
MENTAL REPRESENT A TION whose content is
that Pegasus is winged. Now, there are dif-
ferent kinds of representations: pictures,
maps, models, and words - to name only
some. Exactly what sort of REPRESENT A-
TION is mental representation? (See
IMAGER Y; CONNECTIONISM.) Sententialism
distinguishes itself as a version of rep-
resentationalism by positing that mental
representations are themselves linguistic
expressions within a 'language of thought'
(FODOR, 1975, 1987; Field, 1978; Maloney,
1989). While some sententialists conjecture
that the language of thought is just the
thinker's spoken language internalized
(Harman, 1982), others identify the lan-
guage of thought with Mentalese, an unar-
ticulated, internal language in which the
computations supposedly definitive of cogni-
tion occur. Sententialism is certainly a bold
and provocative thesis, and so we turn to
the reasons that might be offered on its
behalf.
SEMANTICS AND REASONING
THOUGHTS, in having contents, possess
semantic properties (see CONTENT; SEMAN-
TIC/SYNTACTIC). Thinking that Pegasus is
winged, Abelard somehow both thinks about
Pegasus, despite the horse's non-existence,
and thinks of the horse as winged.
Thoughts appear, then, to denote and attri-
bute, denotation and attribution thus being
among the semantic properties of thoughts.
What kind of mental representation might
L
support denotation and attribution if not
linguistic representation? Perhaps, when
thinking of Pegasus, Abelard deploys an
actual Mentalese name for that actually
non-existent horse. After all, names evi-
dently can denote even what wants exis-
tence. And if Abelard's thinking of Pegasus
is his mentally naming Pegasus, then
Abelard's thinking of Pegasus as winged
would be Abelard's concatenating the Men-
talese adjective meaning winged with Pega-
sus's Mentalese name. The Mentalese
sentence 'PEGASUS IS WINGED' would
then serve to bring Pegasus to mind and to
portray him as winged. Also, in thinking of
Pegasus simply as winged, Abelard might
well ignore other of Pegasus's character-
istics, including his colour. In this sense,
thoughts are surgically precise in just the
way that sentences are and, say, pictures
are not. 'PEGASUS IS WINGED' predicates
only wingedness of Pegasus while remain-
ing silent on his colour. However, a picture
of Pegasus that depicts his wings will strug-
gle to conceal his colour. If, then, thoughts
denote and precisely attribute, sententialism
may be best positioned to explain how this
is possible.
Beliefs are true or false. If, as representa-
tionalism would have it, beliefs are relations
to mental representations, then beliefs must
be relations to representations that have
truth values among their semantic proper-
ties. Sentences, at least declaratives, are
exactly the kind of representations that
have truth values, this in virtue of denoting
and attributing. So, if mental representa-
tions are as sententialism says, we could
readily account for the truth valuation of
mental representations.
Beliefs serve a function within the mental
economy. They play a central part in
401
LANGUAGE OF THOUGHT (1)
reasoning and, thereby, contribute to the
control of behaviour (see BELIEF; RATION-
ALITY). Reason capitalizes on various
semantic and evidential relations among
antecedently held beliefs (and perhaps other
attitudes) to generate new beliefs to which
subsequent behaviour might be tuned.
Apparently, reasoning is a process that
attempts to secure new true beliefs by
exploiting old (true) beliefs. By the lights of
representationalism, reasoning must be a
process defined over mental representations.
Sententialism tells us that the type of repre-
sentation in play in reasoning is most likely
sentential - even if mental - representation.
(See ARTIFICIAL INTELLIGENCE; COGNITIVE
PSYCHOLOGY; COMPUTATIONAL MODELS.)
Abelard believes both that Eloise is ashen
and that if she is ashen, then she is ill.
Remarkably, he concludes that Eloise is ill.
How does this happen? Suppose that both
the beliefs he exploits and the one he pro-
duces in concluding that Eloise is ill all
essentially involve mental sentences. If that
were the case, then we would have a route,
even if largely uncharted, to understanding
how reasoning unfolds in Abelard. Possibly,
in reasoning mental representations stand
to one another just as do public sentences
in valid formal derivations. Reasoning would
then preserve truth of belief by being the
manipulation of truth-valued sentential
representations according to rules so selec-
tively sensitive to the syntactic properties of
the representations as to respect and pre-
serve their semantic properties. The senten-
tialist hypothesis is thus that reasoning is
formal inference; it is a process tuned pri-
marily to the structures of mental sen-
tences. Reasoners, then, are things very
much like classically programmed com-
puters (Fodor, 1980; Searle, 1980).
Would that the story could be so tidily
told! Arguably we have infinitely many
beliefs. Yet certainly the finitude of the
brain or relevant representational device
defies an infinity or corresponding repre-
sentations. So preserving sententialism
requires disavowing the apparent infinitude
of beliefs in favour of something like a dis-
tinction between (finitely many) actual
402
beliefs - these being relations to actual
Mentalese sentences - and (infinitely many)
dispositional beliefs - these being the unac-
tualized but potential consequences of their
actual counterparts. But. this distinction in
hand, we will now need to know how the
mind - as a sentential processor - is able so
elegantly to manage and manipulate its
actual beliefs so as regularly to produce the
new beliefs rationally demanded of it in
response to detectable environmental fluc-
tuations. This and other related matters
lead to notoriously difficult research prob-
lems whose solution certainly bears upon
the plausibility of the language of thought.
The sententialist must admit that if these
problems finally prove intractable, then
whatever warrant sententialism might
otherwise have had will have evaporated
(Fodor, 1983). But this aside, there are
additional reasons in abductive support of
sententialism.
SYSTEMATICITY AND PRODUCTIVITY
Thinking is also systematic and productive
(Fodor and Pylyshyn, 1988; Dennett, 1987).
Abelard wonders whether William is taller
than Roscelin. This implies that Abelard is
capable of considering that Roscelin is taller
than William. More generally, the fact that
Abelard can have some thoughts entails
that he can have certain other semantically
related thoughts. How is this systematicity
possible?
Suppose that Abelard's thought that
William is taller than Roscelin involves the
registration of 'WILLIAM IS TALLER THAN
ROSCELIN'. This Mentalese sentence is itself
a complex representation containing simpler
representations, mental words in the most
rudimentary case. As complex mental
representations, mental sentences result
from processes ultimately defined on Menta-
lese words and expressions. So, if Abelard
can produce 'WILLIAM IS TALLER THAN
ROSCELIN', he must have access to
'WILLIAM', 'ROSCELIN' and 'IS TALLER
THAN'. And if he has these mental repre-
sentations, he is capable of producing
'ROSCELIN IS TALLER THAN WILLIAM'
and, thereby, considering that Roscelin is
taller than William. Sententialism posits
that mental representations are linguisti-
cally complex representations whose
semantic properties are determined by the
semantic properties of their constituents. If
this should prove correct, then we would be
poised to explain the systematicity of
thought (Schiffer, 1987).
Closely related to thought's systematicity
is its productivity. Thought is productive in
that we appear to have a virtually un-
bounded competence to think ever more
complex novel thoughts having certain
clear semantic ties to their less complex pre-
decessors. Abelard's thinking that Eloise is
ashen suffices for his being able to think
that Eloise is not ashen, which in turn
secures his ability to think that Eloise is not
not ashen. And so on, until the brain's phy-
sical resources are exhausted. Systems of
mental representation apparently exhibit
the sort of productivity distinctive of spoken
languages. Sententialism accommodates
this fact by identifying the productive
system of mental representation with a lan-
guage of thought, the basic terms of which
are subject to a productive grammar.
Opacity
The opacity of thought also legislates in
favour of sententialism. While thought
respects some semantic relations among
mental representations, it can be utterly
blind to others. Abelard believes that Eloise
is ashen. Eloise is, in fact, the most literate
woman in Paris, though Abelard is ignorant
of that. He fails, then, to believe that the
most literate woman in Paris is ashen. How
is this opacity possible? (See INTENTION-
ALITY.) Sententialism answers: To believe
that Eloise is ashen is distinctively to be
related to 'ELOISE IS ASHEN'. To believe
that the most literate woman in Paris is
ashen is to be related to the different mental
sentence, 'THE MOST LITERATE WOMAN
IN PARIS IS ASHEN'. The sheer fact that
Abelard is related to the first does not suffice
for his being similarly related to the second,
even though the two mental sentences
happen to agree in truth value in virtue of
denoting the same woman and attributing
to her the same property. Thinking, accord-
ing to sententialism, may then be like
quoting. To quote an English sentence is to
issue, in a certain way, a token of a given
English sentence type; it is certainly not
similarly to issue a token of every semanti-
cally equivalent type. Perhaps thought is
much the same. If to think is to token a
sentence in the language of thought, the
sheer tokening of one mental sentence need
not insure the tokening of another formally
distinct equivalent. Hence, thought's
opacity.
OBJECTIONS TO THE LANGUAGE OF
THOUGHT
Objections to the language of thought come
from various quarters. Some will not toler-
ate any edition of representationalism,
including sententialism; others endorse
representationalism while denying that
mental representations could involve any-
thing like a language. We first remark on
general objections to representationalism
and then glance at complaints specific to
sententialism. This done, we conclude with
a peek at the naturalization of the semantics
of the language of thought.
Adverbialism
Representationalism is launched by the
assumption that psychological states are
relational. that being in a psychological
state minimally involves being related to
something. But perhaps psychological states
are not at all relational (Churchland, P.M.,
1981). Might not the logical form of
'Abelard thinks that Eloise is ashen' be the
same as 'Abelard argues with eloquence'
rather than 'Abelard argues with William'?
'Abelard argues with eloquence' is equiva-
lent to 'Abelard argues eloquently', a
monadic predication assigning the monadic
property of arguing eloquently. Similarly,
'Abelard thinks that Eloise is ashen' may
attribute to Abelard not a relation to any-
thing but simply the monadic property of
thinking in a certain way, namely 'Eloise-is-
ashen-Iy'. Adverbialism begins by denying
403
that expressions of psychological states are
relational, infers that psychological states
themselves are monadic and, thereby,
opposes classical versions of representation-
alism, including sententialism.
Adverbialism aspires to ontological sim-
plicity in eschewing the existence of entities
as theoretically recondite as mental rep-
resentations. Nevertheless. it is hard
pressed plausibly and simply to explain
what is intuitively semantically common to
Abelard's thoughts that Eloise is ashen and
that Roscelin is ashen. The supposed
monadic properties of thinking Eloise-is-
ashen-Iy and thinking Roscelin-is-ashen-Iy
are, apparently, no more mutually similar
than either is to the property of thinking
William-is-tall-ily. It is, after all, only an
orthographic accident and totally without
significance that the predicates for the first
two properties have portions of their spel-
ling in common. Thus, unless adverbialism
allows for internally complex properties - in
which case it seems to have no metaphysi-
cal advantage over its relational rival - it
seems unable to meet the psychological
facts.
Instrumentalism
Others urge that talk of mental representa-
tions not be construed realistically but only
as a sort of useful. predictive psychological
calculus (Dennett, 1987). Possibly ascrip-
tions of thoughts are simply attempts to
explain behaviour in the face of massive
ignorance of the relevant internal dynamics.
A small child may speak grammatically
correct English, and we may say of her that
she knows that 'corn' is a noun. However.
from this it would be quite a long leap to
infer that the child actually deploys a spe-
cific mental representation that itself lit-
erally means that 'corn' is a noun. No.
better to say that installed in the child is
some unspecified cognitive architecture that
warrants our saying only that she implicitly
knows 'corn' to be a noun and not that she
manipulates any representation explicitly
representing that 'corn' is a noun. The
child's knowledge is perhaps best viewed as
simply a state supervenient on any of many
404
quite disparate cognitive architectures (see
SUPERVENIENCE).
A representationalist will reply that. yes,
not all thought ascriptions do point to speci-
fic mental representations, but those that do
not nevertheless do ride piggy-back on those
that do. Strictly, the child does not think
that 'corn' is a noun. Rather, she literally
thinks, say, that it is permissible to utter
'Please pass the corn' but not 'Please corn
the plate'. When we ascribe to the child the
thought that 'corn' is a noun, we do not
speak literally but only acknowledge that
she is disposed to use 'corn' in certain ways.
And if we are representationalists, we will
hold that the child's way with 'corn' is the
result of her processing specific mental
representations in certain ways. As casual
observers of ourselves and others, we
cannot cavalierly suppose that our psycho-
logical attributions correctly capture the
targeted internal dynamics. Nonetheless,
when scientists of the mind finally manage
to isolate the states over which psycho-
logical processes are defined. the isolated
states will prove to be relations to mental
representations, indeed mental sentences.
Attitudes and Conceptual Roles
Other anti-representationalists, for the sake
of reductio, begin from the hypothesis of
representationalism and suppose that
Abelard's (propositional attitude of) believ-
ing Mars to be a planet consists in his being
related in a certain way to a particular
mental representation, R, meaning that
Mars is a planet. If it is just such an occur-
rence of R in Abelard that constitutes his
believing Mars to be a planet, then if R were
to be copied in, say, Roscelin. Roscelin
would thereby believe Mars to be a planet.
But this is absurd. For Roscelin need not.
simply by registering R. have any attitude
that must be said to be about Mars. He
might. despite the copying. know nothing at
all about Mars. He might have never seen it.
He might not know that it is to be found in
the night sky or have any notion whatso-
ever of what distinguishes planets from
stars. Deprived of this information. Roscelin
could hardly be said to believe anything
about Mars, much less that it is a planet.
despite the fact that he happens, by artifice,
to register R. Thus, belief is not just a rela-
tion to a representation and, a fortiori, not a
relation to a mental sentence (Dennett,
1987).
This objection to representationalism
generally and sententialism particularly
shows at most that belief cannot simply be a
relation to an isolated mental representa-
tion. Believing that Mars is a planet may
typically require that a certain mental
representation play a designated conceptual
role in a production involving a cast of atti-
tudes and a chorus of representations. Yet
so much is consistent with representations -
mental sentences - figuring prominently in
holistic doxastic systems (see HOLISM).
Perhaps to believe anything one must be
disposed to deploy many representations in
lots of ways. It may well be that the
example at hand simply shows that the
belief relation is itself to be functionally con-
strued (see FUNCTIONALISM). That is, Rosce-
lin's being related by belief to R may entail
that Roscelin be disposed to issue R in
certain contexts and, in other contexts, to
use R in the production of other representa-
tions and behaviour. Nevertheless, this is
consistent with Roscelin's mentally repre-
senting - as opposed to believing - that Mars
is a planet sheerly by tokening, and not
otherwise using, R.
Non-verbal Thinkers
Still, some representationalists want noth-
ing of sententialism. Most simply put, their
complaint is this: sententialism puts the cart
before the horse. First comes the ability to
think, then the ability to use language.
Non-linguistic animals and neonates can
think, but neither can speak. Therefore,
thinking must involve some form of non-
linguistic mental representation (Church-
land, P.S., 1980).
Put so baldly. the complaint is plainly a
non sequitur. Yes, some thinking creatures
do not speak; some do not have a public lan-
guage. Still, that does not imply that they
lack an unspoken, internal, mental lan-
guage. Sententialism need not insist that the
language of thought be any natural spoken
language like Chinese or English. Rather, it
simply proposes that psychological states
that admit of the sort of semantic properties
scouted above are likely relations to the sort
of structured representations commonly
found in, but not isolated to, public lan-
guages. This is certainly not to say that all
psychological states in all sorts of psycho-
logical agents must be relations to mental
sentences. Rather the idea is that thinking -
at least the kind Abelard exemplifies -
involves the processing of internally
complex representations. Their semantic
properties are sensitive to those of their parts
much in the manner in which the meanings
and truth conditions of complex public sen-
tences are dependent upon the semantic fea-
tures of their components. Abelard might
also exploit various kinds of mental repre-
sentations and associated processes. A sen-
tentialist may allow that in some of his
cognitive adventures Abelard rotates mental
images or recalibrates weights on connec-
tions among internally undifferentiated net-
worked nodes. Sententialism is simply the
thesis that some kinds of cognitive phenom-
ena are best explained by the hypothesis of a
mental language. There is, then, no prin-
cipled reason to suppose that the cognitive
prowess of non-verbal creatures precludes
the language of thought.
NATURALISTIC SEMANTICS AND
SENTENTIALISM
Such are among sententialism's replies to
some classical objections. Nonetheless,
while sententialism may be able to with-
stand these objections, if it hopes to triumph
it must finally march on the fundamental
semantic question: How do expressions in
the language of thought take on meaning
(Field, 1978)? Minimally, it is necessary to
specify sufficient conditions for Mentalese
expressions coinciding in meaning or
content. Here sententialism might dragoon
its structured representations and con-
jecture that Mentalese expressions with the
same structure are the same in content
(Fodor, 1980). This hypothesis is radically
405
individualistic; it asserts that sameness of
mental meaning is determined by non-
relational properties intrinsic to Mentalese
expressions (see SUPERVENIENCE). Individu-
alism asserts that mental sentences that are
physically, formally or functionally the
same are semantically the same.
Certainly, individualism is disconfirmed if
Mentalese should include indexicals. 'I AM
THE WINNER' in the minds of both Abelard
and Eloise cannot be true in each case. But
if the truth value of this mental sentence
can vary with its tokening, then copies of
the same need not mean the same. How-
ever, consideration of mental indexicals to
one side, individualism is subject to another,
now classic, challenge (Putnam, 1975;
Burge, 1979). Might not an extrinsic differ-
ence in physical, social or historical setting
establish a semantic difference in intrinsi-
cally identical mental representations?
Assume that Twin-Abelard is a synchro-
nized molecular duplicate of Abelard. So, all
and only the mental sentences in Abelard
occur in Twin-Abelard. The Abelards
inhabit numerically distinct but otherwise
similar worlds, with the only difference
between the worlds residing in a non-
evident, deep theoretical difference in the
phenomenally identical stuffs called 'water'
in both worlds. Strictly, then, whereas
Abelard's world contains water, Twin-
Abelard's world does not. Both Abelards
happen to token the Mentalese sentence
'WATER IS WET'. They therefore satisfy the
condition individualism takes to suffice for
sameness of thought content. Yet do they
really think the same? Abelard thinks of
water; Twin-Abelard does not (see TWIN
EAR TH).
BROAD AND NARROW CONTENT
It is universally agreed that, since the truth
conditions of the Abelards' thoughts differ,
there is a sense in which the mental sen-
tences of the Abelards do indeed differ in
content - broad content. Still, there is
another sense, one oblivious to differences
in truth conditions, in which the Abelards'
mental representations are the same in
406
content - narrow content. From the view-
point of Abelard - from the inside, as it
were - his thought is semantically indis-
tinguishable from that of his twin. If, by
some artifice, Abelard should be instanta-
neously put in the situation of Twin-
Abelard but otherwise unaltered, Abelard's
thinking would continue just as it otherwise
would have. But then there must be a sense
in which contextual differences do not
determine semantic differences. Thus, dupli-
cate mental sentences have the same
narrow content even if they should have
different broad content.
How, then, do broad and narrow con-
tents mesh? Minimally, what secures
water's being wet as the truth condition for
Abelard's, but not Twin-Abelard's, token of
'WATER IS WET'? An answer here certainly
would serve to sanction the notion of
narrow mental content. For narrow content
is evidently a function from the context in
which a mental representation occurs to its
truth (or satisfaction) conditions. That is,
mental representations are the same in
narrow content just in case they have the
same broad content or truth conditions in
all the same contexts (Fodor, 1990). Thus,
what contextually determines the truth
conditions of Abelard's token of 'WATER IS
WET'? Since this mental representation is
itself complex, it is natural to suppose
that its truth conditions are somehow influ-
enced by the semantic properties of its
constituents. Accordingly, what - if any-
thing - naturalistically determines that, in
Abelard's actual context, his 'WATER'
denotes water (Field, 1978; Schiffer, 1987)?
HOLISM AND ATOMISM
Holism (Harman, 1982; Field, 1978); and
atomism (Dretske, 1988; Fodor, 1990; Milli-
kan, 1984) compete as strategies for
answering this last question. Holists hold
that 'WATER' denotes as it does because it
plays the specific inferential (or causal) role
it does in Abelard's cognitive economy. This
role must be construed in terms embracing
facts extrinsic to Abelard's skin under pain
of preserving the broad semantic distinction
between the Abelards' mental representa-
tions (Block, 1986).
Holistic theories of mental content are
terribly brittle. They entail that agents in
the same environment cannot have
thoughts with the same content unless they
have perfectly isomorphic representational
systems. This brittleness mocks the fact that
mental content can remain constant over
variation in conceptual organization, in-
cluding learning. Also, one wonders
whether holists, in relying on the inferential
connections among representations in order
to specify their meanings, will find it neces-
sary to distinguish between those inferences
that are central to a representation's
meaning and those that are not (Fodor and
LePore, 1992). For in that case holists
appear to be required to resurrect the dis-
tinction between the analytic and synthetic
long ago buried by QUINE.
Atomists hope to pin the denotation of a
mental representation on the manner in
which it nomic ally depends on its environ-
ment. Here too there is competition. Some
atomists are satisfied to fuss with the natur-
alized notions of information (Dretske,
1988). Others want to complement that
with attention to the asymmetric depend-
encies among relations between mental
representations and their various external
causes (Fodor, 1990). Others still look to the
historical or evolutionary niche a mental
representation occupies (Millikan, 1984). In
any case, it yet remains on the research
agenda of representationalists generally and
sententialists particularly to explain how,
exactly, mental symbols accrue content.
See also An Essay on Mind section 2.1.3;
CONCEPTS; CONCEPTUAL ROLE SEMANTICS;
CONTENT; DENNETT; DRETSKE; FODOR; FOLK
PSYCHOLOGY; PROPOSITIONAL ATTITUDES;
BIBLIOGRAPHY
Block, N. 1986. Advertisement for a semantics
for psychology. Midwest Studies in Philoso-
phy: Studies in the Philosophy of Mind, Vol.
10,615-78.
Midwest Studies in Philosophy: Studies in
Metaphysics, Vol. 4, 73-122.
Churchland, P.M. 1981. Eliminative materi-
alism and the propositional attitudes. Journal
of Philosophy, 78, 67-90.
Churchland, P.S. 1980. Language, thought
and information processing. Noils, 14, 147-
70.
Dennett, D. 1987. The Intentional Stance. Cam-
Dretske, F. I. 1988. Explaning Behavior. Cam-
Field, H. 1978. Mental representation.
Erkenntnis, 13, 9-61.
Fodor, J. 1980. Methodological solipsism con-
sidered as a research strategy in cognitive
psychology. Behavioral and Brain Sciences, 3,
63-110.
Fodor, J. 1983. Modularity of Mind. Cambridge,
Fodor, J., and LePore, E. 1992. Holism: A
Shopper's Guide. Oxford: Basil Blackwell.
Fodor, J., and Z. Pylyshyn. 1988. Connection-
ism and cognitive architecture: A critical
analysis. Cognition, 28, 3-71.
Harman, G. 1982. Conceptual role semantics.
Notre Dame Journal of Formal Logic, 23,
432-3.
Maloney, J.C. 1989. The Mundane Matter of the
Mental Language. Cambridge: Cambridge
University Press.
Millikan, R.G. 1984. Language, Thought and
Other Biological Categories. Cambridge, MA:
Putnam, H. 1975. The meaning of 'Meaning'.
In Mind. Language and Reality. Cambridge:
Schiffer. S. 1987. Remnants of Meaning. Cam-
Searle. J. 1980. Minds, brains and programs
Behavioral and Brain Sciences. 3,417-24.
nitive Science: The Case Against Belief. Cam-
J. CHRISTOPHER MALONEY
407
language of thought (2) There is no
denying it: the Language of Thought (LOT)
hypothesis has a compelling neatness about
it. A thought is depicted as a structure of
internal representational elements, com-
bined in a lawful way, and playing a
certain functional role in an internal pro-
cessing economy (see FUNCTIONALISM;
REPRESENT ATION). Relations between
thoughts (e.g. the semantic overlap
between the thought that John loves wine
and the thought that John loves food)
consist in the recurrence of some of the
inner representational elements. Novel
thoughts and the much-vaunted systemati-
city of thought (the fact that beings who
can think 'John loves wine' and 'Mary loves
food' can always think 'John loves food'
and 'Mary loves wine' - see Fodor and
Pylyshyn, 1988) are accounted for in the
same way. Once the representational ele-
ments and combinatoric rules are in place,
of course such interanimations of potential
contents will occur. The predictive success
of propositional attitude talk (the ascription
of e.g. beliefs and desires such as 'John
believes that the wine is good') is likewise
explained, on the hypothesis that the public
language words pick out real inner rep-
resentational complexes which are causally
potent and thus capable of bringing about
actions. And finally, what distinguishes an
intentional action from a mere reflex is, on
this model the fact that intervening
between input and action there is, in the
intentional case, an episode of actual token-
ing of an appropriate symbol string. 'No
intentional causation without explicit repre-
sentation', as the rallying cry (see Fodor,
1987, p. 25) goes. A pretty package indeed.
And all for the price of a language of
thought. My advice to the consumer,
however, is 'beware': all that neatness
masks some hidden costs.
COST A: CONSERVATIVE ATTITUDE
TOWARDS REPRESENTATIONAL CHANGE
It is fair to ask where we get the powerful
inner code whose representational elements
need only systematic construction to
408
express, for example, the thought that
cyclotrons are bigger than black holes (see
INNATENESS). But on this matter, the
Language of Thought theorist has little to
say. All that CONCEPT learning could be
(assuming it is to be some kind of rational
process and not due to mere physical
maturation or a bump on the head),
according to the LOT theorist, is the trying
out of combinations of existing repre-
sentational elements to see if a given
combination captures the sense (as evi-
denced in its use) of some new concept.
The consequence is that concept learning,
conceived as the expansion of our repre-
sentational resources, simply does not
happen. What happens instead is that we
work with a fixed, innate repertoire of ele-
ments whose combination and construction
must express any content we can ever
learn to understand (see e.g. Fodor, 1975).
And note that this is not the (trivial) claim
that in some sense the resources a system
starts with must set limits on what knowl-
edge it can acquire. For these are limits
which flow not, for example, from sheer
physical size, number of neurons, con-
nectivity of neurons, etc., but from a base
class of genuinely representational ele-
ments. They are more like the limits that
being restricted to the propositional calcu-
lus would place on the expressive power of
a system than, say, the limits that having a
certain amount of available memory
storage would place on one.
But this picture of representational stasis,
in which all change consists in the re-
deployment of existing representational
resources, is one that is fundamentally
alien to much influential theorizing in
DEVELOPMENTAL PSYCHOLOGY. The prime
example of a developmentalist who believed
in much stronger forms of representational
change is probably Piaget (see e.g. Piaget,
1955). But much contemporary work is
likewise committed to studying kinds of
change which a staunchly nativist Fodor-
ian must deem impossible. One example is
work by Annette Karmiloff-Smith (e.g.
Karmiloff-Smith, 1992) which places epi-
sodes in which there occurs a genuine
expansion of representational power at the
very heart of a model of human develop-
ment. In a similar vein, recent work in the
field of connectionism (see CONNECTIONISM,
and below) seems to open up the possibility
of putting well-specified models of strong
representational change back at the centre
of the cognitive scientific endeavour (see e.g.
Bates and Elman, 1992; Plunkett and
Sinha, 1991).
Someone might, of course, adopt the
Language of Thought hypothesis as a model
of the structure of adult human cognition
yet reject the strong nativist version of its
origins. The combinatorial inner code would
then be treated as a product of an extended
developmental process. The question that
would then arise is: how deep is the role of
this product, and how are we to explain its
development? For example, it might be (see
e.g. DENNETT, 1991) that the kind of com-
binatorial inner code thus isolated is devel-
oped only (at best) by language-users and
that much of (non-human) animal cogni-
tion, as well as much of what goes on in
other kinds of human problem solving,
requires understanding in very different
terms. To the extent that this was so, the
Language of Thought would not play quite
the central role in cognitive science that
Fodor imagines (see cognitive psychology).
Instead such a quasi-linguistic code would
be one developmental product among
many, whose explanatory role was limited
to a particular sub-area of human cogni-
tion. Such a sub-area might be so small
as to include only those episodes of reason-
ing and planning which involve the con-
scious rehearsal of sentences (as in
Dennett's (1987) idea of 'linguistically in-
fected cognition') or it may be large enough
to include many other kinds of cognitive
episode (such as musical thought). But
either way, the understanding of how the
underlying combinatoric code develops
might actually be more crucial, in terms of
the deep understanding of cognitive process,
than understanding the structure and use of
the code itself (though, doubtless, the two
projects would need to be pursued hand-in-
hand).
COST B: REPRESENTATIONAL ATOMISM
The Language of Thought story depicts
thoughts as structures of CONCEPTS, which
in turn exist as elements (for any basic
concepts) or concatenations of elements (for
the rest) in the inner code (see Fodor,
1986). One upshot of this is that little can
be said about intrinsic relations between
basic representational items. Even bracket-
ing the (difficult) question of which, if any
words in our public language may express
contents which have as their vehicles
atomic items in the Language of Thought
(an empirical question on which I assume
Fodor to be officially agnostic), the question
of semantic relations between atomic items
in the Language of Thought remains. Are
there any such relations? And if so, in
what do they consist? Two thoughts are
depicted (we saw) as semantically related
just in case they share elements from the
language of thought. But the elements
themselves (like the words of public lan-
guage on which they are modelled) seem to
stand in splendid isolation from one
another. An advantage of some connec-
tionist approaches (see CONNECTIONISM)
lies precisely in their ability to address
questions of the interrelation of basic repre-
sentational elements (in fact, activation
vectors) by representing such items as loca-
tions in a kind of semantic space. In such a
space related contents are always expressed
by related representational elements. The
connectionist's conception of significant
structure thus goes much deeper than the
Fodorian's. For the connectionists repre-
sentations need never be arbitrary. Even
the most basic representational items will
bear non-accidental relations of similarity
and difference to one another. The Fodor-
ian, having reached representational
bedrock, must explicitly construct any such
further relations. They do not come for free
as a consequence of using an integrated
representational space. Whether this is a
bad thing or a good one will depend, of
course, on what kind of facts we need to
explain. But my suspicion is that repre-
sentational atomism may turn out to be a
409
conceptual economy that a science of the
mind cannot afford.
COST C: NO MODEL OF GLOBAL PROCESSES
Consider FODOR (1983). Fodor (1983) is,
surprisingly, a compact and (to my mind)
decisive demonstration of the ultimate
inadequacy of the LOT story. For the LOT
story, as Fodor often reminds us, gets its
clearest computational expression in so-
called classical ARTIFICIAL INTELLIGENCE
(AI) approaches. Specifically, in such classi-
cal AI approaches as posit a fixed set of
representational primitives and a kind of
grammar defined over them, and which go
on to model mental processes as involving
sequences of legal transitions among the
stored representational complexes. Yet the
argument in Fodor (1983) is that this kind
of model faces its greatest challenge in
trying to account for the nature of so-called
Central Processing. Central Processing, in
this usage, means the system or systems
involved in belief-fixation and rational
thought and inference (see RATIONALITY),
viz. the very systems which would be the
locus of what, in his (1987) were called
'episodes of mental causation'. The claim, in
1983, was that classical, symbolic AI had
made great progress in the understanding of
non-central systems - these were largely
domain-specific input systems (see e.g.
Fodor, 1983, p. 103). But two properties
distinguished the processing achieved by
such systems from that achieved by the
central systems, and these properties were
shown to be highly resistant to any classical
treatment. These were the properties of
being Quinean and being isotropic. A
Quinean system is one in which 'the degree
of confirmation assigned to any given
hypothesis is sensitive to properties of the
entire belief system' (Fodor, 1983, p. 107).
An isotropic system is one in which any
part of the knowledge encoded can turn out
to be relevant to the system's decisions
about what to believe (see Fodor, 1983,
p. 105). The two properties are obviously
closely bound up. Together, they character-
ize systems in which information processing
410
is profoundly global. And much of what
goes on in Central Processing, according to
Fodor, has just this global character. When
we choose whether or not to accept a new
belief, we do so by allowing both that any
other belief we hold could in principle be
relevant to the decision (isotropic) and by
allowing the decision to turn also on the
collective impact of the sum of our other
beliefs (Quinean; see QUINE). Moreover all
forms of analogical reasoning, by effecting
the 'transfer of information among cognitive
domains previously assumed to be mutually
irrelevant' (Fodor, 1983, p. 107), are them-
selves evidence of the fundamentally iso-
tropic nature of Central Processing. These
kinds of global information processing,
Fodor believes, are distinctive of 'higher
cognition'. Yet they have not succumbed to
the advances of classical AI and cognitive
science. And for a very good reason. It is
that such globally sensitive processing runs
classical systems very quickly into well-
known problems of combinatorial explosion.
For there are no fixed sets of beliefs, mark-
able out in advance, amongst which the
relevant ones can be assumed to hide. But
the problem of searching amongst the con-
tents of the entire belief set is simply intract-
able in classical models; the amount of time
and/or computation required increases
exponentially with the number of items to
be taken into account. One instance of this,
noted by Fodor, is the so-called frame
problem in AI, viz. how to update the right
sub-set of a system's beliefs as new informa-
tion is received. The frame problem, Fodor
claims, is just one instance of the general
inability of classical AI to model globally
sensitive information processing. Indeed, so
pessimistic does Fodor become about the
whole situation that he proclaims his in-
famous 'First law of the non-existence of
Cognitive Science', viz.: 'The more global
... a cognitive process is, the less anybody
understands it' (Fodor, 1983, p. 107).
Classical AI, to sum this all up, is depicted
by Fodor (1983) as a research programme
which has done well in helping us model a
variety of input systems and peripheral,
modular processing. But one which has
failed to illuminate the domain of real
thought, i.e. belief-fixation and central pro-
cessing. How strange, then, that the upshot
of Fodor and Pylyshyn (1988) seems to be
that although connectionist models might
(perhaps) help us understand various per-
ipheral. perceptual processing devices, the
classical approach must be preferred in the
domain of 'real' thought, since real
thoughts (like the thought that Mary loves
John) form a systematic set and the best
explanation of such systematicity lies in our
supposing them to be underpinned by clas-
sical processing strategies. Strange too, that
Fodor (1987, pp. 143-7) stresses the com-
mitments of actual cognitive models to
symbol-manipulating approaches (using
examples drawn from the processing of lan-
guage parsing input devices) as part of an
argument in favour of a symbolic model of
real thought, i.e. central processing. There
is no mention of the fact that all the suc-
cessful models, it seems, are targeted on the
fundamentally different class of non-global
computational processes. Strangeness is not,
of course, to be confused with incon-
sistency. It is quite consistent to hold that
(1) classical AI faces fundamental prob-
lems in dealing with the global
processes characteristic of Central Pro-
cessing,
and
(2) classical AI is especially well placed to
deal with the systematicity of contents
characteristic of Central Processing.
Still, to completely reject connectionism for
its alleged failure to explain systematicity
whilst simultaneously believing that the
classical alternative fails to explain an
equally deep feature of central processing
seems a trifle partisan. The more so since (to
a degree) the acknowledged strengths of the
connectionist approach lie in its ability to
make computationally tractable the task of
globally sensitive information processing.
The fixation of belief, insofar as it is to be
sensitive to both the individual and the col-
lective properties of a large set of simulta-
neously held beliefs, is surely a prime case
for a connectionist treatment. Such systems
excel at simultaneously satisfying large
bodies of constraints, and the time taken to
perform such a task does not increase ex-
ponentially with the number of constraints
to be respected (for a nice discussion, see
Oaksford and Chater, 1991).
To sum up, then, I am not convinced that
the benefits of the LOT story outweigh its
costs. The key benefit is, I suppose, the neat
account of systematic and productive
thought. The key costs are: the association
with a strong nativist model of our repre-
sentational resources (and the associated
antipathy towards the idea of strong repre-
sentational change); the commitment to
representational atomism and consequent
inability to pursue the question of semantic
structure beyond a given (and all-too-
rapidly reached) point; and the self-
acknowledged intractability of globally
sensitive information processing to (at least
the classical AI incarnation of) the LOT
approach.
My own belief, as the reader will surely
have guessed, is that something a bit like a
language of thought may well exist. But it is
the symbolic problem-solving tip of a large
and developmentally extended iceberg.
Beneath the symbolic waters, and reaching
back across our individual developmental
history, lie the larger, less well-defined
shapes of our basic cognitive processes. To
understand these, and the episodes of
genuine representational change in which
they figure, is to address the fundamentals
of cognition.
See also COMPUTATIONAL MODELS; CON-
CEPTUAL ROLE SEMANTICS; MODULARITY;
PSYCHOLOGY AND PHILOSOPHY; THOUGHTS;
BIBLIOGRAPHY
Bates, E., and Elman, J. 1992. Connectionism
and the study of change. Technical Report
9202, Center for Research and Language,
University of California, San Diego.
411
LEWIS. DAVID: REDUCTION OF MIND
Dennett. D. 1991. Mother Nature versus the
walking encyclopedia. In Philosophy and Con-
nectionist Theory. ed. W. Ramsey. S. Stich.
and D. Rumelhart. Hillsdale. N.J.: Erlbaum.
Fodor. J. 1975. The Language of Thought. New
York: Crowell.
Fodor. J. 1983. The Modularity of Mind. Cam-
Fodor. J. 1986. The present status of the
innateness controversy. In Representations.
Sussex: Harvester.
Fodor. J. 1987. Psychosemantics: The Problem of
Meaning in the Philosophy of Mind. Cam-
Fodor. J .. and Pylyshyn. Z. 1988. Connection-
ism and cognitive architecture. A critical
analysis. Cognition. 28. 3-71.
KarmilotT-Smith. A. 1992. Beyond Modularity:
A Developmental Perspective on Cognitive
Science. Cambridge. MA: MIT Press/Bradford
Books.
Oaksford. M .. and Chater. C. 1991. Against
logicist cognitive science. Mind and Lan-
guage. 6: 1. 1-38.
Plunkett. K.. and Sinha. C. 1991. Connection-
ism and developmental theory. Psykologisk
Skriftserie Aarhus. 16: 1. 1-34.
Piaget. J. 1955. The Child's Construction of
Reality. Routledge & Kegan Paul.
ANDY CLARK
Lewis, David: Reduction of Mind I am a
realist and a reductive materialist about
mind. I hold that mental states are con-
tingently identical to physical - in parti-
cular. neural - states. My position is very
like the 'Australian materialism' of Place.
Smart, and especially Armstrong. Like
Smart and Armstrong. I am an ex-Rylean.
and I retain some part of the Rylean legacy.
In view of how the term is contested, I do
not know whether I am a 'functionalist'.
(See FUNCTIONALISM; IDENTITY THEORIES;
PHYSICALISM; RYLE.)
SUPERVENIENCE AND ANALYSIS
My reductionism about mind begins as part
of an a priori reductionism about every-
412
thing. This world. or any possible world.
consists of things which instantiate funda-
mental properties and which. in pairs or
triples or .... instantiate fundamental rela-
tions. Few properties are fundamental: the
property of being a club or a tub or a pub,
for instance. is an unnatural gerrymander,
a condition satisfied by miscellaneous things
in miscellaneous ways. A fundamental, or
'perfectly natural', property is the extreme
opposite. Its instances share exactly some
aspect of their intrinsic nature. Likewise for
relations (see Lewis. 1983a and 1986a. pp.
59-69). I hold. as an a priori principle, that
every contingent truth must be made true.
somehow. by the pattern of coinstantiation
of fundamental properties and relations. The
whole truth about the world. including the
mental part of the world. supervenes on this
pattern. If two POSSIBLE WORLDS were
exactly isomorphic in their patterns of co-
instantiation of fundamental properties and
relations. they would thereby be exactly
alike simpliciter (Lewis. 1992. p. 218). (See
SUPERVENIENCE.)
It is a task of physics to provide an inven-
tory of all the fundamental properties and
relations that occur in the world. (That's
because it is also a task of physics to dis-
cover the fundamental laws of nature. and
only the fundamental properties and rela-
tions may appear in the fundamental laws;
see Lewis. 1983a. pp. 365-70). We have no
a priori guarantee of it. but we may reason-
ably think that present-day physics goes a
long way toward a complete and correct
inventory. Remember that the physical
nature of ordinary matter under mild condi-
tions is very well understood (Feinberg.
1966). And we may reasonably hope that
future physics can finish the job in the same
distinctive style. We may think. for
instance, that mass and charge are among
the fundamental properties; and that what-
ever fundamental properties remain as yet
undiscovered are likewise instantiated by
very small things that come in very large
classes of exact duplicates. We may further
think that the very same fundamental prop-
erties and relations. governed by the very
same laws. occur in the living and the dead
parts of the world, and in the sentient and
the insentient parts, and in the clever and
the stupid parts. In short: if we optimisti-
cally extrapolate the triumph of physics
hitherto, we may provisionally accept that
all fundamental properties and relations
that actually occur are physical. This is the
thesis of materialism.
(It was so named when the best physics
of the day was the physics of matter alone.
Now our best physics acknowledges other
bearers of fundamental properties: parts of
pervasive fields, parts of causally active
spacetime. But it would be pedantry to
change the name on that account, and
disown our intellectual ancestors. Or worse,
it would be a tacky marketing ploy, akin to
British Rail's decree that second class pas-
sengers shall now be called 'standard class
customers'.)
If materialism is true, as I believe it is,
then the a priori supervenience of every-
thing upon the pattern of coinstantiation of
fundamental properties and relations yields
an a posteriori supervenience of everything
upon the pattern of coinstantiation of fun-
damental physical properties and relations.
Materialist supervenience should be a con-
tingent matter. To make it so, we supply a
restriction that makes reference to actuality.
Thus: if two worlds were physically iso-
morphic, and if no fundamental properties
or relations alien to actuality occurred in
either world, then these worlds would be
exactly alike simpliciter. Disregarding alien
worlds, the whole truth supervenes upon
the physical truth. In particular, the whole
mental truth supervenes. So here we have
the common core of all materialist theories
of the mind (Lewis, 1983a, pp. 361-5).
A materialist who stops here has already
said enough to come under formidable
attack. An especially well-focused version of
the attack comes from Frank Jackson
(1982). Mary, confined in a room where all
she can see is black or white, studies the
physics of colour and colour vision and
colour experience (and any other physics
you might think relevant) until she knows
it all. Then she herself sees colour for the
first time, and at last she knows what it's
like to see colour. What is this knowledge
that Mary has gained? It may seem that she
has eliminated some possibilities left open
by all her previous knowledge; she has dis-
tinguished the actual world from other pos-
sible worlds that are exactly like it in all
relevant physical respects. But if materialist
supervenience is true, this cannot be what
happened. (See CONSCIOUSNESS; QUALIA.)
Materialists have said many things about
what does happen in such a case. I myself,
following Nemirow (1990), call it a case of
know-how: Mary gains new imaginative
abilities (Lewis, 1990). Others have said
that Mary gains new relations of acquaint-
ance, or new means of mental representa-
tion: or that the change in her is just that
she has now seen colour. These suggestions
need not be taken as rival alternatives. And
much ink has been spent on the question
whether these various happenings could in
any sense be called the gaining of 'new
knowledge', 'new belief', or 'new informa-
tion'. But for a materialist, the heart of the
matter is not what does happen but what
doesn't: Mary does not distinguish the actual
world from other worlds that are its physi-
cal duplicates but not its duplicates simpli-
citer.
Imagine a grid of a million tiny spots -
pixels - each of which can be made light or
dark. When some are light and some are
dark, they form a picture, replete with inter-
esting intrinsic gestalt properties. The case
evokes reductionist comments. Yes, the
picture really does exist. Yes, it really does
have those gestalt properties. However the
picture and the properties reduce to the
arrangement of light and dark pixels. They
are nothing over and above the pixels. They
make nothing true that is not made true
already by the pixels. They could go
unmentioned in an inventory of what there
is without thereby rendering that inventory
incomplete. And so on.
Such comments seem to me obviously
right. The picture reduces to the pixels. And
that is because the picture supervenes on
the pixels: there could be no difference in the
picture and its properties without some dif-
ference in the arrangement of light and dark
413
pixels. Further. the supervenience is asym-
metric: not just any difference in the pixels
would matter to the gestalt properties of the
picture. And it is supervenience of the large
upon the small and many. In such a case,
say I. supervenience is reduction. And the
materialist supervenience of mind and all
else upon the arrangement of atoms in the
void - or whatever replaces atoms in the
void in true physics - is another such case.
Yet thousands say that what's good
about stating materialism in terms of super-
venience is that this avoids reductionism!
There's no hope of settling this disagree-
ment by appeal to some uncontested defini-
tion of the term 'reductionism'. Because the
term is contested, and the aim of some con-
testants is to see to it that whatever position
they may hold, 'reductionism' shall be the
name for something else.
At any rate, materialist supervenience
means that for anything mental, there are
physical conditions that would be sufficient
for its presence, and physical conditions
that would be sufficient for its absence.
(These conditions will include conditions
saying that certain inventories are com-
plete: an electron has only so-and-so
quantum numbers, for instance, and it
responds only to such-and-such forces. But
it's fair to call such a condition 'physical',
since it answers a kind of question that
physics does indeed address.) And no matter
how the world may be, provided it is free of
fundamental properties or relations alien to
actuality, a condition of the one sort or the
other will obtain. For all we know so far,
the conditions associated with a given
mental item might be complicated and mis-
cellaneous - even infinitely complicated and
miscellaneous. But so long as we limit our-
selves just to the question of how this
mental item can find a place in the world of
fundamental physics, it is irrelevant how
complicated and miscellaneous the condi-
tions might be.
It may seem unsatisfactory that physical
conditions should always settle whether the
mental item is present or absent. For
mightn't that sometimes be a vague ques-
tion with no determinate answer? A short
414
reply to this objection from vagueness is
that if it did show that the mental was irre-
ducible to fundamental physics despite
supervenience, it would likewise show that
boiling was irreducible to fundamental
physics - which is absurd. For it is a vague
matter just where simmering leaves off and
boiling begins.
A longer reply has three parts. (1) If the
physical settles the mental insofar as any-
thing does, we still have materialist super-
venience. Part of what it means for two
physically isomorphic worlds to be just alike
mentally is that any mental indeterminacy
in one is exactly matched by mental inde-
terminacy in the other. (2) Whenever it is a
vague question whether some simplistic
mental classification applies, it will be deter-
minate that some more subtle classification
applies. What's determinate may be not
that you do love him or that you don't, but
rather that you're in a certain equivocal
state of mind that defies easy description.
(3) If all indeterminacy is a matter of
semantic indecision (Lewis, 1986a, pp.
212-l3), then there is no indeterminacy in
the things themselves. How could we
conjure up some irreducible mental item
just by failing to decide exactly which redu-
cible item we're referring to?
It may seem that when supervenience
guarantees that there are physical condi-
tions sufficient for the presence or absence
of a given mental item, the sufficiency is of
the wrong sort. The implication is necessary
but not a priori. You might want to say, for
instance, that black-and-white Mary really
did gain new knowledge when she first saw
colour; although what she learned followed
necessarily from all the physics she knew
beforehand, she had remained ignorant
because it didn't follow a priori.
A short reply to this objection from
necessity a posteriori is that if it did show
that the mental was irreducible to funda-
mental physics, it would likewise show that
boiling was irreducible to fundamental
physics - which is absurd. For the identity
between boiling and a certain process
described in fundamental physical terms is
necessary a posteriori if anything is.
(A longer reply, following Jackson
(1992), is founded upon the 'two-dimen-
sional' analysis of necessity a posteriori put
forward by Stalnaker (1978), Davies and
Humberstone (1980), and TichY (1983).
Two-dimensionalism says that there is no
such thing as a necessary a posteriori pro-
position. However, one single sentence e
may be associated in two different ways
with two different propositions, one of them
necessary and the other one contingent;
and the contingent one can be known only
a posteriori. Suppose we choose to adopt a
conception of meaning under which our
conventions of language sometimes fix
meanings only as a function of matters of
contingent fact - for example, a conception
on which the meaning of 'boils' is left
dependent on which physical phenomenon
turns out to occupy the boiling-role. Then if
we interpret a sentence e using the mean-
ings of its words as fixed in world WI, we
get proposition HI; using the meanings as
fixed in W 20 we get H
2
; and so on. Call these
the propositions horizontally expressed by e
at the various worlds: and let H be the pro-
position horizontally expressed by e at the
actual world. The proposition diagonally
expressed by e is the proposition D that holds
at any world W iff the proposition horizon-
tally expressed by e at W is true at W. So if
we know D, we know that e horizontally
expresses some truth or other, but we may
not know which truth. Sentence e is neces-
sary a posteriori iff H is necessary but D is
knowable only a posteriori. Likewise, a pro-
position P necessarily implies that e iff P
implies H; but P a priori implies that e iff P
implies D. Our worry was that when e was
about the mind, and P was a premise made
true by fundamental physics, P might imply
that e necessarily but not a priori. But if so,
and if you think it matters, just take
another proposition Q; let Q be true at
exactly those worlds where e horizontally
expresses the same proposition H that it
actually does. Q is true. Given the materi-
alist supervenience of everything, Q as well
as P is made true by fundamental physics. P
and Q together imply a priori that e. So the
gap between physical premises and mental
conclusion is closed. Anyone who wants to
reopen it - for instance, in order to square
materialist supervenience with Mary's sup-
posed ignorance - must somehow show
that the two-dimensional analysis of neces-
sity a posteriori is inadequate.)
If we limit ourselves to the question how
mind finds a place in the world of physics,
our work is done. Materialist supervenience
offers a full answer. But if we expand our
interests a little, we'll see that among the
supervenient features of the world, mind
must be very exceptional. There are count-
less such features. In our little toy example
of the picture and the pixels, the super-
venient properties number 2 to the power: 2
to the millionth power. In the case of mate-
rialist supervenience, the number will be far
greater. The infinite cardinal beth-3 is a
conservative estimate. The vast majority of
supervenient features of the world are given
only by miscellaneously infinite disjunctions
of infinitely complex physical conditions.
Therefore they are beyond our power to
detect, to name, or to think about one at a
time. Mental features of the world, however,
are not at all beyond our ken. Finite as-
semblies of particles - us - can track them.
Therefore there must be some sort of simpli-
city to them. Maybe it will be a subtle sort
of simplicity, visible only if you look in just
the right way. (Think of the Mandelbrot set:
its overwhelming complexity, its short and
simple recipe.) But somehow it must be
there. Revealing this simplicity is a job for
conceptual analysis.
Arbiters of fashion proclaim that analysis
is out of date. Yet without it, I see no pos-
sible way to establish that any feature of the
world does or does not deserve a name
drawn from our traditional mental vocabu-
lary. We should repudiate not analysis itself,
but only some Simplistic goals for it. We
should allow for semantic indecision: any
interesting analysandum is likely to turn
out vague and ambiguous. Often the best
that anyone analysis can do is to fall safely
within the range of indecision. And we
should allow for semantic satisficing: ana-
lysis may reveal what it would take to
415
deserve a name perfectly, but imperfect
deservers of the name may yet deserve it
well enough. (And sometimes the perfect
case may be impossible.) If so. there is
bound to be semantic indecision about how
well is well enough.
I offer not analyses, but a recipe for ana-
lyses. We have a very extensive shared
understanding of how we work mentally.
Think of it as a theory: FOLK PSYCHOLOGY.
It is common knowledge among us; but it is
tacit, as our grammatical knowledge is. We
can tell which particular predictions and
explanations conform to its principles. but
we cannot expound those principles system-
atically. (Pace Lewis, 1972, p. 256, eliciting
the general principles of folk psychology is
no mere matter of gathering platitudes.)
Folk psychology is a powerful instrument of
prediction. We are capable of all sorts of
behaviour that would seem bizarre and
unintelligible. and this is exactly the beha-
viour that folk psychology predicts, rightly.
will seldom occur. (But we take a special
interest in questions that lie beyond the pre-
dictive power of folk psychology: wherefore
ingrates may fairly complain of a lack of
interesting predictions!) Folk psychology has
evolved through thousands of years of close
observation of one another. It is not the last
word in psychology, but we should be con-
fident that so far as it goes - and it does go
far - it is largely right.
Folk psychology concerns the causal rela-
tions of mental states, perceptual stimuli,
and behavioural responses. It says how
mental states, singly or in combination, are
apt for causing behaviour; and it says how
mental states are apt to change under the
impact of perceptual stimuli and other
mental states. Thus it associates with each
mental state a typical causal role. Now we
have our recipe for analyses. Suppose we've
managed to elicit all the tacitly known
general principles of folk psychology. When-
ever M is a folk-psychological name for a
mental state, folk psychology will say that
the state M typically occupies a certain
causal role: call this the M-role. Then we
analyse M as meaning 'the state that typi-
cally occupies the M-role'. Folk psychology
416
implicitly defines the term M, and we have
only to make that definition explicit.
Since the causal roles of mental states
involve other mental states, we might fear
circularity. The remedy is due in its essen-
tials to Ramsey (1931a, pp. 212-236) and
Carnap (1963, pp. 958-66); see also Lewis
(1970, 1972). Suppose, for instance, that
folk psychology had only three names for
mental states: L, M, N. We associate with
this triplet of names a complex causal role
for a triplet of states, including causal rela-
tions within the triplet: call this the LMN-
role. Folk psychology says that the states L,
M, N jointly occupy the LMN-role. That
implies that M occupies the derivative role:
coming second in a triplet of states that
jointly occupy the LMN-role. Taking this as
our M-role, we proceed as before. Say that
the names L, M, N are interdefined. The
defining of all three via the LMN-role is a
package deal.
We might fear circularity for another
reason. The causal roles of mental states
involve responses to perceptual stimuli. But
the relevant feature of the stimulus will
often be some secondary quality - for
instance, a colour. We cannot replace the
secondary quality with a specification of the
stimulus in purely physical terms, on pain
of going beyond what is known to folk psy-
chology. But if we analyse the secondary
quality in terms of the distinctive mental
states its presence is apt to evoke, we close a
definitional circle. So we should take inter-
definition further. Let folk psychology
include folk psychophysics. This will say, for
instance, that the pair of a certain colour
and the corresponding sensation jointly
occupy a complex causal role that consists
in part, but only in part, of the former being
apt to cause the latter. Now we have a de-
rivative role associated with the name of the
colour, and another associated with the
name of the sensation: the role of coming
first or coming second, respectively, in a
pair that jointly occupies this complex role.
We might worry also about the beha-
viour that mental states are apt for causing.
Often we describe behaviour in a mentally
loaded way: as action. To say that you
kicked the ball to your team-mate is to
describe your behaviour. But such a
description presupposes a great deal about
how your behaviour was meant to serve
your desires according to your beliefs; and
also about the presence of the ball and the
playing surface and the other player, and
about social facts that unite players into
teams. More threat of circularity? More need
for interdefinition? 1 don't know how such
further interdefinition would work; and
anyway, it would be well to call a halt
before folk psychology expands into a folk
theory of the entire LebensweItl
Describing the behaviour in purely physi-
cal terms - the angle of the knee, the velo-
city of the foot - would get rid of those
presuppositions. But, just as in the case of
the stimuli, it would go beyond what is
known to folk psychology. Further, these
descriptions would never fit the behaviour
of space aliens not of humanoid shape; and
yet we should not dismiss out of hand the
speculation that folk psychology might
apply to aliens as well as to ourselves.
Fortunately there is a third way to
describe behaviour. When you kicked the
ball, your body moved in such a way that if
you had been on a flat surface in Earth-
normal gravity with a suitably placed ball
in front of you and a suitably placed team-
mate some distance away, then the impact
of your foot upon the ball would have pro-
pelled the ball onto a trajectory bringing it
within the team-mate's reach. That descrip-
tion is available to the folk. They wouldn't
give it spontaneously, but they can recog-
nize it as correct. It presupposes nothing
about your mental states, not even that you
have any; nothing about whether the ball
and the playing field and the gravity and
the team-mate are really there; nothing
about your humanoid shape, except that
you have some sort of foot. It could just as
well describe the behaviour of a mindless
mechanical contraption, in the shape of a
space alien (with a foot), thrashing about in
free fall.
(I don't say that we should really use
these 'if-then' descriptions of behaviour.
Rather, my point is that their availability
LEWIS, DAVID: REDUCTION OF MIND
shows how to unload the presuppositions
from our ordinary descriptions.)
If M means 'the state that typically occu-
pies the M-role' and if that role is only
imperfectly occupied, what are we to do? -
Satisfice: let the name M go to a state that
deserves it imperfectly. And if nothing
comes anywhere near occupying the M-
role? - Then the name M has no referent.
The boundary between the cases is vague.
To take an example from a different term-
inducing theory, 1 suppose it to be in-
determinate whether 'dephlogisticated air'
refers to oxygen or to nothing. But folk
psychology is in far better shape than the
phlogiston theory, despite scare stories to
the contrary. We can happily grant that
there are no perfect deservers of folk-
psychological names, but we shouldn't
doubt that there are states that deserve
those names well enough.
What to do if the M-role, or the LMN-role,
turns out to be doubly occupied? I used to
think (Lewis, 1970, 1972) that in this case
too the name M had no referent. But now 1
think it might be better, sometimes or
always, to say that the name turns out to
be ambiguous in reference. That follows the
lead of Field (1973): and it is consistent
with, though not required by, the treatment
of Carnap (1963). Note that we face the
same choice with phrases like 'the moon of
Mars'; and in that case too I'd now lean
toward ambiguity of reference rather than
lack of it.
My recipe for analyses, like Rylean analy-
tic BEHAVIOURISM, posits analytic truths
that constrain the causal relations of
mental states to behaviour. (We have no
necessary connections between distinct
existences, of course; the necessity is
verbal. The state itself could have failed to
occupy its causal role, but would thereby
have failed to deserve its mental name.)
But the constraints are weak enough to be
credible. Because the state that typically
occupies a role need not occupy it invari-
ably, and also because a state may deserve
a name well enough in virtue of a role that
it occupies imperfectly, we are safe from
the behaviourist's bugbears. We have a
417
place for the resolute deceiver, disposed
come what may to behave as if his mental
states were other than they really are. We
have a place for the total and incurable
paralytic with a rich mental life and no
behavioural dispositions whatever. We even
have a place for a madman whose mental
states are causally related to behaviour and
stimuli and one another in a totally
haywire fashion (Lewis, 1980). And yet not
anything goes. At some point - and just
where that point comes is a matter of
semantic indecision - weird tales of mental
states that habitually offend against the
principles of folk psychology stop making
sense; because at some point the offending
states lose all claim to their folk-psycho-
logical names. To that extent, analytic
behaviourism was right. To quote my
closest ally in these matters, ' ... outward
physical behaviour and tendencies to
behave do in some way enter into our
ordinary concept of mind. Whatever theory
of mind is true, it has a debt to pay, and a
peace to be made, with behaviourism'
(Armstrong, 1968, p. 68).
When we describe mental state M as the
occupant of the M-role, that is what Smart
(1959) calls a topic-neutral description. It
says nothing about what sort of state it is
that occupies the role. It might be a non-
physical or a physical state, and if it is
physical it might be a state of neural activ-
ity in the brain, or a pattern of currents
and charges on a silicon chip, or the jang-
ling of an enormous assemblage of beer
cans. What state occupies the M-role and
thereby deserves the name M is an a pos-
teriori matter. But if materialist super-
venience is true, and every feature of the
world supervenes upon fundamental
physics, then the occupant of the role is
some physical state or other - because
there's nothing else for it to be. We know
enough to rule out the chip and the cans,
and to support the hypothesis that what
occupies the role is some pattern of neural
activity. When we know more, we shall
know what pattern of neural activity it is.
Then we shall have the premises of an
argument for psychophysical identification:
418
mental state M = the occupant of the
M-role (by analYSiS),
physical state P = the occupant of the
M-role (by science),
therefore M = P.
(See Lewis, 1966, 1972; and see Armstrong.
1968, for an independent and simultaneous
presentation of the same position, with a
much fuller discussion of what the definitive
causal roles might be.)
That's how conceptual analysis can
reveal the simple formula - or anyway, the
much less than infinitely complicated
formula - whereby, when we know
enough, we can pick out a mental feature of
the world from all the countless other fea-
tures of the world that likewise supervene
on fundamental physics.
The causal-role analyses would still hold
even if materialist supervenience failed.
They might even still yield psychophysical
identifications. Even if we lived in a spook-
infested world. it might be physical states
that occupied the causal roles (in us, if not
in the spooks) and thereby deserved the
folk-psychological names. Or it might be
non-physical states that occupied the roles.
Then, if we knew enough parapsychology,
we would have the premises of an argu-
ment for psycho-non-physical identification.
When our argument delivers an identifica-
tion M = P, the identity is contingent. How
so? - All identity is self-identity, and
nothing could possibly have failed to be self-
identical. But that is not required. It's con-
tingent, and it can only be known a poster-
iori, which physical (or other) states occupy
which causal roles. So if M means 'the
occupant of the M-role' it's contingent
which state is the referent of M: it's con-
tingent whether some one state is the
common referent of M and P: so it's con-
tingent whether M = P is true.
Kripke (1972) vigorously intuits that
some names for mental states, in particular
'pain', are rigid deSignators: that is, it's not
contingent what their referents are. I myself
intuit no such thing, so the non-rigidity
imputed by causal-role analyses troubles me
not at all.
Here is an argument that 'pain' is not a
rigid designator. Think of some occasion
when you were in severe pain, unmistak-
able and unignorable. All will agree, except
for some philosophers and faith healers,
that there is a state that actually occupies
the pain role (or near enough); that it is
called 'pain'; and that you were in it on that
occasion. For now, I assume nothing about
the nature of this state, or about how it
deserves its name. Now consider an unac-
tualized situation in which it is some differ-
ent state that occupies the pain role in place
of the actual occupant; and in which you
were in that different state; and which is
otherwise as much like the actual situation
as possible. Can you distinguish the actual
situation from this unactualized alternative?
I say not, or not without laborious investi-
gation. But if 'pain' is a rigid designator,
then the alternative situation is one in
which you were not in pain, so you could
distinguish the two very easily. So 'pain' is
not a rigid designator.
Philosophical arguments are never incon-
trovertible - well, hardly ever. Their
purpose is to help expound a position, not to
coerce agreement. In this case, the con-
troverter might say that if the actual occu-
pant of the pain role is not a physical state,
but rather is a special sort of non-physical
state, then indeed you can distinguish the
two situations. He might join me in saying
that this would not be so if the actual occu-
pant of the role were a physical state - else
neurophysiology would be easier than it is -
and take this together with intuitions of
rigidity to yield a reductio against materi-
alism. Myself, I don't see how the physical
or non-physical nature of the actual occu-
pant of the role has anything to do with
whether the two situations can be dis-
tinguished. Talk of 'phenomenal character'
and the like doesn't help. Either it is loaded
with question-begging philosophical doc-
trine, or else it just reiterates the undisputed
fact that pain is a kind of experience.
(The controverter just imagined would
agree with the discussion in Kripke, 1972,
pp. 344-42. But I don't mean to suggest
that Kripke would agree with him. At any
rate, the words I have put into his mouth
are not Kripke's.)
If there is variation across worlds with
respect to which states occupy the folk-
psychological roles and deserve the folk-
psychological names (and if this variation
doesn't always require differences in the
laws of nature, as presumably it doesn't)
then also there can be variations within a
single world. For possibility obeys a prin-
ciple of recombination: roughly, any possi-
ble kind of thing can coexist with any other
(Lewis, 1986a, pp. 86-92). For all we
know, there may be variation even within
this world. Maybe there are space aliens,
and maybe there will soon be artificial intel-
ligences, in whom the folk-psychological
roles are occupied (or near enough) by
states very different from any states of a
human nervous system. Presumably, at
least some folk-psychological roles are occu-
pied in at least some animals, and maybe
there is variation across species. There
might even be variation within humanity. It
depends on the extent to which we are
hard-wired, and on the extent of genetic
variation in our wiring.
We should beware, however, of finding
spurious variation by overlooking common
descriptions. Imagine two mechanical cal-
culators that are just .alike in design. When
they add columns of numbers, the amount
carried goes into a register, and the register
used for this purpose is selected by throwing
a switch. Don't say that the carry-seventeen
role is occupied in one machine by a state of
register A and in the other by a state of reg-
ister B. Say instead that in both machines
alike the role is occupied by a state of the
register selected by the switch. (Equival-
ently, by a state of a part of the calculator
large enough to include the switch and both
registers.) If there is a kind of thinking that
some of us do in the left side of the brain
and others do in the right side, that might
be a parallel case.
If M means 'the occupant of the M-role'
and there is variation in what occupies the
419
M-role, then our psychophysical identities
need to be restricted: not plain M = P, but
M-in-K = P, where K is a kind within which
P occupies the M-role. Human pain might
be one thing, Martian pain might be some-
thing else (Lewis, 1980). As with con-
tingency, which is variation across worlds,
so likewise with variation in a single world:
the variability in no way infects the identity
relation, but rather concerns the reference
of the mental name.
The threat of variation has led many to
retreat from 'type-type' to 'token-token'
identity. They will not say that M = p,
where M and P are names for a state that
can be common to different things at differ-
ent times - that is, for a property had by
things at times. But they will say that
m = p, where m and p are mental and
physical names for a particular, unrepea-
table event. Token-token identities are all
very well, in their derivative way, but the
flight from type-type identities was quite
unnecessary. For our restricted identities, of
the form M-in-K = p, are still type-type.
But don't we at least have a choice?
Couldn't our causal role analyses be recast
in terms of the causal roles of tokens, and if
they were, would they not then yield token-
token identities? After all, the only way for
a type to occupy a causal role is through
the causes and effects of its tokens. The
effects of pain are the effects of pain-events.
- I think, following Jackson, Pargetter, and
Prior (1982), that this recasting of the ana-
lyses would not be easy. There are more
causal relations than one. Besides causing,
there is preventing. It too may figure in
folk-psychological causal roles; for instance,
pain tends to prevent undivided attention to
anything else. Prevention cannot straight-
forwardly be treated as a causal relation of
tokens, because the prevented tokens do not
exist - not in this world, anyway. It is better
taken as a relation of types.
If a retreat had been needed, a better
retreat would have been to 'subtype-
subtype' identity. Let MK name the con-
junctive property of being in state M and
being of kind K; and likewise for PK. Do we
really want psychophysical identities of the
420
form MK = PK? - close, but I think not
quite right. For one thing, M-in-K is not the
same thing as MK. The former but not the
latter can occur also in something that isn't
of kind K. For another thing, it is P itself,
not PK. that occupies the M-role in things of
kind K.
Non-rigidity means that M is different
states in different possible cases; variation
would mean that M was different states in
different actual cases. But don't we think
that there is one property of being in the
state M - one property that is common to
all, actual or possible, of whatever kind,
who can truly be said to be in state M? -
There is. It is the property such that, for any
possible X, X has it just in case X is in the
state that occupies the M-role for X's kind at
X's world. (In Lewis, 1970, I called it the
'diagonalized sense' of M.) The gerund
'being in M' can be taken, at least on one
good disambiguation, as a rigid designator
of this property. However, this property is
not the occupant of the M-role. It cannot
occupy that or any other causal role
because it is excessively disjunctive, and
therefore no events are essentially havings
of it (Lewis, 1986c). To admit it as causally
efficacious would lead to absurd double-
counting of causes. It would be like saying
that the meat fried in Footscray cooked
because it had the property of being either
fried in Footscray or boiled in Bundoora -
only worse, because the disjunction would
be much longer and more miscellaneous.
Since the highly disjunctive property of
being in M does not occupy the M-role, I
say it cannot be the referent of M. Many
disagree. They would like it if M turned out
to be a rigid designator of a property
common to all who are in M. So the prop-
erty I call 'being in M', they call simply M;
and the property that I call M, the occupant
of the M-role, they call 'the realization of
M'. They have made the wrong choice,
since it is absurd to deny that M itself is
causally efficacious. Still, their mistake is
superficial. They have the right properties in
mind, even if they give them the wrong
names.
It is unfortunate that this superficial
question has sometimes been taken to mark
the boundary of 'functionalism'. Sometimes
so and sometimes not - and that's why I
have no idea whether I am a fuctionalist.
Those who take 'pain' to be a rigid desig-
nator of the highly disjunctive property will
need to controvert my argument that 'pain'
is not rigid, and they will not wish to claim
that one can distinguish situations in which
the pain-role is differently occupied. Instead,
they should controvert the first step, and
deny that the actual occupant of the pain-
role is called 'pain'. I call that denial a
reductio.
CONTENT
A mind is an organ of REPRESENTATION.
Many things are true according to it; that is,
they are believed. Or better, they are more
or less probable according to it; that is, they
are believed or disbelieved to varying
degrees. Likewise, many things are desired
to varying positive or negative degrees.
What is believed, or what is desired, we call
the CONTENT of BELIEF or DESIRE.
(I think it an open question to what extent
other states with content - doubting, won-
dering, fearing, pretending, ... - require
separate treatment, and to what extent they
can be reduced to patterns in belief and desire
and contentless feeling. Be that as it may, I
shall ignore them here (see EMOTION ).)
What determines the content of belief and
desire? - The occupation of folk-psycho-
logical roles by physical states, presumably
neural states; and ultimately the pattern of
coinstantiation of fundamental physical
properties and relations. But to say just that
is to say not much. Those who agree with it
can, and do, approach the problem of
content in very different ways.
I can best present the approach I favour
by opposing it to an alternative. A crude
sketch will suffice, so in fairness I name my
opponent Strawman. I doubt there is anyone
real who takes exactly the position that
Strawman does - but very many are to be
found in his near vicinity.
Strawman says that folk psychology says
- and truly - that there is a LANGUAGE OF
THOUGHT. It has words, and it has syntactic
constructions whereby those words can be
combined into sentences. Some of these sen-
tences have a special status. Strawman says
they are 'written in the belief box' or 'in the
desire box', but even Strawman doesn't take
that altogether literally. There are folk-psy-
chological causal roles for the words, for the
syntactic constructions, and for the belief
and desire boxes. It is by occupying these
roles that the occupants deserve their folk-
psychological names.
The question what determines content
then becomes the question: what deter-
mines the semantics of the language of
thought? Strawman says that folk psychol-
ogy specifies the semantic operations that
correspond to syntactic constructions such
as predication. As for the words, Strawman
says that folk psychology includes, in its
usual tacit and unsystematic way, a causal
theory of reference (more or less as in
Kripke, 1972). There are many relations of
acquaintance that connect the mind to
things, including properties and relations, in
the external world. Some are relations of
perceptual acquaintance. Others are less
direct: you are acquainted with the thing by
being acquainted with its traces. Often, you
are acquainted with the thing by way of its
linguistic traces that is, you have heard of it
by name. Somehow, in virtue of the differ-
ent causal roles of different words of the
language of thought, different words are
associated with different relations of
acquaintance, which connect them to differ-
ent external things. Whatever a word is
thus connected to is the referent of that
word.
Once the words of the language of
thought have their referents, the sentences
have their meanings. These are structures
built up from the referents of the words in a
way that mirrors the syntactic construction
of the sentences from the words. Take pre-
dication - Strawman's favourite example. A
word F of your language of thought is con-
nected by one relation of acquaintance to
the property of being French. Another word
A is connected by another relation of
acquaintance to the man Andre. (The first
421
relation might be linguistic, for instance,
and the second perceptual.) The syntactic
construction of predication builds a sen-
tence F(A). Its meaning is the ordered pair
of the property of being French and Andre.
Such a pair is a 'singular proposition', true
just in case its second element instantiates
its first. (Other singular propositions are
triples, quadruples, ... , with relations in
the first place.) If you have F(A) written in
your belief box, you thereby believe that
Andre is French.
Strawman's account of content is
sketchy, as I said it would be. Even with
help from all his allies, I doubt he will find it
easy to fill the gaps. I especially wonder
what he can say about how the words get
hooked up to the right relations of
acquaintance. A causal theory of reference
for public language might usefully mention
mutual expectations among language-users,
intentions to instill beliefs, semantic inten-
tions, or other such instances of mental
content. But even if we had corresponding
expectations or intentions about our own
language of thought, Strawman could not
without circularity use them in a general
account of mental content.
Suppose, all the same, that Strawman's
account could be completed successfully by
its own lights. I would still have four objec-
tions.
First, I don't believe that folk psychology
says there is a language of thought. Rather,
I think it is agnostic about how MENTAL
REPRESENT A TION works - and wisely so.
What is the issue? Of course everybody
should agree that the medium of mental
representation is somehow analogous to
language. A raven is like a writing-desk.
Anything can be analogized to anything.
And of course nobody thinks the head is full
of tiny writing.
A serious issue, and one on which I take
folk psychology to be agnostic, concerns the
relation between the whole and the parts of
a representation. Suppose I have a piece of
paper according to which, inter alia, Colling-
wood is east of Fitzroy. Can I tear the paper
up so that I get one snippet that has exactly
422
the content that Collingwood is east of
Fitzroy, nothing more and nothing less? If
the paper is covered with writing, maybe I
can; for maybe 'Collingwood is east of
Fitzroy' is one of the sentences written
there. But if the paper is a map, any snippet
according to which Collingwood is east of
Fitzroy will be a snippet according to which
more is true besides. For instance, I see no
way to lose the information that they are
adjacent, and that a street runs along the
border. And I see no way to lose all infor-
mation about their size and shape.
(A hologram, or famously a connectionist
network (see CONNECTIONISM), differs even
more from a paper covered with writing. If
we make a hologram of the map and break
it into snippets, detail will be lost in blur.
But the arrangement of all the suburbs, pro-
vided it was shown with sufficient promi-
nence on the original map, will remain to
the last.)
Mental representation is language-like to
the extent that parts of the content are the
content of parts of the representation. If our
beliefs are 'a map ... by which we steer', as
Ramsey said (1931b, p. 238), then they are
to that extent not language-like. And to
that extent, also, it is misleading to speak in
the plural of beliefs. What is one belief? No
snippet of a map is big enough that, deter-
minately, something is true according to it,
and also small enough that, determinately,
nothing is true according to any smaller
part of it. If mental representation is map-
like (let alone if it is hologram-like) then
'beliefs' is a bogus plural. You have beliefs
the way you have the blues, or the mumps,
or the shivers.
But if mental representation is language-
like, one belief is one sentence written in the
belief box, so 'beliefs' is a genuine plural.
Whether the plural is bogus or genuine is
not settled by rules of grammar. Rather, it is
an empirical question, and a question that
folk psychology leaves open. 'The shivers'
might be a parallel case. Is there such a
thing as one shiver? - Maybe and maybe
not. I don't think one cycle of vibration
should be called 'one shiver', but there
might be a better candidate. What if one
firing of a control neuron would set you
shivering for four seconds, and prolonged
shivering is caused by this neuron firing
every two seconds? If so, I think the shiver-
ing set off by one firing could well be called
'one shiver', and then it is right to say that
shivering consists of a sequence of over-
lapping shivers. Under this hypothesis, the
plural is genuine. Under other hypotheses,
the plural is bogus.
Of course you might say, under the
hypothesis that mental representation is
map-like, that any proposition true accord-
ing to the mental map is one belief. Or you
might say that the one belief that Colling-
wood is east of Fitzroy is the highly dis-
junctive state of having some mental map
or other according to which that inter alia is
true. Say so if you like. But I only insist that
if you say either thing, then you may not
also assume that 'one belief' is the sort of
thing that can occupy a causal role. You
may still say '. . . because he believes that
Collingwood is east of Fitzroy', but only if
you mean by it ' ... because he has beliefs'
- bogus plural! - 'according to which inter
alia Collingwood is east of Fitzroy'.
If Strawman heeds the advice of some of
his allies, he will respond by changing his
position. He will give away conceptual ana-
lysis and folk psychology, and market his
wares as 'cognitive science' (see COGNITIVE
PSYCHOLOGY). No problem, then, if the folk
are agnostic about the language of thought.
Let it be a new hypothesis, advanced
because it best explains ... What? Well-
known facts about belief? - But 'belief' is a
folk-psychological name for a kind of state
posited by folk psychology. If Strawman
leaves all that behind him, where shall he
find his evidence? He can never again set up
thought experiments and ask us what we
want to say about them. That would only
elicit our folk-psychological preconceptions.
He can make a fresh start if he really wants
to - I assume he will not want to - but he
cannot have his cake and eat it too. (See
Jackson, 1992.)
If Strawman stands his ground, on the
other hand, he will insist that folk psychol-
ogy is far from agnostic about the language
of thought. It has plenty to say, after all,
about our 'concepts' (or 'ideas') of things.
Our concept of a concept, says Strawman, is
just our concept of a word of the language
of thought. - I doubt it. I haven't much of
any concept of Elsternwick. I have little idea
what the place looks like, what sort of
people live there, . . . All I know is that
there is a place of that name, and roughly
where it is. But I do have the word. (At
least. I have the word 'Elsternwick' of our
public language. If I have a language of
thought, presumably this word has been
borrowed into it.) My lack of a concept isn't
lack of a word; rather, I lack any very rich
cluster of associated descriptions.
Strawman can reply that even if I haven't
much concept of Elsternwick, still I have
enough of one that I can think about
Elsternwick (for instance, when I think how
little I know about it). It is this minimal
concept of a concept, he says, that is our
folk-psychological concept of a word of the
language of thought. - Yes, I have a
concept of Elsternwick in the minimal sense
that I have whatever it takes to be able to
think about it. But must the basis of such
an ability, in general or even in this case, be
the possession of a word? On that question,
the folk and I remain agnostic.
My second objection to Strawman's account
is that it delivers only wide content. Which
singular propositions you believe depends
upon which external things are suitably
connected by relations of acquaintance to
the words of your language of thought.
Strawman holds that all content is wide
because he has learned the lesson of TWIN
EARTH. Recall the example (Putnam, 1975,
pp. 139-42). Oscar the Earthling believes
that water often falls from clouds. Twoscar
on Twin Earth is in no way acquainted with
water, that is, with H
2
0. Rather, Twoscar is
acquainted with XYZ, a superficially similar
liquid that is abundant on Twin Earth, in
exactly the way that Oscar is acquainted
with H
2
0. There is no other relevant differ-
ence between Twoscar and Oscar. We are
invited to agree that Twoscar does not
believe that water falls from clouds, and
423
believes instead that XYZ falls from clouds.
Strawman does agree.
And so do r. but with many reservations.
For one thing, I think agreement is not com-
pulsory. Like any up-to-date philosopher of
1955, I think that 'water' is a cluster
concept. Among the conditions in the
cluster are: it is liquid, it is colourless, it is
odourless, it supports life. But, pace the phi-
losopher of 1955, there is more to the
cluster than that. Another condition in the
cluster is: it is a natural kind. Another con-
dition is indexical: it is abundant here-
abouts. Another is metalinguistic: many call
it 'water'. Another is both metalinguistic
and indexical: I have heard of it under the
name 'water'. When we hear that XYZ off
on Twin Earth fits many of the conditions in
the cluster but not all, we are in a state of
semantic indecision about whether it
deserves the name 'water'. (See Unger,
1984, pp. 79-104. But while I agree with
Unger about what happens in various cases,
I don't endorse all the morals he draws.)
When in a state of semantic indecision, we
are often glad to go either way, and accom-
modate our own usage temporarily to the
whims of our conversational partners
(Lewis, 19 79b). So if some philosopher, call
him Schmutnam, invites us to join him in
saying that the water on Twin Earth differs
in chemical composition from the water
here, we will happily follow his lead. And if
another philosopher, Putnam (1975),
invites us to say that the stuff on Twin Earth
is not water - and hence that Twoscar does
not believe that water falls from clouds - we
will just as happily follow his lead. We
should have followed Putnam's lead only for
the duration of that conversation, then
lapsed back into our accommodating state of
indecision. But, sad to say, we thought that
instead of playing along with a whim, we
were settling a question once and for all.
And so we came away lastingly misled.
The example half suceeds. It is not com-
pulsory, but certainly it is permissible, to
say that Oscar does believe that water falls
from clouds and differently acquainted
Twoscar does not. Therefore wide content
does serve a purpose. It enters into the ana-
424
lysis of some sentences that are about belief,
or at least partly about belief; or at least it
does so under some permissible disambigua-
tions of these sentences.
Other examples are similar. Twoscar is
acquainted with molybdenum as Oscar is
with aluminium; with a disease of bone as
Oscar is with a disease of joints; with spy
robots as Oscar is with cats; and so on. It
seems to matter little whether Twoscar is
our neighbour, or whether he lives on a
remote planet, or whether he lives in a dif-
ferent possible world. In each case we find
that the difference in what Twoscar and
Oscar are acquainted with makes a differ-
ence to the truth value, under some dis-
ambiguation, of some sentences that are at
least partly about belief. But that is all we
find. There is nothing here to support
Strawman's thesis that wide content is the
only kind of content; or that it is in any
way pre-eminent or basic.
We should not jump to the conclusion
that just any belief sentence is susceptible to
Twin Earth examples. Oscar thinks that
square pegs don't fit round holes; I don't
think you can tell an even halfway convin-
cing story of how Twoscar, just by being
differently acquainted, fails to think so too.
Oscar believes there's a famous seaside place
called 'Blackpool'; so does differently
acquainted Twoscar, though of course it
may not be Blackpool - not our Blackpool -
that he has in mind. Oscar believes that the
stuff he has heard of under the name
'water' falls from clouds. So does Twoscar -
and so does Twoscar even if you alter not
only his acquaintance with water but his
relations of acquaintance to other things as
well. You know the recipe for Twin Earth
examples. You can follow it in these cases
too. But what you get falls flat even as an
example of how content is sometimes wide,
let alone as evidence that content is always
wide.
The famous brain in a bottle is your exact
duplicate with respect to brain states and
their typical causal roles; but is acquainted
only with aspects of the computer that
fabricates its virtual reality. You and the
brain share no objects of acquaintance. So,
according to Strawman, you and the brai:n
share no common beliefs whatever.
Newborn Swampman, just this moment
formed by an unlikely chance assembly of
atoms, also is your exact duplicate with
respect inter alia to brain states and their
typical causal roles (Davidson, 1987). But
so far, he hasn't had time to become
acquainted with much of anything. There-
fore, according to Strawman, he believes
not much of anything.
Strawman and his allies may think that
we have here two remarkable philosophical
discoveries. I think, rather, that Strawman's
thesis that all content is wide has here met
with a twofold reductio ad absurdum. Granted,
the brain in a bottle shares no wide content
of belief with you. Granted, Swampman has
no wide content of belief at all. Yet there
must be some good sense in which both the
brain and Swampman are your mental
twins; some good sense in which they
believe just what you do. (And in our less
extreme cases, there must be some good
sense in which Twoscar believes just what
Oscar does.) Strawman's position is unac-
ceptable. Not because it posits wide content;
but because it omits narrow content,
content independent of what one is acquain-
ted with. It omits the sort of content that you
and the brain and Swampman, and likewise
Oscar and Twoscar, have in common.
(Narrow content is independent of what
you are acquainted with, but that does not
mean that it is altogether intrinsic to you.
For it still depends on the causal roles of
your brain states; causation depends on the
laws of nature; and if some sort of regu-
larity theory of lawhood is true, living
under such-and-such laws is not intrinsic to
you. Further, it is the typical causal roles of
your brain states that matter. But you may
be an atypical member of your kind; hence
what is typical of your kind is not intrinsic
to you. So I can say only this: if X and Yare
intrinsic duplicates, and if they live under
the same laws of nature, and if they are the
same in kind, then they must be exactly
alike in narrow content.)
In insisting on the existence of narrow
content, I am not guided by any preconcep-
tion about what sort of properties may
figure in causal explanation, or in truly sci-
entific explanation. I dare say the funda-
mental laws of physics must concern
perfectly natural. intrinsic properties. But
that's irrelevant, since causal and scientific
explanation seldom consists in subsumption
under these fundamental laws. Rather, it is
a matter of giving information about how
things are caused (Lewis, 1986b). Such
information can come in many forms, both
within science and without. and there is no
reason to proscribe extrinsic classifications.
(Lynne Baker told me a nice example: the
science of economics is all about extrinsic
properties like poverty and debt. Yet there is
nothing wrong, and nothing unscientific, in
saying that Fred stays poor because of his
burden of debt.)
I am guided, rather, by my tacit mastery
of the principles of folk psychology. I said:
Oscar believes that the stuff he has heard of
under the name 'water' falls from clouds;
and so does Twoscar. (And so do you, and
so does the brain in a bottle, and so does
Swampman.) These are ordinary folk-
psychological belief sentences; but narrow
ones, as witness the fact that they are not
susceptible to Twin Earth examples.
This narrow content is content, rightly
so-called: something is true according to the
belief-system in question. The content is
true on condition that the stuff the believer
has heard of under the name 'water' does
indeed fall from clouds; otherwise false. It is
not 'purely syntactic content' - something I
take to be a contradiction in terms. Nor is it
a mere function that delivers genuine
content as output when given circum-
stances of acquaintance as input. Nor is it
merely phenomenalistic content, restricted
in subject matter to the believer's experi-
ence.
However, it is not content that can be given
by a singular proposition, and that leads to
my third objection against Strawman's
account.
Strawman's singular propositions suffice
to specify which things have which proper-
ties. If all else supervenes upon the pattern
425
of coinstantiation of fundamental properties.
that in turn will suffice to specify the way
the world is. But much of the content of our
knowledge and belief is de se: it concerns
not the world but oneself. (See Perry, 1977;
Lewis, 1979a; Chisholm, 1979.) However
much I may know about the things that
make up the world, their properties and
their arrangement, it is something extra to
know which one of all these things is me.
This is de se knowledge, whereby I locate
myself in the world and self-ascribe the
properties I think myself to possess, but is
not knowledge of how the world is. Its
content cannot be captured by singular pro-
positions. What singular proposition is
expressed when I say, or I think, 'I am DL'?
- Just the proposition that DL = DL. And
when I self-ascribe the property F? - Just the
proposition that DL is F. But I can know
these propositions without knowing who I
am, or whether I am F. (And you can know
them too.) Strawman's only recourse is to
say that de se knowledge is characterized
not by its de se content but some other way
- and if he says that, he confesses that his
account of content is inadequate. Belief that
falls short of knowledge can likewise have
de se content. If you take yourself to be DL,
your false belief and my true belief have
their de se content in common. Desire also
has de se content. If you desire to be F and I
believe myself to be F, again the two atti-
tudes have their de se content in common.
There is also tensed content. The world is
spread out over many times; but we can
have knowledge, or belief or desire, about
which of these times is now. Again, this is
not knowledge of how the whole spread-out
world is. It is something extra. Some would
speak of content de se et nunc, but I would
subsume de nunc under de se. For I think we
persist through time by consisting of many
time-slices, or momentary selves; and in the
last analysis, it is these momentary selves
that do our thinking. So when I think 'It's
now time for lunch', that's one of my
momentary selves self-ascribing de se the
property of being located at lunchtime.
The 'propositions', if we may call them
that, which make up de se content are true
426
or false not absolutely, as singular proposi-
tions are, but relative to a subject. (Or to a
subject at a time, if you don't believe in
momentary selves.) The content of my
knowledge de se that I am DL is something
that is true for me but not for you. Its
linguistic expression requires a first-person
pronoun, or some equivalent device. We
could call it an 'egocentric proposition' (or
'egocentric and tensed'). Or we can simply
identify it with the property that I self-
ascribe: the property of being DL. Likewise
the de se content of my belief that I have F is
just the property of F itself; the de se content
of my belief that it's lunchtime is the prop-
erty (possessed not by the whole of me but
by some of my momentary selves) of being
located at lunchtime; and so on. A de se self-
ascription of a property is true just on con-
dition that the self-ascriber possesses the
self-ascribed property.
(May I say, then, that de se belief has
'truth conditions'? Not if Strawman has his
way. He goes in for terminological piracy.
He transforms one term after another into a
mere synonym for 'singular proposition'. He
has taken 'object of thought'. He has taken
'content'. He has taken 'proposition'. He is
well on the way to taking 'truth condition'.
When he has taken all the terms for his
own, dissident thoughts will be unsayable.)
Since Strawman has no place for de se
content, it makes sense that he overlooks
narrow content as well. For narrow content
is very often de se. To revisit our previous
example: Oscar self-ascribes having heard,
under the name of 'water', of a liquid that
falls from clouds. He also self-ascribes the
property of being at a place (and time) in
the vicinity of which the most abundant
liquid is one that falls from clouds. Differ-
ently acquainted Twoscar self-ascribes these
same two properties, and in this way Oscar
and Twoscar share the same de se narrow
content of belief.
On my own view, it is just such de se
narrow content that underlies wide content.
The semantics of the alleged language of
thought needn't enter into it. To the extent
that language enters my story at all, it is
not by way of the language of thought, but
rather by way of thought about language -
about the ordinary public language,
whereby, for instance, Oscar heard of some-
thing under the name 'water'.
Here is one recipe (Lewis, 1979a, pp. 538
-43): if R is a relation of acquaintance, and
subject S self-ascribes being R-acquainted
uniquely with something that has property
F (the narrow part), and if Sis R-acquainted
uniquely with A (the wide part), S thereby
widely believes the singular proposition that
A has F. There are variants on the recipe.
Our example of French Andre was a case in
which property F as well as individual A
enters indirectly as an object of acquain-
tance; we must of course let in cases where
the property F gives way to a relation with
two or more relata; maybe sometimes we
should drop the qualification 'uniquely'; and
maybe sometimes the relation R is not, or
not entirely, a matter of acquaintance. But
in every case, wide belief in a singular pro-
position derives from narrow de se self-
ascription plus facts about what the subject
is related to.
Often we know a lot about which sin-
gular propositions someone believes in this
wide and derivative way; but we know less
about how - in virtue of just which self-
ascriptions and relations of acquaintance -
he believes those singular propositions. So
it's no surprise to find that our ordinary-
language belief sentences often seem to be
ascriptions of wide content. Often; but not
always. In these last few paragraphs I've
been talking about de se narrow content,
and I've been talking about it in plain
English. (Such bits of jargon as I used were
first explained in plain English.)
There are still other dimensions to the
semantic complexity and the multifarious
ambiguity of ordinary-language belief sen-
tences. Think of the belief sentences that
show up as test cases in articles advocating
one semantic analysis or another. I always
want to say: 'in a sense that's true, in a
sense false'. One complication is that we get
direct-quotational effects even in what is
ostensibly indirect quotation (see Rieber,
1992). An example: Fred knows perfectly
well that the house he lives in is made of
wood, but Fred also thinks that 'abode' is
the English word for a house made of mud-
brick. 'Fred believes that he has an abode -
yes or no?' In at least some contexts (this
isn't one of them) I'd be prepared to insist
on 'no'. Wouldn't you? Moral: if you hope
to understand the folk psychology of belief
by studying the linguistic phenomenology of
ordinary belief sentences, you're in for big
trouble.
I've said that narrow content is very often
de se, but by resorting to a cheap trick I can
change 'often' to 'always'. Take an apparent
exception: the narrow belief that square
pegs won't fit in round holes. Take this to be
the de se self-ascription of the property of
inhabiting a world wherein square pegs
won't fit in round holes. A peculiar prop-
erty, since either all the inhabitants of the
world share it or else none do; and, like
many other self-ascribed properties, very far
from fundamental: but in a broad enough
sense of the word, a property all the same.
Likewise you can self-ascribe the property of
inhabiting a world where there's a famous
seaside place called 'Blackpool'. And so on,
until all narrow content has been included
as de se. Hoky, but maybe worthwhile for
the sake of uniform treatment.
My final objection is that Strawman ignores
large parts of the folk psychology of belief
and desire: the parts that characterize
aspects of our RATIONALITY. Folk psychol-
ogy says that a system of beliefs and desires
tends to cause behaviour that serves the
subject's desires according to his beliefs.
Folk psychology says that beliefs change
constantly under the impact of perceptual
evidence: we keep picking up new beliefs,
mostly true, about our perceptual surround-
ings; whereupon our other beliefs (and our
instrumental desires) change to cohere with
these new beliefs. Folk psychology sets pre-
sumptive limits to what basic desires we can
have or lack: de gustibus non disputandum,
but still a bedrock craving for a saucer of
mud would be unintelligible (Anscombe,
1958, pp. 69-71). Likewise it sets limits to
our sense of plausibility: which hypotheses
we find credible prior to evidence, hence
427
which hypotheses are easily confirmed
when their predictions come true. And it
sets presumptive limits on what our con-
tents of belief and desire can be. Self-
ascribed properties may be 'far from funda-
mental', I said - but not too far. Especially
gruesome gerrymanders are prima facie
ineligible to be contents of belief and desire.
(See Lewis, 1983a, pp. 370-7: Lewis,
1986a, pp. 38-9 and 105-8.) In short, folk
psychology says that we make sense. It
credits us with a modicum of rationality in
our acting, believing, and desiring.
(Beware. 'Rationality' is an elastic word,
and here I've stretched it to cover a lot. If
you'd rather use it more narrowly - just for
the serving of desires according to beliefs,
say - no harm done. So long as you don't
just ignore the several other departments of
rationality that I listed, it doesn't matter
what you call them.)
If mental states are to be analysed as
occupants of folk-psychological roles, and if
the folk psychology of belief and desire has a
lot to say about rationality, and if what it
says is framed in terms of content, then it
seems that constraints of rationality are
constitutive of content. Yet Strawman's
account of content makes no place for con-
stitutive rationality. Why not?
Perhaps Strawman thought, wisely, that
it would be better to say too little than too
much. It wouldn't do to conclude that, as a
matter of analytic necessity, anyone who
can be said to have beliefs and desires at all
must be an ideally rational homo economicus!
Our rationality is very imperfect, Strawman
knows it, and he knows that the folk know
it too. Of course we overlook options and
hypotheses, we practice inference to the
third-best explanation, we engage in double
think, and so on, and on, and on.
But there is no cause for alarm. Folk psy-
chology can be taken as a theory of imper-
fect, near-enough rationality, yet such
rationality as it does affirm can still be con-
stitutive, And even if folk psychology did set
too high a standard - even if, to take the
worst case, it were a theory of ideal ration-
ality - still an imperfect but near-enough
occupant of a folk-psychological role could
428
thereby be an imperfect but near-enough
deserver of a folk-psychological name.
Remember also that the typical occupant of
a role needn't occupy it in every case. In
short, constitutive rationality leaves plenty
of room for human folly.
(I think that systematic theories of ideal
rationality - decision theory, for instance,
and the theory of learning from experience
by conditionalizing a subjective probability
distribution - are severely idealized versions
of parts of folk psychology. They are
founded upon our tacit knowledge of folk
psychology, elicited in the guise of 'intu-
ition'. But folk psychology also supplies the
grains of salt to be applied to these idealiza-
tions. Sometimes it supplies complementary
pairs of opposite idealizations: a quantitative
theory of subjective probabilities and util-
ities precise to however many decimal
places, and alongside it a non-quantitative
theory of beliefs and desires that don't admit
of degree at all.)
Constitutive rationality is part of the
legacy of behaviourism, and that is a second
reason why Strawman mistrusts it. A beha-
viourist analysis might say, roughly, that a
subject's beliefs and desires are those beliefs
and desires, attribution of which would best
make sense of how the subject is disposed to
behave, and of how his changing beha-
vioural dispositions depend on the changing
perceptible features of his surroundings, But
Strawman is a robust realist about beliefs
and desires. He takes them to be genuine
inner states, and causes of behaviour. He
won't like an analysis that dispenses with
efficacious inner states in favour of mere
patterns of dispositions. Still less would he
like it if the behaviourist went on to say that
attributions of belief and desire governed by
constitutive rationality were instrumentally
useful. or warranted by rules of assertability,
but not straightforwardly true.
I applaud these misgivings. I too am a
robust realist about beliefs and desires.
(About whole systems of beliefs and desires,
anyway, though maybe not about all the
little snippets - the sentences written in the
belief and desire boxes - of which these
systems mayor may not be composed.) But
I say the proper remedy is not to shun con-
stitutive rationality, but to apply it differ-
ently. The behaviourist applies it directly to
the subject: I say we should apply it to the
subject's inner state. The behaviourist says
that the subject has that system of beliefs
and desires that best makes sense of how
the subject is disposed to behave. Whereas
I'd say that the inner state is that system of
beliefs and desires that best makes sense of
the behaviour which that state is apt for
causing in subjects. Thus I'd use con-
stitutive rationality not to dispense with
causally efficacious inner states, but rather
to define their content.
A third reason why Strawman shuns
constitutive rationality is that sometimes it
needs to be applied not to the singular pro-
positions that are the wide content of belief
and desire, but instead to the underlying de
se narrow content. The furniture of the
Lebenswelt which presents us with our prob-
lems of decision and learning consists, in
the first instance, of objects given qua
objects of acquaintance, and individuated
by acquaintance. (See Hintikka, 1972:
Lewis, 1983b.) That is a matter of narrow
content. If you are lucky, and you're never
wrong or uncertain about whether you're
really R-acquainted with something, and
you're never wrong or uncertain about
whether the thing you're Rracquainted
with is or isn't the same as the thing you're
Rracquainted with, then we can talk about
your beliefs and desires entirely in terms of
wide content. We can safely let things sim-
pliciter stand in for things-qua-objects-of-
acquaintance. But if you're not so lucky,
that won't work. Take unlucky Pierre
(Kripke, 1979). He self-ascribes being Rr
acquainted with a pretty city and being Rr
acquainted with an ugly city. But in fact he
is R1-acquainted and Rracquainted with
the same city, London. Thereby he believes
both that London is pretty and that
London is ugly. (Kripke derives this conclu-
sion from certain premises, but I find the
conclusion at least as obvious as the pre-
mises.) I take this to be a conflict in wide
content: Pierre widely believes two singular
propositions that predicate conflicting prop-
erties of the same thing. Folk psychology
says that by careful attention we can detect
and eliminate conflicts in our beliefs -
especially if we're good at logic, as Pierre is.
But plainly that was never meant to apply
to Pierre's conflict of Singular propositions.
Mere thought can't save him. What he
needs is the information de se that he is Rr
acquainted and Rracquainted with the
very same thing. (See Lewis, 1981.)
And suppose Pierre believes that by
boarding the bus before him, he can be
taken to London for a week of sight-seeing.
Would boarding that bus serve his desires
according to his beliefs? It helps not at all to
know that he widely believes both that
London is pretty and that London is ugly.
What does help is the information already
given about his narrow self-ascriptions, plus
one further thing: he also self-ascribes
having a bus before him that would take
him to the place he is Rracquainted with.
(See Lewis, 1986a, p. 58.)
(On constitutive rationality, see Stalnaker,
1984, pp. 1-42; Lewis, 1974, 1986a, pp.
27-40. But see Lewis, 1974, with caution:
it began as a conversation with Donald
Davidson, and I went rather too far in
granting undisputed common ground. (1) I
gave an important place to the subject
Karl's beliefs as expressed in Karl's own lan-
guage: that certainly suggests language-of-
thoughtism, though I hope I committed
myself to nothing more than the safe thesis
that Karl's medium of mental representation
is somehow analogous to language. (2) I was
too individualistic: I ignored the possibility
that deviant Karl might believe something
in virtue of the causal role of his inner state
not in Karl himself but in others who are
more typical members of Karl's kind. (3) I
had not yet come to appreciate the role of de
se content. Also see Lewis, 1986a, pp. 27-
40, with caution: besides endorsing con-
stitutive rationality, I also stated it within a
controversial framework of realism about
unactualized possibilia. I still think that's a
good way to state it: but I never said it was
the only way. Constitutive rationality and
realism about possibilia needn't be a package
deal!)
429
This completes my list of objections
against Strawman's program for explaining
content. Doubtless you can think of ever so
many ways of amending Strawman's theses
to get around my objections. Some lists of
amendments would take us to the positions
really held by real people. Of course I can't
show that no version of Strawman-
amended can work. But for myself, I pin my
hopes on a more radical reversal of Straw-
man's position.
With Strawman for a foil, my own
approach can be summed up quickly. The
contentful unit is the entire system of beliefs
and desires. (Maybe it divides up into con-
tentful snippets, maybe not.) That system is
an inner state that typically causes beha-
viour, and changes under the impact of per-
ception (and also spontaneously). Its
content is defined, insofar as it is defined at
all, by constitutive rationality on the basis
of its typical causal role. This content is in
the first instance narrow and de se (or de se
et nunc if you'd rather steer clear of momen-
tary selves). Wide content is derivative, a
product of narrow content and relationships
of acquaintance with external things.
See also DAVIDSON; DENNETT; FODOR; PHI-
LOSOPHY AND PSYCHOLOGY; PROPOSI-
TIONAL ATTITUDES; REASONS AND CAUSES;
THOUGHTS; THOUGHT AND LANGUAGE.
BIBLIOGRAPHY
Blackwell.
Mind. London: Routledge & Kegan Paul.
Carnap, R. 1963. Replies and expositions. In
The Philosophy of Rudolf Carnap, ed. P. A.
Schilpp. Cambridge University Press, 859-
1Ol3.
Chisholm, R.M. 1979. The indirect reflexive.
In Intention and Intentionality: Essays in
Honour of G. E. M. Anscombe, ed. C.
Diamond and J. Teichman. Brighton:
Harvester.
Davidson, D. 1987. Knowing one's own mind.
Proceedings and Address of the American
Philosophical Association, 60, 441-58.
430
Davies, M.K., and Humberstone, 1.1. 1980.
Two notions of necessity. Philosophical
Studies, 38, 1-30.
Feinberg, G. 1966. Physics and the Thales
problem. Journal of Philosophy, 66, 5-l3.
Field, H. 1973. Theory change and the inde-
terminacy of reference. Journal of Philosophy,
70,462-481.
Hintikka, J. 1972. Knowledge by acquaintance
- individuation by acquaintance. In Ber-
trand Russell: A Collection of Critical Essays,
ed. D. Pears. Garden City, N.J.: Doubleday.
Jackson, F.C. 1982. Epiphenomenal qualia.
Philosophical Quarterly, 32, 127-36.
Jackson, F .C. 1992. Armchair metaphysics.
Presented at a conference on the place of
philosophy in the study of the mind, Uni-
versity of New South Wales.
Jackson, F.C., Pargetter, R., and Prior, E.W.
1982. Functionalism and type-type identity
theories. Philosophical Studies, 42, 209-25.
Kripke, S. 1972. Naming and necessity. In
Semantics of Natural Language, ed. D. David-
son and G. Harman. Dordrecht: Reidel.
Kripke, S., 1979. A puzzle about belief. In
Meaning and Use, ed. Avishai Margalit.
Lewis, D. 1966. An argument for the identity
theory. Journal of Philosophy, 63, 17-25.
Reprinted with additions in Lewis, 1983c.
-- 1970. How to define theoretical terms.
Journal of Philosophy, 67,427-46. Reprinted
in Lewis, 1983c.
-- 1972. Psychophysical and theoretical
identifications. Australasian Journal of Philo-
sophy, 50, 249-58.
-- 1974. Radical interpretation. Synthese,
23, 331-44. Reprinted with postscripts in
Lewis, 1983c.
-- 1979a. Attitudes de dicto and de se. Philo-
sophical Review, 88, 5l3-43. Reprinted with
postscripts in Lewis, 1983c.
-- 1979b. Scorekeeping in a language
game. Journal of Philosophical Logic, 8, 339-
59. Reprinted in Lewis, 1983c.
-- 1980. Mad pain and Martian pain. In
Readings in Philosophy of Psychology. Vol. 1,
ed. N. Block. Cambridge, MA: Harvard Uni-
versity Press. Reprinted with postscript in
Lewis, 1983c.
-- 1981. What puzzling Pierre does not
believe. Australasian Journal of Philosophy,
59,283-89.
-- 1983a. New work for a theory of uni-
versals. Australasian Journal of Philosophy,
61, 343-77.
-- 1983b. Individuation by acquaintance
and by stipulation. Philosophical Review, 92,
3-12.
-- 1983c. Philosophical Papers, Vol. 1.
-- 1986a. On the Plurality of Worlds. Oxford:
Basil Blackwell.
-- 1986b. Causal explanation. In Lewis,
Philosophical Papers, Vol. 2. Oxford Uni-
versity Press.
-- 1986c. Events. In Lewis, Philosophical
Papers, Vol. 2. Oxford University Press.
-- 1990. What experience teaches. In Mind
and Cognition: A Reader, ed. W. G. Lycan.
-- 1992. Critical notice of D. M. Armstrong.
A Combinatorial Theory of Possibility. Aus-
tralasian Journal of Philosophy. 70. 211-24.
Nemirow. L. 1990. Physicalism and the cogni-
tive role of acquaintance. In Mind and Cogni-
tion: A Reader. ed. W.G. Lycan. Oxford: Basil
Blackwell.
Perry. J. 1977. Frege on demonstratives. Phi-
losophical Review. 86. 474-97.
Putnam. H. 1975. The meaning of 'meaning'.
In Language, Mind and Knowledge, ed. K.
Gunderson. Minneapolis, MN: University of
Minnesota Press.
Ramsey, F.P. 1931a. Theories. In Ramsey. The
Foundations of Mathematics. London: Rout-
ledge & Kegan Paul.
Ramsey, F.P. 1931b. General propositions and
causality. In Ramsey, The Foundations of
Mathematics. London: Routledge & Kegan
Paul.
Rieber, S. 1992. A test for quotation. Philo-
sophical Studies, 68, 83-94.
Smart, J,J.C. 1959. Sensations and brain pro-
cesses. Philosophical Review, 68. 141-156.
Stalnaker, R. 1978. Assertion. Syntax and
Semantics. 9, 315-32.
Stalnaker, R. 1984. Inquiry. Cambridge, MA.:
M.LT. Press.
Tichy, P. 1983. Kripke on necessity a poster-
iori. Philosophical Studies, 43. 225-41.
Unger. P. 1984. Philosophical Relativity. Min-
neapolis, MN: University of Minnesota Press.
Thanks to the Boyce Gibson Memorial
Library and the philosophy department of
Birkbeck College; and to the editors, Ned
LEIBNIZ'S LAW
Block. Alex Byrne, Mark Crimmins. Allen
Hazen. Ned Hall. Elijah Millgram. Thomas
Nagel, and especially Frank Jackson.
DAVID LEWIS
Leibniz's Law Named after the seven-
teenth-century philosopher and logician
G. W. von Leibniz. this so-called 'law' is
better thought of as a logical principle
which governs our use of the identity rela-
tion. Perhaps more than other such prin-
ciples, it has been subject to various
challenges, but many regard it as funda-
mental to the explanation of identity. The
law is most easily stated against the back-
ground of some dispute about identity. For
example, suppose that you have lost your
treasured fountain pen, and notice. some
weeks later. someone who works in your
office sporting a pen just like yours in his
pocket. Being of a suspicious character. you
wonder to yourself whether the pen in his
pocket is the pen you lost. What you
wonder of course is not whether the pen is
the same type as yours, but whether it is the
very same pen. Using 'a' as a name for the
pen that has gone missing, and 'b' for the
one seen in the pocket, what you wonder is
whether the following statement (expressed
in logical symbols) is true:
a=b
In it the equality sign is read as: 'is one and
the same as'. What Leibniz's Law says is
that if this identity is true, then any prop-
erty of a is also a property of b (and vice
versa). And many think this is virtually a
truism about identity, since if the above
identity is true, there is really only one pen,
and things cannot have and lack the same
property at the same time. In somewhat
more formal terms, Leibniz's Law is usually
stated this way:
For any x, any y and any property 0.
IF x=y THEN 0x if and only if 0y
The importance of this principle for the
philosophy of mind comes from its role in
431
LEIBNIZ'S LAW
discussions of IDENTITY THEORIES of the
mind. For, using this principle, one can at
least attempt to refute specific versions of
the identity theory by finding some property
of a mental item not shared by the physical,
or vice versa. If some such property can be
found then the part of the principle follow-
ing the 'THEN' will be false, and, by contra-
position, the identity itself will be false,
However, two things make this strategy less
than straightforward: first, as was men-
tioned earlier, not everyone thinks that
Leibniz's Law is true (e.g. that it doesn't
432
apply in modal contexts), even though it
has the air of a truism; and, secondly, ver-
sions of the identity theory in the philo-
sophy of mind are PROPERTY identity
theories, whereas the above formulation is
appropriate to individual things like pens.
This is important because it seems pos-
sible to formulate Leibniz's Law for proper-
ties in ways that make it less of a weapon
against the type identity theory.
See also DA VIDSON; LEWIS.
SAMUEL GUTTENPLAN
M
materialism see LEWIS; PHYSICALISM.
memory We are acutely aware of the
effects of our own memory, its successes
and its failures, so that we have the impres-
sion that we know something about how it
works. But, with memory, as with most
mental functions, what we are aware of is
the outcome of its operation and not the
operation itself. To our introspections, the
essence of memory is language based and
intentional. When we appear as a witness
in court then the truth as we are seen to
report it is what we say about what we
intentionally retrieve. This is, however, a
very restricted view of memory albeit with a
distinguished history. William James (1890)
said 'Memory proper is the knowledge of a
former state of mind after it has already
once dropped from consciousness; or rather
it is the knowledge of an event, or fact, of
which meantime we have not been think-
ing, with the additional consciousness that
we have thought or experienced it before'
(p. 648).
One clue to the underlying structure of
our memory system might be its evolu-
tionary history. We have no reason to
suppose that a special memory system
evolved recently or to consider linguistic
aspects of memory and intentional recall as
primary. Instead, we might assume that
such features are later additions to a much
more primitive filing system. From this per-
spective one would view memory as having
the primary functions of enabling us (the
organism as a whole, that is, not the con-
scious self) to interpret the perceptual
world and helping us to organize our
responses to changes that take place in
that world.
SOME USES OF THE TERM 'MEMORY'
Before continuing, it is necessary to refer to
certain distinctions made concerning
memory so that it will be clear what I am
not talking about. The first distinction is
that between Short Term Memory (STM)
and Long Term Memory (LTM). These are
highly misleading terms, in that they imply
that there are two, and just two such enti-
ties, separate and integral with STM char-
acterized as having limited capacity and
being available for limited amounts of time
only, in contrast with LTM where capacity
is unlimited and memories are permanent.
In practice, STM and LTM are oper-
ationalized in terms of particular tasks.
Experiments supposedly testing the proper-
ties of these two mental entities, in fact,
initially distinguished them solely on the
basis of the time interval between presenta-
tion of the stimulus and the recall. Sub-
sequently, there has been much research
aimed at establishing the distinction
between the two - other than, that is, in
terms of long and short - that has tested
such propositions as STM is subject to pho-
nological interference and LTM to semantic
interference. The straightforward memory
span task was taken as being equivalent to
STM. Finding that this task was subject to
semantic influences, and that tasks indexing
LTM could be subject to phonological influ-
ences, led some writers to abandon the dis-
tinction on the grounds of economy.
Unfortunately, most theories of memory
tend to say nothing about the processes
required for speech, language and meaning
although most memory experiments involve
words. All theories of language processing,
however, require processors that have the
ability to integrate over a number of items
433
MEMORY
and. thus. information must be preserved
over short amounts of time specifically for
the purpose of language-oriented computa-
tions. The structures within which this
takes place are called buffer memories.
While we are performing tasks involving
the recall of information over short inter-
vals. the contents of these buffer memories
will be available. Notions of economy in the
theory of memory do not apply. Thus a
contentious claim such as STM is the route
of entry to LTM for new memories could be
reformulated as the non-contentious prior to
the creation of a permanent memory. the rel-
evant material will be found in one or more
structures having the function of acting as
buffers.
The second distinction commonly made is
that between implicit memory and explicit
memory. Implicit memory is revealed when
performance on a task is facilitated in the
absence of conscious recollection. Alter-
natively. the abilities of densely amnesic
patients to learn while having no conscious
recollection at all of the episodes during
which the learning took place (or. indeed. of
anything else since the onset of the
amnesia) are seen as evidence of the distinc-
tion. A typical experimental paradigm is
that of repetition priming. where previous
study of a list of words will help in a follow-
ing task to make a word/non-word judg-
ment of a letter string or to complete word
fragments such as A-A-IN or U-V-SE. Such
priming is quite generally found with
amnesic subjects and also for normal sub-
jects in the case that they do not remember
having seen the words in the study period.
Now. nearly all theories of word recognition
require there to be structures responsible for
decoding the stimulus at a low level (which
may or may not be related to the buffer
memories already referred to). These struc-
tures are termed input lexicons - or some
such name. From studies of the effects of
priming in word recognition it is reasonable
to suppose that input lexicons carry traces
of their activity over time. There is also
good reason to believe that we have no con-
scious control over the activity of the input
lexicons and are unaware of activity in
434
them. Much of the evidence for implicit
memory can be accounted for on the basis
of the operation of such structures or those
required to organize verbal or other output.
Note that the implicit/explicit distinction
has nothing to do with the usual conscious/
unconscious distinction. Implicit memory
could be seen as non-conscious rather than
unconscious. More important is that influ-
ences on performance in experiments
testing implicit memory will derive from dif-
ferent sources than those involved in other
memory performance.
MEMOR Y PROPER
The memory we will focus on was defined
above as that enabling us to interpret the
perceptual world and helping us to organize
our responses to changes that take place in
that world. For both of these functions we
have to accumulate experience in a memory
system in such a way as to enable the pro-
ductive access of that experience at the
appropriate times. The memory we are
interested in here. then. can be seen as the
repository of experience. Of course. beyond
a certain age. we are able to use our mem-
ories in different ways. both to store infor-
mation and to retrieve it. Language is vital
in this respect and it might be argued that
much of socialization and the whole of
schooling are devoted to just such an exten-
sion of an evolutionary (relatively) straight-
forward system. It will follow that most of
the operation of our memory system is pre-
conscious. That is to say. consciousness
only has access to the product of the
memory processes and not to the processes
themselves. The aspects of memory that we
are conscious of can be seen as the final
state in a complex and hidden set of opera-
tions.
How should we think about the structure
of memory? The dominant metaphor is that
of association. Words. ideas. and. indeed.
emotions are seen as being linked together
in an endless. amorphous net. That is.
indeed. the way our memory can seem to us
if we attempt to reflect on it directly. But. as
already indicated. it would be a mistake to
dwell too much on the products of con-
sciousness and imagine that they represent
the inner structure. For a cognitive psycho-
logist interested in natural memory phenom-
ena there were a number of reasons for
being deeply dissatisfied with theories based
an associative nets. One ubiquitous class of
memory failure seemed particularly trouble-
some. This is the experience of being able to
recall a great deal of what we know about an
individual, other than their name. On one
such occasion, I was discussing someone's
research with a colleague. The main results
were familiar to us both. We knew where
the man worked, where he lived, what his
wife was called and the last time we had
heard him talk. We felt certain we would
instantly recognize his name if it were pro-
duced by someone else and that if we had
started with his name we would have been
able to retrieve the rest of the information.
How might various theories of memory
account for this phenomenon? First we can
take an associative network approach (e.g.
Anderson, 1976, 1983; Anderson and
Bower, 1973; Raajmaker and Shiffrin,
1980). In the idealized associative network,
concepts, such as the concept of a person,
are represented as nodes, with associated
nodes being connected through links. Gen-
erally speaking, the links define the nature
of the relationship between nodes, e.g. the
subject-predicate distinction (Anderson,
1976; Anderson and Bower, 1973). Let us
suppose that the name of the person we are
trying to recall is Bill Smith. We would have
a BILL SMITH node (or a node correspond-
ing to Bill Smith) with all the available
information concerning Bill Smith being
linked to form some kind of propositional
representation. Now, failure to retrieve Bill
Smith's name, while at the same time being
able to recall all other information concern-
ing Bill Smith, would have to be due to an
inability to traverse the links to the BILL
SMITH node. However, this seems contra-
dictory to one principle of associative net-
works: content addressability (e.g. Anderson
and Bower, 1973; Anderson, 1976). That
is to say, given that anyone constituent
of a propositional representation can be
MEMORY
accessed, the propositional node, and conse-
quently all the other nodes linked to it,
should also be accessible. Thus, if we are
able to recall where Bill Smith lives, where
he works, whom he is married to, then, we
should, in principle, be able to access the
node representing his name. To account for
the inability to do so, some sort of tempor-
ary 'blocking' of content address ability
would seem to be needed. Alternatively,
directionality of links would have to be spe-
cified (Anderson, 1976, 1983), though this
would have to be done on an ad hoc basis.
Next we can consider schema approaches
(Bartlett, 1932; Rumelhart, 1980; Schank,
1980; Schank and Abelson, 1977). Schema
models stipulate that there are abstract
representations, i.e. schemata, in which all
invariant information concerning any par-
ticular thing are represented. So we would
have a person schema for Bill Smith that
would contain all the invariant information
about him. This would include his name,
personality traits, attitudes, where he lived,
whether he had a family, etc. It is not clear
how one would deal with our example
within a schema framework. Since some-
one's name is the quintessentially invariant
property, then, given that it was known, it
would have to be represented in the schema
for that person. From our example, we
knew that other invariant information, as
well as variant, non-schematic information
(e.g. the last talk he had given) were avail-
able for recall. This must be taken as evi-
dence that the schema for Bill Smith was
accessed. Why, then, were we unable to
recall one particular piece of information
that would have to be represented in the
schema we clearly had access to? We would
have to assume that within the person-
schema for Bill Smith are sub-schemata (see
e.g. Schank, 1980) one of which contained
Bill Smith's name, another containing the
name of his wife, and so on. We would
further have to a s ~ u m e that access to the
sub-schemata was independent and that, at
the time in question, the one containing
information about Bill Smith's name was
temporarily inaccessible. Unfortunately the
concept of temporary inaccessibility is
435
MEMORY
without precedent in schema theory and
does not seem to be independently motiv-
ated.
There are two other classes of memory
problem that do not fit comfortably into the
conventional frameworks. One is that of not
being able to recall an event in spite of most
detailed cues. This is commonly found when
one partner is attempting to remind the
other of a shared experience. Finally, we all
have the experience of a memory being trig-
gered spontaneously by something that was
just an irrelevant part of the background for
an event. Common triggers of such experi-
ences are specific locales in town or
country, scents and certain pieces of music.
What we learn from these three kinds of
event are that we need a model which
readily allows the following three properties:
(1) not all knowledge is directly retrievable;
(2) the central parts of an episode do not
necessarily cue recall of that episode;
(3) peripheral cues, which are non-essen-
tial parts of the context, can cue recall.
In response to these requirements, we pro-
posed the Headed Records model of memory
(Morton, Hammersley and Bekerian, 1985;
Morton and Bekerian, 1986; Morton, 1990,
1991). The framework within which the
model is couched is that of information pro-
cessing. In trying to solve the problems, we
first supposed, following Norman and
Bobrow (1979), that memory consists of
discrete units, or Records, each containing
information relevant to an 'event', an event
being, for example, a person or a personal
experience. Information contained in a
Record could take any number of forms,
with no restrictions being placed on the way
information is represented, on the amount
being represented or on the number of
Records that could contain the same
nominal information. Attached to each of
these Records would be some kind of access
key. The function of this access key, we sug-
gested, is singular: it enables the retrieval of
the Record and nothing more. Only when
the particular access key is used can the
Record, and the information contained
436
therein, be retrieved. As with the Record,
we felt that any type of information could be
contained in the access key. However, two
features would distinguish it from the
Record. First, the contents of the access key
would be in a different form to that of the
Record, e.g. represented in a phonological or
other sensory code rather than a semantic
or other central code. Second,. the contents
of the access key would not be retrievable.
The access key was termed the Heading, and
the framework became Headed Records
(henceforth HR).
Following Norman and Bobrow (1979),
the first thing that happens when memory
is interrogated is that a Description is
formed. This is the information used in the
search. What is needed is something which
is suitable as a memory probe. To give an
example, if I asked you the question 'Could
you give me the name of your best friend's
wife?', it would not make much sense for
you to interrogate your memory for (best
friend's Wife). What you have to do is to split
the task up into its component parts. The
first part is (best friend). Once that has been
decided then (X's Wife) can be found. The
Description is formed from currently avail-
able information from external sources
(such as an explicit question), internal
sources (a Record that has just been
retrieved) and a Task Specification. The
Task Specification contains a list of the
current goals. A Description can comprise a
number of independent fields. These will
include some environmental information
and some internal state variables as well as
lexical, propositional or other content.
The nature of the match required
between the Description and a Heading will
be a function of the type of information in
the Description. If the task is to find the
definition of a word or information on a
named individual then a precise match may
be required at least for the verbal part of the
Description. We assume that the Headings
are searched in parallel. On many occasions
there will be more than one Heading that
matches the Description. However, we
require that only one Record be retrieved at
a time (see also Anderson, 1976; Anderson
and Bower, 1973; Rumelhart, 1980;
Schank, 1980). Evidence in support of this
assumption is summarized in Morton, Ham-
mersley and Bekerian (1985). The data
indicate that the more recent of two possible
Records is retrieved. We conclude first that
once a match is made the search process
terminates and secondly that the matching
process is biased in favour of the more
recent Heading (see also Hasher et aI., 1981;
Martin, 1971; Postman and Underwood,
1973). There is, of course, no guarantee
that the retrieved Record will contain the
information that is sought. The Record may
be incomplete or wrong. In such cases, or in
the case that no Record had been retrieved,
there are two options: either the search is
continued or it is abandoned. If the search
is to be continued then a new Description
will have to be formed since searching
again with the same Description would
result in the same outcome as before. Thus,
there has to be a list of criteria upon which
a new Description can be based.
Retrieval depends upon a match between
the Description and the Heading. The rela-
tionship between the given cue and the
Description is open. It is clear that there
needs to be a process of Description forma-
tion which will pick out the most likely
descriptors from the given cue. Clearly, for
the search process to be rational the set of
descriptors and the set of Headings should
overlap. Indeed, the only reasonable state of
affairs would be that the creation of Head-
ings and the creation of Descriptions is the
responsibility of the same mechanism.
ACCOUNTING FOR PHENOMENA
The model contrasts with most other
models of memory in that it is explicitly not
freely content addressable. In the model,
search only occurs on Headings. Informa-
tion that is central to an event memory will
serve as a cue for the recall of that memory
only if it is found in the Heading. If such
information is only present in the Record, it
cannot be used as a retrieval cue. The con-
verse of this is that information in the
Heading need not be present in the linked
MEMORY
Record. Thus, something which would be a
reliable cue for a set of knowledge might be
unretrievable if that set of knowledge were
accessed by other means. Not being able to
retrieve people's names is a common
example of such a principle operating in
practice.
The account of this phenomenon in HR
terms is that the name forms a part of a
Heading. Since the Headings have a
number of components, and it is not neces-
sary for the match between Heading and
Description to be complete, it would be pos-
sible for the Record to be accessed by some
other cue, such as the place where the
subject of the Record, Bill Smith in the pre-
vious analysis, had last been encountered.
The information in the Record would be
retrieved, but there would be no way of
retrieving the contents of the Headings. For
another individual, of course, the name
could be in the Record and the situation
would not arise. Such variability in memory
organization is as much a burden to the
theorist as it is to the owner of the memory.
An experimental way of determining the
components of Headings is through a com-
parison of the relative effectiveness of vari-
ables on recognition memory compared
with recall. The reason for this lies in the
difference between the way these two tasks
map onto the HR framework. Recognition
memory involves the subject's judging
whether or not the presented material had
previously been experienced. This requires
that the material forms a Description which
matches a Heading and that the Record
that is retrieved contains information which
enables the evaluation system to decide
whether or not the task demands have been
satisfied. In recall, on the other hand, the
subject is given only some notion of the
topic and the circumstances of the previous
encounter. The material itself has to be
found in a Record.
The data indicate that the literal form of
the stimulus serves as a cue in recognition
memory. In recognition memory for text,
high- and low-level propositions (defined by
their centrality to the theme of the passage)
are equally well recognized (Yekovitch and
437
MEMORY
Thorndyke, 1981). In free recall, the
higher-level propositions are better reported
in spite of instructions for literal recall
(Kintsch, 1974). The HR interpretation of
this would be that the literal form of each
sentence is directly addressable - Le. con-
stitutes a Heading, whereas only what is
evaluated as most important finds its way
into the Record. Equally, the sensitivity of
recall to state and context variables con-
trasted with the relative insensitivity of
recognition memory also indicates that such
variables are to be found in Headings. There
are other experimental ways of determining
what kind of information gets into the
Headings. Thus, an experiment by Godden
and Baddeley (1975, 1980) showed that
the context is important. They asked deep-
sea divers to learn material on the sea bed,
and showed that subsequent recall was
worse on land compared with back under
water. Further data of interest to us here
indicate that state variables such as mood
and drug state affect recall and not recogni-
tion (Bower, 1981; Eich, 1980), which in
HR terms means that they are to be found
in Headings.
In the course of our development we will
have built up a large number of routines to
guide our behaviour in different circum-
stances. When such a Record is accessed it
serves as a referent Record. The basic cycle
involves first of all the default retrieval of
the control Record used to select a salient
feature of the environment as a suitable
descriptor. Next a Description is formed and
the search process leads to the retrieval of a
referent Record which is used to interpret
the environment and guide our actions. An
evaluation process must be operating con-
tinuously during the last of these stages to
ensure that the referent Record continues to
be appropriate. Such a process is not pecu-
liar to the present model but would also be
required in some forms of schema theory.
An important question is how we move
from one record to another, Le. how the
whole system operates in interpreting situa-
tional experience; being in a restaurant, for
example. A switch of referent Record may
be necessitated by a change in the environ-
438
ment (if, for example, someone starts a
brawl in the restaurant), a change in the
demands imposed in the same environment
(e.g. starting to discuss business with one's
dinner companions after having ordered
dinner), or some change requiring a general
problem-solving routine to be retrieved (as
when one's spouse arrives unexpectedly in
the same restaurant). In all these cases the
sequence of events would be:
(1) detection of the inadequacy of the refer-
ent Record by the evaluation routine;
(2) formation of a new Description;
(3) retrieval of a different referent Record.
We have assumed that only one referent
Record can be used at a time. This is the
simplest assumption. In any case we could
not expect a very direct relationship
between the dynamics of behaviour and the
underlying representation. The processes
that mediate between the representations in
the Records and actual behaviour will have
the effect of smoothing over the underlying
joints, much as the underlying multi-
layered structure of an utterance is dis-
guised by the time it becomes speech.
One consequence of a Headed Records
system is that there will be multiple repre-
sentations both of knowledge and of skills.
This will occur for a number of reasons.
First of all, procedures corresponding to
mental or physical skills will be represented
in Records. Whenever such a developing
skill is used and is changed as a result of
this practice then there will be a new
Record of the improved form of the proce-
dure. However, the old form will remain,
since there is no overwriting in the system.
The new form will be used next time the
procedure is called, following the recency
principle, but the old form would be re-used
if, for example, the context of use of the two
forms had been different and the context for
the old form were reinstated. This may seem
to go against a body of evidence that has
accumulated over the last 20 years con-
cerning the unreliability and changeability
of memory especially in the work of Loftus
(1979) and Loftus and Ketcham (1991).
Loftus and her colleagues have demon-
strated quite conclusively that memory per-
formance is subject to change following
false or misleading information. However,
Bekerian and Bowers (1983) showed that at
least one of Loftus's results could be
reversed by a change in the testing condi-
tions. The reconciliation is that, under ques-
tioning or testing, new Records are set up.
To the extent to which the Heading on the
new Record is the same as that on the old
one, the more recent will be retrieved.
Functionally, then, in many cases it will be
as if the old memory is inaccessible.
AMNESIA
The most common cause of amnesia in
adults is Korsakoff's syndrome. It is char-
acterized by loss of ability to recall incidents
from the past (retrograde amnesia), inability
to recall current activities after intervals
sometimes as short as a minute (antero-
grade amnesia) and often an almost normal
performance in short-term memory tasks.
Learning can take place for a variety of
materials in spite of the patient denying that
they have ever been in the situation before.
My assumption is that the breakdown in
organic amnesia is multifaceted and vari-
able.
Baddeley (1984) has come the nearest to
a characterization of amnesia that I have
seen that approaches the complexity of its
target. The starting point was what amne-
sics can learn. He points out that a wide
range of amnesic patients show apparently
normal learning both on verbal tasks and
on complex and apparently semantically
based perceptual tasks. What the tasks have
in common is that they allow the patient to
reflect learning without having to consider
the provenance of the information that was
used.
In an attempt to give this characteriza-
tion of amnesic learning some theoretical
force, Baddeley invokes a distinction
between relatively automatic retrieval pro-
cesses and the active problem-solving aspect
of recall that he terms 'recollection'.
Suppose, then, that amnesic patients lack
MEMORY
the ability to recollect, although they can
still build up and run off 'procedures'. The
consequences of the disability would
include:
(1) an inability to use incidental detail as
confirmation of the correctness of some-
thing retrieved;
(2) an inability to reject incorrect associ-
ations produced by automatic pro-
cedures;
(3) an inability to iterate a retrieval cycle to
follow clues and check them through
memories with episodic characteristics.
Baddeley points out that the concept of
recollection only goes a little way towards
accounting for amnesia. Nonetheless it has
the outstanding advantage that it is
dynamic and makes contact with real
memory phenomena.
Within an HR model there are a number
of ways that apparent forgetting can take
place. First, the material may not be laid
down in either Headings or in Records.
These problems would affect recognition
and recall respectively. Secondly, particular
kinds of information may not be represented
in new Records. This could have the effect
of the Record being rejected by the evalu-
ation process on the basis of particular task
specifications. This could also account for
the massive increase in proactive inter-
ference in list learning with amnesics except
where the contexts are exaggeratedly dis-
tinct (Winocur and Kinsbourne, 1978).
The third possibility is that Heading for-
mation or Description formation might be
altered. If both are altered then there would
be no anterograde amnesia but there would
be retrograde amnesia. This is the principle
I have used to account for infantile amnesia
(Morton, 1990). If Heading formation alone
is changed then there could be no retro-
grade amnesia but there would be antero-
grade amnesia. Then, of course, these
factors, together with others, could co-occur
in ways which are unique, and we will find
the need to define a number of subtypes.
Amnesia characterizes certain non-
organic states such as multiple personality
439
MEMORY
disorder (MPD). The manifestation of MPD
is variable. but it seems possible to char-
acterize the condition as one where part of
the memory system is divided up into a
number of mutually exclusive sets. Each per-
sonality can be seen as being able to access
only one such set plus a common set of
linguistic and other general information. In
HR terms. this would be achieved by having
a self marker in the Headings that had to be
matched by a component of the Description
during memory search. Whichever person-
ality dominated the processing systems. the
corresponding self marker would auto-
matically be part of the Description. and
Records lacking a matching component in
the Heading could not be retrieved.
The opposite. of forgetting is the creation
of false memories. Loftus (1993) has de-
scribed an ingenious method of inducing
such memories under experimental condi-
tions. Specifically. she and her colleagues
showed that it was quite easy to make
people believe that they had been lost on a
particular occasion when they were five
years old. Their technique involves getting a
relative to recall a number of events that
occurred in the subject's past. including the
fictitious event. Sometimes after a few days.
the subject comes to believe in the event.
even retrieving confirmatory visual images
and other of the indices we use in assessing
our own recollection of real events.
There is other evidence to suggest that
some of our recall of the past is reconstruc-
tive. without any awareness on our part. as
Bartlett (1932) demonstrated. However. it is
also clear that the demands of the question-
ing are quite crucial and that under some
circumstances people can assess the reli-
ability of their recall.
In sum. memory can be seen as a phe-
nomenon that results from the interpreta-
tion of material selected from a retrieved
Record which contains a selection of an
individual's sensory experience of an exter-
nal event in the past plus such bits of prior.
related memories and default values that
happened to get attached in the original
process of Record formation or filled in at
the time of recall. If it is a conceptual truth
440
that no one can remember that p when p is
false. it is a psychological truth that you
can never remember.
See also ARTIFICIAL INTELLIGENCE; COGNI-
TIVE PSYCHOLOGY; COMPUT ATION AL
MODELS; CONSCIOUSNESS; REPRESENT A-
TION.
BIBLIOGRAPHY
Anderson. J.R. 1976. Language. Memory and
Thought. Hillsdale. NJ: Erlbaum.
Anderson. J.R. 1983. Architecture of Cognition.
Cambridge. MA.: Harvard University Press.
Anderson. J.R .. and Bower. G. 1973. Human
Associative Memory. Washington. D.C.:
Winston.
Baddeley. A.D. 1984. Neuropsychological
evidence and the semantic/episodic distinc-
tion. Behavioral and Brain Sciences. 7. 238-
9.
Bartlett. F.C. 1932. Remembering. Cambridge
University Press.
Bekerian. D.A.. and Bowers. J.M. 1983. Eye-
witness testimony: Were we misled? Journal
of Experimental Psychology: Learning Memory
and Cognition. 9. 139-45.
Bower. G.H. 1981. Mood and memory. Amer-
ican Psychologist. 36. 129-48.
Eich. J.E. 1980. The cue-dependent nature of
state-dependent retrieval. Memory and Cog-
nition. 8.157-73.
Godden. D.R.. and Baddeley. A.D. 1975.
Context-dependent memory in two natural
environments: On land and under water.
British Journal of Psychology. 66. 325-32.
Godden. D.R .. and Baddeley. A.D. 1980. When
does context influence recognition memory?
British Journal of Psychology. 71. 99-104.
Hasher. L .. Attig. M .. and Alba. J. 1981. I
knew it all along: Or did I? Journal of Verbal
Learning and Verbal Behavior. 20. 86-96.
James. W. 1890. Principles of Psychology. vol.
1. New York: Holt.
Kintsch. W. 1974. The representation of
meaning in memory. Hillsdale. N.J.: Lawrence
Erlbaum.
Loftus. E.F. 1979. Eyewitness Testimony. Cam-
bridge. MA.: Harvard University Press.
Loftus. E.F. 1993. The reality of repressed
memories. American Psychologist. 48. 518-
37.
Loftus, E.F., and Ketcham, K. 1991. Witness
for the Defence. New York: St Martin's Press.
Martin, E. 1971. Verbal learning theory and
independent retrieval phenomena. Psycho-
logical Review, 78, 314-32.
Morton, J. 1990. The development of event
memory. The Psychologist, 3, 3-10.
Morton, J. 1991. Cognitive pathologies of
memory: a headed records analysis. In
Memories, Thoughts, and Emotions: Essays in
Honor of George Mandler, ed. W. Kessen, A.
Ortony, and F. Craik. Hillsdale, NJ: Erlbaum.
Morton, J., and Bekerian, D.A. 1986. Three
ways of looking at memory. In Advances in
Cognitive Science 1, ed. N. E. Sharkey. Chi-
chester: Ellis Horwood.
Morton, J., Hammersley, R.H., and Bekerian,
D.A. 1985. Headed records: A model for
memory and its failures. Cognition, 20,1-23.
Norman D.A. and Bobrow, D.G. 1979. Descrip-
tions: an intermediate stage in memory
retrieval. Cognitive Psychology, 11, 107-23.
Postman, L., and Underwood, B. 1973. Critical
issues in interference theory. Memory and
Cognition, 1, 19-40.
Raajmakers, J., and Shiffrin, R. 1980. SAM: A
theory of probabilistic search of associative
memory. In The Psychology of Learning and
Motivation, vol. 14, ed. G. Bowes. New York:
Academic Press.
Rumelhart, D. 1980. The building of blocks of
cognition. In Theoretical Issues in Reading
Comprehension, ed. R. Spiro, B. Bruce and
W. Brewer. Hillsdale, NJ: Erlbaum.
Schank, R.C. 1980. Language and memory.
Schank, R.C., and Abelson, R.P. 1977. Scripts
Plans, Goals and Understanding: An Inquiry
into Human Knowledge Structures. Hillsdale,
NJ: Erlbaum.
Winocur, G., and Kinsbourne, M. 1978. Con-
textual cuing as an aid to Korsakoff am-
nesics. Neuropsychologia, 16,671-82.
Yekovitch, F.R., and Thorndyke, P. 1981. An
evaluation of alternative functional models
of narrative schemata. Journal of Verbal
Learning and Verbal Behavior, 20, 454-69.
JOHN MORTON
mental representation When we think
about the Biffel Tower we can be said to
represent it in our thought. In slightly dif-
MODULARITY
ferent terminology, we can be said to
possess a mental representation of the Biffel
Tower, and to differ in this from someone
who lacks the means to think about that
famous iron structure. So understood, a
mental representation is simply a species of
REPRESENT ATION. However, deep and
vexing problems arise when one tries to go
beyond this minimal description and say
more fully what kind of thing a mental
representation is. Are thoughts somehow
made up of mental presentations? Do
mental representations have a causal and/
or functional presence in the individual
human mind or brain? And if they do have
some such presence, are they like images, or
more like linguistic signs? These questions
set the agenda for a large part of con-
temporary philosophy of mind. See CON-
CEPTS; DENNETT; DRETSKE; FODOR;
FUNCTIONALISM; IMAGERY; LANGUAGE OF
THOUGHT.
SAMUEL GUT TEN PLAN
mind-body problem see An Essay on Mind
section 3; DAVIDSON; DUALISM; FODOR;
FUNCTION ALISM; IDENTITY
LEWIS; PHYSICALISM; RYLE;
SUPERVENIENCE.
modality see POSSIBLE WORLD.
THEORIES;
SEARLE;
modularity One could stretch the use of
the term 'modularity', and argue that even
Plato held a doctrine of the modularity of
mind (in the Republic) or that he explicitly
rejected the modularity of long-term
MEMOR Y (in the Thaeatetus). FODOR himself
(1983) credits nineteenth-century phrenolo-
gist Gall with the idea of the modularity of
mind. But in fact the modularity debate as it
is currently framed in Cognitive Science
derives straightforwardly from the argu-
ment advanced in Fodor's book The Modu-
larity of Mind (1983) and the literature it
inspired. Subsequent to the appearance of
this influential volume, scholars and scien-
tists in philosophy, COGNITIVE PSYCHO-
LOGY. ARTIFICIAL INTELLIGENCE and
linguistics have all actively pursued
441
MODULARITY
research aimed at discovering which of the
mind's subsystems are modular, and the
respects and degrees to which they are so.
Broadly speaking, a module is a relatively
autonomous component of the mind - one
which, while it interacts with, receives
input from and sends output to, other cog-
nitive processes or structures, performs its
own internal information processing unper-
turbed by external systems. The theory of a
single module and its operation could
hence, with the exception of mentioning its
input and output interfaces, in the ideal
case. be developed without ever mentioning
the remainder of the mind. Following Fodor.
for a process to be modular is for it to satisfy
these eight conditions. the first four of
which played the greatest role in the
empirical assessment of the modularity
hypothesis: (1) domain specificity; (2) man-
datoriness; (3) informational encapsulation;
(4) speed; (5) shallow output; (6) lack of
access of other processes to intermediate
representations; (7) neural localization; (8)
succeptibility to characteristic breakdown.
Typically. input and output modules -
those responsible for PERCEPTION and
ACTION - are candidates for modularity. It
is much less plausible to argue that central
processes such as those recruited in induc-
tive reasoning or interpreting Platonic dia-
logues are subserved by modules. For one
thing. such central processes demand broad
access to a wide range of knowledge. For
another. it would be bizarre to suggest that
evolutionary processes would issue in the
existence of special neural structures
devoted to these tasks. Perception and
motor control. on the other hand. are typi-
cally highly data-driven. and the processes
subserving them plausibly respond fairly
directly to selection pressure. Debates in
contemporary cognitive science concerning
the modularity hypothesis typically focus on
one or more of the following questions. (1)
Which cognitive faculties. if any. exhibit all
or most of these characteristics? (2) Does
this list of characteristics in fact cluster and
determine a cognitive 'natural kind'. or is
there a more felicitous characterization of
what are intuitively modular systems?
442
(3) Are cognitive modules part of the innate
cognitive architecture of the human mind
or is modularity. to the extent that it is
present. an artefact of learning or other
developmental processes? (See DEVELOp
MENTAL PSYCHOLOGY.)
Let us consider each of these character-
istics and the way they figure in current
debates concerning modularity in cognitive
science. When we ask of a cognitive capacity
whether it is domain specific. we are asking
whether it has as its object a unique and
idiosyncratic domain. So. for instance. the
claim that the phonological analysis system
is modular would be the claim that there are
specific auditory processes or capabilities
that are brought into play for speech analy-
sis that are not recruited for any other kind
of hearing - for instance for listening to
music. and that these processes are triggered
by and exploit unique features of human
speech. Domains that have been argued by
modularists to have such unique features
and for the analysis of which dedicated
modules have evolved include human lan-
guage. human speech. medium-sized visual
objects. depth relations. and motor control.
Critics of general modularity claims or of
claims to the domain specificity of a parti-
cular candidate module typically argue that
the processes to which modularists advert
are in fact instances of more general cogni-
tive processes that are recruited across
domains. or that even plausibly central pro-
cesses can be highly domain specific. For
instance Fodor and others argue that on-
line language processing recruits domain-
specific processes, since the mechanisms
called upon to transduce speech sounds into
linguistic representations - whether these
outputs are characterized as phonological or
syntactic representations - do not operate
on. say. automobile noise or bird songs.
Arbib (1972. 1979. 1989). Arbib and Han-
son (1987). Arbib. Boylls and Deb (1974)
and Stillings (1989) among others point
out. on the other hand. that the fact that
the sighted can learn to read type. that the
blind can learn to read braille and that
the deaf can learn to use sign suggests
that the parsing mechanism can hardly be
specialized to the domain of speech signals.
Moreover, contemporary connectionist
models of language learning (Rumelhart
and McClelland, 1986; Rumelhart, Hinton
and Williams, 1986) challenge the idea that
even phonological and syntactic parsing are
idiosyncratic processes 'wired in' to the cog-
nitive architecture for the purposes of lin-
guistic processing (see CONNECTIONISM).
These models suggest that they may only be
specific instances of general learning and
perceptual processing strategies fine-tuned
by training to linguistic processing. This line
of argument gains further support from the
observation that skill acquisition results in
performance having exactly the character-
istics Fodor and others ascribe to modules:
speed, mandatory operation, encapsulation,
and perhaps even localization. This is a
quite general feature of models of skill
acquisition, despite other architectural dif-
ferences, and so is apparent in production
systems as well as connectionist systems.
(see Stillings (1989a, 1989b). Karmiloff-
Smith (1979, 1985, 1986, 1990, 1993)
however, points out that much of this may
be accounted for by the propensity of the
mind to acquire modules that are not
innately specified.
Modularists argue that modules are man-
datory in operation. We don't have a choice
regarding whether to bring our scene recog-
nition mechanisms to bear on incoming
visual data. Objects emerge from the optic
array whether we would prefer a 'bloomin,
buzzin confusion' or not. Similarly, we have
no choice about whether to regard a
sequence of words uttered in our native lan-
guage as a linguistic expression or a non-
linguistic soundstream. But we can choose
what to think about - modularity or cricket,
for instance. Mandatoriness makes good
sense as Fodor (1989) notes for systems that
are designed to detect features of the
immediate environment that have signifi-
cant implications for survival, and for
systems designed to act on that environment
when speed makes a difference. Where it is
important to act quickly, it is an advantage
to have a system that acts automatically. On
the other hand, central cognitive processes
MODULARITY
must, if they are to be flexible enough to
serve their function - presumably that of
allowing us to adapt our behaviour to the
variable demands of our physical and social
environments - be subject to voluntary
control. Modularists point to the phenomena
to which we have just adverted to argue
that perceptual and motor processes typi-
cally are mandatory. (The mandatory char-
acter of motor processes is actually a bit
delicate to spell out: we certainly have the
ability to decide to act, say, to type, to walk,
or to speak, and motor action is voluntary in
that sense. But we don't have the ability,
unless we really work at it, to determine via
conscious control the trajectories of our
fingers over the keyboard, the precise move-
ments of our legs, or the positions our
tongues and lips adopt as we speak. To place
these under conscious control. we cease to
simply move, and begin to act, in the thea-
trical sense of that term, at least according
to modularity theory.)
Critics of modular claims, on the other
hand, point out that many central processes
seem to be mandatory in the same sense
that modularists claim that modular periph-
eral processes must be. Marslen-Wilson and
Tyler (1989), for instance, point out that
mapping linguistic input onto discourse
representation is as mandatory as speech
recognition or syntactic parsing. But inas-
much as discourse representation recruits
central processes, this observation would
seem to undermine the modularity thesis.
Moreover, many obviously central processes
seem to be mandatory. While we have a
great deal of voluntary control over what
we think about, we often find what we
think about is forced upon us, and we have
no reason to believe that we have any more
control over the microprocesses of abstract
thought than over those involved in loco-
motion or speech.
When modularists argue that modular
processes are informationally encapsulated
they mean that cognitive modules have no
access to information from elsewhere in the
cognitive system (except of course for the
initial input into an output system). All of
the information available to a module
443
MODULARITY
comes directly from its own subsystems or
from their dedicated input devices. So,
according to modularists, when processing
incoming speech, my parser has access only
to its internally represented grammar and to
the acoustic properties of the speech signal;
when viewing the Miiller-Lyer illusion, my
visual system has access only to the visible
properties of the drawing, and not, import-
antly, to my own knowledge that the main
lines are in fact of equal length. Central pro-
cesses, on the other hand, share knowledge
freely. When I reason about cognitive
science, for instance, it is important that my
knowledge about the history of philosophy
is able to interact with my ability to reason
deductively and my knowledge of cognitive
psychology. But none of this knowledge
helps me a bit, according to the modularist.
when my visual recognition system or my
parser is called upon to interpret incoming
data. After all, the assignment of a semantic
value to a sentence requires first determin-
ing its syntactic structure and logical form,
and the information needed to do this is
entirely represented in the phonological and
syntactic portion of the grammar. Informa-
tional encapsulation is at the very heart of
the modularity thesis. For this characteristic
of a system is what ensures that it is really
autonomous in operation from the remain-
der of the mind.
Anti-modularists cite a variety of phe-
nomena as putative counterexamples to
the claim that input and output modules
are in fact encapsulated. Altmann (1986),
Marslen-Wilson (1975, 1980) and Marslen-
Wilson and Tyler (1989), for instance,
argue that general knowledge and context
penetrate the syntactic parsing of sentences.
Miller (1977, 1981) presents evidence that
syntactic information can drive phonologi-
cal processing. And even plausibly encapsu-
lated movements can be penetrated by
intentions to act differently. Arbib (1989)
suggests as well that much visual computa-
tion, even at fairly low levels, is influenced
by representations that penetrate the visual
system from more central cognitive pro-
cesses.
Modular processes are, according to
444
modularity theory, very fast. In fact speed is
one good diagnostic characteristic of a
system in determining whether it is
modular, and the speed of input analysis
and output execution is often used as a
premise in arguments for modularity. Fodor
(1983, 1987) for instance, argues that
since perceptual processes are fast, and
since it takes time to decide which of a vast
store of general knowledge to bring to bear
on problem solving, modular processes must
be encapsulated. He also argues that the
speed of modular processes derives in part
from their mandatoriness: the fact that
there is no need to decide whether to bring
a modular process into play or how to do so
eliminates planning time from total
response time. And of course speed is
achieved by the evolutionary tailoring of
the modular system to its particular
domain. Whereas a general-purpose system
achieves its generality at the expense of effi-
ciency in anyone domain, modular systems
are useless in domains other than their
proper ones precisely because they have
evolved to be optimal in their particular
domains. And indeed the contrasts between
on-line sentence processing in one's native
tongue and deliberate parsing of a sentence
in an unfamiliar language using explicit
grammatical knowledge, or between
walking naturally and imitating someone
else's walk are dramatic.
The argument from the speed with which
allegedly modular systems operate to their
modularity is, however, non-demonstrative.
It relies on the important premise that speed
in perceptual and motor processing is
achieved by serial computations utilizing
efficient algorithms that operate on a con-
strained range of data. That is indeed one
way to achieve speed. But there are others.
As Fodor acknowledges, Bruner argued -
and more recent connectionist work on per-
ception has supported this view - that speed
can also be bought by the efficient exploita-
tion of a wide range of relevant knowledge if
that knowledge is felicitously represented.
Connectionist perceptual networks accom-
plish highly efficient and plausibly psycho-
logically realistic recognition and
categorization using distributed representa-
tion systems that make all of the system's
knowledge relevant to each discrimination.
These systems violate encapsulation, and,
when suitably integrated or generalized,
may turn out to violate domain specificity as
well (McClelland, 1988; McClelland and
Rumelhart, 1981; Jacobs et ai., 1991).
There is another line of criticism against
the argument from speed to other modular
properties. Highly skilled behaviour becomes
quite rapid. The classic work of de Groot
(1966) and Chase and Simon (1973) on
chess perception, as well as the phenom-
enon of athletic or musical skill develop-
ment show that it is possible to develop
highly rapid processing of phenomena to
which it is highly implausible that there is a
domain-specific module dedicated. These
phenomena raise the possibility that such
phenomena as sentence parsing or visual
scene analysis may be so fast simply
because they are so well-practised.
It is essential to the view of the mind as
comprising a non-modular central proces-
sing system together with a set of modular
input/output faculties that the input fac-
ulties deliver relatively shallow representa-
tions - that is, that this information be
given in a rather 'raw' state. For if the input
presented by the perceptual faculties were to
be too highly processed, those faculties
would require far more knowledge than is
permitted by encapsulation. Correspond-
ingly, the instructions delivered to output
modules must be highly processed motor
programs, requiring no interpretation on
the part of those modules that would
demand the use of central processes or more
general data structures.
Arguments concerning the interface
between the centre and the periphery are
vexed, and are hard to assess independently
of the fate of the remainder of the modu-
larity hypothesis. For if the modular view of
the mind is accepted, then the shallow
output condition follows simply from the
fact that so much of language understand-
ing - in particular discourse representation
- is driven by background knowledge and
reasoning, and from the fact that we can
MODULARITY
plan our movements. On the other hand,
arguments such as those of Marslen-Wilson
and Tyler (1989), Forster (1980,1981) and
others suggest that discourse representa-
tion, or other forms of semantic evaluation,
are as fast and mandatory as any 'purely'
perceptual or syntactic processes, and even
casual observation of athletic performance
suggests that very complex movements can
be planned by a skilled actor. These con-
siderations tend to push the boundaries of
candidate modules towards the centre to a
degree that must make modularists uncom-
fortable. For then the boundary between
non-modular central processes and modules
becomes theoretically insignificant, and the
distinction becomes one of degree rather
than of kind.
Another important consequence of the
modular view is that central processes will
have no access to the intermediate repre-
sentations produced or utilized by the
modules. For the modules by definition
interact with central processes, including
attentional processes, only at their proper
interfaces. This is a happy consequence for
the modularity view, since it accords both
with naive intuition and with the striking
disparity between the deliverances of
psychological and psycho linguistic research
on the one hand and our introspective
awareness on the other: It seems over-
whelmingly plausible that underlying even
the simplest cognitive operations, to which
we have only the vaguest introspective
access, there are countless cognitive pro-
cesses running automatically. As you read
this sentence, for instance, there are stages
of visual analysis, lexical access, syntactic
representation, logical form analysis and
semantic evaluation which we can only
hypothesize based upon our best theories,
but which we can never observe. The
modularity hypothesis gives us a tidy
explanation of the lack of introspective
awareness of these stages of processing (see
INTROSPECTION): they occur inside
modules, and central processes have access
only to the outputs of modules, and
awareness is a central process.
On the other hand, critics of modularity
445
MODULARITY
will point out that we lack introspective
awareness of many of our cognitive pro-
cesses, prominently including many that
the modularity theory would regard as
central. As Kahneman and Tversky and
their associates have demonstrated, our
own views about the processes we use in
logical and statistical inference are often far
off the mark, and we have no reliable intro-
spective access to the processes we in fact
use. So while it is true that modularity cor-
rectly predicts our inability to introspect too
far into our perceptual and motor processes,
the fact that we also need an explanation of
our inability to introspect our central pro-
cesses, in the view of critics of modularity,
diminishes the confirmatory virtues of this
consequence.
The final two characteristics of modular
system are tightly connected, and may best
be discussed together: modular systems are,
according to standard modularity theory,
neurally localized and are subject to char-
acteristic and isolated patterns of break-
down. This second characteristic is, of
course, a straightforward consequence of
the first. If each cognitive module is sub-
served by a specific local neural structure
then it should be possible to selectively
impair the function of a single module
through injury to the specific structure in
which it is realized. This is an important
empirical prediction for modularity theory,
and represents an area of research that has
been on balance highly favourable to the
modularist viewpoint. For a wealth of recent
neurophysiological evidence supports the
view that such systems as vision, audition,
olfaction, language perception, propriocep-
tion, motor control and speech production
are largely localized in specific areas of the
brain, and that, modulo some neural plasti-
city that allows variable recovery of func-
tion, injury to these specific areas causes
significant degradation in the function of the
corresponding cognitive modules.
The fact that the brain has evolved in
such a way as to incorporate specific 'dedi-
cated processors' supports the general view-
point of modularity theory: it suggests that
there are anatomical structures that are
446
domain-specific in their function, at least
with respect to some specification of domain.
The verifiable fact that these structures are
active whenever their appropriate stimulus
conditions are present supports the manda-
toriness thesis. The presence of such dedi-
cated structures explains the speed of
modular processes in a way harmonious
with that in which modularism seeks to
explain such speed for independent reasons.
The satisfaction by the candidate modules of
the remaining conditions - encapsulation,
shallowness of output, and lack of access of
central processes to intermediate representa-
tions - are neither directly confirmed nor
disconfirmed by these data. But the fact of
localization suggests that further neuroana-
tomical study of the connectivity of these
modular sections of the brain to those sec-
tions subserving general cognition and the
other modules might shed light on these
questions.
Neuroanatomical data however do not at
this stage settle the modularity question.
For one thing, localization is not as neat as
this picture suggests. Many areas of the
brain are activated in the course of, for
example, visual perception or language
understanding, including areas thought to
be involved with memory and reasoning.
Such phenomena may lend comfort to anti-
modularists. Moreover, on some partition of
the brain, even semantic memory is going
to turn out to be local. The choice of what
to consider a theoretically relevant bound-
ary in the brain depends to some degree on
what one is looking for. And no modularist
would want such a localization to count in
favour of such a central process as semantic
memory counting as modular. In that case
everything would be a module, and the
theory would collapse. Finally, even if there
is relative localization of function, and even
if these locales should be modular in many
senses, the patterns of connectivity in the
brain might be such that so much informa-
tion passes between so many levels of so
many modules that their distinctively
modular character is lost. All of these, of
course, are empirical matters, and all
remain to be settled.
Before closing it is worth considering the
relationship between connectionism and
modularity. For connectionist models of cog-
nitive processes are becoming increasingly
prominent in cognitive science, and it is often
claimed (Churchland, 1986, 1989) that
these models are incompatible with 'clas-
sical' views of cognitive processes including,
presumably, modularist views. Connection-
ism might seem especially threatening to
modularity theory in that connectionist
models typically emphasize the broad con-
nectivity of their networks and the dis-
tributed character of their representations,
whereas modular models typically emphasize
the insular character of peripheral proces-
sing and the localized character of repre-
sentations and the processes operating
thereon. But though the general outlooks
that motivate connectionism and modularity
may be at odds, it would be incorrect to con-
clude that connectionist models are inher-
ently non-modular. As Tannenhaus, Dell
and Carlson (1989) have shown, it is quite
possible to construct modular connectionist
networks. And it could well turn out that the
arguments and programmes of modularists
and connectionists could converge in a
model of a connectionist but modular cogni-
tive architecture in which distinct regions of
a vast network are dedicated to specific
cognitive tasks and are only weakly con-
nected to other regions, save for strong local
'output' or 'input' connections.
This survey in no way exhausts the litera-
ture or debates concerning modularity, and
should suggest no correct answer to any
specific modularity debate, let alone to the
debate as a whole. Most if not all of the inter-
esting questions are empirical, and the
current state of our understanding of human
cognition leaves all of them unsettled.
See also COMPUTATIONAL MODELS.
BIBLIOGRAPHY
Altmann, G.T. 1986. Reference and the resolu-
tion of local syntactic ambiguity: The effect
of context during human sentences proces-
sing. Ph.D. Dissertation, University of Edin-
burgh.
MODULARITY
Arbib, M.A. 1972. The Metaphorical Brain: An
Introduction to Cybernetics as Artificial Intelli-
gence and Brain Theory. Wiley-Interscience.
Arbib, M.A. 1989. Modularity and interaction
of brain regions underlying visoumotor
coordination. In J .L. Garfield 1989.
Arbib, M.A., and A.R. Hanson, eds. 1987.
Vision, Brain, and Cooperative Computation.
Arbib, M.A., C.C. Boylis, and P. Dev. 1974.
Neural models of spatial perception and the
control of movement. In Keidel et al.
Chase, W.G., and H.A. Simon. 1973. The
Mind's eye in chess. In Chase 1973.
Churchland, Paul M. 1989. A Neurocomputa-
tional Perspective: The Nature of Mind and the
Structure of Science. Cambridge, MA. MIT
Press.
Churchland, P.S. 1986. Neurophilosophy
Toward a Unified Science of the Mind- Brain.
de Groot, Andriaan D. 1966. Perception and
memory versus thought: Some old ideas and
recent findings. In Problem SolVing, ed. B.
Kleinmuntz.
Fodor, J.A. 1975. The Language of Thought.
Crowell.
Fodor, J.A. 1979. Superstrategy. In Cooper
and Walker.
Fodor, J.A. 1981. Methodological solipsism
considered as a research strategy in psy-
chology. In Representation: Philosophical
Essays on the Foundations of Cognitive Science,
ed. J. A. Fodor. Cambridge, MA. MIT Press.
Fodor, J.A. 1983. The Modularity of Mind.
Fodor, J.A. 1989. Modules, frames, fridgeons,
sleeping dogs, and the music of the spheres.
In J. L. Garfield 1989.
Fodor, J.A., T.G. Bever. 1965. The psychologi-
cal reality of lingUistic segments. Journal of
Verbal Learning and Verbal Behavior, 4, 414-
20.
Fodor, J.A., and 1. Sag. 1982. Referential and
quantificational indefinites. Linguistics and
Philosophy,S, 344-89.
Fodor, J.A., T.G. Bever, and M.F. Garret. 1974.
The Psychology of Language: An Introduction
to Psycholinguistics and Generative Grammar,
McGraw-Hill.
Forster, K.1. 1980. Absence of lexical and
orthographic effects in a same-different task.
Memory and Cognition, 8, 210-15.
447
MODULARITY
Forster, K.1. 1981. Priming and the effects of
sentence and lexical contexts on naming
time: Evidence for autonomous lexical pro-
cessing. Quarterly Journal of Experimental
Psychology, 33, 465-95.
Garfield, J.L. 1989. The Modularity of Mind.
Jacobs, R.A., Jordan, M.I., and Barton, A.G.
1991. Task decomposition through compe-
tition in a modular connectionist archi-
tecture: The what and where vision tasks.
Cognition, 15, 195-212.
Karmiloff-Smith, A. 1979. A functional
approach to child language. Cambridge Uni-
versity Press.
Karmiloff-Smith, A. 1985. Language and cog-
nitive processes from a developmental per-
spective. Language and Cognitive Processes, 1.
61-85.
Karmiloff-Smith, A. 1986. From meta-
processes to conscious access: evidence from
children's drawing. Cognition, 23, 95-147.
Karmiloff-Smith, A. 1990. Constraints on
representational change: evidence from
children's drawing. Cognition, 34, 57-83.
Karmiloff-Smith, A. 1993. Beyond modularity:
A developmental perspective on cognitive
science. Cambridge, MA.: MIT Press.
Marslen-Wilson, W.D. 1975. The Limited
compatibility of linguistics and perceptual
explanations. In CLS Papers from the Para-
session on Functionalism.
Marslen-Wilson, W.D. 1980. Sentence percep-
tion as an interactive parallel process.
Science, 198, 226-8.
Marslen-Wilson, W.D., and A. Welsh. Proces-
sing interactions and lexical access during
world recognition in continuous speech.
Cognitive Psychology, 10, 29-63.
Processing structure of sentence perception,
Nature, 257,784-6.
The temporal structure of spoken language
understanding. Cognition, 8, 1-71.
Against modularity. In J.L. Garfield 1989.
McClelland, J.L. 1988. Connectionist models
and psychological evidence. Journal of
Memory and Language, 27, 429-39.
McClelland, J.L., and Rumelhart, D.E. 1981.
An interactive activation model of context
effects. In letter perception. I: An account of
448
basic findings. Psychological Review, 88,
357-407.
Miller, J.L., ed. 1977. Systems Neuroscience.
Academic Press.
Miller, J.L. 1981. Effects of speaking late on
segmental distinctions. In Ellis, 1985.
Plunkett, K., and Marchman, V. 1990.
Regular and irregular morphology and the
psychological status of rules of grammar. In
Proceedings of the 17th Annual Meeting of the
Berkeley Linguistics Society. Berkeley, CA.:
Berkeley Linguistics Society.
Plunkett, K., and Marchman, V. 1991. U-
multilayered perception: implications for
child language acquisition. Cognition, 38,
43-102.
Rumelhart, D.E., and McClelland, J.L. 1986.
On learning the past tenses of English verbs.
In Parallel distributed Processing: Explorations
in the Microstructure of Cognition. Vol. 2:
Psychological and Biological Models, ed. J. L.
McClelland, D. E. Rumelhart, and the PDP
Research Group. Cambridge MA.: MIT
Rumelhart, D.E., Hinton, G.E., and Williams,
D. 1986. Learning internal representations
by error propagation. In Parallel Distributed
of Cognition. Vol. 1: Foundations, ed. D.E.
Rumelhart, J.L. McClelland, and the PDP
Stillings, N. 1989a. Introduction. In J.L. Gar-
field 1989.
Stillings, N. 1989b. Modularity and nat-
uralism in theory of vision. In J.L. Garfield
1989.
Tanenhaus, M.A., Gary, S. Dell., and Greg
Carlson. 1989. Context effects and lexical
processing: A connectionist approach to
modularity. In J.L. Garfield 1989.
Tversky, A. 1982. Judgement Under Uncer-
tainty: Heuristics and biases. Cambridge
University Press.
JAY L. GARFIELD
monadic see PROPER TY.
multiple realization see FUNCTIONALISM;
IDENTITY THEORIES; SUPER VENIENCE.
N
narrow/broad content see An Essay on
Mind section 2.1.3; CONTENT; EXTERNAL-
ISM/INTERN ALISM.
naturalism Naturalism with respect to
some realm is the view that everything that
exists in that realm, and all those events
that take place in it, are empirically acces-
sible features of the world. Sometimes nat-
uralism is taken to imply that some realm
can be in principle understood by appeal to
the laws and theories of the natural scien-
ces, but one must be careful here since nat-
uralism does not by itself imply anything
about reduction. Historically, 'natural' con-
trasts with 'supernatural', but in the
context of contemporary philosophy of mind
where debate centres around the possibility
of explaining mental phenomena as part of
the natural order, it is the non-natural rather
than the supernatural that is the contrast-
ing notion. The naturalist holds that they
can be so explained, whilst the opponent of
naturalism thinks otherwise, though it is
not intended that opposition to naturalism
commits one to anything supernatural.
'Naturalism' is often used inter-
changeably with 'PHYSICALISM' and 'mate-
rialism', though each of these hints at more
specific doctrines. Thus, 'physicalism' sug-
gests that, among the natural sciences,
there is something especially fundamental
about physics. And 'materialism' has con-
notations going back to eighteenth- and
nineteenth-century views of the world as
essentially made of material particles whose
behaviour is fundamental for explaining
everything else. Moreover, as noted above,
one should not take naturalism in regard to
some realm as committing one to any sort
of reductive explanation of that realm, and
there are such commitments in the use of
'physicalism' and 'materialism'.
See also DENNETT; FODOR; LEWIS.
SAMUEL GUTTENPLAN
natural kind The collection or set of
things that are, have been, or will be in
someone's briefcase is a heterogeneous one;
there is probably nothing which the
members of this set have in common except
their presence in the briefcase. In contrast,
there are sets of things that have a more
unified or homogeneous membership, e.g.
the set of typewriters and the set of tigers.
The members of each of these have features
or PROPERTIES in common (or at least over-
lapping) that give the sets their unity and
make it appropriate to call them 'kinds'. To
understand the concept of a natural kind,
one must focus on the difference between
kinds as represented by the set of type-
writers and those as represented by the set
of tigers.
Typewriters and tigers are certainly kinds
of thing, but there is an important differ-
ence. Tigers have evolved to become a
species whose offspring are caused geneti-
cally to resemble one another. In contrast,
typewriters, being an invention of ours, do
not by nature come to share properties. If
anything it is the other way around: we use
our knowledge of nature to turn bits of
metal and plastic into devices that perform
the typewriting function of making our
words visible. Tigers thus form a natural
kind, whereas typewriters form an artificial
kind.
Though tiger is a central case, there are
features of this example which could
mislead someone into thinking of natural
449
NORMATIVE
kinds too restrictively. For one thing, philo-
sophers have spoken of natural kinds where
evolution and genetics have no role: gold is
often cited as a natural kind and the
members of this kind - atoms of gold - cer-
tainly do not have a common biological
background, though they do share certain
physical properties. Yet it is not merely the
sharing of properties - whether biological or
physical - that makes the idea of a natural
kind philosophically important. What is
crucial to the notion is that the shared
properties have an independence from any
particular human way of conceiving of the
members of the kind. Thus, we think of
tigers as having stripes and living in
jungles, and we think of gold as yellow,
malleable and used in making jewellery. But
insofar as each is a natural kind, we must
allow that our usual ways of thinking of
these things might be wrong. For example,
it is intelligible that something belong to the
kind tiger though it lacked stripes - perhaps
we just haven't yet come across tigers like
that; or that gold might be found which
was not yellow and malleable. This is
because what makes something a tiger or
gold is a matter of what is sometimes called
its 'real essence' - that set of properties
shared by members of the kind, whether we
have discovered them or not. Gold is a par-
ticularly good example here because we
have only recently discovered that what is
crucial to this kind is that its members are
atoms with atomic number 79. Before the
atomic theory had been worked out, people
knew that this or that substance was gold,
but they didn't properly know what made
something a member of the kind.
It is this feature of natural kinds that
gives the concept a certain prominence in
philosophy of mind. For example, a
common way of discussing the status of
beliefs is to ask: do beliefs form a natural
kind? Those who answer 'yes' think that
there are discoverable, though not yet dis-
covered, features of beliefs which give them
the required unity. Among the possibilities
here are that beliefs are a certain type of
brain state, a certain type of functional
state, or a state that has evolved to play a
450
certain role in human endeavour. (See
CONTENT; FUNCTIONALISM; IDENTITY THE-
ORIES.) But however one spells it out, the
idea is that, in being a natural kind, belief
has a real essence which could be revealed
by the right kind of research. Those who
demur tend to think of belief more as some-
thing that human beings have invented for
certain purposes, but that does not have
anything like an elusive real essence.
SAMUEL GUTTENPLAN
neurophysiology see P A IN.
normative As a rule, satellites follow ellip-
tical orbits, and, also as a rule, drivers in
the UK keep to the left side of the road. But
there is a big difference between these two
sorts of rule-guided behaviours: satellites are
governed by the rules or laws of motion;
whereas drivers choose to follow the rules of
the road. In this second case, one speaks of
the rules as norms. Moreover, though this is
more controversial, it has been maintained
that one can speak of normativity even
when there is no exercise of explicit choice
in the relevant behaviour. Thus, many lin-
guists, following CHOMSKY, think of our use
of language as a case of rule-following even
though speakers are not usually aware of
the relevant norms.
The notion of normativity, in a slightly
different guise, figures in other areas of the
philosophy of mind. Firstly, it is often said
that the attribution of propositional atti-
tudes is normative, though it may not be
immediately apparent what rules or norms
are at issue in this case. What is meant is
roughly this: there are standards of
RATIONALITY that govern our attribution of
attitudes to each other even though it may
well be impossible to spell these standards
out in terms of specific rules or norms. For
example, it seems plausible that we cannot
attribute beliefs about atoms and electrons
to a child of three. The 'cannot' here marks
the fact that it would not be rationally
explicable or intelligible that such a young
child could have acquired the concepts
necessary to have such beliefs. (See
DAVIDSON.)
Secondly, it has been maintained that the
very possession of CONCEPTS is a normative
matter. The computerized robot which can
sort nuts and bolts by size and shape oper-
ates according to rules. But these rules are
those the programmers have built in - the
robot itself no more aims to conform to
rules than do satellites in earth orbit. For
this reason, though the robot has a dis-
criminative ability in respect of nuts and
bolts, one might well resist the idea that the
robot has the concepts of a nut or a bolt. To
have a concept one has to have the idea
that one is justified in making the relevant
NORMATIVE
discriminations, and such talk of justifica-
tion is of a piece with talk of rationality and
intelligibility - it is a matter of being guided
by rules in a fully normative sense.
The above brief discussion may make it
seem as if there is a clear line between rule-
governed and rule-guided behaviour, but in
fact it is a matter of great controversy how
one is to spell out this difference (see
WITTGENSTEIN). And it is also a matter of
great importance for our understanding of
such things such as language use, concept
possession and the attribution of proposi-
tional attitudes.
SAMUEL GUTTENPLAN
451
o,P
ontology Ontology is the branch of meta-
physics centrally concerned with determin-
ing what there is. (The name comes from the
present participle of the Greek verb corre-
sponding to the English verb 'to be'.) Thus, if
one asks whether there are numbers and
other abstract objects, or whether there are
PROPER TIES, one is asking ontological ques-
tions. Given the fundamental nature of these
questions, ontology plays a part in virtually
all areas of philosophical investigation, but it
has a specific importance to certain debates
within the philosophy of mind. For example,
suppose one agrees that, besides particular
things such as books and tables, there are
also properties of these things (such as being
made of wood or paper) and relations among
them (such as the book's lying on the table).
Allowing properties and relations the same
sort of reality or existence as particular
things is by no means uncontroversial.
Many philosophers think that one should
keep one's ontological commitments to the
minimum, and these philosophers - known
as 'nominalists' - would count only parti-
cular physical objects as onto logically sui-
table. But even if you are willing to accept
properties and relations into your ontology,
it is still a further question whether you
would count, e.g. beliefs as properties of
persons and/or as relations between persons
and belief contents. This sort of question
about belief is ontological, and such ques-
tions figure widely in most areas of philo-
sophy of mind. Discussions of consciousness
and action are often cast as debates about
the ontological status of such things as
pains, sensations of colour, qualia and parti-
cular instances of action. (See An Essay on
Mind section 2.1: DENNETT: FODOR: LEWIS.)
SAMUEL GUTTENPLAN
452
pain (perceptual properties and neural
mechanisms) Pain is an unpleasant
sensory experience that is typically asso-
ciated with bodily injury and/or is described
by people using terms that imply bodily
injury or damage. Pain is virtually universal
amongst humans and probably all other
mammalian species. Though common, the
experience is personal. Furthermore, the
language used to communicate the experi-
ence to others encompasses an enormous
variety of words, phrases, and metaphors.
Despite its subjective nature and complex-
ities, neural and psychological studies have
provided insights into the mechanisms of
pain. In this chapter we will focus upon
what is known of the nature of bodily pain
and the neural mechanisms that are con-
sidered to subserve it.
THE NATURE OF PAIN
Although pain has the general properties of
a sensory experience, it has, in addition,
features beyond sensation that make it both
more complex and of interest to a range of
people other than sensory physiologists. The
most important distinguishing feature of
pain is its affective-motivational aspect. In
contrast to most other sensations, the pain
experience necessarily includes a quality of
unpleasantness and the wish for its immedi-
ate termination. Thus pain is one of the
major forces, along with pleasure, that can
shape behaviour.
THE SUBJECTiVE EXPERIENCE OF BODILY
PAIN
(1) 'Pain' refers to a subjective experience
(see SUBJECTIVITY), an unpleasant sensa-
tion, that is felt in a particular location
within the body. In addition to location, the
other simple sensory properties of pain
include intensity and duration. Location
and intensity may vary with time.
(2) In common with all somatic sensa-
tions, pain has the property of sensory
quality (see QUALIA). Quality is a compound
property that distinguishes a specific type of
pain from non-painful sensations and from
different types of pain. For example, aching
and burning are different qualities. The
quality of a pain is often described in terms
of a stimulus that might elicit it (Le.
burning, pricking or tearing). These terms
often convey the sense of penetration, intru-
sion and assault upon the body. The quality
of a pain is in part determined by the tem-
poral and spatial variation of its primary
properties (e.g. a brief sharp throbbing pain
that radiates into the wrist).
(3) In addition to the intensity of the
stimulus that elicits it, the intensity of per-
ceived pain is influenced by powerful mod-
ifying factors. These factors include the
attention, expectation and state of arousal
of the subject. For example, when two
stimuli are applied simultaneously at differ-
ent sites on the body, one stimulus may
enhance or suppress the sensation resulting
from the other stimulus. The effect of one
stimulus on the sensation evoked by a
second stimulus depends on the proximity
of the two stimuli and their relative inten-
sities (e.g. biting one's lip may ease the pain
of a sprained ankle). Another example is
that identical noxious stimuli, when repeat-
edly applied at the same site, evoke pain
sensations that progressively increase in
intensity and area.
(4) The experience of pain character-
istically includes a negative EMOTION. This
emotion is experienced by human subjects
as the desire to escape, to terminate the
sensation. When the sensation is intense
and/or prolonged or its duration uncertain,
the experience includes anxiety and/or
depression. This negative emotional com-
ponent is called the affective-motivational
dimension of pain to distinguish it from the
sensory-discriminative dimensions described
in (1) and (2) above.
PAIN
(5) The negative effect of pain confers
upon it the power, along with pleasure, to
shape behaviour. This motivational power
assures pain a place of great and unique
importance relative to other sensations.
Obviously, better understanding of learning,
MEMOR Y and the human personality
requires a fuller understanding of pain. The
reverse is also true. Thus pain is a fascinat-
ing object of study not only for neuroscient-
ists but for medical scientists, psychologists,
philosophers and theologians.
(6) As with all other sensory phenomena,
pain has a cognitive-evaluative component.
This component represents both an abstrac-
tion and synthesis of the sensory and affect-
ive dimensions. Thus you might be aware of
a severe pain in your heel which forces you
to stop walking. The cognitive-evaluative
aspect of pain may involve remembering
how far you will have to walk and weigh-
ing the decision to endure the unpleasant
sensations against not getting to work on
time.
This dimension of pain includes its
meaning. In some situations the meaning of
a pain is by far its most important dimen-
sion to the individual. For example, the
development of even mild pain in a patient
being treated for cancer may be terrifying
and depressing if it is believed to signify
recurrence of a malignant tumour.
(7) The sensory, affective and evaluative
dimensions of pain have lawful inter-
relationships. The examples cited in (6)
illustrate how affect can be closely tied to
meaning. In humans, psychological studies
confirm that the affective dimension of pain
can be powerfully reduced or increased by
factors such as psychological set and per-
sonality traits or by manipulations such as
hypnosis or distraction.
(8) Duration is a critical factor that influ-
ences pain. Clinically, the persistence of
pain is associated with profound changes in
the affective and evaluative dimensions of
pain. Whereas acute pain (minutes to
hours and days) is associated with rest-
lessness, arousal and fear, chronic pain
(weeks to months and years) is associated
with resignation, depression, reduced activ-
453
PAIN
ity, and preoccupation with all bodily sen-
sations.
NEURAL MECHANISMS AND
EXPERIMENTAL APPROACHES TO THE
STUDY OF PAIN
The nervous system can be envisaged as the
organ of CONSCIOUSNESS, and its properties
constrain what is accessible to the con-
scious mind. When certain noxious stimuli
are applied to appropriately innervated parts
of the body, a coded message is 'delivered'
to the brain. The result is a series of tissue-
protective responses plus the subjective
experience of pain. In this section we will
briefly outline what is known of the neural
elements involved in generating the differ-
ent dimensions of pain: their location, con-
nectivity and coding properties.
Since a major goal of pain research is to
explain the experience of pain in terms of
the activity of neurons, parallel psycho-
physical and neurophysiological studies
have been very informative. This approach
makes it possible to compare subjective
responses reported by human subjects with
neuronal responses using identical stimuli.
Such an approach allows neuroscientists to
determine how neural information related
to pain is processed and transformed at
various levels of the nervous system.
It is important to point out that pain is
the subjective accompaniment of activity in
specific nervous system pathways. A major
function of activity in these pathways is to
protect the tissues of the body from damage.
Thus, stimuli which elicit the sensation of
pain characteristically produce tissue-pro-
tective responses including withdrawal or
escape from the offending stimulus,
increased blood flow to the affected area and
increased vigilance. Furthermore, such
stimuli produce changes that lead to our
learning to avoid situations similar to those
previously associated with pain.
Primary Afferent Nociceptors, Adequate
Stimuli, Access to the Nervous System
( 1) There are nerve endings in most bodily
tissues that selectively respond to stimuli
454
(e.g. pinching, burning) that usually cause
pain. These nerve endings are called noci-
ceptors. They respond to mechanical distor-
tion, damage, deformation, or intense
thermal stimulation of the tissues they
innervate. Nociceptors also respond to local
increases in the concentration of chemicals
that are produced by tissue damage.
(2) Nociceptors are transducers, which
means that they transform one form of
energy of (Le. thermal, mechanical or
chemical stimuli) to another form, electro-
chemical impulses, which can be trans-
mitted and interpreted by the central
nervous system.
(3) Each nociceptor is connected to a cell
process called the axon through which it
communicates to the central nervous
system. Together, the nerve ending, its axon
and the cell body that supports it are called
the primary afferent nociceptor (see figure
1). In the usual course of events, a noxious
stimulus leads to a transduction event that
produces electrochemical impulses in
primary afferent nociceptors. The electro-
chemical impulses (called spikes), propagate
along the axon of the primary afferent noci-
ceptor from the site of the stimulus to the
central nervous system. In the central
nervous system, the primary afferent noci-
ceptor activates other neurons that are part
of a network whose activity eventuates in
the sensation of pain.
The Criteria for Identification of Neurons and
Pathways Involved in Pain
Neuroscientists studying pain have used a
variety of criteria to identify particular
neurons as candidates for a role in pain
transmission:
(1) They have compared the response of
the neurons to reproducible noxious stimuli
with the response of human subjects to the
same stimulus. For example, one can apply
thermal stimuli of increasing intensity to
the skin of human subjects and have them
report their perceived pain intensity using a
quantifiable rating scale. With this
approach subjects report that perceived pain
intenSity increases exponentially with tem-
peratures above a threshold value of about
PAIN
THE PRIMARY AFFERENT NOCICEPTOR
TRANSDUCTION
impulses
CENTRAL NERVOUS
SYSTEM
Stimulus
----.
----.. --. ----..

"'axon
sensitive
peripheral
terminals
neurotransmitter
release
Figure 1. The primary afferent nociceptor. These nerve cells have three major components. The cell
body, p, which provides all the requisite energy and structural components of the nerve cell. The
axon is a narrow tube which bifurcates, sending one branch into the central nervous system and
another into the bodily tissues. A noxious stimulus activates the nociceptor by a transduction
process (see text) resulting in impulses which propagate along the axon to the central nervous
system where they communicate with second-order neurons by release of neurotransmitter.
45C. In parallel electrophysiological experi-
ments, primary afferent nOciceptor impulse
frequency has been shown to increase with
stimulus intensity in a manner that paral-
lels the human psychophysical intensity
function. This neural-psychological parallel
is a major criterion by which to identify a
neuron that could be involved in pain
transmission.
(2) Imposed (e.g. artificial electrical)
activation of that type of neuron evokes a
sensation identified as painful by human
subjects. For example, electrical stimulation
of the axons of primary afferent nociceptors
in a peripheral nerve reliably elicits reports
of pain in human subjects. Conversely,
blocking primary afferent nociceptors blocks
pain (for example, when a dentist blocks a
nerve with a local anaesthetic).
(3) The neurons in question have appro-
priate anatomical connections for a role in
pain. In other words, given our under-
standing of the anatomy of pain pathways.
a neuron should fulfil criteria (1) and (2)
above and send its axon to a region that
has been implicated in pain transmission.
(4) Manipulations that reduce perceived
pain intensity also reduce the response of
these neurons to a constant noxious stimu-
lus. For example. morphine. which is a
powerful pain reliever. reliably inhibits the
activity of central nervous system neurons
that fulfil criteria (1)-(3). Thus the inhibi-
tion of a neuron by an analgesic dose of
morphine can be used as a criterion to
implicate the neuron in pain transmission.
Central Nervous System Pathways for Pain
The sensory neurons that innervate bodily
tissues. that is. the primary afferent noci-
ceptors. have been described above. Under
typical conditions. tissue injury sufficient to
cause pain will produce activity in a popu-
lation of primary afferent nociceptors. The
particular population of active nociceptors
will be determined by the stimulus type (e.g.
a burn). its intensity. duration and location.
The coding properties will be discussed in
more detail below; however. it is obvious
that this population of primary afferent
nociceptors will generate a wave of impulses
that is conducted through the peripheral
nerves to the central nervous system.
Each primary afferent nociceptor is con-
nected to many neurons in the central
nervous system at structures called
455
PAIN
SYNAPTIC TRANSMISSION
G)
---+- ----.. ----
- - - - - - - - - - ~ ~ - . - - ~ ~ ~ ~ : :
Primary
afferent
nociceptor
Central Nervous System
neurotransmit er /
release
Second order neuron
Figure 2. Synaptic transmission. Impulses invading the central nervous system terminals of primary
afferent nociceptors cause release of neurotransmitter which diffuses across the synaptic cleft to
activate the second-order neuron.
synapses. At each synapse. the axon of the
primary afferent nociceptor terminates on a
secondary neuron. As the electrochemical
impulse in the afferent reaches the termina-
tion of the axon. it causes the release of
molecules called neurotransmitters from the
terminal. These neurotransmitters diffuse
across the narrow synaptic gap to activate
the neuron contacted by the terminal (see
figure 2).
Each central neuron receives multiple
contacts from a variety of primary afferents.
including. but not limited to. nociceptors.
Furthermore. the neurons that receive input
from primary afferents also receive signi-
ficant input from other cells in the central
nervous system. Some of these inputs cause
an excitatory response. other inputs cause
inhibitory responses. The major point is that
the second-order neuron contacted by the
primary afferent nociceptor does not simply
relay the message it receives. The second-
order neuron performs an integrative trans-
formation of inputs from multiple primary
afferent nociceptors and from other central
nervous system neurons (see figure 3).
When the balance between its excitatory
and inhibitory inputs is positive the central
456
neuron is activated and begins to generate
impulses that propagate along its axon to its
terminals. which contact other central
neurons both locally and in other regions of
the central nervous system.
Clearly. the barrage of primary afferent
impulses entering the central nervous
system initiates a cascade of activity in suc-
cessive neuronal relays. Shortly after a
noxious stimulus. there is a distinct spatio-
temporal pattern of activated (and inhibited)
neurons throughout the nervous system.
Somehow. this pattern of activity gives rise
to the subjective experience of pain.
Although the pattern of activity elicited by
a noxious stimulus is widespread in the
central nervous system. the distribution is
discrete and has important regularities. For
example. the axons of primary afferent noci-
ceptors activated by a noxious stimulus to
the leg enter the central nervous system at
the part of the spinal cord that lies in the
lower part of the spinal canal. In human
beings. this part of the spinal cord lies in a
canal within the vertebral column of the
lower back. The axons of the spinal cord
neurons which conduct the information
about noxious stimuli to particular destin a-
CONVERGENCE, DIVERGENCE
AND MODULATION
2
modulatory
neuron
PAIN
To
Perceptual
Centers
peripheral nervous system
central nervous system
pain transmission neurons
Figure 3. Convergence, divergence and modulation. Primary afferent nociceptor 1 branches to supply
several second-order neurons. This is divergence. Primary afferents 1 and 2 converge on two of the
three second-order cells. The axons of second-order cells are gathered in a fascicle called a tract
which ascends to perceptual centres. Second-order neurons are subject to input from modulatory
neurons which control their response to primary afferent nociceptors.
tions in the brain are gathered together in a
circumscribed bundle (figure 3). Such a
spatially discrete bundle ofaxons of common
origin and destination within the central
nervous system is called a tract. Most of the
axons in this 'pain' tract ascend to terminate
in a restricted region deep in the cerebral
hemisphere called the thalamus. The regions
of the thalamus that receive these terminals
are each reciprocally connected to a discrete
region of the cerebral cortex. Damage to this
pathway anywhere between the spinal cord
and the thalamus can impair a person's
ability to detect stimuli that produce pain.
Furthermore, artifically imposed electrical
stimulation of the pathway can induce pain-
like sensations that the subject reports to be
located in a particular body part (see the fol-
lowing section, para. (2) for a discussion of
projected sensation).
There are limits to what objective meas-
urement can tell us about the pain expe-
rience. Thus, although we can determine by
experiment which parts of the nervous
system are necessary for a subject to identify
a stimulus as painful and to gauge its inten-
sity, it is not clear that it is possible to deter-
mine the site or sites in the brain where the
subjective experience actually 'occurs'. In
fact it is arguable whether a subjective
experience has a spatial location. What we
can determine is how neural networks iden-
tify stimuli as noxious and communicate
this information to other neurons.
How Neurons Encode Information Related to
Pain Sensation
As discussed above, neuronal communica-
tion has two major components, impulses
and synaptic transmission. In the central
nervous system neurons are chemically
activated by synaptic input, the information
is then encoded in impulses which propa-
gate along the axon of the neuron to a
synapse upon the next neuron in the
pathway. Neurons have limited possibilities
for encoding information; these include fre-
quency and duration of discharge, the
number of neurons activated and the par-
ticular subset of neurons activated.
457
PAIN
(1) Coding of pain intensity. The simplest
neuronal code is that of impulse frequency.
As illustrated above in the example of the
primary afferent nociceptor. stimulus inten-
sity is clearly encoded by impulse frequency.
In fact. many studies have shown that fre-
quency coding of stimulus intensity is a
general characteristic of neurons implicated
in transmitting information about pain. In
addition. there is evidence that intensity
may be encoded by the number of nocicep-
tive neurons activated at each successive
level along the pain pathway.
(2) Coding of pain location. projected
sensation. The location of a noxious stimu-
lus is encoded by the location of the
neurons that it activates. At each level of
the pain pathway. neurons are arranged in
an orderly topographical representation
(map) of the body. Thus noxious stimuli to
adjacent body areas activate neurons in
adjacent regions of each representation.
There is direct evidence that these maps
encode information relevant to the sub-
jectively experienced location of the sensa-
tion. Thus. in awake human subjects.
artificial electrical stimulation at a brain site
corresponding to the neural representation
of a body region (e.g. the hand area of the
cerebral cortex) elicits a sensation that is
reported by the subject to be located in that
body region. These experiments illustrate
quite clearly the phenomenon of projected
sensation. They demonstrate that activity in
a specific brain location is sufficient to
produce a sensation that is perceived as
occurring at a specific location in the
body. No actual stimulus (or injury) to the
body part is required. In fact. the presence
of the body part itself is not required.
Following amputation of a limb. virtually
all amputees report having the sensation
that the limb is still present. even though
they can see that it is no longer present.
Frequently. patients with pain in a limb
before amputation continue to experience
pain after the limb is amputated. This
'phantom limb' pain is notoriously difficult
to relieve.
(3) Coding of pain quality. The quality of
a somatic sensation is a complex function
458
related to stimulus intensity and location
and the temporal variation of these para-
meters. Pain quality will be encoded by the
temporal pattern of activity produced in the
relevant set of neurons by the stimulus. It is
important to emphasize that it is not just
the nociceptors that contribute to the
quality of pain. An intense stimulus will
activate receptors that respond to light
stimuli (e.g. touch receptors) in addition to
receptors that only respond to noxious
stimuli (e.g. nociceptors). Thus. stimuli that
produce pain activate neurons that in
themselves do not encode 'painfulness' but
that do contribute an aspect of the quality
of the sensory experience that is labelled
'pain' by the subject.
The Modulation of Pain Transmission
The experience of pain is characteristically
variable. Thus. stimuli that are virtually
identical may be reported as painful in one
situation and innocuous in another situ-
ation. We now know that a significant part
of this variability is due to activity in spe-
cific neuronal networks that can selectively
control pain transmission neurons. The
activity of these pain modulating networks
partly depends on the subject's expectation.
focus of attention and level of arousal. Thus
fear. anger or heightened awareness may
change the perceived intensity of pain pro-
duced by an injury. The point is that the
brain does not respond passively to informa-
tion about stimuli. it actively controls its
own input through these modulatory net-
works (see figure 3).
One of the best known of the modulatory
networks controls neurons that receive
direct connections from primary afferent
nociceptors. This modulating network is
activated by a certain class of pain-relieving
drugs. the narcotics (e.g. morphine and
codeine). It has also been shown that endo-
genous morphine-like compounds (e.g.
endorphins) are present in many of the
neurons of this pain-modulating network. It
seems likely that narcotics relieve pain at
least partly by mimicking the action of
endogenous opioids. Interestingly. one of the
most reliable activators for this endorphin-
mediated pain-modulating network is pro-
longed, inescapable pain.
It is also important to point out that there
is good evidence that modulating networks
can enhance pain perception under some
circumstances. It is thus possible in theory
for modulating networks to generate a pain
signal in the absence of an intense stimulus
to the body.
SUMMARY
In this chapter we have discussed the
nature of pain from the neurobiological and
psychophysical perspective with an empha-
sis on the neural mechanisms underlying
pain perception. For the sake of clarity and
brevity we have focused on the sensory
aspects of pain because this is what is
subject to the most straightforward
mechanistic analysis. We have said very
little about the neurological basis of the
affective-motivational aspect of pain, which
is an essential feature. Even less is known
about those distinctively human evaluative
aspects of pain that deal with its meanings
and the suffering associated with them.
These are all crucial and characteristic parts
of the pain experience, however, and their
study requires an experiential approach that
is distinctly different from the methods cur-
rently used to study the purely sensory
aspects of pain. As our conceptual frame-
work and methodological sophistication
grow, so will our understanding of these
complex phenomena.
BIBLIOGRAPHY
Fields, H.L. 1987. Pain. New York: McGraw-
Hill.
Price, D.D. 1988. Psychological and neural
mechanisms of pain. New York: Raven Press.
Wall, P.D., and Melzack, R. 1994. Textbook of
Pain, 2nd Edn. Churchill Livingstone.
HOW ARD L. FIELDS
DONALD D. PRICE
perception Philosophical issues about
perception tend to be issues specifically
PERCEPTION
about sense-perception. In English (and the
same is true of comparable terms in many
other languages) the term 'perception' has
a wider connotation than anything that has
to do with the senses and sense-organs,
though it generally involves the idea of
what may imply, if only in a metaphorical
sense, a point of view. Thus it is now
increasingly common for news-commenta-
tors, for example, to speak of people's per-
ception of a certain set of events, even
though those people have not been wit-
nesses of them. In one sense, however, there
is nothing new about this; in seventeenth-
and eighteenth-century philosophical usage,
words for perception were used with a
much wider coverage than sense-perception
alone. It is, however, sense-perception that
has typically raised the largest and most
obvious philosophical problems.
Such problems may be said to fall into
two categories. There are, first, the epistem-
ological problems about the role of sense-
perception in connection with the acquisi-
tion and possession of knowledge of the
world around us. These problems - does
perception give us knowledge of the so-
called 'external world'?, how and to what
extent? - have become dominant in
epistemology since Descartes because of his
invocation of the method of doubt, although
they undoubtedly existed in philosophers'
minds in one way or another before that. In
early and middle twentieth-century Anglo-
Saxon philosophy such problems centred on
the question whether there are firm data
provided by the senses - so-called sense-
data - and if so what is the relation of such
sense-data to so-called material objects.
Such problems are not essentially problems
for the philosophy of mind, although certain
answers to questions about perception
which undoubtedly belong to the philo-
sophy of mind can certainly add to epistem-
ological difficulties. If perception is
assimilated, for example, to SENSATION,
there is an obvious temptation to think that
in perception we are restricted, at any rate
initially, to the contents of our own minds.
The second category of problems about
perception - those that fall directly under
459
PERCEPTION
the heading of the philosophy of mind - are
thus in a sense prior to the problems that
exercised many empiricists in the first half
of this century. They are problems about
how perception is to be construed and how
it relates to a number of other aspects of the
mind's functioning - sensation, concepts
and other things involved in our under-
standing of things, BELIEF and judgment,
the IMAGINATION, our ACTION in relation
to the world around us, and the causal pro-
cesses involved in the physics, biology and
psychology of perception. Some of the last
were central to the considerations that
Aristotle raised about perception in his De
Anima.
CONSCIOUSNESS
It is obvious enough that sense-perception
involves some kind of stimulation of sense-
organs by stimuli that are themselves the
product of physical processes, and that sub-
sequent processes which are biological in
character are then initiated. Moreover, only
if the organism in which this takes place is
adapted to such stimulation can perception
ensue. Aristotle had something to say about
such matters, but it was evident to him that
such an account was insufficient to explain
what perception itself is. It might be
thought that the most obvious thing that is
missing in such an account is some refer-
ence to CONSCIOUSNESS. But while it may
be the case that perception can take place
only in creatures that have consciousness in
some sense, it is not clear that every case of
perception directly involves consciousness.
There is such a thing as unconscious per-
ception and psychologists have recently
drawn attention to the phenomenon which
is described as 'blindsight' - an ability, gen-
erally manifested in patients with certain
kinds of brain-damage, to discriminate
sources of light, even when the people con-
cerned have no consciousness of the lights
and think that they are guessing about
them. It is important, then, not to confuse
the plausible claim that perception can take
place only in conscious beings with the less
plausible claim that perception always
460
involves consciousness of objects. A similar
point may apply to the relation of percep-
tion to some of the other items I mentioned
earlier, e.g. concept-possession.
SENSATION
Historically, it has been most common to
assimilate perception to sensation on the
one hand and judgment on the other. The
temptation to assimilate it to sensation
arises from the fact that perception involves
the stimulation of an organ and seems to
that extent passive in nature. The tempta-
tion to assimilate it to judgment on the
other hand arises from the fact that we can
be said to perceive not just objects but that
certain things hold good of them, so that
the findings, so to speak, of perception may
have a propositional character (Dretske,
1969, ch. 2). But to have a sensation, such
as that of a pain, by no means entails per-
ceiving anything or indeed having aware-
ness of anything apart from itself. Moreover,
while in looking out of the window we may
perceive (see) that the sun is shining, this
may involve no explicit judgment on our
part, even if it gives rise to a belief, and
sometimes knowledge. (Indeed, if 'see that'
is taken literally, seeing-that always implies
knowledge; to see that something is the case
is already to apprehend, and thus know,
that it is so.)
The point about sensation was made
admirably clear by Thomas Reid in the
eighteenth century (Reid, 1941, Essays 1
and 2). Reid said that sensation involved an
act of mind 'that hath no object distinct
from the act itself'. Perception, by contrast,
involved according to Reid a 'conception or
notion of the object perceived', and a
'strong and irresistible conviction and belief
of its present existence', which, moreover,
are 'immediate, and not the effect of reason-
ing'. Reid also thought that perceptions are
generally accompanied by sensations and
offered a complex account of the relations
between the two. Whether all this is correct
in every detail need not worry us at present,
although it is fairly clear that perceiving
need not be believing. Certain illusions,
such as the Miiller-Lyer illusion, are such
that we may see them in a certain way, no
matter what our beliefs may be about them
or whether we have any beliefs about them.
Once again, however, it is arguable that
such (mis)perceptions could only take place
in believers, whether or not beliefs about
the objects in question occur in the actual
perception.
CONCEPTS AND ACTIVITY
Similar considerations apply to concept-
possession (Reid's 'conception or notion').
(See CONCEPTS.) It is certainly not the case
that in order to perceive a cyclotron I must
have the (or a) concept of a cyclotron; I
may have no idea of what I am perceiving,
except of course that it is something. But to
be something it must have some distin-
guishable characteristics and must stand in
some relation to other objects, including
whatever it is that constitutes the back-
ground against which it is perceived. In
order to perceive it I must therefore have
some understanding of the world in which
such objects are to be found. That will, in
the case of most if not all of our senses, be a
spatial world in which things persist or
change over time. Hence, perception of
objects presupposes forms of awareness that
are spatiotemporal. It is at least arguable
that that framework would not be available
were we not active creatures who are
capable of moving about in the world in
which we live (Hamlyn, 1990, ch. 5). Once
again, it is not that every perception
involves some activity on our part,
although some may do so, but that percep-
tion can take place only in active creatures,
and is to that extent, if only that extent, not
a purely passive process.
It must be evident in all this how far we
are getting from the idea that perception is
simply a matter of the stimulation of our
sense-organs. It may be replied that it has
long been clear that there must be some
interaction between what is brought about
by stimulation of sense-organs and sub-
sequent neural. including cortical. pro-
cesses. That, however, does not end the
PERCEPTION
problem, since we are now left with the
question of the relation between all that and
the story about sensations, beliefs, concepts
and activity that I have mentioned. Some of
that issue is part of the general MIND-BODY
PRO B L EM, but there is also the more specific
problem of how these 'mental' items are to
be construed in such a way as to have any
kind of relation to what are apparently the
purely passive causal processes involved in
and set up by the stimulation of sense-
organs.
INFORMATION-PROCESSING AND
REPRESENT ATIONS
One idea that has in recent times been
thought by many philosophers and psycho-
logists alike to offer promise in that con-
nection is the idea that perception can be
thought of as a species of information-
processing, in which the stimulation of the
sense-organs constitutes an input to sub-
sequent processing, presumably of a com-
putational form. The psychologist J. J.
Gibson suggested that the senses should be
construed as systems the function of which
is to derive information from the stimulus-
array, indeed to 'hunt for' such information
(Gibson, 1966). He thought, however, that
it was enough for a satisfactory psycho-
logical theory of perception that his
account should be restricted to the details
of such information pick-up, without refer-
ence to other 'inner' processes such as
concept-use. Although Gibson has been
very influential in turning psychology
away from the previously dominant sensa-
tion-based framework of ideas (of which
gestalt psychology was really a special
case), his claim that reliance on his notion
of information is enough has seemed
incredible to many. Moreover, his notion of
'information' is sufficiently close to the
ordinary one to warrant the accusation
that it presupposes the very ideas of, for
example, concept-possession and belief that
he claimed to exclude. The idea of informa-
tion espoused by him (though it has to be
said that this claim has been disputed) is
that of 'information about', not the techni-
461
PERCEPTION
cal one involved in information theory or
that presupposed by the theory of computa-
tion.
The most influential psychological theory
of perception has in consequence been that
of David Marr, who has explicitly adopted
the 'computational metaphor' in a fairly
literal way (Marr, 1982). He distinguished
three levels of analysis: (1) the description of
the abstract computational theory involved,
(2) the account of the implementation of
that theory in terms of its appropriate algo-
rithm, and (3) the account of the physical
realization of the theory and its algorithm.
All this is based on the idea that the senses
when stimulated provide representations on
which the computational processes can
work (see COMPUTATIONAL MODELS;
REPRESENT ATION). Other theorists have
offered analogous accounts, if differing in
detail. Perhaps the most crucial idea in all
this is the one about representations. There
is perhaps a sense in which what happens
at, say, the level of the retina constitutes, as
a result of the processes occurring in the
process of stimulation, some kind of repre-
sentation of what produces that stimulation,
and thus some kind of representation of the
objects of perception. Or so it may seem if
one attempts to describe the relation
between the structure and character of the
objects of perception and the structure and
nature of the retinal processes. One might
indeed say that the nature of that relation is
such as to provide information about the
part of the world perceived, in the sense of
'information' presupposed when one says
that the rings in the sectioning of a tree's
trunk provide information of its age. This is
because there is an appropriate causal rela-
tion between the two things, which makes
it impossible for it to be a matter of chance.
Subsequent processing can then be thought
to be one carried out on what is provided in
the representations in question.
One needs to be careful here, however. If
there are such representations, they are not
representations for the perceiver. Indeed it is
the thought that perception involves rep-
resentations of that kind which produced
the old, and now largely discredited,
462
philosophical theories of perception which
suggested that perception is a matter, pri-
marily, of an apprehension of mental states
of some kind (e.g. sense-data) which are
representatives of perceptual objects, either
by being caused by them or in being in
some way constitutive of them. Also, if it be
said that the idea of information so invoked
indicates that there is a sense in which the
processes of stimulation can be said to have
content, but a non-conceptual content
(Evans, 1982, chs 5 and 6; Peacocke, 1983,
ch. I), distinct from the content provided by
the subsumption of what is perceived under
concepts, it must be emphasized that that
content is not one for the perceiver. What
the information-processing story provides is,
at best, a more adequate categorization
than previously available of the causal pro-
cesses involved. That may be important but
more should not be claimed for it than there
is. If in perception in a given case one can
be said to have an experience as of an
object of a certain shape and kind related to
another object it is only because there is
presupposed in that perception the posses-
sion of concepts of objects, and more par-
ticularly, a concept of space and how
objects occupy space.
THE REQUISITES OF PERCEPTION AND
THE NATURE OF EXPERIENCE
Perception is always concept-dependent at
least in the sense that perceivers must be
concept possessors and users, and almost
certainly in the sense that perception entails
concept-use in its application to objects. It is
at least arguable that those organisms that
react in a biologically useful way to some-
thing but that are such that the attribution
of concepts to them is implausible, should
not be said to perceive those objects,
however much the objects figure causally in
their behaviour. Moreover, in spite of what
was said earlier about unconscious percep-
tion and blindsight, perception normally
involves consciousness of objects. Moreover,
that consciousness presents the objects in
such a way that the experience has a
certain phenomenal character, which is
derived from the sensations which the
causal processes involved set up. This is
most evident in the case of touch (which
being a 'contact sense' provides a more
obvious occasion for speaking of sensations
than do 'distance senses' such as sight). Our
tactual awareness of the texture of a surface
is, to use a metaphor, 'coloured' by the
nature of the sensations that the surface
produces in our skin, and which we can be
explicitly aware of if our attention is drawn
to them (something that gives one indica-
tion of how attention too is involved in per-
ception).
It has been argued (Millar, 1991b) that
the phenomenal character of an experience
is detachable from its conceptual content in
the sense that an experience of the same
phenomenal character could occur even if
the appropriate concepts were not available.
Certainly the reverse is true - that a
concept-mediated awareness of an object
could occur without any sensation-
mediated experiences - as in an awareness
of something absent from us. It is also the
case, however, that the look of something
can be completely changed by the realiza-
tion that it can be thought of in a certain
way, so that it is to be seen as X rather than
Y. To the extent that that is so, the phe-
nomenal character of a perceptual experi-
ence should be viewed as the result of the
way in which sensations produced in us by
objects blend with our ways of thinking of
and understanding those objects (which, it
should be noted, are things in the world
and should not be confused with the sensa-
tions which they produce).
Seeing things in certain ways also some-
times involves the imagination (Wittgen-
stein, 1953, II, xi) (and in what is perhaps a
special sense of 'imagination' (Strawson,
1974, ch.3) it perhaps always does). In
imagination we may bring to bear a way of
thinking about an object which may not be
the immediately obvious one, and being
visually imaginative, as an artist may have
to be, is at least a special case of our general
ability to see things as such and suches. But
that general ability is central to the faculty
of visual perception and, mutatis mutandis of
PERCEPTUAL CONTENT
the faculty of perception in general. What
has been said may be enough to indicate
the complexities of the notion of perception
and how many different phenomena have
to be taken into consideration in elucidating
that notion within the philosophy of mind.
But the crucial issue, perhaps, is how they
are all to be fitted together within what
may still be called the 'workings of the
mind'.
See also COGNITIVE PSYCHOLOGY; CONTENT;
INTENTIONALITY; PERCEPTUAL CONTENT;
PSYCHOLOGY AND PHILOSOPHY.
BIBLIOGRAPHY
Dretske, F. 1969. Seeing and Knowing London:
Evans, G. 1982. Varieties of Reference. Oxford:
Clarendon Press.
Gibson, J.J. 1966. The Senses Considered as Per-
ceptual Systems. Boston, MA.: Houghton
Mifflin.
Hamlyn, D.W. 1990. In and Out of the Black
Box. Oxford: Basil Blackwell.
Marr, D. 1982. Vision. San Francisco: W. H.
Freeman.
Millar, A. 1991a. Concepts, experience and
inference. Mind, C. 4, 495-505.
Millar, A. 1991b. Reasons and Experience.
Peacocke, C. 1983. Sense and Content. Oxford:
Clarendon Press.
Strawson, P.F. 1974. Freedom and Resentment.
London: Methuen.
Reid, T. 1941. Essays on the Intellectual Powers
of Man, ed. A. D. Woozley. London: Mac-
millan. Originally published 1785.
D. W. HAMLYN
perceptual content Where the term
'content' was once associated with the
phrase 'content of consciousness' to pick
out the subjective aspects of mental states,
its use in the phrase 'perceptual content' is
intended to pick out something more closely
akin to its old dual 'form', the objective and
publicly expressible aspects of mental states.
463
PERCEPTUAL CONTENT
There are nevertheless important links
between these diverse uses. We might call a
theory which attributes to perceptual states
a content in the new sense, 'an intentional
theory' of perception. On such a view, per-
ceptual states represent to the subject how
her environment and body are. The content
of perceptual experiences is how the world
is represented to be. Perceptual experiences
are then counted as illusory or veridical
depending on whether the content is correct
and the world is as represented. In as much
as such a theory of perception can be taken
to be answering the more traditional pro-
blems of perception, it will be addressing the
same issues as led earlier writers to discuss
the content of consciousness. In the first
part of this essay, I raise the question of
how a theory of perceptual content may be
seen to answer the traditional problems; in
the second, I look to some of the debates
about the nature of content in the light of
this answer.
APPEARANCE AND CONTENT
Why attribute an intentional content to
perceptual states at all? To see perceptual
states as intentional is to assimilate them
to other mental phenomena which are
paradigm examples of INTENTIONALITY:
BELIEFS and DESIRES, and also occurrent
mental episodes, such as acts of judgment
and consciously entertaining a thought.
The basic justification for thinking of per-
ceptual states as intentional will be the
same as that for these other mental states:
their general functional role within our
mental economies - that is, how they inter-
act with other mental states, in particular
in the fixation of belief and control of
ACTION - can only be understood in terms
of attributing to them such content. A
subject gains information about the state of
her body and her immediate environment
through perception. We can explain why
the subject's perceptual beliefs have the
content that they do, and why her inten-
tional actions are performed in the way that
they are, by positing a corresponding
content to the perceptual states which give
464
rise to those beliefs and control those
actions.
This idea contrasts with the once domi-
nant sense-datum theory of perception. This
is often defined solely in terms of the objects
of perception: as the claim that there are
certain mental entities, sense data, whose
existence and qualitative nature depend
upon a subject's awareness of them, and
which are the only immediate objects of
perception (cf. Jackson, 1977; Perkins,
1983). Yet this traditional view has impli-
cations not only for an account of the
objects of perceptual states, but also for that
of their nature and how they relate to other
mental states. A sense-datum theory
explains the phenomenological character of
perceptual states in terms of the properties
that these mental entities have: on this
view, how things appear to the subject will
correspond to how these entities are. If the
view seeks to give an account of perceptual
experience purely in terms of the immediate
awareness of sense data, it must hold that
the phenomenological character of percep-
tual experience can be completely described,
if at all, in terms independent of the sub-
ject's immediate, physical environment.
That consequence seems at odds with the
introspectible evidence of what it is like for
us to perceive (see INTROSPECTION), as has
been noted by many critics of sense-datum
theories (cf. Strawson, 1979). When you
look at the page of the book in front of you,
the scene before your eyes looks some way
to you. The most natural way to describe
what it is like to be looking at the book,
introspecting one's current experience, is to
employ the very same vocabulary as one
would use to describe the scene perceived.
One might put this point by saying that the
experience strikes one as being of the book
in front of one or, allowing for the fact that
one may be deluded as to the nature of
what is there, that it is at least as if such a
book is there - as of a book, in shorthand -
but certainly not of any mental surrogate of
that book. Not only is this the most natural
way to describe the experience, no adequate
description of what the experience is like for
one could leave this out.
In turn, it is difficult to see how, on a
purely sense-datum theory of perception,
the immediate awareness of mental entities
alone could explain the fixation of beliefs
about the environment or actions upon that
environment (see CONTENT; EXTERNALISM/
INTERNALISM). Such experiences could only
figure in the rational production of such
beliefs and actions against a background
of beliefs of the subject which link the
presence to her of a mental entity with the
presence in her environment of some object,
such as beliefs about what normally causes
such mental states. It is questionable
whether all perceivers do have such beliefs,
or that such beliefs are operative in the fixa-
tion of their perceptual beliefs.
In contrast, an intentional theory of per-
ception does not conflict with the intro-
spective claims of common sense and is able
to attribute to perceptual experience an
explanatory role in the fixation of belief and
control of action without appeal to such
sophisticated background beliefs. On such a
view, one's current visual experience may
represent the presence of a book in one's
immediate environment. One's visual
experience will then have a certain phe-
nomenological character in virtue of how it
represents the immediate environment to
be. Where the sense-datum theory claims
that one is immediately aware of mental
entities, the intentional theory can simply
claim that one is aware of the objects in
one's environment, and aware of them in
virtue of the content of one's experience. In
normal circumstances, where perceptual
experience fixes belief, the content of the
experience, which will be how things
appear to the subject, will be matched, in
part, by the content of the belief formed.
While the above expresses a disagreement
between sense-datum theories and inten-
tional theories, there is a further matter on
which they agree. It is common to assume
that the mental state one is in when per-
ceiving, the perceptual experience, is of a
kind which could occur even if one has an
illusion or hallucination. This assumption is
often appealed to in the so-called argument
from illusion. If you could have had the
PERCEPTUAL CONTENT
same state of mind as you now do while
looking at this page even though you were
suffering an hallucination, then it cannot be
essential to your being in such a state of
mind that any object independent of you
should have certain qualities or be perceived
by you. A sense-datum theory assumes,
nevertheless, that there must be some object
with the requisite qualities perceived by you
and hence concludes that the object in
question is a mental one, dependent for its
existence entirely on one's awareness of it.
Commonly intentional theories accept the
same claim about the commonality of per-
ceptual experience, but respond to it in a
radically different way, drawing an analogy
between experience and belief. It is possible
to believe that Tucson is north of Boston
even though that is not so. No one supposes
that because there is no appropriate objec-
tive fact in the world for one's belief to cor-
respond to, it must instead be a relation to
some subjective fact. Instead it is in the
nature of beliefs that they can be true or
false, and the intentional content of the
belief specifies how the world would have to
be for the belief to be true. Similarly, if
perceptual experience has an intentional
content, that experience can be of some
physical state of affairs, even if the state of
affairs in question is not actual.
An alternative response to the problem of
illusion is to reject the assumption shared
by both views above that experience forms
a common kind among perception, illusion
and hallucination. On this view, a report of
one's experience such as 'It looks to me as if
there is a book before me' does not describe
a single kind of state of mind present
whether or not I am perceiving, but would
be made true by a disjunction of distinct
types of mental state - either a state of ver-
idical perception or one which although not
veridical perception is nevertheless indis-
tinguishable from it (cf. McDowell, 1982).
Where the negative component of this view
is to deny that there is a common account
of how things appear to a perceiver across
the three cases - veridical perception, per-
ceptual illusions and hallucinations - the
positive element is to allow an alternative
465
PERCEPTUAL CONTENT
account of the one case, veridical percep-
tion. In contrast to the sense-datum the-
ory's claim that we are only immediately
aware of mental objects, this view will claim
that we immediately perceive physical
objects. To this extent the position will be in
accord with the intentional theory and the
introspective support for it. But it will argu-
ably be at odds with the intentional theory
on another ground. The sense-datum theory
assumes that the objects and qualities in
virtue of which perceptual experience has
its character must exist or be realized in
order for the experience to be so, and this
assumption will likewise be made on this
view: the kinds of physical object and
qualities that one's experience is of, or as of,
must exist or be realized in order for one to
have an experience of the kind one does
when veridic ally perceiving. One's percep-
tual experience having the PHENOMEN-
0LoGIcAL character it does is constituted,
according to this view, by one's physical
environment being so. In contrast to this
the intentional theory claims that the phe-
nomenological character of experience is
constituted by the experience having a
certain intentional content, which content
it could have were the environment not so.
The intentional theory denies that there are
any intermediary objects between us and
the external world of which we are aware,
but it may yet not accommodate the
thought that we are immediately aware of
the external world, if that thought is best
expressed in terms of this third view -
'naIve realism', as we may label it.
NaIve realism is a form of extreme extern-
alism and consequently comes with a high
price. The thought that perceptual experi-
ences form a common kind among the three
cases is seemingly supported by appearances
- what better evidence could there be for
the common nature of the three kinds of
case than that they are indistinguishable to
the subject of those mental states? It is also
supported by facts about the possible
immediate causes and behavioural effects of
perceptual states, since it is plausible that
the same immediate stimulation of the
retina could in one case produce a visual
466
perception of an apple and in another the
hallucination as of such an apple; and plau-
sible to suppose that each state could
produce the same kind of physical move-
ment. Since many philosophers are pre-
pared to endorse some form of externalism
about the mind, the consequence that the
same physiological causes and effects corre-
spond to distinct mental states may not be
too high a price for them. But the phenom-
enological considerations mark the percep-
tual case out from other areas and may be
thought to give specific reasons for resisting
externalism. The issue may then turn on
whether naIve realism can make good its
claims concerning the common-sense
thought that we are immediately aware of
the external world. For if one can only
make sense of the immediacy of experience
in naIve realist terms, then any perceptual
experience which genuinely gives one
immediate access to the external world will
be constituted by the world being so. In as
much as illusions and hallucinations are
indistinguishable from perceptions, they too
will, on this view appear to possess this
property even though they lack it. If naIve
realism is correct about the nature of per-
ceptual experience, at least some perceptual
experiences will be genuinely immediate
while others will merely appear to be so.
However, even if naIve realism is wrong
about the actual nature of experience, its
claims about common sense might still be
correct. It would then be the best account
not of the nature of experience but what
that nature appears to us to be.
The above discussion presents the debate
as one between three contrasting theories.
But it is best to see the fundamental divi-
sions here as concerning the different types
of aspects or features that the conscious
character of perceptual experience can
have. For the intentional theory and the
naIve realist both suppose that experience
has features that are as of objects in the
subject's physical environment which can
exist independently of her, while the sense-
datum theory claims that there are features
of experience individuated by reference to
sense data and their qualities which exist
only in as much as the subject is aware of
them. On the other hand, the sense-datum
theory and the naive realist both claim that
experience has features that require that the
objects or qualities in virtue of which it is
individuated should exist or be realized;
while the intentional theory allows that
experience may have a feature as of some
object or quality, without that existing or
being realized. These differences can be seen
in terms of the answers to two questions
about the relation between features of
experience and the objects or qualities in
terms of which those features are indi-
viduated. Can that object or quality exist or
be realized independently of the experience
having that feature? Can the experience
have that feature independently of the
object or quality existing or being realized?
We can represent this on a matrix:
PERCEPTUAL CONTENT
commitment to sense data but still claim
that there are subjective features of experi-
ence, often called QUALIA, where they take
these to be modifications of the experiential
act itself rather than any object of that act.
An intentional theory of perception, which
ascribes perceptual content to sensory
states, attributes intentional properties to
experience. The naive realist, on the other
hand, supposes that veridical perception
possesses naive realist features. There is a
fourth option on the matrix here left un-
filled. This is not to claim that there are
no such properties of experience, but if we
assume that the argument from illusion
motivates the positing of either intentional
or subjective features, it is not surprising
that no theory would appeal to such qual-
ities. Note also that on some views of inten-
tional content, the view here labelled
Can the object/quality exist/be realized independently of the experiential feature?
Yes No
Intentional Feature ?
Can the experience
Yes have this feature
without the object/
quality existing/
being realized?
Naive Realist Subjective Feature
Feature
There are three types of feature that the
theories of perception are appealing to.
However, an experience possessing one of
these features is not precluded from posses-
sing either of the others. Rather the debate
between the different theories concerns
whether the phenomenological character of
experience can be explained without posit-
ing one or other of these features. Theories
of perception differ according to which
attributes they claim perceptual experience
to have. A sense-datum theory claims that
they have at least some subjective features;
there are other views which seek to avoid
No
'naive realist' will be taken to be a variety of
the intentional theory, and the debate
between them seen as one internal to that
theory.
Intentional theories of perception have
come to be accepted where it is agreed that
we must posit intentional properties of
experience; and alternative approaches have
been rejected either where they are thought
unnecessary or inadequate. The main focus
of debate among intentional theorists has
been over the nature of perceptual content
and the nature of the mental state that has
that content.
467
PERCEPTU AL CONTENT
VARIETIES OF CONTENT
Content and Consciousness
What relation is there between the content
of a perceptual state and conscious experi-
ence? One proponent of an intentional
approach to perception notoriously claims
that it is 'nothing but the acquiring of true
or false beliefs concerning the current state
of the organism's body or environment'
(Armstrong 1968, p. 209). Many critics of
Armstrong have supposed that it is a defect
of Armstrong's own account and even of
any intentional theory that it cannot give
an adequate account of what conscious
experience is like. Beliefs need not be con-
scious states of mind, and there is no reason
to think that the acquiring of beliefs need be
conscious either. Armstrong claims, not
implausibly, that there can be unconscious
perception, but the complaint remains that
he cannot give an adequate account of con-
scious perception, given the 'nothing but'
element of his account. However, an inten-
tional theory of perception need not be allied
with any general theory of CONSCIOUSNESS,
one which explains what the difference is
between conscious and unconscious states.
If it is to provide an alternative to a sense-
datum theory, the theory need only claim
that where experience is conscious, its
content is constitutive, at least in part, of
the phenomenological character of that
experience. This claim is consistent with a
wide variety of theories of consciousness,
even the view that no account can be given.
An intentional theory is also consistent
with either affirming or denying the pre-
sence of subjective features in experience.
Among traditional sense-datum theorists of
experience, H. H. Price (1932) attributed in
addition an intentional content to percep-
tual consciousness. Among recent accounts,
Peacocke (1983, ch. 1) and Shoemaker
(1990) both attribute subjective properties
to experience - in the former case labelled
sensational properties, in the latter qualia -
as well as intentional content. One might
call a theory of perception that insisted that
all features of what an experience is like are
determined by its intentional content, a
468
purely intentional theory of perception.
Harman (1990) is an example of one who
holds such a view.
Experience and Belief
Armstrong not only sought to explain per-
ception without recourse to sense-data or
subjective qualities but also sought to
equate the intentionality of perception with
that of belief. There are two aspects to this:
the first is to suggest that the only attitude
towards a content involved in perceiving is
that of believing, and the second is to claim
that the only content involved in perceiving
is that which a belief may have. The former
suggestion faces an immediate problem,
recognized by Armstrong, of the possibility
of having a perceptual experience without
acquiring the corresponding belief. One
such case is where the subject already pos-
sesses the requisite belief - here Armstrong
talks of perception maintaining, rather than
leading to the acquisition of, belief. The
more problematic case is that of disbelief in
perception, where a subject has a perceptual
experience but refrains from acquiring the
corresponding belief. For example, someone
familiar with the Miiller-Lyer illusion, in
which lines of equal length appear unequal,
is unlikely to acquire the belief that the lines
are unequal on encountering a recognizable
example of the illusion. Despite that, the
lines may still appear unequal to them.
Armstrong seeks to encompass such cases
by talk of dispositions to acquire beliefs and
talk of potentially acquiring beliefs. On his
account this is all we need say of the psy-
chological state enjoyed. However, once we
admit that the disbelieving perceiver still
enjoys a conscious occurrent experience,
characterizing it in terms of a disposition to
acquire a belief seems inadequate. There are
two further worries. One may object that
the content of perceptual experiences may
play a role in explaining why a subject dis-
believes in the first place: someone may fail
to acquire a perceptual belief precisely
because how things appear to her is incon-
sistent with her prior beliefs about the
world. Secondly some philosophers have
claimed that there can be perception
without any corresponding belief. Cases of
disbelief in perception are still examples of
perceptual experience that impinge on
belief; where a sophisticated perceiver does
not acquire the belief that the Miiller-Lyer
lines are unequal, she will still acquire a
belief about how things look to her. Dretske
(1969) argues for a notion of non-epistemic
seeing, on which it is possible for a subject
to be perceiving something while lacking
any belief about it because she has failed to
notice what is apparent to her. If we
assume that such non-epistemic seeing
nevertheless involves conscious experience
it would seem to provide another reason to
reject Armstrong's view and admit that if
perceptual experiences are intentional states
then they are a distinct attitude-type from
that of belief. However, even if one rejects
Armstrong's equation of perceiving with
acquiring beliefs or dispositions to believe,
one may still accept that he is right about
the functional links between experience and
belief, and the authority that experience has
over belief, an authority which can never-
theless be overcome.
CONCEPTUAL AND NON-CONCEPTUAL
CONTENT
Increasingly, proponents of the intentional
theory of perception argue that perceptual
experience is to be differentiated from belief
not only in terms of attitude, but also in
terms of the kind of content that experience
is an attitude towards (see CONTENT).
According to them we need to accept that
there are varieties of content, that we need
to distinguish between what has been called
'conceptual' and 'non-conceptual' content.
It is claimed that one cannot accommodate
the nature of perceptual experience solely in
terms of conceptual content, the only kind
of content that beliefs and other PROPOSI-
TIONAL ATTITUDES possess, but must
ascribe to it a non-conceptual content.
If thought is structured, having a thought
is to exercise the conceptual abilities corre-
sponding to the concepts that compose that
thought (see CONCEPTS). An individual's
possession of a given concept is revealed in
PERCEPTUAL CONTENT
the range of thoughts that she is able to
have that contain that concept (cf. Evans,
1982, ch. 4). The conditions on any of her
conceptual mental states possessing a
content will then derive from the conditions
of her possessing the concepts that compose
that content. If perceptual experiences
possess certain contents where the corre-
sponding conditions on concept-possession
are not met, they thereby possess a non-
conceptual content. The various arguments
used to support the claim that experience
has a non-conceptual content tend to
include two components: the first is to
argue that a given condition on possessing
a concept has not been met by a given indi-
vidual; the second is to give grounds for
nevertheless attributing a given content to
the experience.
For example, it may be argued that only
relatively sophisticated creatures such as
ourselves, who employ language and are
able to think about our own states of mind
genuinely possess concepts, and so have
conceptual thought. On the other hand, it
may be claimed, other animals and infants,
although not sapient, are sentient; and their
experiences, no less than ours, purvey infor-
mation to them and control their beha-
viour. It is not implausible to claim that
there are some ways in which we experi-
ence the world in which they can too. If
that is so, then there is a common content
between some of their mental states and
ours, even though only we are concept-
possessors. It cannot be a constraint on that
content that one possess the requisite con-
cepts, hence that content is non-conceptual
(Evans, 1982, ch. 5; Dretske, 1982, ch. 6).
Another argument derives from the phe-
nomenological character of experience. It is
often claimed that the character of a per-
ceptual experience is much richer and more
fine-grained than is the corresponding pro-
positional attitude to which it may give rise.
Confronted by a beech tree, a subject may
be unable to put into words the specific
outline that she sees the tree to have, or the
various shades its leaves possess. Further-
more, outside of the context of actually
viewing the tree, she may be unable to
469
PERCEPTUAL CONTENT
recognize the same shape again, or to think
about it. Nevertheless she may be able to
discriminate between that particular outline
of the tree or shade of leaf and others that it
could have had. Here it is tempting to claim
that how things look to her outrun the con-
ceptual capacities that she then possesses,
as evidenced by her verbal descriptions of
what is before her, or her later ability to
think about it.
One response to this line of reasoning is
to point out that she is at least able to
attend to and indicate the outline when pre-
sented with it, and think, 'That outline is
particularly striking' - so it is not true that
how things appear to her cannot have a
bearing on her thoughts. This is a nice
issue, and it is not clear that the response
answers the full force of the original intui-
tion. The fact that a perceiver may, through
attending to features of her experience,
come to be able to demonstrate that feature,
or even acquire a recognitional capacity for
it, certainly supports the claim that each
aspect could be matched by a corresponding
concept. That does not yet show that in
order for the perceiver to have an experi-
ence with that content, she must thereby
possess the relevant concept. Rather it
seems more plausible to say that we can
explain the demonstrative concept she pos-
sesses in that context, or the recognitional
capacity she acquires, in terms of the
content of the experience. This would
require us to suppose that the experience
has the content independently of the con-
ceptual capacities she actually possesses.
What forces us, an opponent of this view
might press, to view the extra element as
intentional but non-conceptual, rather than
non-intentional and possibly subjective?
Subjectivists have commonly argued for the
need for some subjective features to experi-
ence on precisely the ground that we need
to recognize the phenomenological differ-
ences between sensing and mere thinking.
Likewise they have often treated this sub-
jective aspect as the 'given' in experience
which plays a part in explaining the acqui-
sition and application of empirical concepts.
Within the present context, the non-con-
470
ceptual content appealed to by intentional
theories seems to have taken over both
roles.
In defence, the proponent of a plurality of
contents may seek support in two direc-
tions. On the one hand, she may insist that
introspection supports the view that even
those aspects of experience that outrun the
subject's current conceptual capacities seem
to be as of the objective world, and do not
seem to be subjective. On the other, the the-
orist may appeal outside of philosophical
debates, to the notions of content at play
within psychological theories of perception
to show the need for notions of content
which are not obviously tied to the subject's
conceptual powers.
Whether this response is finally sufficient
is perhaps undecided. What is clear is that
where the problems of perception are them-
selves ancient, debates about perceptual
content are as yet in their infancy.
BIBLIOGRAPHY
D.M. Armstrong 1968. A Materialist Theory of
the Mind. London: Routledge & Kegan Paul.
T.M. Crane, ed. 1992. The Contents of Experi-
ence. Cambridge University Press.
F. Dretske, 1969. Seeing and Knowing. London:
F. Dretske, 1982 Knowledge and the Flow of
Information: Oxford: Basil Blackwell.
G. Evans 1982. The Varieties of Reference.
G. Harman, 1990. The intrinsic quality of
experience. Philosophical Perspectives, 4, 31-
52.
F. Jackson, 1977, Perception: A Representative
Theory. Cambridge University Press.
J. McDowell, 1982. Criteria, defeasibility and
knowledge. Proceedings of the British
Academy, 68, 455-79.
C.A.B. Peacocke, 1983. Sense and Content.
C.A.B. Peacocke, 1989. Perceptual content. In
Themes from Kaplan, ed J. Almog et al. New
York: Oxford University Press.
C.A.B. Peacocke, 1992. A Study of Concepts.
Cambridge MA: MIT Press.
Perkins, M. 1983. Sensing the World. Indiana-
polis: Hackett Publishing Co.
H.H. Price. 1932. Perception. London:
Methuen.
S. Shoemaker, 1990. Qualities and qualia:
what's in the mind?' Philosophy and Phenom-
enological Research, 50, 109-31.
P.F. Strawson, 1979. Perception and its
objects. In Perception and Identity, ed. G. F.
McDonald. London: Macmillan.
M. G. F. MARTIN
phenomenal/phenomenological Strictly
speaking, a phenomenon is that which is
present to the mind when we exercise our
senses. Put in a slightly different way, it is
whatever is in consciousness when some-
thing is seen, touched, heard, etc. When
one hears it said that scientists have
'observed some phenomenon such as the
passage of a comet', this is a deviation from
the strict usage: the phenomenon is the
appearance to us of the comet and not the
'external' fact of the comet's passage.
In the philosophy of mind, the adjectival
form 'phenomenal' tends to be used in the
strict sense to describe that which passes in
our consciousness in PERCEPTION, but there
is a no consensus at all about whether there
really are such things as phenomenal states
or events. (Philosophers, like most current
English speakers, are not so strict with the
nominal form 'phenomenon'.) Phenomenal
states or events often trade under other
names: 'SENSATIONS', 'sense-data', 'QUALIA'
(though many think of qualia as properties
of phenomenal states and not the states
themselves).
As is typical of 'ology' forms, 'phenomen-
ology' is the study of phenomena under-
stood in the strict sense. Thus, if you
investigate what appears to us when we
perceive the world rather than the world
that so appears, you are engaged in phe-
nomenology. Through the writings of
Edmund Husserl (1859-1938), phenomen-
ology came to be regarded as a branch of
philosophy, though as such it has a com-
plexity not revealed by the above-mentioned
simple perceptual example. As with 'phe-
nomenon', there tends to be a certain care-
lessness in the use of 'phenomenological'.
PHYSICALISM (1)
One often hears reference to 'phenomen-
ological' states when what is at issue are
phenomenal states. This use of the 'study
of' form has probably arisen because of the
felt need to have a word not likely to be
confused with the now more outward
looking 'phenomenon'. (See also CONSCIOUS-
NESS; SUBJECTIVITY.)
SAMUEL GUTTENPLAN
physicalism (1) The term 'physicalism' has
been used in a variety of family-resemblance
related ways in recent philosophy of mind.
The same is true of 'materialism'. Often the
two terms are used interchangeably,
although there has also been some tend-
ency to employ 'materialism' more gener-
ically than 'physicalism' - in particular, to
use 'physicalism' for psychophysical IDEN-
TITY THEORIES, while employing 'materi-
alism' in a more inclusive way (as in
'eliminative materialism' and 'non-reductive
materialism'). Here I will survey the range
of philosophical doctrines that have been
placed under the rubrics 'physicalism' and
'materialism' from the late 1950s to the
present, in an effort to give a sense for
the differences among these doctrines, the
underlying similarities, and some of the
reasons why there has been disagreement
about what should count as a physicalist or
materialist conception of mentality.
On this topic, as with many topics in
philosophy, there is a distinction to be made
between (i) certain vague, partially inchoate,
pre-theoretic ideas and beliefs about the
matter at hand; and (ii) certain more precise,
more explicit, doctrines or theses that are
taken to articulate or explicate those pre-
theoretic ideas and beliefs. There are various
potential ways of precisifying our pre-theo-
retic conception of a physicalist or materi-
alist account of mentality, and the question
of how best to do so is itself a matter for
ongoing, dialectical. philosophical inquiry.
(In order to emphasize this conceptual slack,
I will hereafter use 'materialism' for the body
of pre-theoretic ideas, and will treat the
various potential precisifications as alter-
native referents of 'physicalism'.)
471
PHYSICALISM (1)
I begin by formulating, as generically and
neutrally as possible, some central ideas
constitutive of a materialist conception of
human nature:
(1) Humans are, or are fully constituted by,
entities of the kind posited in (an ideally
completed) physics. There are no in-
corporeal Cartesian souls, or vital spirits,
or entelechies (see DU ALISM; HISTOR Y).
(2) The human body is a causally complete
physico-chemical system: although the
body is highly susceptible to external
causal influence, all physical events in
the body, and all bodily movements, are
in principle fully explainable in physico-
chemical terms.
(3) Any instantiation of any property by, or
within, a human being is ultimately
explainable in physico-chemical terms.
These first three assertions are typically
viewed as highly confirmed empirical
hypotheses, not as a priori assumptions.
They do not yet mention mentality, and in
fact are compatible with the contention that
humans never undergo any mental events
or states, and never instantiate any mental
properties, at all. (Here and henceforth, the
terms 'event' and 'state' are used for con-
crete, spatiotemporally located, occurrences
and conditions - i.e. token events and
states.) Standardly, however, materialism
takes mentality seriously:
(4) Mentality is real; humans undergo
mental events and states, and instanti-
ate mental properties.
(5) Much of human behaviour is describ-
able as ACTION, not merely as raw
motion; and action is mentalistic ally
explainable.
(6) Much of human mental life is mental-
istically explainable.
And standardly, though not invariably,
mentalistic explanation is taken to be
causal:
(7) Mentalistic explanation is a species of
causal explanation; mentality is caus-
472
ally efficacious, both intra-mentally and
in the aetiology of behaviour.
In embracing claims (4)-(7) in addition to
(1)-(3), materialism embodies the idea that
mentality can be accommodated within a
conception of human beings as complex
physico-chemical systems whose behaviour,
whose internal events and states, and
whose properties are all explainable in
physico-chemical terms.
The various candidate precisifications of
these generic materialist ideas - i.e. the
various versions of physicalism - differ
largely with respect to what they say about
the status of mental properties. I turn now
to typology.
REDUCTIVE PHYSICALISM
The simplest, most straightforward, versions
of physicalism approach theses (1)-(7) in
the context of discussions of inter-theoretic
reduction in philosophy of science. Human
psychology is construed as a theory, or
perhaps a body of theories; and materialism
is explicated by way of the contention that
human psychology is 'micro-reducible' to
neuroscience, in much the same manner
that classical thermodynamics was micro-
reduced to statistical molecular mechanics.
The Type/Type Identity Theory
According to traditional accounts in philo-
sophy of science, micro-reduction is a form
of explanation: the higher-level theory is
shown to be derivable from the lower-level
theory together with 'bridge laws' asserting
the nomic coextensiveness of higher-level
theoretical properties of structurally
complex entities (e.g. temperature, a prop-
erty of a gas), and lower-level theoretical
properties involving micro-components (e.g.
mean kinetic energy, a property involving
the gas's component molecules). But in
addition, the received view is that these so-
called bridgelaws, although empirically
established, actually reveal the identity of
the properties they cite. A gas's tempera-
ture, for instance, just is its mean molecular
kinetic energy. Bridge laws that did not
reflect underlying property identities. it is
claimed. would be insufficient for full-
fledged reduction: they would be funda-
mental, unexplainable. theoretical principles
over and above the theoretical principles of
the reducing science.
The type/type identity theory thus expli-
cates claims (3) and (4) of materialism in the
simplest possible way: mental properties are
real, and mental property-instantiations in
humans are physico-chemically explainable.
because mental properties just are physico-
chemical properties. Likewise. mutatis
mutandis. for mental events and states.
Early papers defending the identity theory
stressed both the ever-mounting evidence
for systematic micro-explain ability of the
phenomena described in the various special
sciences on the basis of more basic sciences
(ultimately physics). and the ontological
queerness that would attend unexplainable
correlations between physical and non-
physical properties. Smart (1962) put it this
way:
That everything should be explicable in
terms of physics ... except the occur-
rence of sensations seems to be frankly
unbelievable. Such sensations would be
'nomological danglers,' to use Feigl's
expression. ... I cannot believe that
ultimate laws of nature could relate
simple constituents to configurations
consisting of perhaps billions of
neurons. . . . Such laws would be like
nothing so far known in science.
They have a queer 'smell' to them. (pp.
161-2)
Some defenders of the type/type identity
theory also invoke a causal account of
mental terms and concepts. As Lewis
(1966) summarizes the reasoning:
The definitive characteristic of any (sort
of) experience as such is its causal role.
its syndrome of most typical causes and
effects. But we materialists believe that
these causal roles which belong by
analytic necessity to experiences.
belong in fact to certain physical states.
PHYSICALISM (1)
Since those physical states possess the
definitive characteristics of experience.
they must be the experiences. (p. 17)
Lewis (1980) has refined the causal account
of mental notions by claiming that our
terms for mental state-types denote in a
population-relative way. Thus. Martians
could have mentality even if they are
drastically different. physico-chemically.
from humans: mental-state terms would
denote different physical properties relative
to the Martian population than they do
relative to the human population (see
LEWIS).
Reduction without Identity
A few philosophers reject the type/type
identity theory while still espousing. or at
least allowing the possibility of. the micro-
reducibility of psychology to natural
science. Shaffer (1963). for instance. argues
that even if concrete mental events are
identical to physical events. nevertheless the
properties of mental events by which we
introspectively notice them and categorize
them are non-physical:
Let us take the case where a person
reports the having of some mental
event, the having of an after-image. a
thought. or a sensation of pain. Now
such a person has surely noticed that
something has occurred. and he has
surely noticed that this something has
some features .... Now it seems to me
obvious that. in many cases at least.
the person does not notice any physical
features. . . . Yet he does notice some
feature. . .. The noticing of some non-
physical feature is the only way to
explain how anything is noticed at all.
(p. 163)
Shaffer maintains. as does Kim (1966). that
genuine micro-reduction can be under-
written by universally quantified bicondi-
tional bridge laws. even if these bridge laws
express nomic correlations between distinct
properties. In effect. they hold that this
brand of property dualism still does justice
473
PHYSICALISM (1)
to materialism, since it is still a reductionist
conception of mentality.
However, if the bridge laws are them-
selves unexplainable, hence metaphysically
fundamental, then a serious prima facie
doubt arises whether this brand of property
dualism really comports with tenet (3) of
materialism - and whether it really counts
as a reductionist position. In genuine reduc-
tive explanation, presumably, any unex-
plained laws should either be part of the
reducing theory or else should reside at a
still more fundamental explanatory level.
This judgment, I take it, is one principal
reason for the received view that the soc al-
led 'bridge laws' in classic models of inter-
theoretic reduction really must reflect inter-
theoretic property identities in order to effect
a genuine reduction.
ELIMINATIVE PHYSICALISM
Eliminative physicalism embraces claims
(1)-(3) of materialism, while arguing that
claims (4)-(7) are very likely false (see
ELIMIN A TIVISM). Eliminative physicalists
typically share reductive physicalism's con-
ception of how theses (4)-(7) of materialism
are to be construed. That is, the elim-
inativists too maintain that in order for
mentality to be accommodatable within a
materialist metaphysics, psychology would
have to be reducible to physical theory; and,
in accordance with the received view of
inter-theoretic reduction, they maintain
that such reduction would involve system-
atic type/type psychophysical identities. But
they then point out the serious empirical
possibility that this reducibility requirement
cannot be met - that our mental categories
will turn out not to correlate neatly with
the properties and natural kinds of natural
science. They argue that such type/type
matchups are rather unlikely - among
other reasons, because commonsense
mentalistic psychology (so-called FOLK
PSYCHOLOGY) predates modern science. As
Churchland (1981) puts it:
If we approach homo sapiens from the
perspective of natural history and the
474
physical sciences, we can tell a coher-
ent story of his constitution, develop-
ment, and behavioral capacities which
encompasses particle physics, atomic
and molecular theory, organic chem-
istry, evolutionary theory, biology,
physiology, and materialistic neuro-
theory. That story, though still radic-
ally incomplete, is already extremely
powerful. ... And it is deliberately and
self-consciously coherent with the rest
of our developing world picture ....
But FP [folk psychology] is no part of
this growing synthesis. Its intentional
categories stand magnificently alone,
without visible prospect of reduction to
that larger corpus. (p. 75)
Eliminative physicalists sometimes offer
further reasons in support of the falsity of
folk psychology, in addition to its likely irre-
ducibility to natural science. For instance,
they sometimes argue (i) that folk psychol-
ogy is committed to the existence of
language-like mental representations in
homo sapiens; and (ii) that in light of recent
theoretical developments in cognitive
science and neuroscience, this LANGUAGE
OF THOUGHT hypothesis is probably false.
NON-REDUCTIVE PHYSICALISM
Much recent philosophy of mind, while still
remaining true to materialism, has ques-
tioned in one way or another the conten-
tion that in order for theses (4)-(7) to be
true, mentalistic psychology must be reduc-
ible to physical science via type/type
psychophysical bridge laws expressing
either property identities or nomic coexten-
siveness of distinct properties. The rubric
'non-reductive physicalism' covers a range
of philosophical positions that meet this
description. I will give a broad typology of
these views. Let me remark at the outset,
though, that for the term 'reduction' (just
as for 'phYSicalism'), one should really dis-
tinguish between (i) a vague, pre-theoretic
meaning, and (ii) proposed precisifications
or explications of the pre-theoretic notion.
Although the positions to be described are
'non-reductive' relative to the traditional,
received, standard explication of reduction
in philosophy of science, they also can be -
and often are - called 'reductive' anyway.
As the traditional explications of materi-
alism get replaced by less-restrictive concep-
tions, there is a natural tendency for the
received explication of reduction to begin
undergoing the same fate.
Anomalous Monism
DA VIDSON (1970) adopts a position that
explicitly repudiates reductive physicalism,
yet purports to be a version of materialism
nonetheless. Davidson holds that although
token mental events and states are identical
to token physical events and states, mental
types - i.e. kinds, and/or properties - are
neither identical to, nor nomic ally coex-
tensive with, physical types. His argument
for this position relies largely on the conten-
tion that the correct assignment of mental
and actional properties to a person is
always a holistic matter, involving a global,
temporally diachronic, 'intentional inter-
pretation' of the person.
Davidson's reasoning suggests that the
demand for psychophysical type/type iden-
tities, or for nomic coextensions, is an
excessive demand for a materialist to make
- that even though mental properties do not
conform to this demand, there is nothing
inherently objectionable or mysterious, from
a materialistic perspective, about the fact
that humans are susceptible to intentional
interpretation, and hence that mental kinds
and attributes can be accommodated within
a materialistic metaphysics in some other
way. As regards claims (4)-(7) of materi-
alism, he is sometimes interpreted as main-
taining that the psychophysical identity
theory suffices, by itself, to underwrite these
claims: mental events are physical, hence
are part of the physico-chemical causal
nexus, hence are causally explanatory.
But as many philosophers have in effect
pointed out, accommodating claims (4)-(7)
of materialism evidently requires more than
just token mental/physical identities.
Mentalistic explanation presupposes not
merely that mental events are causes, but
PHYSICALISM (1)
also that they have causal/explanatory rele-
vance as mental - i.e. relevance insofar as
they fall under mental kinds or types. Thus
a prima facie question arises whether David-
son's position, which denies there are strict
psychophysical or psychological laws, can
accommodate the causal/explanatory relev-
ance of the mental qua mental. If not, then
his view would amount, in effect, to EPI-
PHENOMEN ALISM with respect to mental
properties.
Apart from anomalism per se, the problem
of causal/explanatory relevance evidently
arises for any form of non-reductive physic-
alism (and, indeed, for non-identity versions
of reductive physicalism too). For, if mental
properties are not identical to physical prop-
erties, and if claims (1)-(3) of materialism
are true, then there is a prima facie worry
that physico-chemical properties 'do all the
real causal/explanatory work', and that
mental properties are epiphenomenal. Non-
reductive physicalists of all stripes, whether
or not they espouse Davidson's anomalism,
thus bear the burden of arguing that mental
properties, despite being irreducible to phy-
sical ones, nevertheless have genuine
causal/explanatory relevance even though
the human body is a causally complete
physico-chemical system. The task is to
explain why, and how, there can be
genuine causal explanation at higher levels
of description (the mental level in parti-
cular), rather than 'explanatory exclusion'
of higher levels by lower levels. Substantial
recent attention in philosophy of mind has
been devoted to this task (see REASONS AND
CAUSES).
Although Davidson himself said rather
little by way of a positive, non-reductive,
account of the interrelations between physi-
cal and mental properties, he did make
these suggestive remarks:
Although the position I describe denies
there are psychophysical laws, it is
consistent with the view that mental
characteristics are in some sense
dependent, or supervenient, on physi-
cal characteristics. Such supervenience
might be taken to mean that there
475
PHYSICALISM (1)
cannot be two events alike in all physi-
cal respects but different in some
mental respect, or that an object
cannot alter in some mental respect
without altering in some physical
respect. Dependence or supervenience
of this kind does not entail reducibility
through law or definition. (Davidson,
1970, p. 88)
This passage has inspired a substantial
amount of philosophical effort, in philos-
ophy of mind and also elsewhere in meta-
physics, seeking to harness the notion of
SUPERVENIENCE to undergird claims (4)-(7)
of materialism. Supervenience has been
invoked, for instance, in efforts to account
for the causal/explanatory efficacy of
mental properties qua mental (e.g. Kim,
1984). And it has often been regarded in
the manner intimated by Davidson in the
above-quoted passage: as a relation that
makes mental properties dependent on the
physical, and hence 'materialistically
respectable', without reducing them to
physical properties.
Functionalism
FUNCTIONALISM denies that mental proper-
ties are identical to specific physico-
chemical properties, and asserts instead that
they are multiply realizable properties
whose essence is their typical causal role
within creatures that instantiate them.
Mental properties are like the properties
being locked and being unlocked, instantiable
by those physical devices that are locks.
These properties are not narrowly physical
(since there are innumerably many kinds of
locks, physically quite different from one
another), but rather are causal/functional.
To be locked (unlocked) is to be in a state
which prevents the lock from opening
(causes the lock to open) when subjected to
the appropriate opening-action. Likewise,
the functionalist claims, the essence of each
mental property is its syndrome of typical
causal connections to sensory properties,
behavioural properties, and other such
mental properties. Although mental proper-
ties are not narrowly physical, on this view,
476
they are materialistically kosher even so.
For, on any occasion when a mental prop-
erty is instantiated, there is an explanation
available for why it is being instantiated: viz.
a physical property is instantiated whose
causal role, in the given creature, qualifies
this property as a realizer of the higher-
order, mental, property.
Functionalists typically invoke the physi-
cal possibility of multiple realization as a
reason to reject reductive physicalism and
instead identify mental properties with
functional properties. Yet as I remarked
earlier, some reductive physicalists also
advocate a causal account of mental terms
and concepts, and employ this account in
arguing in favour of psychophysical prop-
erty identities. The reductionists propose to
accommodate the physical possibility of
Martian mentality not by allowing that
mental properties are differently realized in
Martians than in humans, but instead by
treating mental property-terms as species-
relative non-rigid designators: these terms
denote physical properties, all right, but dif-
ferent ones relative to Martians than relat-
ive to humans.
But the appropriate functionalist rejoinder
is to point out the apparent physical possi-
bility of a single species of creatures whose
individual members are so constituted phys-
ically that mental properties are realizable,
within a single creature, by a radical multi-
plicity of distinct physical properties. (For all
we now know, humans themselves might
be like this.) Thus, once one appreciates the
full extent of the possibilities concerning
multiple realization, reductive physicalism
comes to look excessively strong. This point
applies not just to psychology, but to the
special sciences generally (cf. FODOR, 1974).
'Naturalizing' Projects
Functionalism has begun to wane in popu-
larity in philosophy of mind, in part because
of the relevance of TWIN EARTH thought
experiments: it seems that an Earthling and
his Twin Earth duplicate could be in ident-
ical functional states but distinct mental
states. (The Earthling is thinking about
water; the duplicate is thinking about XYZ.)
(See EXTERNALISM INTERNALISM; INTER-
NATIONALITY.) Nevertheless, there is still
widespread acceptance of the contention
that mental properties are multiply realiz-
able physically, rather than being identical
to physical properties or nomic ally coex-
tensive with them. So in the wake of func-
tionalism there have arisen a variety of
philosophical positions and projects which
seek to accommodate theses (1)-(7) in some
way other than via traditional reductive
physicalism. Typically these projects get
placed together under the rubric 'natur-
alization'. The project of naturalization
involves giving a tractable specification, in
non-intentional and non-semantic terms -
though not necessarily in narrowly physical
terms - for the instantiation of any mental
property with specific intentional CONTENT.
'Tractability' is my own term; and the
demand for it is usually implicit in recent
philosophical work, rather than being on
the surface. Roughly, a tractable specifica-
tion is a relatively compact, relatively non-
baroque, non-disjunctive, cognitively sur-
veyable, formulation of sufficient conditions
(for some philosophers, sufficient and neces-
sary conditions). Recent philosophical pro-
posals for 'naturalizing' intentional mental
properties have centred, for instance, on
systematic correlation between MENTAL
REPRESENTATIONS and what they represent
(e.g. Fodor, 1990); on the biological notion
of relational proper function (e.g. Millikan,
1984; DRETSKE, 1988); or on the computer
scientist's notion of simulation (e.g.
Cummins, 1989).
Post-Analytic Physicalism
A number of philosophers (e.g. Baker,
1987; Stich, 1992; Tye, 1992) have
recently argued that projects for 'naturaliz-
ing' mentality probably cannot succeed. For
one thing, counterexamples keep surfacing,
and the accumulating inductive evidence
suggests that they always will. For another,
it looks increasingly likely that our mental
concepts (indeed, most of our concepts) are
essentially vague, and have some kind of
'prototype' structure - rather than having
sharply delimited applicability conditions,
PHYSICALISM (1)
expressible via 'conceptual analyses' of the
kind traditionally sought during the heyday
of High Church analytic philosophy. So
there are reasons to seek out some new
way of explicating claims (4)-(7) of materi-
alism still further removed from the tradi-
tional type/type psychophysical identity
theory. Let me now sketch such a post-
analytic physicalism (cf. Horgan, 1992,
1993).
One key tenet is that mental properties
and facts are supervenient on physical prop-
erties and facts. It is important to appreciate
that this supervenience thesis could well be
true even if there is no way to tractably
specify the non-mental conditions that
suffice for mental phenomena. Perhaps, for
instance, the physical supervenience base
for an intentional mental property, on a
given occasion of instantiation, generally
involves a good-sized chunk of spacetime
extending well beyond the cognizer's own
body and well beyond the time at which the
mental event occurs; perhaps it involves
a rather gargantuan number of physico-
chemical goings-on within that extended
spatiotemporal region; and perhaps there
isn't any simple way to describe, in non-
intentional vocabulary, all the relevant
aspects of this hugely complex super-
venience base. Perhaps, in addition, the
supervenience of the mental is largely a
holistic matter - with mental properties
supervening collectively, as part of the
correct global intentional interpretation
of a cognizer - or perhaps the cognizer's
whole community or whole species. In
short, it might be that the search for tract-
ably specifiable, cognitively surveyable, non-
intentional sufficient conditions for mental-
ity is utterly hopeless - and yet that the
mental supervenes on the physical none-
theless.
A second tenet of post-analytic physical-
ism, as with other versions of non-
reductive physicalism, is that mental
properties have genuine causal/explanatory
relevance even though the human body is a
causally complete physico-chemical system.
Although mental properties always have
their efficacy via the physical properties
477
PHYSICALISM (1)
that realize them on particular occasions of
instantiation - rather than being funda-
mental force-generating properties, on a par
with the force-generating properties of
physics - nevertheless they are causally and
explanatorily efficacious anyway. (A physi-
cal property that realizes a mental property,
on a given occasion of the mental property's
instantiation, does not necessarily constitute
the full supervenience base for that mental
property.)
A third tenet is that inter-level super-
venience facts are in principle explainable,
rather than being fundamental and sui
generis. This tenet is evidently dictated by
contention (3) of materialism. For, unless
psychophysical supervenience facts are
themselves explainable, then the instanti-
ation of mental properties is not explainable
on the basis of physico-chemical facts, but
only on the basis of such facts plus meta-
physically fundamental inter-level, super-
venience facts.
Even if there are no tractably specifiable
non-semantic and non-intentional sufficient
conditions for intentional mental states, the
supervenience of the mental on the physical
may well be susceptible, in principle, to
explanation anyway. Exploring what kinds
of criteria these explanations should meet,
and making a case that psychophysical
supervenience relations are indeed suscep-
tible to such explanations (cf. Horgan and
Timmons, 1992), are thus important pro-
jects for post-analytic physicalism. To be
sure, a position that asserts the explain-
ability of psychophysical supervenience
relations, but does not offer any tractable
specification of non-intentional sufficient
conditions for mentality, would differ from
the versions of physicalism that have domi-
nated the recent philosophical landscape.
But this would be yet another phase in the
ongoing dialectical interplay between philo-
sophers' pre-theoretic understanding of
materialism and their attempts to give that
understanding an adequate theoretical
articulation.
See also An Essay on Mind section 3;
PHYSICALISM (2).
478
BIBLIOGRAPHY
Baker, L.R. 1987. Saving Belief: A Critique of
Physicalism. Princeton University Press.
Churchland, P.M. 1981. Eliminative materi-
alism and propositional attitudes. Journal of
Cummins, R. 1989. Meaning and Mental Rep-
resentation. Cambridge, MA.: MIT Press/
Bradford Books.
ence and Theory, ed. 1. Foster and J. W.
setts Press.
Dretske, F. 1988. Explaining Behavior: Reasons
in a World of Causes. Cambridge, MA.: MIT
Fodor, J. 1974. Special sciences (or: the dis-
unity of science as a working hypothesis).
Synthese, 28, 97-115.
Essays. Cambridge, MA.: MIT Press/Bradford
Books.
Horgan, T. 1992. From cognitive science to
folk psychology: computation, mental rep-
resentation, and belief. Philosophy and Phe-
nomenological Research, 52,447-84.
Horgan, T. 1993. Nonreductive materialism
and the explanatory autonomy of psychol-
ogy. In Naturalism: A Critical Appraisal, ed.
S. J. Wagner and R. Warner. Notre Dame:
University of Notre Dame Press.
Horgan, T., and Timmons, M. 1992. Troubles
on moral twin earth: moral queerness
revived. Synthese, 92, 221-60.
Kim, J. 1966. On the psycho-physical identity
theory. American Philosophical Quarterly, 3,
227-35.
venient causation. Midwest Studies in Philo-
sophy, Vol. 9, Minneapolis: University of
Minnesota Press.
Lewis, D. 1980. Mad pain and Martian pain.
Readings in the Philosophy of Psychology, Vol.
1, ed. N. Block. Cambridge, MA.: Harvard
University Press.
Millikan, R. 1984. Language, Thought, and
Other Biological Categories. Cambridge, MA.:
Shaffer, J. 1963. Mental events and the brain.
Journal of Philosophy, 60, 160-66.
Smart, J.J.C. 1962. Sensations and brain pro-
cesses. In Philosophy of Mind, ed. V. C.
Chappell. Englewood Cliffs: Prentice-Hall.
Stich, S. 1992. What is a theory of mental
representation? Mind, 101, 243-61.
Tye, M. 1992. Naturalism and the mental.
Mind, 101. 421-41.
TERENCE E. HORGAN
physicalism (2): against physicalism 'How
the purer spirit is united to this clod' said
Joseph Glanvill, 'is a knot too hard for
fallen humanity to untie'. Although some
contemporary philosophers share this pessi-
mism about the MIND-BODY PROBLEM,
there is in general a consensus that some
form of physicalism is the solution to the
problem. Physicalism is the thesis that all
entities - whether objects, events, proper-
ties, relations or facts - are, or are reduci-
ble to, or are ontologically dependent on,
physical entities. Physicalists might differ in
their other metaphysical commitments:
they may deny that there are any events,
properties or facts at all, for example. But
common to all forms of physicalism is the
view that whatever exists is in some sense
physical. My aim here is to outline the
main physicalist positions, and express
some scepticism about the physicalist con-
sensus.
Opposition to physicalism standardly
arises from reflection on the irreducible
nature of the conscious properties of the
mind (see CONSCIOUSNESS; QUALIA). These
properties are claimed to be essentially sub-
jective, something that a purely physical
world must lack, because knowledge of
these properties is supposed to essentially
involve subjective experience in a way that
knowledge of the physical does not (see the
celebrated 'knowledge argument' of
Jackson, 1986).
However, this sort of argument is more
properly directed against naturalism - the
thesis that the empirical world can be
entirely accounted for by the natural sci-
ences, including empirical psychology -
rather than physicalism as such. (So nat-
uralism is distinct from physicalism if there
PHYSICALISM (2): AGAINST PHYSICALISM
are 'natural' features of the world that are
not physical.) The knowledge argument, if
successful, would refute any view that
claims that everything in the world is
objective, and thus the potential subject-
matter of a natural science. Such a view
need not be physicalism. In this essay I
shall be concerned only with arguments
specifically against physicalism proper, not
those against the weaker thesis of nat-
uralism.
THE DEFINITION OF 'PHYSICALISM'
Before assessing physicalism, we need to
know what it is. But the definition of 'phys-
icalism' gives rise to a problem even before
we consider the difficulties posed, for
example, by consciousness. The problem is
how to formulate a conception of the phys-
ical that is not so strong as to collapse into
an implausible ELIMIN A TIVISM - the out-
right denial of mental phenomena - but not
so weak as to let in the mental as physical
by definition. In defining the physical, phys-
icalists must steer a path between the two
horns of this dilemma: the physical must
not be defined in such a way as to make
physicalism obviously false, nor trivially
true.
I shall have to assume here without
argument that the first horn of this dilemma
- eliminative physicalism - is unacceptable.
The second horn is a problem because
physicalism needs a principled distinction
between the mental and the physical if it is
not to be vacuous. It needs a characteriza-
tion of the essential marks of the mental,
such as consciousness and INTENTION-
ALITY. But it also needs a characterization
of the essential marks of the physical. And
many attempts to give such a characteriza-
tion either fail to include obviously physical
phenomena, "Of fail to exclude obviously
mental phenomena.
As an example, consider the claim made
by a number of philosophers that the physi-
cal is the spatial (see Charles, 1992). This
captures many things that are obviously
physical: as well as the medium-sized dry
goods of everyday experience, the atoms,
479
electrons and quarks of microphysics come
out as physical on this definition. But does
the definition rule out the mental? Surely
nothing is more natural than the idea that
my thoughts and sensations occur where
my body is. So how can this definition of
the physical fail to count the mental as
being trivially physical? There is of course a
philosophical conception of the mental on
which mental phenomena are not spatial -
this is what some dualists believe (see
DUALISM). But this conception is part of a
specific and contentious theory of the
mental, not a constraint on any such
theory.
In any case, not every dualist is com-
mitted to the non-spatiality of the mental
(see Hart, 1988). A dualist could coherently
hold that I am not the same thing as my
body or brain, even though I occupy the
same space as it. The physicalist could
respond that even if the mental were
spatial, on this dualist conception it would
fail to be physical if it could occupy the
same space as a physical thing without
sharing that thing's parts (see Charles,
1992, p. 281). So a 'ghostly' conception of
the mental would be ruled out by this defi-
nition of the physical. But now the defini-
tion rules out too much, since rainbows and
clouds will turn out not to be physical
objects: a bird can pass through a cloud
without sharing any of its parts.
The proper response for a physicalist is to
give up the attempt to define the physical in
a a priori terms, and to define it instead in
terms of the entities postulated by physical
science. After all, physicalism is standardly
supposed to be an empirical conjecture, not
an a priori truth. Understood in this way,
physicalism is the claim that physical
science has a certain ontological authority.
Different kinds of physicalism say different
things about this authority: some say that
the only entities there are are those physical
science talks about. Others say that if there
are entities other than those physical
science talks about, these will be ultimately
dependent on physical entities.
I shall discuss a couple of these different
options below. But the prior question is:
480
what is physical science? Obviously physical
science includes all the many branches of
physics as now conceived: mechanics, elec-
tromagnetism, thermodynamics, gravity,
particle physics, etc. But no physicalist
thinks that this is the full extent of physical
science. For one thing, chemistry and
molecular biology are normally considered
unproblematically physical. For another, no
one believes that physics proper is now
finished: physics may discover many new
entities and laws in order to adequately
account for reality.
But why are chemistry and molecular
biology thought to be physical sciences? One
possibility is that these sciences reduce to
physics in some sense (see REDUCTION). But
this is not based on the fact that reductions
between (say) all theories in chemistry and
physics have actually been achieved; the
idea is rather that they reduce to physics 'in
principle'. But what is the principle? It
cannot be physicalism itself, if reducibility to
the physical is supposed to tell us the
boundaries of physical science. Yet if some
physical theories are not reducible to
physics proper, then what prevents an irre-
ducible psychology from being a physical
science in this sense?
What about the development of future
physics? If current physics develops in
certain unforseeable ways, then how can
physicalism be defined now in such a way as
to rule out the mental? Perhaps physical
science can be defined as the final theory of
everything. But if it turns out that irre-
ducible psychological properties are
appealed to in the final theory of every-
thing, then the mental will once again
count as physical by mere definition.
There are therefore real difficulties in
defining the physical which the physicalist
must address (see Crane and Mellor, 1990).
For the purposes of this article, however, we
need to fix on a rough definition of physical
science, in order to assess current physical-
ist proposals. Let this (admittedly vague)
notion of the physical be: the entities
appealed to by the true completed science
whose explananda are uncontroversially
physical (see Papineau, 1990).
ARGUMENTS FOR PHYSICALISM
So what do physicalists claim to be the rela-
tion between the mental and the physical.
so conceived? Physicalist answers can be
divided into two kinds: those that say that
mental phenomena are strictly identical to
certain physical phenomena (see IDENTITY
THEORIES), and those that say mental phe-
nomena are not identical with, but de-
pendent or supervenient upon certain physical
phenomena (see SUPERVENIENCE). These
two views can be seen as two ways of
reading the claim that the mental is physi-
cal: the first view says that the 'is' in ques-
tion is the 'is' of identity, the second that it
is the 'is' of constitution.
There are a number of arguments for the
various forms of identity theory that have a
common structure. First a claim is made
about the causal nature of the physical
world: every physical effect has a complete
physical cause. Purely physical causes
suffice to fix (or to fix the chances of) every
physical effect. For brevity, call this claim
'Completeness'. Then a claim is made about
mental causation: some or all mental states
have some physical effects. We then have a
conflict: if physical effects are not massively
overdetermined by distinct physical and
mental causes, then how is mental causa-
tion possible, given Completeness (see
REASONS AND CA USES)? The physicalist
solution is to identify the mental cause with
some physical state of the brain. (For the
argument in this form, see Papineau, 1990;
for other arguments that derive an identity
theory from versions of such a conflict - in
very different ways - see LEWIS, 1966 and
DAVIDSON,1980.)
Notice that if there is to be a conflict
between mental causation and Complete-
ness, the notion of causation as applied to
the mental must be the same as that applied
to the physical. For if mental causation
were utterly different from physical causa-
tion, then mental and physical causes
would not be competitors, and there would
be no motivation for the identity claim. I
shall call this assumption the 'Homogeneity'
of mental and physical causation. Accord-
ing to this Homogeneity claim, the terms
'physical' and 'mental' in 'physical/mental
causation' are really transferred epithets:
what is physical or mental are the relata of
causation, not the causation itself.
The general argument from reconciling
Completeness with mental causation is an
argument for the identity of mental phe-
nomena with certain physical phenomena.
There are two kinds of identity theory: the
reductive type identity theory which says
that mental properties are physical proper-
ties (Lewis, 1966); and the token identity
theory which says merely that mental par-
ticulars are physical particulars (Davidson,
1980).
What can be said for this general argu-
ment? It might be thought that someone
looking for a reason for believing in physic-
alism as a solution to the mind-body
problem would not be impressed by the
appeal to Completeness. For Completeness
seems just too close to the physicalist con-
clusion to be persuasive. Why should
someone unpersuaded of the truth of physi-
calism about the mind be swayed by the
claim that all physical effects have complete
physical causes?
The response will be that Completeness,
while not a law of physics itself, is a high-
level empirical generalization about the
practice and content of physical science.
David Lewis formulates a version of Com-
pleteness in terms of the idea of explanation:
there is some unified body of scientific
theories of the sort we now accept,
which together provide a true and
exhaustive account of all physical phe-
nomena. They are unified in that they
are cumulative: the theory governing
any physical phenomenon is explained
by theories governing phenomena out
of which that phenomenon is com-
posed and by the way it is composed
out of them. The same is true of the
latter phenomena, and so on down to
fundamental particles or fields gov-
erned by a few simple laws, more or
less as conceived in present-day theo-
retical physics. (Lewis, 1966, p. 105)
481
Lewis here appeals to two ideas: first, that
entities are exhaustively composed out of
physical parts; and second, that the theories
that explain the whole are themselves
explained by theories that explain the parts
out of which the whole is composed.
Can Completeness be derived from these
ideas? I think not: non-physicalists can
accept these ideas without accepting Com-
pleteness. So for one thing, certain non-
physicalists can accept that non-physical
objects have physical parts, since they will
want to formulate their thesis in terms of
non-physical properties, not parts. So the
idea that macroscopic objects are composed
out of physical parts will not trouble this
kind of non-physicalist, so long as these
objects have properties that are non-
physical. And if properties are implicated in
causation, it is therefore open for this non-
physicalist to deny that all effects have com-
plete physical causes.
In any case, the thesis about explanation
is questionable. What it implies is that all
macro-phenomena can be explained by the
laws governing their micro-parts. But many
physicalists are happy to accept that psy-
chological explanations have a degree of
autonomy from physical explanations - this
is why I framed Completeness as a claim
about causation, rather than explanation.
And even in physics, it is not always true
that all explanations are micro-reductive in
the above way. For example: consider a gas
sample at a constant temperature whose
voume is suddenly halved. If the gas is
ideal. then according to Boyle's law, it will
reach an equilibrium pressure that is twice
its initial pressure. The law does not explain
what all the gas's molecules must do before
it reaches equilibrium, except that they must
be such as to double the sample's pressure.
The law thus explains the molecules' beha-
viour in terms of the behaviour of the whole
sample (see Crane and Mellor, 1990, p.
190).
To respond to this that macro-laws like
these are ultimately reducible to micro-laws
only raises the question-begging assump-
tions about reducibility 'in principle' men-
tioned above. And this move also introduces
482
a stronger physicalist commitment than
that involved in Completeness; so it would
be unwise to employ it in a defence of Com-
pleteness.
Nor does Completeness follow from philo-
sophical claims about causation, even if the
'real essence' of causation is the flow of
energy (see Crane and Mellor, 1990, pp.
191-192; Hart, 1988, ch. 9). So with
'physical' understood in a non-trivial way,
Completeness is a substantial empirical con-
jecture that seems at least to lack decisive
support. In the face of this, it is not absurd
for a non-physicalist to deny Completeness,
and hold that some physical effects have
irreducible mental causes. The non-
physicalist does not thereby have to hold
that actions break the laws of physics - if
there are laws linking mental and physical
phenomena, these will have to be consistent
with the laws of physics. But this is just a
special case of the truism that all truths
must be consistent.
A less extreme non-physicalist response is
to allow Completeness but hold that actions
are overdetermined by mental and physical
causes. Physicalists cannot, and usually do
not, deny that overdetermination is possible;
their standard objection is rather to the
extent of the overdetermination involved in
the psychophysical case. As Stephen Schiffer
puts it, 'it is hard to believe that God is such
a bad engineer' (Schiffer, 1987, p. 148). But
non-physicalism can make its position more
palatable by emphasizing the nomic links
between the mental and physical causes.
Massive overdetermination of action will
thus be shown to be explicable, and not a
mere coincidence.
NON-REDUCTIVE PHYSICALISM
SO far, I have discussed only the identity
theories. However, many physicalists think
that identity theories are not essential to
physicalism, and are objectionable even on
physicalist grounds. Type identity is thought
to be too strong to be plausible, because of
the well-known 'variable realization' objec-
tion: it seems empirically unlikely that every
instance of a mental property is an instance
of the same physical property (see TYPE/
TOKEN).
Token identity, on the other hand, is con-
sidered too weak to explain the relation
between the mental and the physical. To
say that every mental particular (object or
event) is a physical particular does not tell
us what it is about the physical nature of
the particular that makes it the mental par-
ticular it is. An analogy: it is hardly an
explanation of why all US presidents have
been white males to simply assert the token
identity claim that each particular US pre-
sident is identical with some particular
white male.
For this reason, no physicalist holds a
token identity theory without supplement-
ing it with some claim about the relation
between mental and physical properties.
The claim is standardly that the physical
properties of a mental particular determine
its mental properties: physical type identity
guarantees mental type identity, but not
vice versa. To put it another way, the
mental supervenes upon the physical (see
SUPERVENIENCE). SO all mental particulars
are physical. without mental properties
being identical with physical properties: this
is non-reductive physicalism.
However, supervenience alone does not
give you physicalism, since any theory that
asserts a nomic dependence of the mental
on the physical (e.g. epiphenomenalist
dualism) can accept supervenience (see also
Charles, 1992). To get physicalism, what
has to be added to supervenience is the
claim that the mental is 'nothing over and
above' its supervenience base. If you take
away the physical base, you take away the
mental too, since the mental is in some
sense 'composed' out of or 'realized' by the
physical. For some philosophers, this claim
is the essence of physicalism (see Snowdon,
1989).
The non-reductive physicalist now owes
us an account of composition or realization.
But whatever the account, I think that a
serious problem will nonetheless arise for
non-reductive physicalism. Again the issue
turns on Completeness. If Completeness is
true, and massive overdetermination is
false, what happens to mental causation?
Non-reductive physicalism faces a dilemma:
either mental states or events are genuine
causes, in which case it has to embrace
massive overdetermination or deny Com-
pleteness - since the mental and the physi-
cal causes are not identical. Or the mental is
causally impotent, in which case non-
reductive physicalism embraces EPIPHE-
NOMENALISM.
Physicalists typically respond to this
dilemma in one of two ways. The first is to
say that supervenience allows mental prop-
erties to inherit the causal efficacy of their
subvening physical bases (see Segal and
Sober, 1991). Completeness is preserved at
the physical level. and given supervenience,
mental properties 'ride on top' of the funda-
mental physical causes. The second
response distinguishes between the notions
of causal explanation as applied to the
mental and as applied to the physical (see
Jackson and Pettit, 1990). Again, Complete-
ness is preserved at the level of micro-
physical particles, but some sort of causal
role for the mental is preserved by the fact
that mental explanations have a different
structure than physical explanations.
Neither of these responses, it seems to me,
avoids epiphenomenalism. The first response
must deprive the mental - and indeed, any
supervening features - of genuine causal
efficacy, on pain of violating Completeness.
(Some physicalists agree: see Schiffer, 1987,
p. 154.) The second response is epiphenom-
enalist in a somewhat different sense: in dis-
tinguishing between the kinds of causation
applied to the mental and the physical. it is
committed to the denial of Homogeneity.
But if this is abandoned, we no longer have
the problem to which physicalism was sup-
posed to be the solution.
And this is the reason why the physicalist
should be worried by epiphenomenalism: it
deprives physicalism of its only plausible
motivation. The physicalist assumption that
was driving the arguments for the Identity
Theories was Completeness. The mind has
to be physical because otherwise we could
not account for its effects in the physical
world. But now if we deny that the mind
483
POSSIBLE WORLD
has effects, then we have no reason for
saying it is physical (see Lewis, 1966, esp.
pp. 105-6). For what exactly would be the
content of insisting that the epiphenomenal
mental properties were 'really' or 'funda-
mentally' physical, once it is admitted that
they are not the same as physical proper-
ties, nor reducible to them? Once mental
causation is out of the picture, it seems a
mere terminological issue whether we call
ourselves epiphenomenalist dualists or epi-
phenomenalist physicalists. Calling these
epiphenomena 'physical' would be to rele-
gate the term 'physical' once again to a
merely honorific label.
Certainly many physicalists would feel
the need to insist on the physical nature of
these epiphenomena, just to rule out any-
thing that looks, by their lights, too peculiar
or weird. But if there is no conflict with
Completeness, this requirement seems
unmotivated. This objection to non-physical
epiphenomena is reminiscent of J. L.
Mackie's well-known 'argument from
queerness' against the existence of objective
moral values. And the reaction to both
arguments should be the same: the world
might turn out to be stranger than we
think. Which is, after all, something that we
have learned from physics itself.
See also An Essay on Mind section 3; PHY-
SICALISM (1).
BIBLIOGRAPHY
Charles, D. 1992. Supervenience, composition
and physicalism. In Reduction, Explanation
and Realism, ed. D. Charles and K. Lennon.
Oxford: Clarendon Press. pp. 265-296.
Crane, T., and Mellor, D.H. 1990. There is no
question of physicalism. Mind, 99,185-206.
Davidson, D. 1980. Mental events. In Essays
on Actions and Events. Oxford University
Press.
Hart, W.D. 1988. The Engines of the Soul. Cam-
Jackson, F. 1986. What Mary did not know.
Journal of Philosophy, 83, 291-95.
Jackson, F., and Pettit, P. 1990. Causation in
the philosophy of mind. Philosophy and Phe-
nomenological Research, 50, 195-214.
484
Reprinted in Lewis, D., Philosophical Papers
Volume I, Oxford University Press, 1983.
Papineau, D. 1990. Why supervenience? Ana-
lysis, 50, 66-71.
Schiffer, S. 1987. Remnants of Meaning. Cam-
Segal, G., and Sober, E. 1991. The causal effi-
cacy of content. Philosophical Studies, 63,
1-30.
Snowdon, P. 1989. On formulating materi-
alism and dualism. In Cause, Mind and
Reality. Dordrecht: Kluwer, pp. l37-58.
TIM CRANE
possible world 'England might win the
next test match in cricket.' This sentence
says something about the fortunes of the
English team, but it certainly does not assert
that they will win. Instead, it makes what is
called a 'modal' claim, where the specific
mode is that of possibility. The sentence
would mean just the same if it took this
more explicit modal form: 'It is possible that
England will win the next test match.' In
addition to the mode possibility, there is the
correlative mode of impossibility. One could
have said: 'It is not impossible that England
will win the next test match', and, though
this has an undeniably pessimistic flavour,
it strictly and literally says the same thing
as the sentence cast in terms of possibility.
Impossibility is a form of the mode necessity:
a claim is said to be impossible when it is
necessarily false. An example of an affirm-
ative claim of necessity would be: 'Bachelors
must be unmarried'. This last could also be
rendered in the more explicit modal idiom
as: 'It is necessary that bachelors are
unmarried.'
Modal claims are fundamental to the
ways we characterize the world, but, since
they are not claims about how things actu-
ally are or will be, it is far from obvious
what makes them true. And it is here that
the notion of a possible world has a part to
play. If you think of the totality of how
things are as one world - the actual world -
then one could conceive of lots of other
worlds simply by imagining different one or
more features of the actual world. Thus,
there will be worlds almost exactly like
ours, but containing one more grain of sand
on a certain beach, or lacking one mole hill
on a certain lawn. And of course there will
be many worlds much more radically differ-
ent from the actual world.
The complete set of these worlds (includ-
ing the actual world) make up the infinite
set of possible worlds. Depending on your
ontological leanings (see ONTOLOGY), you
can think of this set as genuinely existing,
or existing only as an experiment in
thought. But in either case they have a real
use in providing straightforward truth con-
ditions for modal claims. Take the earlier
example: 'It is possible that England will
win the next test match.' We can under-
stand this as the following claim about the
set of possible worlds:
There is at least one possible world in
which it is true that England will win
the test match.
Notice that, in this construal. we have
traded the explicit use of modality for lan-
guage that is directly assertoric. As one
says, we are 'quantifying' over possible
worlds (,there is at least one possible
world .. .') and the claim made about
these worlds does not employ the language
of modality ('it is true that .. .'). Correla-
tively, we can use the set of possible worlds
to give us a direct characterization of the
truth conditions for claims of necessity:
In all possible worlds bachelors are
unmarried.
The usefulness of possible worlds in provid-
ing truth conditions for modal claims is
undeniable, though the examples given are
only the beginning of the story. Moreover,
since modality is integral to notions such as
SUPERVENIENCE and identity (see IDENTITY
THEORIES), there is a widespread use of
possible worlds talk in the philosophy of
mind.
SAMUEL GUTTENPLAN
PRACTICAL REASONING
practical reasoning From the premises:
Socrates is a man,
All men are mortal.
one can justifiably draw the conclusion
that:
Socrates is mortal.
This typical example of what is called 'the-
oretical' reasoning consists in moving from
one set of things which are thought true to
another, and a great deal of human enter-
prise depends upon correctly making such
transitions in thought. However, there is
another kind of reasoning which is no less
important, but whose object is not merely
transition from one set of thoughts to
another. Following Aristotle, who first
described it, this kind of reasoning is called
(a little misleadingly) 'practical'; it is rea-
soning whose conclusion is not some
further thought but the undertaking of
some action. For example, if one thinks:
I am thirsty,
The water in front of me will quench
my thirst,
then the conclusion likely to be drawn is
not some further thought, but the action of
reaching for the water.
Whilst it is easy to characterize practical
reasoning as in this way parallel to theor-
etical reasoning, real difficulties emerge
when one tries to exploit the parallel
further. The canons of theoretical reasoning
one finds in branches of logic are not
obviously matched by practical reasoning,
not least because actions, unlike thoughts,
are neither true nor false. One cannot thus
say that a good piece of practical reasoning
preserves truth or even a high probability of
truth. Yet, there is no doubt that some sort
of standards apply. (See INTENTION;
RATIONALITY; REASONS AND CAUSES.)
SAMUEL GUTTENPLAN
privacy (of the mental) see FIRST-PERSON
AUTHORITY.
485
PROPOSITION
private language argument see An Essay
on Mind section 2.2.4; WITTGENSTEIN.
property A thorough discussion would
take one deep into metaphysical and onto-
logical issues, but, in the context of philo-
sophy of mind, it is important to have
some grasp of this notion. The best way to
appreciate what is meant by a property is
by contrast with two others: predicate and
concept. Consider first the sentence: 'John
is bearded.' The word 'John' in this sen-
tence is a bit of language - a name of
some individual human being - and no
one would be tempted to confuse the word
with what it names. Consider now the
expression 'is bald'. This too is a bit of lan-
guage - philosophers call it a 'predicate' -
and it brings to our attention some prop-
erty or feature which, if the sentence is
true, is possessed by John. Understood in
this way, a property is not itself linguistic
though it is expressed, or conveyed, by
something that is, namely a predicate.
What might be said is that a property is a
real feature of the world, and that it should
be contrasted just as sharply with any pre-
dicates we use to express it as the name
'John' is contrasted with the person
himself. However, it is a matter of great
controversy just what sort of ontological
status should be accorded to properties (see
ONTOLOGY), so one should treat this parti-
cular formulation of the contrast with
caution. What is important, though, is to
recognize that, in terms of the natural
divide between language and what lan-
guage is about, predicates fall on the lan-
guage side and properties on the other.
It should be mentioned here that proper-
ties (and predicates) can be more compli-
cated than the above example allows. For
instance, in the sentence, 'John is married
to Mary', we are attributing to John the
property of being married, and, unlike the
property of being bald, this property of John
is essentially relational. The predicate ('is
married to') which expresses this property
requires two names (for example, 'John' and
'Mary') in order to make up a complete sen-
486
tence, whilst 'is bald' only requires a single
name. (Predicates like 'is bald' are called
monadic, and those like 'is married to'
dyadic. Moreover, it is commonly said that
'is married to' expresses a relation, rather
than a property, though the terminology
is not fixed here. Some authors speak of
relations as different from properties in
being more complex but like them in being
non-linguistic, though it is more common
to treat relations as a sub-class of proper-
ties.)
The second notion to contrast with prop-
erty is CONCEPT, but one must be very
careful here, since 'concept' has been used
by philosophers and psychologists to serve
many different purposes. One use has it that
a concept is something mental, it is a
certain way of conceiving of some aspect of
the world. As such, concepts have a kind of
subjectivity: two different individuals might,
for example, have different concepts of birds,
one thinking of them primarily as flying
creatures and the other as feathered. Con-
cepts in this sense are often described as a
species of MENT AL REPRESENT ATION, and as
such they stand in sharp contrast to the
notion of a property, since a property is
something existing in the world. Also, given
what was said above about the non-
linguistic nature of properties, any under-
standing of concepts as linguistic would
make it impossible to identify concepts and
properties. However, it is possible to think of
a concept as neither mental nor linguistic,
and this would allow, though it doesn't
dictate, that concepts and properties are the
same kind of thing.
SAMUEL GUTTENPLAN
proposition When an English speaker says
'it is raining', and a French speaker says 'il
pleut' , it is natural to claim that they are
saying the same thing, but it is not all that
clear what that 'thing' is. It can scarcely be
the sentence that each utters, for these are
markedly different. Also, whilst it is true
that these sentences mean the same, it is
odd to say that they are 'saying the same
meanings'. Indeed, aside from oddness, any
appeal to meaning here seems to put the
cart before the horse. For one would think
that two sentences mean the same when
they are used to say the same thing. and
this brings us back to the question of what
thing it is that is said.
Almost universally, the first step in clear-
ing up the problem is to use the word 'pro-
position': we can say that the speakers in
the above example used sentences in their
own languages to express the same proposi-
tion. But now we need to be told what a
proposition is, and this is no simple matter.
Two ways of going about it can be found in
the philosophical literature. The first is
based on the notion of a POSSIBLE WORLD.
The actual world consists of all actual
things, properties and relations. But one can
imagine other worlds which differ from the
actual world in respect of the things they
contain and/or in respect of properties and
relations characterizing those things. Thus,
there are possible worlds in which I am
sitting with both feet on the floor now
instead of with legs crossed, as well as
others in which I don't exist at all. Indeed,
it is obvious that once you think of possible
worlds in this way, there will be an infinite
number of them, each slightly or greatly dif-
ferent from the others and from the actual
world. Consider now the sentence: 'Grass is
green.' In some worlds - including the
actual world - this will be true. In others, it
will be false - either grass will be some
other colour, or, perhaps, there will be no
such thing as grass. Thus, the sentence par-
titions all possible worlds into two sorts:
those in which the sentence is true, and
those in which it is false. And we can define
the proposition expressed by the sentence
'grass is green' as the set of possible worlds
in which the sentence is true.
The second way to deal with propositions
is to regard them as having a structure that
mirrors the structure of sentences. On this
conception, a proposition is an ordered set
of things and their properties and/or rela-
tions. Moreover, the specific constituents of
a given proposition structurally correspond
to the sentence expressing the proposition.
Thus the proposition expressed by 'grass is
PROPOSITION
green' is a special sort of abstract entity
consisting of grass, the property of being
green, and the relation between these that
makes it true that grass is green. It is felt by
some writers that a notion of proposition
that allows such structure has advantages
in certain cases over the possible-worlds
conception. For example, consider the sen-
tence 'The number three is odd.' Numbers
are plausibly regarded as having their prop-
erties essentially, Le. as keeping them in all
possible worlds. After all, how could some-
thing be the number three without being
odd? For this sort of reason, the sentence
'The number three is odd' will be true in
exactly the same worlds as the sentence:
'The number three is prime', and thus, on
the possible worlds conception these two
sentences will express the same proposition.
But this seems unacceptable: someone could
surely assert one and not the other - the
sentences just don't seem to express the
same thing. However, on the structured-
entity conception of, proposition, this is
easily handled: 'odd' and 'prime' pick out
different properties, so the constituents of
each relevant proposition will be different.
In the philosophy of mind, propositions
understood in one way or another have
figured centrally in attempts to understand
INTENTION ALITY. As described above, a
proposition, on one or another construal, is
what a sentence expresses. However, atti-
tudes have CONTENTS and it is natural to
regard these contents as propositions.
Indeed, attitudes are often simply called
PROPOSITION AL ATTITUDES. Of course,
when someone believes or desires some-
thing, there need be no sentence uttered,
and one cannot therefore say that a propo-
sition is what a belief or a desire expresses.
Instead it is usual to regard the attitudes as
mental states directed either toward a sen-
tence that expresses a proposition, or
toward a proposition itself. Of course, estab-
lishing the relations between attitudes and
propositions is only the first step in attempt-
ing to describe intentionality and this whole
area is fraught with intricate difficulties.
SAMUEL GUTTENPLAN
487
PROPOSITION AL ATTITUDES
propositional attitudes Our everyday con-
ception of mentality bristles with notions
like BELIEF, DESIRE, INTENTION, hope, fear,
wish. Philosophers group these together
and call them 'propositional attitudes'.
What links these attitudes is the fact that
they are identified by their propositional
contents: a belief that snow is white is iden-
tified by the proposition that snow is white.
Since propositional contents are attributed
in English by 'that'-clauses, two people's
beliefs, say, are counted as different if they
are correctly identified by non-equivalent
'that' -clauses.
Propositional attitudes are woven into
almost all aspects of ordinary life. Legal,
social, political and economic practices
would be unthinkable without the attitudes
(see Baker, 1987). For example, nothing
would be a contract if the parties to it
lacked attitudes - such as belief that they
were incurring certain obligations. No one
could be held responsible for anything in
the absence of attitudes about, among other
things, what one is doing. In general, we
make sense of behaviour in terms of atti-
tudes: Smith turned up at 4 o'clock, because
she thought that the meeting had been
scheduled for that time. Reasons for action
are intimately tied to attitudes (see ACTION).
Indeed, some philosophers (such as Donald
Davidson) take the attitudes to be definitive
of the mental.
On the one hand, propositional attitudes
are pervasive in our descriptive, explan-
atory, and justificatory practices. On the
other hand, as mental states, they seem
problematic to contemporary philosophers
seeking insurance against DUALISM. Such
philosophers, perhaps a majority, think that
propositional attitudes stand in need of
'vindication' either by materialistic meta-
physics or by science. In pursuit of such
vindication, they have raised a number of
questions about propositional attitudes, of
which I shall discuss three that remain
unresolved after intense investigation. (1)
How are attitudes identified by content
related to presumably less problematic phys-
ical states? (2) Given that brain states cause
bodily movements, can propositional atti-
488
tudes also have a causal role in behaviour?
(3) In what way, if at all, will propositional
attitudes figure in a comprehensive scientific
psychology?
THE NATURALIZATION PROJECT
The first question - How are attitudes iden-
tified by content related to physical states? -
has given rise to efforts to provide condi-
tions, in naturalistic terms, for an internal
state to have propositional content.
Roughly, to require that the conditions be
given in naturalistic terms is to require that
they be statable without the use of 'that'-
clauses and without the use of semantic
terms like 'denotes', 'refers to', or 'means
that'. Those engaged in naturalization are
investigating a special case of the general
question: how can one physical state repres-
ent another? Under what physical condi-
tions does a state have a certain content?
How can a state be 'directed upon' a state of
affairs - one that may not even obtain? (See
CONTENT; INTENTION ALITY.)
Not only are contents intentional-with-a-
't' (Le. directed upon a state of affairs that
mayor may not obtain), but also attribu-
tions of contents are intensional-with-an-'s':
substitution of co-referring terms may
change the truth value of the attribution.
For example, from the attribution to Smith
of a belief that the robber fled the scene, and
the fact that the robber is the mayor, we
cannot infer that Smith believes that the
mayor fled the scene. If Smith is unaware
that the robber is the mayor, she may have
the first belief, but lack the second. Contents
(and hence beliefs) are thus sensitive to the
way things are described.
The naturalizer ties propositional attitudes
to types or tokens of internal states. (A type
is a kind of state; a token is a particular
spatiotemporal instance of a kind of state.)
(See IDENTITY THEORIES; SUPERVENIENCE.)
The idea is that people token internal states
that have certain 'shapes' or other non-
semantic properties, and these states have
content in virtue of correlations of one sort
or another between their tokenings and
external conditions. The naturalization
project is to spell out the kind of correlation
that confers content. There are several ver-
sions, of which I shall describe three. Each
traces content back to what causes tokens of
a certain type (aetiological theories) and/or
to what is the function of tokens of a certain
type (teleological theories).
On Ruth Millikan's (1984) 'teleofunc-
tional' evolutionary account, contents of
attitudes derive from the functions of
mechanisms designed by natural selection
to make and to use abstract 'maps' or 'dia-
grams' or 'mental sentence' pictures of the
world in order to produce actions appro-
priate to that world. These mechanisms
have biologically normal ways of accom-
plishing the abstract map-making and map-
using tasks, but may also fail quite often,
just as the mouse-tracking abilities of the
house-cat may fail in particular cases. Then
maps are produced that are 'wrong', given
how the system is designed to use them. A
question for this approach is whether the
psychological mechanisms of concept- and
belief-formation in humans, as designed by
evolution, are determinate enough, as
regards their historical biologically normal
ways of functioning, to yield the determi-
nacy we find in belief content. (For a differ-
ent kind of teleological functionalism. see
Lycan, 1988.)
A second approach to naturalizing the
attitudes looks to lawful covariation of
tokenings of certain states and certain
external conditions. Certain states are
natural indicators of other states: for
example, the number of tree rings indicates
the age of the tree. There is a general causal
dependency of the indicator on what it indi-
cates. Another way to put the point is to
say that the number of tree rings carries
information about the age of the tree: under
optimal or ideal conditions, the number of
tree rings covaries with the age in years of
the tree. DRETSKE and others have used the
notions of carrying information and of indi-
cating something (Dretske, 1988) to show
how physical states could have content or
meaning.
A difficulty for this approach to natur-
alization is to account for the possibility of
PROPOSITIONAL ATTITUDES
error. Suppose that tokens of a certain type
have regularly been caused by the presence
of dogs; but on some occasion, a token of
that type is caused by a cat. Then, if the
content of a token is determined by what
information it carries (Le. by what caused
it), there is no way for the token to mis-
represent a dog as a cat. Here are two ways
in which 'indicator' theories may try to
meet the problem of error. (i) Specify ideal
conditions. Identify the content of a token
by its cause in ideal conditions; then con-
strue misrepresentation in terms of what
causes tokens of the type in non-ideal con-
ditions. (ii) Specify a learning period, during
which tokens of a certain type acquire a
given content. Identify the content with the
causes of the tokens during the learning
period; then construe misrepresentation in
terms of post-learning-period causes that
differ from learning-period causes of tokens
of the given type. A central difficulty for the
'indicator' theories is to specify non-circu-
larly either ideal conditions or conditions
that define a learning period. Neither non-
circular specification has been forthcoming.
The deepest difficulty for the naturaliza-
tion project is what FODOR has called the
'disjunction problem'. The disjunction
problem is a generalization of the problem
of error: roughly, if the content of a token is
determined by what causes it, then why is
not the content of tokens of a certain type a
disjunction of the indefinitely many kinds
of causes that produce tokens of that type?
For example, if tokens of type T are caused
at different times by cows, horses and
thoughts about cowboys, then why doesn't
T have the content cow-or-horse-or-
thoughts-about-cowboys?
Fodor's solution - and this is the third
approach to naturalizing the attitudes - is
to specify content in terms of asymmetric
dependence. Fodor (1987, 1990), develops
this approach in conjunction with his pos-
tulation of a LANGUAGE OF THOUGHT. The
general idea is that a symbol in the lan-
guage of thought - say, 'c' - expresses the
property of being a cow if and only if (i) it is
a law that cows cause tokens of 'C', (ii)
some tokens of 'C' are caused by non-cows
489
(horses, say); and (iii) if cows did not cause
'C's, then neither would non-cows; but
cows would still cause 'C's even if non-cows
did not cause 'C's.
None of these reductive accounts of the
attitudes has found full acceptance. Many
philosophers simply assume that the atti-
tudes supervene on some physical states or
other, whether or not anybody can specify
any supervenience base for any proposi-
tional attitude, and move on to other issues
(see SUPERVENIENCE).
THE PROBLEM OF MENTAL CAUSATION
One of the issues to which philosophers have
recently turned is the second of the ques-
tions: Given that brain states cause bodily
movements, can propositional attitudes also
have a causal role in behaviour? The
problem of mental causation (Kim, 1988) is
to find a causal role for propositional
attitudes in the aetiology of behaviour.
One way to see the problem is to consider
DA VIDSON 's view of mental events as physi-
cal events described in the vocabulary of
propositional attitudes. This is a kind of
non-reductive materialism, since each
mental token is a physical token, but
mental types are not identified with physical
types. On Davidson's view of causation, pro-
positional attitudes are causes of behaviour,
even though (i) all causal transactions are
governed by strict laws, and (ii) there are no
strict laws between propositional attitudes
and the physical events that constitute the
behaviour. The strict laws that subsume the
mental causes and the physical effects are
wholly physical.
This view has suggested to some a
problem of mental causation. For, on this
view, propositional attitudes (mental events)
have effects in virtue of their physical prop-
erties but not in virtue of their having
content or of their being propositional atti-
tudes. The fact that neural events are
describable as propositional attitudes is ir-
relevant to what they cause, or to the fact
that they cause anything at all; for the
physical properties of a mental event pre-
empt or 'screen off' the mental properties.
490
Mental events, so the charge goes, are caus-
ally impotent.
This line of argument has elicited the
following response. First, one may appeal to
non-strict, or hedged, laws (laws with open-
ended ceteris paribus clauses), and take prop-
erties mentioned in the antecedents of such
laws to be causally relevant to producing
instances of properties mentioned in the
consequents. Hedged laws are common
throughout the special sciences (Fodor,
1987). The existence of laws of the form 'If
S believes that doing A will bring it about
that p, and S wants to bring it about that p,
and ... , then, ceteris paribus, S will do A'
would suffice to secure a causal role for pro-
positional attitudes, on this Fodorean view.
Those worried about mental causation will
point out that even if propositional attitudes
are causes, citing 'ceteris paribus' laws does
not show that they bring about their effects
in virtue of having their contents. (See
REASON AND CAUSES.)
For example, Jaegwon Kim has argued for
a principle of explanatory exclusion, accord-
ing to which there cannot be two or more
complete and independent explanations of a
single event. If there seem to be two com-
plete and independent explanations of an
event (e.g. one physical and one mental),
then either at least one is not complete, or
they are not independent. On the basis of
certain metaphysical assumptions - such as
that every physical event has a complete
physical cause, and that every event is (or
supervenes on) a physical event - Kim
argues that every event has a completely
sufficient physical cause, and hence a com-
plete explanation in terms exclusively of
physical properties. If any 'other' properties
(such as propositional attitudes) seem also
to explain an event, Kim concludes, they
must be identical with, or reducible to,
physical properties. If Kim is right, there is
no 'logical space' for non-reductive materi-
alism of either the Davidsonian or Fodorean
variety.
There are difficult metaphysical issues
here about supervenience and reducibility
(see REDUCTION). For example, Kim's posi-
tion seems to have unwanted consequences.
For the grounds that he gives to show that
there is a problem of mental causation are
equally good grounds to show that there is
a problem of macrocausation generally (see
PHYSICALISM (2)). Suffice it to say that the
verdict is not yet in on the problem of
mental causation.
THE SCIENTIFIC STATUS OF
The third and final question - In what way,
if at all, will propositional attitudes figure in
a comprehensive scientific psychology? -
concerns the scientific status of the proposi-
tional attitudes. Many philosophers hold
that the fate of the propositional attitudes
turns on the outcome of scientific psycho-
logy (see FOLK PSYCHOLOGY). If the explana-
tory kinds of scientific psychology turn out
to be radically different from the attitudes,
such philosophers hold, then the attitudes
should go the way of phlogiston and
witches.
Although a conclusive answer to the
third question must await the outcome of
scientific psychology, there has been im-
portant philosophical work done already on
the issue of individuation. Say that a
scheme of individuation is relational or wide
if it classifies propositional attitudes in part
by reference to the cognizer's environment,
so that cognizers in different environments
may have different attitudes even if they are
molecule-for-molecule duplicates. (see EX-
TERNALISM!INTERNALISM; PUTNAM; TWIN
EARTH). Say that a scheme of individuation
is non-relational or narrow if it classifies
propositional attitudes so that molecule-for-
molecule duplicates necessarily have the
same attitudes, regardless of differences in
their environments.
Are propositional attitudes classified or
individuated in a way that would put them
at odds with entities countenanced by a sci-
entific psychology? Here are four prominent
views on the amenability of attitudes for
incorporation into scientific psychology. (i)
Propositional attitudes are individuated
relationally, and are, therefore, ill-suited for
science (Stich, 1983). (ii) Propositional atti-
tudes are individuated relationally, but still
are suitable for science (Burge, 1986). (iii)
Propositional attitudes are individuated
relationally, but they have narrow contents
('in the head') that are suitable for science
(Fodor, 1987). (iv) Propositional attitudes
are individuated non-relation ally and are
not, therefore, ill-suited for science (Lewis,
1983). Each of these positions remains con-
troversial.
Consider the 'worst case'. Suppose that
we had a comprehensive theory of human
behaviour - say, one based on ARTIFICIAL
INTELLIGENCE models known as 'parallel
distributed processing' (see CONNECTION-
ISM). Suppose further that not only does the
scientific theory fail to postulate internal
states that have propositional or narrow
content, but also that the states that it does
postulate do not even loosely correlate with
attitudes. Although this is extremely vague,
the point is to suppose that 'mature scien-
tific psychology' turns out in such a way
that even under the loosest interpretation of
'vindication', we would all agree that
mature scientific psychology failed to vindi-
cate the attitudes.
The reason that vindication of the atti-
tudes by science is so desired is that, failing
vindication, only two alternatives are envis-
aged: we must either take propositional atti-
tudes to resist incorporation into the
physical world of science, and embrace
dualism (widely considered untenable), or
we must give up propositional attitudes as
fictions, and embrace eliminative materi-
alism.
ELIMIN ATIVE MATERIALISM AND BEYOND
Eliminative materialists, who deny the
reality of propositional attitudes altogether,
take a variety of positions. For example,
QUINE (1960) takes the language of propo-
sitional attitudes to be a 'dramatic idiom'.
Churchland (1989) envisages a new kind of
everyday language that lacks commitment
to propositional attitudes. Typically, elim-
inative materialists hold the view that the
natural sciences are the exclusive arbiter of
what there is, and that propositional atti-
491
tudes fail to measure up to the standards of
science.
An eliminative materialist whose position
rests upon the assumed fact that scientific
psychology will not show how attitudes 'fit'
into the physical world, it would seem,
ought to be an eliminativist about proper-
ties of middle-sized objects (like pillows and
predators) as well. If by 'the physical world'
one means the world described by funda-
mental physics, then we have no idea how
predators fit into the physical world either:
could a characterization of predators be
given in the language of microphysics? If by
'the physical world' one means the world
described non-intentionally and non-
semantically, then pillows and other arte-
facts fail to fit into the physical world:
there is no physically specifiable fact in
virtue of which something is a pillow; what
makes something a pillow depends in part
on facts about its design and use - facts that
we do not know how to specify in a non-
intentional vocabulary. But these are not
great philosophical discoveries that should
motivate eliminativism about pillows or
predators.
DENNETT used to be an 'instrumentalist',
who appreciated the usefulness of belief-
attributions while denying the reality of
beliefs and desires. Such a position has
seemed to many to be unstable, and
Dennett himself has backed away from it
(Dennett, 1991). Now he grants to the atti-
tudes a kind of reality - the reality that pat-
terns of dots may have - and we may look
forward to further elaborations of his views
in the future.
On the reigning pretheoretical conception
of the propositional attitudes, there are
beliefs only if there are internal states that
satisfy the open sentences of the form 'x is a
belief that p'. Putting aside Cartesian
dualism, the idea is that if an attribution of
an attitude is true, there is some particular
neural state (at least partly) in virtue of
which it is true. This conception, call it 'the
standard view', is held by those who take
attitudes (if there are any) to covary in a
systematic way with neural states of believ-
ers. Eliminative materialists endorse the
492
standard view when they infer that there
are no attitudes from the supposed fact that
the best scientific theories do not postulate
internal states that correspond to attribu-
tions of attitudes.
The standard view is responsible, I
believe, not only for eliminative materialism
but also for the three intractable questions
that I have discussed - questions that would
lose their urgency in the absence of the
standard view. The standard view, however,
is neither metaphysically innocent, nor is it
embedded in our ordinary attributions of
attitudes. Perhaps it is time for philosophers
to attend to the standard view itself, and to
rethink what we are doing when we
attribute to each other propositional
attitudes. Shifting focus to the explanatory
(and other) uses to which we put attribu-
tions of propositional attitudes may circum-
vent the metaphysical conundrums just
surveyed.
See also An Essay on Mind section 2.1 ;
FUNCTIONALISM; THOUGHTS.
BIBLIOGRAPHY
Baker, L.R. 1987. Saving Belief: A Critique of
Physicalism. Princeton University Press.
Burge, T. 1986. Individualism and psychol-
ogy. Philosophical Review, 95, 3-45.
Churchland, P.M. 1989. A Neurocomputational
Perspective: The Nature of Mind and the Struc-
ture of Science. Cambridge, MA.: MIT Press/
Bradford Books.
Events. Oxford: Clarendon Press.
Dennett, D.C. 1991. Real patterns. Journal of
Dretske, F. Explaining Behavior: Reasons in a
World of Causes. Cambridge, MA.: MIT
Fodor, J.A. 1987. Psychosemantics: The Problem
of Meaning in the Philosophy of Mind.
Fodor, J.A. 1990. A Theory of Content and Other
Essays. Cambridge, MA: MIT Press/Bradford
Books.
Kim, J. 1988. Explanatory realism, causal
realism and explanatory exclusion. Midwest
Studies in Philosophy, vol. 12, Minneapolis:
University of Minnesota Press. pp. 225-40.
Lewis. D. 1983. Philosophical Papers. Vol. 1.
Lycan. W.G. 1988. Judgement and Justification.
Millikan. R.G. 1984. Language. Thought and
Other Biological Categories: New Foundations
for Realism. Cambridge. MA.: MIT Press/
Bradford Books.
Stich. S.F. 1983. From Folk Psychology to Cog-
nitive Science: The Case Against Belief. Came
bridge. MA.: MIT Press/Bradford Books.
LYNNE RUDDER BAKER
psychoanalytic explanation The task of
analysing psychoanalytic explanation is
complicated initially in several ways. One
concerns the relation of theory to practice.
There are various perspectives on the rela-
tion of psychoanalysis. the therapeutic prac-
tice. to the theoretical apparatus built
around it. and these lead to different views
of psychoanalysis' claim to cognitive status.
The second concerns psychoanalysis' legit-
imation. The way that psychoanalytic
explanation is understood has immediate
implications for one's view of its truth or
acceptability. and this is of course a notor-
iously controversial matter. The third is
exegetical. Any philosophical account of
psychoanalysis must of course start with
Freud himself. but it will inevitably privilege
some strands in his thought at the expense
of others. and in so doing favour particular
post-Freudian developments over others.
A plausible view of these issues is as
follows. Freud clearly regarded psycho-
analysis as engaged principally in the task
of explanation. and held fast to his claims
for its truth in the course of alterations in
his view of the efficacy of psychoanalytic
treatment. Some of psychoanalysis' advo-
cates have. under pressure. retreated to the
view that psychoanalytic theory has merely
instrumental value. as facilitating psycho-
analytic therapy; but this is not the natural
view. which is that explanation is the
autonomous goal of psychoanalysis. and
that its propositions are truth-evaluable.
PSYCHOANALYTIC EXPLANATION
Accordingly. it seems that preference should
be given to whatever reconstruction of
psychoanalytic theory does most to advance
its claim to truth; within. of course. ex-
egetical constraints (what a reconstruction
offers must be visibly present in Freud's
writings).
PSYCHOANALYTIC EXPLANANDA
The explananda of psychoanalysis require
some comment. They may be divided. first.
into primary and secondary explananda.
The latter include art. morality, religion and
other cultural phenomena for which Freud
offered explanations. They are secondary
because psychoanalytic explanation in these
areas depends for its plausibility on the
theory's success in dealing with the psycho-
logical phenomena of individuals which are
psychoanalysis' primary explananda.
Ultimately. the object of psychoanalytic
explanation is nothing less than the entire
shape of a person's life (see Wollheim.
1984). but the theory is formulated initially
in its application to the phenomena that
Freud described as 'gaps' in CONSCIOUS-
NESS. By this phrase Freud meant to indi-
cate those psychological phenomena that
present ordinary psychology with puzzles of
explanation - actions and experiences with
an irrational, or at least non-rational char-
acter.
This last distinction is important. Dreams
and parapraxes (slips of the tongue. bungled
actions such as accidentally dropping an
object, exceptional lapses of memory) are
phenomena about whose explanation ordin-
ary psychology has nothing very helpful to
say. and which provide essential material for
the formulation of psychoanalytic theory.
But such phenomena do not strictly evince
irrationality on the part of the subject. for
they do not stand out in commonsense psy-
chology as violating norms of rationality (see
FOLK PSYCHOLOGY). In this respect they are
merely non-rational, and may be compared
with such phenomena as individual char-
acter or change of mood. which common-
sense psychology also tends to recognize
without endeavouring to explain.
493
Non-rational gaps in commonsense psy-
chology contrast sharply with the irrational,
intrusive symptoms of hysterics and neur-
otics, whose inappropriate emotions and
disorders of thought fill Freud's case histor-
ies (Dora's ailments, Little Hans' phobia, the
obsessions of the Ratman and Wolfman).
Here it is clear that the subject is not just in
error or incompetent, or undergoing an
unusual experience supplementary to ordin-
ary rational self-consciousness, like dream:
rather, she violates rational norms and
encounters opacity in her own, fully self-
conscious experience of herself. Sympto-
matic conditions contrast, moreover, with
the familiar, common-or-garden irration-
ality of SELF-DECEPTION and akrasia (see
WEAKNESS OF WILL), for which failures of
reason ordinary psychology is by and large
able to propose adequate explanations.
Because irrational phenomena of the
symptomatic kind allow us to see psycho-
analytic explanation as motivated by prob-
lems internal to ordinary psychology, there
is reason to conclude that it is irrational
phenomena which provide the principal
warrant for psychoanalytic explanation.
From this starting-point, psychoanalytic
theory may be seen as at once growing in
scope and gaining evidential support: the
theory, developed in one context, is re-
cruited to other contexts, which in turn
suggest its development in certain direc-
tions, and provide data that enable its attri-
butions to be cross-checked from several
angles, as any interpretative procedure
ideally permits (see Wollheim, 1991,
Preface; Hopkins, 1991, 1992). Such a
process of expansion and consolidation may
plausibly be claimed to fit the actual trajec-
tory of Freud's theoretical development.
CAUSAL MECHANISMS OR CONNECTIONS
OF MEANING?
It will now help to take a historical route,
and focus on the terms in which analytical
philosophers of mind began to discuss
seriously psychoanalytic explanation. These
were provided by the long-standing, and
presently un concluded, debate over cause
494
and meaning in psychology (the debate over
psychoanalysis' scientificity is held over for
comment at the end).
It is not hard to see why psychoanalysis
should be viewed in terms of cause and
meaning. On the one hand, Freud's theories
introduce a panoply of concepts which
appear to characterize mental processes as
mechanical and non-meaningful. Included
here are Freud's neurological model of the
mind, outlined in his 'Project for a scientific
psychology'; more broadly, his 'economic'
description of the mental, as having proper-
ties of force or energy (e.g. as 'cathexing'
objects); and his accounts of the mechanism
of repression. So it would seem that psycho-
analytic explanation employs terms logically
at variance with those of ordinary, com-
monsense psychology, where mechanisms
do not play a central role. But on the other
hand, and equally strikingly, there is the
fact that psychoanalysis proceeds through
interpretation, and engages on a relentless
search for meaningful connections in
mental life - something that even a super-
ficial examination of The Interpretation of
Dreams, or The Psychopathology of Everyday
Life, cannot fail to impress upon one. Psy-
choanalytic interpretation adduces mean-
ingful connections between disparate and
often apparently dissociated mental and
behavioural phenomena, directed by the
goal of 'thematic coherence', of giving
mental life the sort of unity that we find in
a work of art or cogent narrative. In this
respect, psychoanalysis would seem to
adopt as its central plank the most salient
feature of ordinary psychology, its insistence
on relating actions to reasons for them
through contentful characterizations of
each that make their connection seem
rational, or intelligible; a goal that seems
remote from anything found in the physical
sciences (see CONTENT; RATIONAUTY;
REASONS AND CAUSES.)
The application to psychoanalysis of the
perspective afforded by the cause-meaning
debate can also be seen as a natural con-
sequence of another factor, namely the
semi-paradoxical nature of psychoanalysis'
explananda. With respect to all irrational
phenomena, something like a paradox
arises (see Davidson, 1982). Irrationality
involves a failure of rational connectedness
and hence of meaningfulness, and so, if it is
to have an explanation of any kind, rela-
tions that are non-meaningful and causal
appear to be needed. And yet, as observed
above, it would seem that, in offering expla-
nations for irrationality - plugging the
'gaps' in consciousness - what psycho-
analytic explanation hinges on is precisely
the postulation of further, albeit non-
apparent, connections of meaning.
For these two reasons, then - the logical
heterogeneity of its explanations and the
ambiguous status of its explananda - it may
seem that an examination in terms of the
concepts of cause and meaning will provide
the key to a philosophical elucidation of
psychoanalysis. The possible views of
psychoanalytic explanation that may result
from such an examination can be arranged
along two dimensions. (1) Psychoanalytic
explanation may then be viewed, after
reconstruction, as either causal and non-
meaningful; or meaningful and non-causal;
or as comprising both meaningful and causal
elements, in various combinations. Psycho-
analytic explanation then may be viewed, on
each of these reconstructions, as either
licensed or invalidated, depending on one's
view of the logical nature of psychology.
So, for instance, some philosophical dis-
cussions infer that psychoanalytic explana-
tion is void, simply on the grounds that it is
committed to causality in psychology. On
another, opposed view, it is the virtue of
psychoanalytic explanation that it imputes
causal relations, since only causal relations
can be relevant to explaining the failure of
meaningful psychological connections. On
yet another view, it is psychoanalysis' com-
mitment to meaning which is its great fault:
it is held that the stories that psychoanalysis
tries to tell do not really, on examination,
add up to coherent wholes, and so do not
explain successfully. And so on.
lt is fair to say that the debates between
these various positions fail to establish any-
thing definite about psychoanalytic expla-
nation. There are two reasons for this. First,
PS YCHOAN AL YTIC EXPLAN A TION
there are several different strands in Freud's
writings, each of which may be drawn on,
apparently conclusively, in support of each
alternative reconstruction. Second, pre-
occupation with a wholly general problem
in the philosophy of mind, that of cause and
meaning, distracts attention from the distin-
guishing features of psychoanalytic expla-
nation. At this point, and in order to
prepare the way for a plausible reconstruc-
tion of psychoanalytic explanation, it is
appropriate to take a step back, and take a
fresh look at the cause-meaning issue in the
philosophy of psychoanalysis.
Suppose, first, that some sort of cause-
meaning compatibilism - such as that of
Donald DAVIDSON - holds for ordinary psy-
chology. On this view, psychological expla-
nation requires some sort of parallelism of
causal and meaningful connections, groun-
ded in the idea that psychological properties
play causal roles determined by their
content. Nothing in psychoanalytic expla-
nation is inconsistent with this picture: after
his abandonment of the early 'Project',
Freud exceptionlessly viewed psychology as
autonomous relative to neurophysiology,
and at the same time as congruent with a
broadly naturalistic world-view (see
NATURALISM). If psychoanalytic explana-
tion gives the impression that it imputes
bare, meaning-free causality, this results
from attending to only half the story, and
misunderstanding what psychoanalysis
means when it talks of psychological
mechanisms. The economic descriptions of
mental processes that psychoanalysis pro-
vides are never replacements for, but rather
themselves always presuppose, characteriza-
tions of mental processes in terms of
meaning. Mechanisms in psychoanalytic
contexts are simply processes whose opera-
tion cannot be reconstructed as instances of
rational functioning (they are what we
might by preference call mental activities,
by contrast with actions; see Wollheim,
1991, Preface). Psychoanalytic explana-
tion's postulation of mechanisms should not
therefore be regarded as a regrettable and
expungable incursion of scientism into
Freud's thought, as is often claimed.
495
PSYCHOANALYTIC EXPLAN A TION
Suppose, alternatively, that hermeneuti-
cists such as Habermas - who follow
Dilthey in regarding the understanding of
human beings as an interpretative practice
to which the concepts of the physical sci-
ences, such as cause, are alien - are correct
in thinking that connections of meaning are
misrepresented through being described as
causal. Again, this does not impact nega-
tively on psychoanalytic explanation since,
as just argued, psychoanalytic explanation
nowhere imputes meaning-free causation.
Nothing is lost for psychoanalytic explana-
tion if causation is excised from the psycho-
logical picture.
The conclusion must be that psycho-
analytic explanation is at bottom indifferent
to the general meaning-cause issue. The
core of psychoanalysis consists in its tracing
of meaningful connections with no greater
or lesser commitment to causality than is
involved in ordinary psychology, (which
helps to set the stage - pending appropriate
clinical validation - for psychoanalysis to
claim as much truth for its explanations as
ordinary psychology). But the discussion
also brings to light what is, surely, the true
key to psychoanalytic explanation: its attri-
bution of special kinds of mental states, not
recognized in ordinary psychology, whose
relations to one another do not have the form
of patterns of inference or practical reasoning.
In the light of this, it is e_asy to under-
stand why some compatibilists and her-
meneuticists assert that their own view of
psychology is uniquely consistent with
psychoanalytic explanation. Compatibilists
are right to think that, in order to provide
for psychoanalytic explanation, it is neces-
sary to allow mental connections that are
unlike the connections of reasons to the
actions that they rationalize, or to the
beliefs that they support; and that, in out-
lining such connections, psychoanalytic
explanation must outstrip the resources of
ordinary psychology, which does attempt to
force as much as possible into the mould of
PRACTICAL REASONING. Hermeneuticists,
for their part, are right to think that it
would be futile to postulate connections
which were nominally psychological but
496
not characterized in terms of meaning, and
that psychoanalytic explanation does not
respond to the 'paradox' of irrationality by
abandoning the search for meaningful con-
nections.
Compatibilists are, however, wrong to
think that non-rational but meaningful
connections require the psychological order
to be conceived as a causal order. The her-
meneuticist is free to postulate psychological
connections that are determined by
meaning but not by rationality: it is coher-
ent to suppose that there are connections of
meaning that are not bona fide rational
connections, without these being causal.
Meaningfulness is a broader concept than
rationality. (Sometimes this thought has
been expressed, though not helpfully, by
saying that Freud discovered the existence
of 'neurotic rationality'.) Although an
assumption of rationality is doubtless neces-
sary to make sense of behaviour in general,
it does not need to be brought into play in
making sense of each instance of behaviour.
Hermeneuticists, in turn, are wrong to
think that the compatibilist's view of psy-
chology as causal signals a confusion of
meaning with causality, or that it must lead
the compatibilist to deny that there is any
qualitative difference between rational and
irrational psychological connections.
THE FORM OF PSYCHOANALYTIC
EXPLANATION: WISH AND PHANTASY
The next task must evidently be to describe
more fully the nature of the non-rational
meaningful connections that figure in psy-
choanalytic explanation.
The naive view of psychoanalysis'
restoration of meaningful connections in
mental life refuses to acknowledge its dis-
tance from ordinary psychology: it sees
psychoanalysis as attributing unconscious
practical syllogisms, in the premises of
which unconscious beliefs and desires are
put to work. In this spirit one might try to
view phobias, for example, as explained by
a false belief about the danger constituted
by an external object, which is falsely
believed to be identical with some (e.g.
oedipal) threat to one's physical safety, and
consequently a proper object of fear and
avoidance. Or, one might view hysterical
symptoms as explained by false beliefs to the
effect that sexual desires may be satisfied by,
or traumatic experiences undone by, the
physical incarnation of those desires or
experiences.
There are some explanatory purposes
which may well be served by the attribution
of unavowed pieces of practical reasoning
(as in 'revealed preference' theory), and
such attributions may indeed have a sub-
ordinate place in the architecture of psycho-
analytic explanation. But the examples
above show lucidly that the core of psycho-
analytic explanation cannot be viewed as
taking this form. The reason for this is
simply that the proposed syllogistic recon-
structions do no more than highlight,
without making any more intelligible, the
real explananda: where do such irrational
desires and beliefs come from, and why are
they not integrated into, and so dissolved
away by rational mental functioning?
(Note that the practical syllogism model
cannot be saved by saying that what figure
as minor premises in psychoanalytic expla-
nations are phantasies instead of beliefs, for
we have no more understanding of how
phantasy, in any ordinary sense of the term,
may combine with DESIRE to produce a
reason for action, than we have of how
something that one merely imagines may
do so.)
The lesson to draw is, once again, that
the true form of psychoanalytic explanation
makes a clean break with practical reason-
ing, and can only be understood in terms of
Freud's complex account of unconscious
mental functioning (see THE UNCONSCIOUS).
There are many, equally cogent ways of
organizing Freud's theories of the uncon-
scious, but they all revolve around a single
fundamental supposition, to the effect that
MENTAL REPRESENTATIONS in the uncon-
scious are formed in direct response to the
person's basic sources of motivation, and
without an interest in truth. Unconscious
processing registers reality only obliquely,
principally insofar as reality fails to allow
the satisfaction of desire or occasions
mental conflict. In response to this aware-
ness of desire as unsatisfied, or of the self as
anxiously conflicted, self and world are pic-
tured Wishfully in the unconscious: as
though the mind were able to remake the
world immediately and without action,
through mere force of Wish, in such a way
that the world is portrayed as meeting
one's needs and one's mental conflict is
resolved (a manner of functioning that
shows the mind to operate unconsciously
as if it took itself to be omnipotent). The
mental representations formed in this way
are manifested in conscious mental life
through various routes and encountered, in
unrecognizable and symbol-laden forms, as
symptoms, dreams, parapraxes and so on;
and take concrete form in the bizarre, grati-
fying, but otherwise futile forms of symbolic
behaviour that psychoanalysis calls 'acting-
out'.
All of this comprises the functional aspect
of unconscious processing. It has also a
formal aspect, which has to do with the
way in which unconscious mental repre-
sentations interact, described by Freud as
primary process. Primary process thinking
is characterized by its sensory and concrete
character, its lack of a firm grasp of identity,
its exaggerated sensitivity to conceptually
irrelevant connections between ideas, and
the consequently associative, metaphoric
routes that it takes. Primary process is not
constrained by the logical conditions of dis-
cursive thought.
It may seem as if these characterizations
of the unconscious leave entirely undecided
its specific contents. These, it may be
thought, ought to be filled in purely a pos-
teriori, as clinical material dictates. So it
may be argued that Freud's stress on infan-
tile experience and the role of biological and
instinctual factors in psychoanalytic expla-
nation is really optional (Jung alleges that
here Freud betrays arbitrariness and reduc-
tionist prejudices). But there are strong
reasons for thinking that the connection of
psychoanalytic explanation with motivation
whose content has an infantile and instinc-
tual character is not just contingent. The
497
PS YCHOAN AL YTIC EXPLAN A TION
functional and formal aspects of uncon-
scious processing are hypothesized with
reference to the facts of adult psychopatho-
logy encountered in clinical work, but
Freud also embeds them in a developmental
theory of the mind. The developmental
theory, which observation of children's
mental life does a great deal to corroborate
(as witnessed in Freud's Three Essays on
Sexuality, and later in Melanie Klein's child
analyses), allows the pre-verbal. unrealistic,
egoistic, and pleasure-directed features of
unconscious processing to be explained in a
deep sense, though the supposition that the
unconscious is an active repository of infan-
tile experience and the medium through
which instinctual forces enter into motiva-
tion. Without the developmental story, the
unconscious's functional and formal aspects
would be left hanging. Equally, without its
developmental dimension, psychoanalytic
explanation would be stuck for determinate
directions in which to interpret. It is part of
the logic of interpretation that diverse phe-
nomena should be shown to derive from
unifying and simplifying sources. Ordinary
psychology's inventory of human motives is
not a resource which psychoanalytic expla-
nation can draw on, for ordinary kinds of
motive are fitted out to explain rational
interactions with the world. From within
any broadly naturalistic conception of
human beings, this leaves a theory which
seeks to extend ordinary psychology's
picture of motivation with effectively no
alternative but to refer itself to infantile
experience and biological givens. A commit-
ment to an infantile-instinctual picture of
the sources of human motivation is needed
for psychoanalytic explanation to avail itself
of causes apt for the production of the
symptoms and other irrational phenomena
which are its principal target.
Freud's own conceptualization of the
unconscious' dynamism centres on the
concepts of wish-fulfilment, repression and
the opposition between the reality and
pleasure principles. The family of Kleinian
concepts constituted by phantasy, the
inner world, and internal objects deepens,
and to some extent subsumes Freud's con-
498
ceptualizations. Klein's concepts have a
simple and powerful rationale. The attribu-
tion of PROPOSITIONAL ATTITUDES in
ordinary psychology may be viewed as
casting a net over the phenomena of overt
behaviour in a way that aims to reproduce
the schemes of classification and object-
individuation adopted by the subject inter-
preted. Roughly, where behaviour reveals
a constant in the subject's apprehension of
the world, interpretation attributes a single
object of thought. Now, exactly the same
strategy can be seen to underlie psycho-
analysis, with the difference that here it is
the spread of propositional attitudes itself,
rather than behaviour, which is the object
of interpretation. Psychoanalysis can be
viewed as embarking on a second wave of
interpretation, based on and giving reappli-
cation to one of the fundamental principles
of ordinary psychology. Just as behaviour
becomes intelligible when set against a
background of propositional attitudes, so a
person's propositional network gains in
intelligibility when it is set against the
kind of unconscious, partly constitutional
background defined by psychoanalytic
attributions. It is essential for this enter-
prise that a different set of objects from
those taken by propositional attitudes be
supposed. For this reason, analysts take
clinical material to reveal thoughts about
internal objects, whose relation to external
objects, onto which they are mapped, is
fluid and elastic. Kleinian theory identifies
the earliest internal objects, which provide
templates for those of later life, with
bodies, or parts of bodies such as the
mother's breast, which the unconscious
represents itself as containing. The con-
junction of ordinary and psychoanalytic
explanation enables us to understand
people as characterizing external objects in
a double fashion: both as real, and as
bearing the significance of phantasized
objects. The phantastic characterization
helps to determine the way in which
external objects are responded to, desired
and so on. In this way, psychoanalysis
functions as a crucial supplement to ordin-
ary psychology: the latter's shortfallings -
the explanation of irrational phenomena,
and the 'giving out' that we inevitably find
when we press our questions (about, for
example, why people desire what they do)
beyond a certain point - are compensated
for when psychoanalytic explanation is
appended.
If it is thought necessary to identify a
deeper philosophical supposition in support
of the idea that our minds are really such
as to be capable of engaging in the kinds of
processes just described - something that
ordinary wakeful self-consciousness might
lead one to doubt - then the answer must
be this: mental states have the sort of
autonomy which disposes them to, simply,
find expression, in whatever form. Expres-
sion is by its nature a function that does
not require rationally appropriate vehicles,
in the sense of actions which have
instrumental value (see Wollheim, 1991,
Preface). This is something that one is
likely to overlook if one concentrates on the
role of practical reasoning in generating
intentional ACTION, although it is in fact a
supposition that may fairly be said to
permeate ordinary psychology, albeit in an
inexplicit form; it is shown in our under-
standing of EMOTION, and underpins our
registration of one another's mental life at
the level of physiognomy. The irrational
wish-fulfilling character of unconscious pro-
cesses results from the mind's natural ten-
dency to leak into outward forms - a
tendency more primitive than its role in
producing instrumental action.
THE LEGITIMATION OF PSYCHOANALYTIC
EXPLAN A TION
Viewed in these terms, psychoanalytic
explanation is an extension of ordinary psy-
chology, one that is warranted by demands
for explanation generated from within
ordinary psychology itself. This has several
crucial ramifications. It eliminates, as ill-
conceived, the question of psychoanalysis'
scientific status (see Hopkins, 1988, 1992)-
an issue much discussed, as proponents of
different philosophies of science have argued
for and against psychoanalysis' agreement
with the canons of scientific method, and its
degree or lack of corroboration. Demands
that psychoanalytic explanation should be
demonstrated to receive inductive support,
commit itself to testable psychological laws,
and contribute effectively to the prediction
of action, have then no more pertinence
than the same demands pressed on ordinary
psychology - which is not very great. When
the conditions for legitimacy are appro-
priately scaled down, it is extremely likely
that psychoanalysis succeeds in meeting
them: for psychoanalysis does deepen our
understanding of psychological laws, im-
prove the predictability of action in prin-
ciple, and receive inductive support in the
special sense which is appropriate to inter-
pretative practices (see Hopkins, 1991.
1992).
Furthermore, to the extent that psycho-
analysis may be seen as structured by and
serving well-defined needs for explanation,
there is proportionately diminished reason
for thinking that its legitimation turns on
the analysand's assent to psychoanalytic
interpretations, or the transformative
power (whatever it may be) of these. Cer-
tainly it is true that psychoanalytic expla-
nation has a reflexive dimension lacked by
explanations in the physical sciences: psy-
choanalysis understands its object, the
mind, in the very terms that the mind
employs in its unconscious workings (such
as its belief in its own omnipotence). But
this point does not in any way count
against the objectivity of psychoanalytic
explanation. It does not imply that what it
is for a psychoanalytic explanation to be
true should be identified, pragmatically,
with the fact that an interpretation may,
for the analysand who gains self-know-
ledge, have the function of translating
their semi-inchoate unconscious mentality
into a properly conceptual form. Nor does
it imply that psychoanalysis' attribution of
unconscious content needs to be under-
stood in anything less than full-bloodedly
realistic terms. Truth in psychoanalysis
may be taken to consist in correspondence
with an independent mental reality, a
reality that is both endowed with SUB-
499
PSYCHOLOGY AND PHILOSOPHY
JECTIVITY and in many respects opaque to
its owner.
See also PSYCHOLOGY AND PHILOSOPHY.
BIBLIOGRAPHY
Cavell, M. 1993. The Psychoanalytic Mind:
From Freud to Philosophy. Cambridge, MA.:
Davidson, D. 1982. Paradoxes of irrationality.
In Philosophical Essays on Freud, ed. R. Woll-
heim and J. Hopkins. Cambridge University
Press.
Freud, S. 1953-74. Standard Edition of the
Complete Psychological Works of Sigmund
Freud, 24 vols, trans. under the general edi-
torship of J. Strachey, in collaboration with
A. Freud, assisted by A. Strachey and A.
Tyson. London: Hogarth Press and the Insti-
tute of Psycho-Analysis.
Gardner, S. 1993. Irrationality and the Philo-
sophy of Psychoanalysis. Cambridge Uni-
versity Press.
Hopkins, J. 1988. Epistemology and depth psy-
chology: critical notes on The Foundations of
Psychoanalysis. Mind, Psychoanalysis and
Science, ed. P. Clark and C. Wright. Oxford:
Basil Blackwell.
Hopkins, J. 1991. The interpretation of dreams.
In The Cambridge Companion to Freud, ed. J.
Neu. Cambridge University Press.
Hopkins, J. 1992. Psychoanalysis, interpreta-
tion, and science. In Psychoanalysis, Mind
and Art: Perspectives on Richard Wollheim, ed.
J. Hopkins and A. Savile. Oxford: Basil
Blackwell.
Klein, M. 1951, 1975. The Writings of Melanie
Klein, 4 vols, ed. R. E. Money-Kyrie. London:
Hogarth Press and the Institute of Psycho-
Analysis.
Wollheim, R. 1984. The Thread of Life. Cam-
Wollheim, R. 1991. Freud, 2nd edn. London:
Harper Collins.
Wollheim, R., ed. 1974. Freud: a Collection of
Critical Essays. New York: Anchor Double-
day. Reprinted as Philosophers on Freud: New
Evaluations. New York: Aronson (1977).
Wollheim, R., and Hopkins, J. eds. 1982.
Philosophical Essays on Freud. Cambridge
University Press.
SEBASTIAN GARDNER
500
psychology and philosophy The last two
decades have been a period of extraordinary
change in psychology. COGNITIVE PSY-
CHOLOGY, which focuses on higher mental
processes like reasoning, decision making,
problem solving, language processing and
higher-level visual processing, has become a
- perhaps the - dominant paradigm among
experimental psychologists, while beha-
viouristically oriented approaches have
gradually fallen into disfavour (see
BEHA VIOURISM). Largely as a result of this
paradigm shift, the level of interaction
between the disciplines of philosophy and
psychology has increased dramatically. The
goal of this article is to sketch some of the
areas in which these interactions have been
most productive, or at least most provoca-
tive. The interactions I will discuss fall into
three categories, though the boundary
between the first two is sometimes rather
fuzzy.
COGNITIVE PSYCHOLOGY AS A SUBJECT
FOR 'DESCRIPTIVE' PHILOSOPHY OF
SCIENCE
One of the central goals of the philosophy of
science is to provide explicit and systematic
accounts of the theories and explanatory
strategies exploited in the sciences. Another
common goal is to construct philosophically
illuminating analyses or explications of
central theoretical concepts invoked in one
or another science. In the philosophy of
biology, for example, there is a rich litera-
ture aimed at understanding teleological
explanations, and there has been a great
deal of work on the structure of evolu-
tionary theory and on such crucial concepts
as fitness and biological function (see
DRETSKE; TELEOLOGY). The philosophy of
physics is another area in which studies of
this sort have been actively pursued. (For
an excellent example in the philosophy of
biology, see Sober, 1984; in the philosophy
of physics, see Sklar, 1974). In undertaking
this work, philosophers need not (and typi-
cally do not) assume that there is anything
wrong with the science they are studying.
Their goal is simply to provide accounts of
the theories, concepts and explanatory
strategies that scientists are using -
accounts that are more explicit, systematic
and philosophically sophisticated than the
often rather rough-and-ready accounts
offered by the scientists themselves.
Cognitive psychology is in many ways a
curious and puzzling science. Many of the
theories put forward by cognitive psycholo-
gists make use of a family of 'intentional'
concepts - like believing that p, desiring that
q, and representing r - which don't appear in
the physical or biological sciences, and
these intentional concepts play a crucial
role in many of the explanations offered by
these theories (see INTENTION ALITY; PROPO-
SITIONAL ATTITUDES). People's decisions
and actions are explained by appeal to their
beliefs and desires. Perceptual processes,
some of which may themselves be repre-
sentational. are said to result in mental
states which represent (or sometimes
misrepresent) one or another aspect of the
cognitive agent's environment (see PERCEP-
TION; REPRESENTATION). While cognitive
psychologists occasionally say a bit about
the nature of intentional concepts and the
explanations that exploit them, their com-
ments are rarely systematic or philosophi-
cally illuminating. Thus it is hardly
surprising that many philosophers have
seen cognitive psychology as fertile ground
for the sort of careful descriptive work that
is done in the philosophy of biology and the
philosophy of physics. Jerry FODOR'S The
Language of Thought (1975) was a pioneer-
ing study in this genre, one that continues
to have a major impact on the field. Robert
Cummins (1983, 1989), Daniel DENNETT
(1978a, 1987) and John Haugeland (1978)
have also done important and widely dis-
cussed work in the what might be called the
'descriptive' philosophy of cognitive psy-
chology.
PHILOSOPHY AS A SOURCE OF
PROPOSALS FOR IMPROVING COGNITIVE
PSYCHOLOGY
The goal of the projects discussed in the pre-
vious section is to provide accurate, illum-
inating descriptions of what is going on in
cognitive psychology. These philosophical
accounts of cognitive theories and the con-
cepts they invoke are generally much more
explicit than the accounts provided by psy-
chologists, and they inevitably smooth over
some of the rough edges of scientists' actual
practice. But if the account they give of
cognitive theories diverges significantly from
the theories that psychologists actually
produce, then the philosophers have just
gotten it wrong. There is, however, a very
different way in which philosophers have
approached cognitive psychology. Rather
than merely trying to characterize what
cognitive psychology is actually doing, some
philosophers try to say what it should and
should not be doing. Their goal is not to
explicate scientific practice, but to criticize
and improve it. The most common target of
this critical approach is the use of inten-
tional concepts in cognitive psychology.
Intentional notions have been criticized on
various grounds. The two that I will con-
sider here are that they fail to supervene on
the physiology of the cognitive agent, and
that they cannot be 'naturalized'.
Perhaps the easiest way to make the
point about SUPERVENIENCE is to use a
thought experiment of the sort originally
proposed by Hilary PUTNAM (1975).
Suppose that in some distant corner of the
universe there is a planet, Twin Earth,
which is very similar to our own. On Twin
Earth there is a person who is an atom for
atom replica of President Clinton. Now the
President Clinton who lives on Earth
believes that Vice President Gore was born
in Tennessee. If you asked him, 'Was Gore
born in Tennessee?' he'd say, 'Yes'. Twin-
Clinton would respond in the same way.
But it is not because he believes that our
Gore was born in Tennessee. Twin-Clinton
has no beliefs at all about our Gore. His
beliefs are about Twin-Gore, and Twin-Gore
was certainly not born in Tennessee. Indeed,
we may even suppose that Twin-Gore was
not born in Twin-Tennessee, and thus that
Clinton's belief is true while Twin-Clinton's
is false. What all this is supposed to show is
that two people can share all their physio-
501
logical properties without sharing all their
intentional properties. To turn this into a
problem for cognitive psychology, two addi-
tional premises are needed. The first is that
cognitive psychology attempts to explain
behaviour by appeal to people's intentional
properties. The second is that psychological
explanations should not appeal to properties
that fail to supervene on an organism's
physiology. (Variations on this theme can
be found in Stich (1978, 1983) and in
Fodor (1987, ch. 2).)
Reactions to this argument have taken a
variety of forms. Perhaps the most radical is
the proposal that cognitive psychology
should recast its theories and explanations
in a way that does not appeal to intentional
properties of mental states but only to their
formal or 'syntactic' properties (Stich,
1983). Somewhat less radical is the sugges-
tion that we can define a species of repre-
sentation or CONTENT - often called 'narrow
content' (see EXTERNALISM/INTERNALISM)-
which does supervene an organism's phy-
siology, and that psychological explanations
that appeal to ordinary ('wide') intentional
properties can be replaced by' explanations
that invoke only their narrow counterparts
(Fodor, 1987). Both of these proposals
accept the conclusion of the argument
sketched in the previous paragraph, and
they go on to propose ways in which cogni-
tive psychology might be modified. But
many philosophers have urged that the
problem lies in the argument, not in the
way that cognitive psychology goes about
its business. The most common critique of
the argument focuses on the normative
premise - the one that insists that psycho-
logical explanations ought not to appeal to
'wide' properties that fail to supervene on
physiology. Why shouldn't psychological
explanations appeal to wide properties, the
critics ask? What, exactly, is wrong with
psychological explanations invoking proper-
ties that don't supervene on physiology?
(See Burge, 1979, 1986.) Various answers
have been proposed in the literature,
though they typically end up invoking
metaphysical principles that are less clear
and less plausible than the normative thesis
502
they are supposed to support. (See, for
example, Fodor, 1987, 1991.)
My own view is that the extensive litera-
ture in this area is mostly a tempest in a
teapot, though I'm afraid I bear some of the
responsibility for provoking it. I know of no
clear or persuasive argument for excluding
wide properties from psychological theories
and explanations. But if you are inclined to
demand that psychology invoke only prop-
erties that supervene on physiology, the
demand is easy enough to satisfy. Given any
psychological property that fails to super-
vene on physiology, it is trivial to character-
ize a narrow correlate property that does
supervene. The extension of the correlate
property includes all actual and possible
objects in the extension of the original prop-
erty, plus all actual and possible physio-
logical duplicates of those objects. Theories
originally stated in terms of wide psycho-
logical properties can be recast in terms of
their narrow correlates with no obvious loss
in their descriptive or explanatory power. It
might be protested that when characterized
in this way, narrow belief and narrow
content are not really species of belief and
content at all. But it is far from clear how
this claim could be defended, or why we
should care if it turns out to be right. (For
more details see Stich (1991) and Stich and
Laurence (forthcoming).)
The worry about the 'naturalizability' of
intentional properties is much harder to pin
down (see NATURALISM). According to
Fodor, the worry derives from 'a certain
ontological intuition: that there is no place
for intentional categories in a physicalistic
view of the world,' (1987, p. 97) and thus
that 'the semantic (and/or the intentional)
will prove permanently recalcitrant to inte-
gration in the natural order' (Fodor, 1984,
p. 32). If intentional properties can't be
integrated into the natural order, then pre-
sumably they ought to be banished from
serious scientific theorizing. Psychology
should have no truck with them. Indeed, if
intentional properties have no place in the
natural order, then nothing in the natural
world has intentional properties, and inten-
tional states do not exist at all. So goes the
worry. Unfortunately, neither Fodor nor
anyone else has said anything very helpful
about what is required to 'integrate' inten-
tional properties into the natural order.
There are, to be sure, various proposals to
be found in the literature. But all of them
seem to suffer from a fatal defect. On each
account of what is required to naturalize a
property or integrate it into the natural
order, there are lots of perfectly respectable
non-intentional scientific or commonsense
properties that fail to meet the standard.
Thus all the proposals that have been made
so far end up throwing out the baby with
the bath water. (For the details, see Stich
and Laurence (forthcoming).)
Now, of course, the fact that no one has
been able to give a plausible account of
what is required to 'naturalize' the inten-
tional may indicate nothing more than that
the project is a difficult one. Perhaps with
further work a more plausible account will
be forthcoming. But one might also offer a
very different diagnosis of the failure of all
accounts of 'naturalizing' that have so far
been offered. Perhaps the 'ontological in-
tuition' that underlies the worry about in-
tegrating the intentional into the natural
order is simply muddled. Perhaps there is no
coherent criterion of naturalizability that all
properties invoked in respectable science
must meet. My own guess is that this diag-
nosis is the right one. Until those who are
worried about the naturalizability of the
intentional provide us with some plausible
account of what is required of intentional
categories if they are to find a place in 'a
physicalistic view of the world' I think we
are justified in refusing to take their worry
seriously.
Recently, John SEARLE (1992) has offered
a new set of philosophical arguments aimed
at showing that certain theories in cognitive
psychology are profoundly wrong-headed.
The theories that are the targets of Searle's
critique offer purely formal or computa-
tional explanations of various psychological
capacities - like the capacity to recognize
grammatical sentences, or the capacity to
judge which of two objects in one's visual
field is further away. Typically these the-
ories are set out in the form of a computer
program - a set of rules for manipulating
symbols - and the explanation offered for
the exercise of the capacity in question is
that people's brains are executing the
program. The central claim in Searle's cri-
tique is that being a symbol or a computa-
tional state is not an 'intrinsic' physical
feature of a computer state or a brain state.
Rather, being a symbol is an 'observer rela-
tive' feature. But, Searle maintains, only
intrinsic properties of a system can play a
role in causal explanations of how they
work. Thus appeal to symbolic or computa-
tional states of the brain could not possibly
play a role in a 'causal account of cogni-
tion'.
There is something quite paradoxical
about Searle's argument. To see the point,
imagine that we find an unfamiliar object
lying on the beach. After playing with it for
a while, we discover that it has some
remarkable capacities. If you ask it math-
ematical questions like, 'How much is 345
times 678?' or 'What is the square root of
1492?' the correct answer appears on a
video display screen. How does the object do
it? In order to find out, we turn it over to a
group of scientists and engineers. After
studying it for a while, they report that the
object is a remarkable computer whose
program includes a sophisticated algorithm
for processing English along with a set of
algorithms for various mathematical tasks.
The report from the research team includes
a detailed specification of the program they
believe the object is using. Now most of us
would be inclined to think that this report
provides deep insight into how the object
manages to produce answers to the ques-
tions we ask it. But if Searle is right, the
program couldn't possibly explain how the
object works. Obviously something has gone
wrong somewhere. As I see it, the problem
with Searle's argument lies in his assump-
tion that cognitive theories of the sort he is
criticizing aim at providing a 'causal
account of cognition'. One of the important
lessons of what I earlier called the 'descript-
ive' philosophy of science is that there are
lots of different strategies of explanation in
503
science. One of the most useful explanatory
strategies is functional decomposition. in
which a complicated capacity is explained
by showing how it can be accomplished by
assembling a number of simpler capacities
in an appropriate way. This strategy is
widely used in biology. And. as Fodor
(1968). Dennett (1978b). Cummins (1983)
and others have argued. it is also central to
the explanatory approach in cognitive psy-
chology. Thus. even if we grant Searle's
claim that being a computational state is
not an intrinsic property of a state. and that
only intrinsic properties can play a role in
causal explanations. we will have no reason
to conclude that the sort of computational
explanations that Searle is criticizing are in
any way problematic.
COGNITIVE PSYCHOLOGY SUGGESTS
WAYS TO RESOLVE PHILOSOPHICAL
PROBLEMS
The last section surveyed some of the philo-
sophical arguments aimed at showing that
cognitive psychology is confused and in
need of reform. My reaction to those argu-
ments was none too sympathetic. In each
case. I maintained. it is the philosophical
argument that is problematic. not the psy-
chology it is criticizing. In this section I
want to turn the tables and consider some
of the proposals that have been made for
using psychological findings to criticize phi-
losophical theories and to resolve traditional
philosophical problems. The tone in this
section will be much more optimistic than
in the previous section. since in this area I
think there is some real progress to report.
Perhaps the most impressive example of
the way in which psychological research
can contribute to the resolution of philo-
sophical disputes is to be found in the vener-
able debate between empiricist and
rationalist accounts of knowledge. Though
this debate is complex and multifaceted. one
central issue has been the extent to which
our knowledge in various domains is derived
from experience. and the extent to which it
is innate (see INNATENESS). Empiricists typi-
cally claim that most or all of our knowl-
504
edge is derived from experience. while
rationalists maintain that important aspects
of what we know are innate. Rationalists
generally recognize that some input from
experience may be needed to activate our
innate knowledge and make it useable.
Without appropriate environmental 'trig-
gers' . our innate knowledge may lie
dormant. To make this point. Leibniz uses
the analogy of a deeply grained block of
marble. The block may have the shape of a
man or a horse within it. though a fair
amount of hammering and chiselling may
be necessary to turn the block into a statue.
In the mid-1960s. Noam CHOMSKY began
developing a set of arguments aimed at
showing how considerations from linguis-
tics and psycho linguistics might be used to
resolve the dispute between the rationalists
and the empiricists. (See. for example.
Chomsky 1965. 1980.) The basic strategy
in Chomsky's argument is obvious enough.
What we should do. he urged. is look at the
input to the process of language acquisition
and the output of that process. If there is a
significant amount of information in the
output that cannot be found in the input.
then the only plausible hypothesis is that
the excess information is innate. What was
novel and striking in Chomsky's argument
was the empirical evidence he offered about
the richness and complexity of the informa-
tion that competent speakers of a language
possess. Chomsky and his followers argued
that there are lots of examples of grammat-
ical rules that people acquire quite reliably.
though the evidence available to them is
not adequate to select the rule actually
acquired over various alternatives that are
not acquired. (For details. see Hornstein and
Lightfoot. 1981.) In the years since
Chomsky first advanced this argument. a
number of other lines of evidence have been
developed that underscore the extent to
which human knowledge is strongly influ-
enced by information-rich. domain-specific
innate mental capacities. Some of the most
impressive studies have demonstrated that
very young children recognize phoneme
boundaries and other subtle features that
are essential for language mastery (Mehler
and Fox, 1984). Studies of visual perception
in young children have also revealed a
great deal of innate structure (Spelke,
1990). Though the details about what is
innate and what is acquired in many
domains remains to be determined, I think
it is now quite clear that the sort of radical
empiricist view often associated with Locke
and with behaviourism is simply untenable.
Another area in which psychological
studies have made important contributions
to philosophy is in the branch of epistemol-
ogy that attempts to characterize the notion
of RATIONALITY and the related notion of
justified inference. Here there are two quite
different lines of influence to report. Some
philosophers have offered accounts of
rationality that build upon the actual infer-
ences that people make and endorse. The
best-known account of this sort is the one
developed by Nelson Goodman (1965).
According to Goodman, the justified infer-
ences are the ones that would be sanctioned
by a certain process. That process begins
with the inferences we are actually inclined
to accept, and it attempts to provide the
simplest and most satisfying set of rules that
will capture those inferences. As the process
proceeds, certain inferences that we are
initially inclined to accept may have to be
thrown out, and certain rules that we initi-
ally find appealing may have to be amended
or rejected entirely. Obviously, the sort of
rules that will ultimately be sanctioned by
this process will depend to a significant
extent on the sorts of inferences that those
using the process are initially inclined to
make. And, while they rarely stress the
point, it is clear that Goodman and others
who are attracted to this account of justifi-
cation typically assume that people's un-
tutored inclinations are generally pretty
good. But during the last twenty years or
so, psychologists studying inference have
accumulated a substantial body of data that
casts serious doubt on this assumption. In
many studies of both deductive and prob-
abilistic reasoning it has been shown that
normal subjects regularly draw inferences
that would be classified as invalid (or worse)
by the prevailing normative theory. (See, for
example, Nisbett and Ross, 1980; Kahne-
man et aI., 1982.) Moreover, there is good
reason to think that at least some of these
patterns of inference are quite robust
enough to survive the pruning process that
Goodman describes. If this is right, it would
not constitute a knock-down argument
against Goodmanian accounts of justified
inference. But the psychological findings do
indicate that resolute Goodmanians are
going to have some unpleasant bullets to
bite. If they want to hang on to their
account of justification, they are going to
have to classify some pretty weird inferences
as 'justified'. (For further details, see Stich,
1990, ch. 4; for a defence of the Good-
manian strategy, see Cohen, 1981.)
Quite a different approach to the assess-
ment of reasoning is one that grows out of
the pragmatist tradition. On this approach,
reasoning is viewed as a tool for achieving
various ends, and good strategies of reason-
ing are those that do a good job in enabling
people to achieve their ends. Which strat-
egies will facilitate which goals is an empir-
ical question, not a matter to be determined
by philosophical argument. So if we want to
know what good reasoning in a given
domain is like, the best way to find out is to
locate people who have been particularly
successful in that domain, and study the
way in which they reason. In recent years,
this strategy has been pursued in a particu-
larly sophisticated way by Herbert Simon
and his co-workers (Langley, Simon, Brad-
shaw and Zytkow, 1987.) Simon and his
colleagues are interested in characterizing
good scientific reasoning. They proceed by
locating clear examples of people who have
been successful in science (the people whose
pictures appear in the science textbooks, as
Simon sometimes puts it) and then trying to
construct computer models that will simu-
late the reasoning of these successful scien-
tists. There is, of course, no guarantee that
this strategy will work, since it might turn
out that the reasoning patterns of successful
scientists have little or nothing in common
- that successful scientific reasoning is not a
'natural kind' in psychology. However,
Simon's work so far suggests that there are
505
important patterns to be discovered in the
thinking of successful scientists. To the
extent that Simon's project succeeds, it will
constitute a particularly exciting sort of
'naturalized epistemology'.
Before closing I want to mention one
other domain in which psychological
research promises to have an important
impact on philosophical theorizing. In ethics
there are a number of views that pre-
suppose substantive theses about human
psychology. Often these psychological theses
are taken to be part of received common
sense about the mind, and thus no defence
is offered. In psychology, however, received
common sense has a distressing tendency to
be mistaken. Consider, for example, those
versions of utilitarian theory that rank
actions on the basis of how well they do in
satisfying people's preferences. These pro-
posals make little sense unless we make the
commonsensical assumption that people
have determinate and reasonably stable
preferences which may be elicited in a
variety of ways. However, much recent
work on the psychology of choice and pre-
ference suggests that this seemingly innoc-
uous assumption may well be mistaken
(Fischhoff et al., 1980; Slovic, 1990.) As
Goldman has noted in a recent discussion of
this literature, 'subtle aspects of how pro-
blems are posed, questions are phrased, and
responses are elicited can have a substantial
effect on people's expressed judgements and
preferences. This leads some researchers to
doubt whether, in general, there are stable
and precise values or preferences antecedent
to an elicitation procedure' (Goldman,
1993). If these doubts turn out to be justi-
fied, then a great deal of work in moral
theory may well turn out to be indefensible,
perhaps even incoherent.
See also ARTIFICIAL INTELLIGENCE; COM-
PUT ATION AL MODELS; DEVELOPMENTAL
PSYCHOLOGY.
BIBLIOGRAPHY
Midwest Studies in Philosophy, Vol.4, 73-121.
506
Burge, T. 1986. Individualism and psycho-
logy. Philosophical Review, 95, 3-46.
Chomsky, N. 1965. Aspects of the Theory of
Syntax. Cambridge, MA.: MIT Press.
Chomsky, N. 1980. Rules and Representations.
New York: Columbia University Press.
Cohen, J. 1981. Can human irrationality be
experimentally demonstrated? Behavioral and
Brain Sciences, 4.
Explanation. Cambridge, MA.: MIT Press
Bradford Books.
Cummins, R. 1989. Meaning and Mental Repre-
sentation. Cambridge, MA.: MIT Press/Brad-
ford Books.
Dennett, D. 1978a. Brainstorms. Cambridge,
Dennett, D. 1978b. Artificial intelligence as
philosophy and as psychology. In Dennett,
1978a.
Dennett, D. 1987. The Intentional Stance. Cam-
Fischhoff, P., Slovic, P., and Lichtenstein, S.
1980. Knowing what you want: Measuring
labile values. In Cognitive Processes in Choice
and Decision BehaVior, ed. T. Wallsten. Hills-
dale, N.J.: Erlbaum.
Fodor, J. 1968. The appeal to tacit knowledge
in psychological explanation. Journal of Phi-
losophy, 65.
Fodor, J. 1984. Semantics, wisconsin style.
Synthese, 59. Reprinted in Fodor, 1990;
page reference is to the latter.
Essays. Cambridge, MA.: MIT Press/Bradford
Books.
Fodor, J. 1991. A modal argument for narrow
content. Journal of Philosophy, 88.
Goldman, A. 1993. Ethics and cognitive
science. In Philosophical Applications of Cogni-
tive Science. Boulder, CO: Westview Press.
Goodman, N. 1965. Fact, Fiction and Forecast.
Indianapolis: Bobbs-Merrill.
Haugeland, J. 1978. The Nature and plaUSi-
bility of cognitivism. Behavioral and Brain
Sciences, 1.
Hornstein, N., and Lightfoot, D. 1981. Expla-
nation in Linguistics. London: Longman.
Kahneman, D., Slovic, P., and Tversky, A., eds
1982. Judgment Under Uncertainty: Heuristics
and Biases. Cambridge University Press.
Langley. P .. Simon. H .. Bradshaw. G .. and
Zytkow. J. 1987. Scientific Discovery: Compu-
tational Explorations of the Creative Processes.
Mehler. J .. and Fox. R .. eds. 1984. Neonate
Cognition: Beyond the Blooming. Buzzing Con-
fusion. Hillsdale. N.J.: 1. Erlbaum Associates.
Nisbett. R .. and Ross. L. 1980. Human Infer-
ence: Strategies and Shortcomings of Social
Judgment. Englewood Cliffs. N.J.: Prentice-
Hall.
Putnam. H. 1975. The meaning of 'Meaning'.
In Language. Mind and Knowledge: Minnesota
Studies in the Philosophy of Science. Vol. 7. ed.
Minnesota Press.
Searle. J. 1992. The Rediscovery of the Mind.
Cambridge. MA.: MIT Press/Bradford Books.
Sklar. 1. 1974. Space. Time and Spacetime. Ber-
keley. CA: University of California Press.
Slovic. P. 1990. Choice. In Thinking. ed. D.
Osherson and E. Smith. Cambridge. MA.:
Sober. E. 1984. The Nature of Selection. Cam-
bridge. MA.: MIT Press/Bradford Books.
Spelke. E. 1990. Origins of visual knowledge.
In An Invitation to Cognitive Science: Visual
Cognition and Action. ed. Hollerbach. D. Osh-
erson. S. Kosslyn and J. Cambridge. MA.:
Stich. S. 1978. Autonomous psychology and
the belief-desire thesis. Monist. 61.
Stich. S. 1983. From Folk Psychology to Cog-
nitive Science. Cambridge. MA.: MIT Press/
Bradford Books.
Stich. S. 1990. The Fragmentation of Reason.
Cambridge. MA.: MIT Press/Bradford Books.
Stich. S. 1991. Narrow content meets fat
syntax. In Meaning in Mind: Fodor and His
Critics. ed. B. Loewer and G. Rey. Oxford:
Basil Blackwell.
Stich. S.. and Laurence. S. Forthcoming.
Intentionality and Naturalism.
STEPHEN STICH
Putnam, Hilary In 1960 I published a
paper titled 'Minds and Machines' (1975.
ch. 18). which suggested a possible new
option in the philosophy of mind. and in
1967 I published two papers (1975. ch. 20
PUTN AM. HILAR Y
and 21) which became. for a time. the
manifestos of the 'functionalist' current.
FUNCTIONALISM (as many of my readers
doubtless already know) holds that we are
analogous to computers, and that our psy-
chological states are simply our 'functional
states'. that is. they are the states that
would figure in an ideal description of our
'program'. In the present 'self-portrait' of
myself as a philosopher of mind I shall
review the reasons that led me to propose
functionalism and the reasons that sub-
sequently led me to abandon it.
ROBOT CONSCIOUSNESS AND THE
PROBLEM OF MAKING FUNCTIONALISM
PRECISE
'Functionalism' views us as automata; that is
as computers that happen to be made of
flesh and blood. According to the function-
alist view. a robot with the same program
as a human being would ipso facto be con-
scious (see CONSCIOUSNESS). Although in a
talk to the American Philosophical Associ-
ation in 1964 (1975. ch. 19). I had drawn
back from that view. arguing that the ques-
tion whether any automaton was conscious
was not really a question of fact but called
for a 'decision' on our part. a decision 'to
treat robots as fellow members of our lin-
guistic community'. when I came to write
the two papers I described as 'functionalist
manifestos'. I considered both the question
as to whether psychological states are really
'functional' (Le. computational) in nature
and the question as to whether an auto-
maton could be conscious to be factual
questions. The earlier talk. I had come to
see, contained an error.
In the 1964 paper. I assumed that if an
'IDENTITY THEORY' (a theory to the effect
that psychological states are identical either
with brain states or with functional states)
were true. then it would have to be true as
a consequence of (1) the meanings of psy-
chological words. and (2) empirical facts
that do not themselves beg the question as
to whether a robot could be conscious. But
the same line of reasoning, I saw in the
1967 papers. if applied to the question
507
PUTN AM, HILAR Y
whether light is electromagnetic radiation of
such-and-such wavelengths, would lead to
the conclusion that this too was not a ques-
tion of fact but called for a 'decision' on our
part, a decision to treat electromagnetic
radiation as light! For persons who,
knowing what we do about the causes and
effects of electromagnetic radiation, still
insisted that light is not identical with elec-
tromagnetic radiation (of such-and-such
wavelengths), but only correlated with the
presence of such radiation would not be
contradicting themselves. That light is elec-
tromagnetic radiation does not follow ana-
lytically from 'non-question-begging' facts
about electromagnetic radiation and about
light.
It is wrong, I now realized, to insist that
statements that make theoretical identifica-
tions of phenomena originally described in
different vocabularies must follow analyti-
cally from 'non-question-begging facts' in
order to be true. If we wish to claim that
light is electromagnetic radiation, we need
only maintain that it is rational to believe
that it is, given what we now know; the
question of analyticity is a red herring.
Thus, I was led to drop my concern with
analyticity entirely, and to return to my
earlier concern (in 1975, ch. 18) with
theoretical identification and with a notion
that I now introduced to go with it, the
notion of synthetic identity of properties. (I
spoke of synthetic identity of properties
because, in the papers published in 1967
and subsequently, I said that not only is
light passing through an aperture the same
event as electromagnetic radiation passing
through the aperture, but that the property
of being light is the very same property as the
property of being electromagnetic radiation of
such-and-such wavelengths.) I rejected the
traditional view that predicates P and Q
correspond to the same property only if P
and Q are analytically coextensive, or if in
some way the necessary coextensiveness of
the predicates is a matter of 'conceptual
analysis'. (See PROPERTY.) In short, I held
(and still hold) that properties can be syn-
thetically identical, and the way in which
we establish that properties are syntheti-
508
cally identical is by showing that identifying
them Simplifies our explanatory endeavours
in certain familiar ways (For a detailed dis-
cussion, see Putnam, 1979, ch.9).
In my earliest paper on the issue of minds
and machines (1975, ch. 18), the theor-
etical identification I predicted we might
come to make in two hundred years was an
identification of human psychological states
with the assumed 'corresponding' brain
states. Although I also suggested in that
paper that we might think of psychological
states as more analogous to what I there
called 'logical states' of machines (states
defined at the 'programming' level) than to
'structural states' of machines (states
defined at the hardware level), thereby
introducing functionalism as an option, I
did not follow up the option at the time. I
now, however, took up the option with a
vengeance, and suggested (particularly in
1975, ch. 21) that just as light is empiri-
cally identical with electromagnetic radi-
ation, so (I proposed as a hypothesis)
psychological states are empirically identical
with functional states. Here is the hypoth-
esis as I stated it (1975, p. 434) (for simpli-
city I stated it only for the case of pain, but I
made clear that it was intended to hold for
psychological states in general):
(1) All organisms capable of feeling pain
are Probabilistic Automata
A Probabilistic Automaton is a device
similar to a TURING Machine, except that
(1) its memory capacity has a fixed finite
limit, whereas a Turing Machine has a
potentially infinite external memory; and
(2) state transitions may be probabilistic
rather that than deterministic. I assumed
that the Probabilistic Automata in question
were equipped with motor organs and with
sensory organs and that certain states cor-
responded to possible 'inputs' and 'outputs'.
(2) Every organism capable of feeling pain
posseses at least one Probabilistic Auto-
maton Description (specifying the func-
tional states of the Automaton and the
transition probabilities between them)
of a certain kind (Le. being capable of
feeling pain is possessing an appropriate
kind of functional organization).
(3) No organism capable of feeling pain
possesses a decomposition into parts
that separately possess Probabilistic
Automaton Descriptions of the kind
referred to in (2). (This rules out a
society of organisms, or a person in a
room running a program.)
(4) For every Probabilistic Automaton
Description of the kind referred to in
(2), there exists a subset of the sensory
inputs such that an organism with that
Description is in pain when and only
when some of its sensory inputs are in
that subset.
I admitted (p. 435), that the hypothesis
was very vague, but I argued that 'in spite
of its admitted vagueness, [it] is far less
vague than the "physical-chemical state"
hypothesis is today, and far more suscep-
tible to investigation of both a mathematical
and an empirical kind.' And then I added
(and this reveals the extent of my 'scient-
ism' at that time), 'Indeed, to investigate
this hypothesis is just to attempt to produce
"mechanical" models of organisms - and
isn't this, in a sense, just what psychology is
all about? The difficult step, of course, will
be to pass from models of specific organisms
(sic) to a normal form for the psychological
description of organisms - for this is what is
required to make (2) and (4) precise. But
this too seems to be an inevitable part of the
program of psychology.'
A paper that ended up being published in
the same year (1975, ch. 20) closes by
allowing that while the functional organiza-
tion of a Turing Machine or a Probabilistic
Automaton is given by the machine table,
the description of the functional organiza-
tion of a human being 'might well be some-
thing different and more complicated'. But
like the paper from which I just quoted, it
does not doubt, that the very raison d'etre of
psychology is to produce mechanical models
of organisms. Nor does it express any reser-
vations about the idea that it is an 'inevit-
able part of the program of psychology' to
PUTNAM. HILARY
provide a normal form for the psychological
description of organisms: not just for the psy-
chological description of human beings,
please note (as if that were not Utopian
enough!), but a normal form for the psycho-
logical description of an arbitrary organism!
Once psychology has progressed far enough
in the pursuit of this 'inevitable program' to
make the hypothesis that mental states are
just functional states precise, it will possible
- or so I claimed - to confirm the hypothesis
in a way analogous to the way in which we
have confirmed theoretical identifications in
physics: the laws of unreduced psychology,
to the extent that they are true, will be
explained by the fact that the psychological
states they speak of are really these func-
tional states, just as the laws of unreduced
optics, to the extent that they were true,
were explained by the fact that the 'light
rays' and 'light waves' they spoke of were
really electromagnetic radiation of certain
wavelengths.
FUNCTIONALISM WITHOUT TURING
MACHINES. AND A PROBLEM FOR
FUNCTION ALISM
At the beginning of 'Philosophy and Our
Mental Life' (1975, ch. 14, p. 292), I
expanded on my qualm about supposing
that psychological states of human beings
are literally Turing Machine states (in these
papers I frequently spoke generally of
Turing Machines, rather than, more pre-
cisely, of Probabilistic Automata). I recog-
nized (1975, ch. 14) that 'the difficulty with
the notion of psychological isomorphism is
that it presupposes the notion of some-
thing's being a functional or psychological
description.' And I went on to say that it is
for that reason that in my previous papers I
explained the notion in terms of Turing
Machines, and I remarked that 'I felt con-
strained, therefore, to defend the idea that
we are Turing Machines.' Turing Machines,
I explained, come with a normal form for
their functional description, the Machine
Table, a standard style of program. 'But it
does not seem fatally sloppy to me, although
it is sloppy,' I wrote, 'if we apply the notion
509
PUTNAM. HILARY
to systems for which we have no detailed
idea at present what the normal form
description would look like - systems like
ourselves.' I claimed that even if we don't
have any idea what a 'comprehensive psy-
chological theory' would look like, we know
enough to point out illuminating differences
between any possible psychological theory,
or even a functional description of an auto-
maton, and a physical description. The most
important of those differences is this: that
systems that are models of the same psycho-
logical theory, systems that are 'psychologi-
cally isomorphic', to use the terminology I
developed in this series of papers, do not
have to be in the same physical state in
order to count as being in the same func-
tional state. States that play the same psy-
chological/functional role are identified at
this level of description. A human being, a
robot with the 'positronic brain' imagined in
Isaac Asimov's science fiction, and a dis-
embodied spirit might be psychologically
isomorphic, and if they were, they could be
in the same psychological states without
ever being in the same physical state.
At this point, although I was still a func-
tionalist, I had begun to be aware of a very
serious problem for the position. Originally
the thesis of functionalism was that our
mental states are identical with (a subset of)
our functional states, where the notion of a
functional state was made clear by identify-
ing it with the notion of a Turing Machine
(or Probabilistic Automaton) State. But in
this paper (1975, ch. 14), I pointed out that
our mental states cannot literally be Turing
Machine States (Turing Machine States
don't have the right sorts of properties; for
details, see 1975, pp. 298-9), and I
replaced the notion of a Turing Machine
Description with the notion of the sort of
description that will be provided by an ideal
psychological theory. What is an ideal psy-
chological theory? We will know what that
is when we have the 'normal form' for psy-
chological theories that I had earlier
claimed it must be 'an inevitable part of the
programme of psychology' to provide.
But is it really any part of the programme
of psychology to provide such a thing?
510
THE UTOPIAN CHARACTER OF
FUNCTION ALISM
A psychological theory, in the ordinary
sense, does not pretend to give a complete
description of all of a human being's, or
even of a rat's, psychological states (even if
we assume that we know what we mean by
talking about all of an organism's psycho-
logical states). Nor does it pretend to give all
of the causal relations between psychologi-
cal states. And this is so whether we think
of Chomskian theories, behaviourist ,the-
ories, Freudian theories, or whatever. No
one has ever claimed to provide a theory in
which so much information about the state
of believing, say, that there are cows in
Romania, and about the connections
between that state and other psychological
states, and between all of these states and
'sensory inputs' and 'behavioural outputs',
is provided as to individuate the state of
believing that there are cows in Romania. A
Machine Table does distinguish a functional
state of a Turing Machine from all other
functional states of that Turing Machine: it
individuates that state. in the sense of pro-
viding a necessary and sufficient condition
for being that state. Even if we are chari-
table, we shall have to admit that the 'ideal
psychological theory' that I envisaged in my
functionalist papers, the kind of theory that
could provide as complete a description of
our psychological states as a Turing
Machine Table provides of the functional
states of a computer, is an utterly Utopian
project (and if we are uncharitable we will
simply say it is a 'we know not what').
This sort of utopianism is also an excel-
lent illustration of what is called 'scientism'.
Scientism is, of course, not the same thing
as a respect for science, or a desire to learn
the results of science, or a conviction that
those results are relevant to philosophical
investigation. But when one is in a frame of
mind (as I was) in which one fails to distin-
guish between science in the sense in which
science is actually done in today's labora-
tories and the most Utopian sort of specula-
tion, then one is indeed in the grip of
scientism. What is wrong with this sort of
Utopianism is not that there is something
wrong with speculating about possibilities
that we are not presently able to realize
when we are able to make clear just what
the hypothetical 'possibilities' are: such
speculation is as old as philosophy itself.
The problem is that it is completely unclear
just what possibility is being envisaged
when one speaks of a 'normal form descrip-
tion of the psychology of an arbitrary
organism'; and the talk about 'the program
of psychology' and about what is 'an inevi-
table part of the program of psychology'
was, I blush to admit, a way of hiding this
sorry state of affairs (in the first instance,
from myself).
The degree of utopianism required to be a
functionalist becomes all the greater when
one recognizes something that I had
emphasized in my writings on the philo-
sophy of language from 1970 on (1970;
1975, ch. 12; 1988), namely that the
meanings of our words (and, I argued, the
content of our thoughts as well) are not
determined simply by our functional organ-
ization in the sense that I talked about in
my functionalist papers, that is, the sense in
which our functional organization is simply
a matter of 'sensory inputs', transitions
from one 'state' to another, and 'motor
outputs'. According to the semantic extern-
alism that I defended (and still defend), the
content of our words and thoughts is partly
determined by our relations with things in
our environment (including other people).
The fact that what causes us to speak of
water is water and not some other liquid
has everything to do with the fact that the
word water refers to water, for example (see
EXTERNALISM/INTERNALISM).
Although I did not discuss this in 'Philo-
sophy and Our Mental Life' (1975, ch. 14), I
was, of course, aware of it, and what I
would have said if someone had asked 'Isn't
your functionalism incompatible with your
semantic externalism?' is that strictly speak-
ing an 'ideal psychological theory' has to be
not a theory of one organism in isolation,
but a theory of a group of organisms, and
has to include a description of their inter-
actions with one another and with their
PUTNAM. HILAR Y
environment, and of the nature of the rel-
evant parts of that environment. (In con-
versation, Richard Boyd suggested the name
'sociofunctionalism' for this position.) But
the stipulation that the ideal psychological
theory must include, properly speaking, also
an ideal sociolinguistic theory makes the idea
of such a theory, if possible, even more
Utopian. Thus it is not surprising that by the
middle 1980s I began seriously to ask myself
'How meaningful. even as a regulative ideal.
is the idea of an ideal psychological (cum
socio-linguistic) theory that individuates all
possible psychological states?'
Let us go back to the robot who was sup-
posed to be 'psychologically isomorphic' to a
human being (call it 'Leslie'). Suppose we
observe that Leslie (or, for that matter, a
human) produces the sound 'sheleg' when
it snows. Especially when the snow is unex-
pected, and Leslie exhibits a 'startled reac-
tion' at the onset of snow, he is likely to
exclaim 'Shelegl' (At least it seems to us like
an exclamation.) It might well be part of the
ideal psychological theory we postulated
that 'under normal conditions an organism
who has the belief that it is snowing in its
conceptual repertoire will be caused to have
that belief by the onset of snow'. It is com-
patible with that piece of 'the ideal psycho-
logical theory' that Leslie may believe that it
is snowing on these occasions, and that
Leslie may be expressing that belief by
saying (or thinking) 'sheleg'. Unfortunately,
there are a host of other possibilities.
Suppose, for example, Leslie and his fellow
robots have a religion of a rather primitive
kind (remember, Leslie's sort of robot is
psychologically isomorphic to human
beings), and it is a peculiarity of this reli-
gion that snow is a sure sign that the gods
are angry. In such a case, 'sheleg' might
well mean 'The gods are angry'.
The reply of the functionalist, of course,
will be that still Leslie's total internal
goings-on will be different in some way
from the internal goings-on of a robot that
thinks 'it's snowing' when it says 'Sheleg'. If
we knew the totality of the robot's internal
goings-on and knew the ideal psychological
theory we would be able to determine that
511
PUTN AM, HILAR Y
the content of 'sheleg' thoughts is 'The gods
are angry' and not 'It's snowing', What
does this require of the ideal psychological
theory?
It requires that the ideal psychological
theory be rich enough to describe the beliefs
of a believer of any possible religion, Or Leslie
may not be a believer in a primitive religion;
the robot may be a superscientist. It may be
saying 'quantum state such-and-such'
when it snows, or making a comment in a
physical theory we don't have yet. In short,
it looks as if an ideal psychological theory, a
theory that would be able to determine the
content of an arbitrary thought, would
have to be able to describe every belief of
every possible kind, or at least every human
belief of every possible kind, even of kinds
that are not yet invented, or that go with
institutions that have not yet come into
existence, More and more, the SuspiCiOn
grows that such a theory is a pure 'we
know not what',
What I think finally pushed me over the
anti-functionalist edge was a conversation I
had one day with Noam CHOMSKY.
Chomsky suggested that the difference
between a rational or a well-confirmed
belief and a belief that is not rational or not
well-confirmed might be determined by
rules that are innate in the human brain
(see INNATENESS), It struck me at once that
it ought to be fairly easy to show, using the
techniques one uses to prove the G6del
theorem, that if Chomsky is right then we
could never discover that he is right; that is
to say, that if what it is rational and not-
rational to believe is determined by a recur-
sive procedure which is specified in our
ideal competence description D, then it
could never be rational to believe that D is
our ideal competence description. And I was
able to show that this indeed is the case
without too much trouble (1985),
I do not want to claim that this argument
applies directly to our present discussion,
since what we are talking about is not
determining what is well confirmed but
with determining what the content of
thoughts is. What did occur to me is this:
what the G6del theorem shows is that
512
human reason is unable to survey itself well
enough to see its own limits, That is to say
that any description of our capacities that
we are able to formalize is a description of a
set of capacities that we are able to go
beyond. But this suggests to me, not that
we can prove that functionalism can't be
made less vague than it presently is, but
that we have no reason to believe that it
could be. In short, it suggests to me that,
first of all, if there is an 'ideal psychological
theory', that is, a theory that does every-
thing that the functionalist wants a
'description of human functional organiza-
tion' to do (let alone a 'normal form for the
description of the functional organization of
an arbitrary organism'), then there is no
reason to believe that it would be within the
capacity of human beings to discover it.
But, secondly, the property of being a
'description of human functional organiza-
tion' is itself such an unclear property that
the idea that there is such a description even if
we cannot recognize it surely goes beyond the
bounds of sense.
POST-FUNCTIONALIST PROGRAMMES
Of course, those who are sympathetic to
functionalism have not given up as a result
of my recantation; there are a number of
what we might call 'post-functionalist' pro-
grammes on the market. One kind of post-
functionalist programme seeks to deal with
the problem which, in my opinion, is fatal
for the 'implicit definition by a theory' idea
which was at the heart of classic function-
alism - an idea that was central not only
to my version of functionalism but also to
the somewhat different version proposed by
David LEWIS (1983), (In Lewis's view we
already have the ideal psychological theory
required to implicity define the content of
an arbitrary thought: it is just folk psycho-
10gyJ) That programme is the one of fixing
the content of thoughts, by relying entirely
on external factors to do that. DRETSKE
(1981) and STALNAKER (1984) fur
example try to define the content of
thoughts as well as of expressions in a lan-
guage by simply looking for probabilistic
relations between the occurrences of
thoughts and expressions and external
states of affairs. However, I believe that
Putnam (1983) and Loewer (1987) have
shown that the information theoretic con-
cepts upon which Dretske and Stalnaker
rely cannot individuate contents finely
enough. FODOR (1990) proposes to rely not
on information theoretic notions, but
rather on the notion of causality. In
Renewing Philosophy (1992, ch. 3) I show
the following. (i) The notion of causality
Fodor employs presupposes intentional
notions; in particular, Fodor needs to
assume a distinction between contributory
causal factors and the cause of an event; I
regard this as an intentional notion
because which is the cause of an event
depends on the interests we have in the
context; it is not something that is in-
scribed in the phenomena themselves. (ii)
The assignments of contents that result if
we look only at the causes of utterances
are the wrong ones. Block (1986) proposes
a 'dual aspect' semantics in which the
content of thoughts is determined by two
factors: one factor, called 'narrow content'
is determined, as in my functionalist the-
ories, by the relations involved in the in-
ternal cognitive processing, in particular by
'inferential roles', and the other factor, the
reference relation, is determined by causal
connections. A difficulty with the first
factor (Putnam, 1988, ch. 3) is that the
inferential roles of concepts change enor-
mously as the result of changes in belief as
well as on account of changes in the
meanings of words, and that there is no
reason to suppose that one can distinguish
between changes in inferential roles that
represent only changes in beliefs from
changes that represent changes in meaning
without explicitly invoking some such
notion as 'meaning'. In addition, the fail-
ures of causal theories of reference and of
information-theoretic accounts of reference
suggest that the other aspect of the 'dual
aspect theory' is in no better shape.
I do not have space here to explain where
I think we should go next. But - and this
PUTNAM. HILAR Y
will be the subject of my next work on
philosophy of mind (my Dewey Lectures (to
be delivered in the spring of 1994) will deal
with the philosophy of mind) - I am struck
by the ways in which key elements of func-
tionalism - in particular, both the causal
theory of perception and the picture of con-
ception as manipulation of symbols - operate
within a seventeenth-century picture of the
mind. I believe that a very different way of
looking at the problems is possible (one sug-
gested, in different ways, by the writings of
Austin and Wittgenstein, as well as certain
earlier philosophers); I hope to develop that
way in future writing.
See also CONTENT; COMPUTATIONAL
MODELS; CONCEPTS; CONCEPTUAL ROLE
SEMANTICS; INTENTIONALITY; RATION-
ALITY; REPRESENTATION.
BIBLIOGRAPHY
Block, N. 1986. An advertisement of a seman-
tics for psychology. Midwest Studies in Philo-
sophy, Vol. 10, 615-78.
Churchland P. 1984. Matter and Consciousness.
Dretske, F. 1981. Knowledge and the Flow, of
Information. Cambridge, MA: MIT Press.
Lewis, D. 1983. Philosophical Papers, Vol. 1.
Loewer, B. 1987. From information to inten-
tionality. Synthese, 70: 2, 287-316.
Putnam, H. 1975: Philosophical Papers, Vol. 2,
Mind, Language and Reality. Cambridge Uni-
versity Press.
-- 1979. Philosophical Papers, Vol. 1. Math-
ematics, Matter and Method, 2nd edn. Cam-
-- 1985. Reflexive reflections. Erkenntnis,
22,143-153.
-- 1988. Representation and Reality. Cam-
-- 1992. Renewing Philosophy. Cambridge,
MA: Harvard University Press.
Stalnaker, R. 1984. Inquiry. Cambridge, MA:
MIT Press.
HILAR Y PUTNAM
513
o
---
qualia Qualia include the ways it feels to
see, hear and smell, the way it feels to have
a pain: more generally, what it's like to
have mental states. Qualia are experiential
properties of sensations, feelings, percep-
tions and, in my view, thoughts and desires
as well. But, so defined, who could deny
that qualia exist? Yet, the existence of
qualia is controversial. Here is what is con-
troversial: whether qualia, so defined, can
be characterized in intentional, functional
or purely cognitive terms. Opponents of
qualia think that the content of experience
is intentional content (like the content of
thought), or that experiences are function-
ally definable, or that to have a qualitative
state is to have a state that is monitored in
a certain way or accompanied by a thought
to the effect that I have that state (see
CONTENT FUNCTIONALISM: INTENTION-
ALITY). If we include the idea that experi-
ential properties are not intentional or
functional or purely cognitive in the defini-
tion of 'qualia', then it is controversial
whether there are qualia.
This definition of 'qualia' is controversial
in a respect familiar in philosophy. A tech-
nical term is often a locus of disagreement,
and the warring parties will often disagree
about what the important parameters of
disagreement are. DENNETT, for example,
has supposed in some of his writings that it
is of the essence of qualia to be non-
relational, incorrigible (to believe one has
one is to have one) and to have no scientific
nature (see Flanagan, 1992, p. 61). This is
what you get when you let an opponent of
qualia define the term. A proponent of
qualia ought to allow that categorizations of
them (beliefs about them) can be mistaken,
and that science can investigate qualia. I
think that we ought to allow that qualia
514
might be physiological states, and that their
scientific nature might even turn out to be
relational. Friends of qualia differ on
whether or not they are physical. In my
view, the most powerful arguments in
favour of qualia actually presuppose a
physicalistic doctrine, the supervenience of
qualia on the brain. (See PHYSICALISM;
SUPER VENIENCE.)
Perhaps the most puzzling thing about
qualia is how they relate to the physical
world. Sometimes this is put in terms of the
explanatory gap, the idea that nothing we
know or can conceive of knowing about the
brain can explain why qualia feel the way
they do. The explanatory gap is closely
related to the thought experiments that
dominate the literature on qualia.
THE KNOWLEDGE ARGUMENT
One of these thought experiments is the
case of Jackson's (1986) Mary, who is
raised in a black and white environment in
which she learns all the functional and
physical facts about colour vision. None-
theless, when she ventures outside for the
first time, she learns a new fact: what it is
like to see red. So, the argument goes, what
it is like to see red cannot be a functional or
physical fact. Dennett (1991) objects that
perhaps she could have figured out which
things are red: but that is beside the point
for two reasons. The question is: does she
know what it is like to see red, not which
things are red? And does she know it simply
in virtue of knowing all the functional and
physical facts about colour vision, whether
or not she is clever enough to figure it out
on the basis of what she knows?
LEWIS denies that Mary acquires any new
knowledge-that, inSisting that she only
acquires knowledge-how, abilities to
imagine and recognize. But as Loar points
out, the knowledge she acquires can appear
in embedded contexts. For example, she
may reason that if this is what it is like to
see red, then this is similar to what it is like
to see orange. Lewis's ability analysis of
Mary's knowledge has the same problem
here that non-cognitive analyses of ethical
language have in explaining the logical
behaviour of ethical predicates.
Here is a different (and in my view more
successful) objection to Jackson (Horgan,
1984b; Peacocke, 1989; Loar, 1990;
Papineau, 1993; van Gulick, 1993). What
Mary acquires when she sees red is a new
PHENOMENAL concept, a recognitional dis-
position that allows her to pick out a
certain type of phenomenal feel. This new
phenomenal concept is a constituent of gen-
uinely new knowledge - knowledge of what
it is like to see red. But the new phenomenal
concept picks out old properties, properties
picked out by physical or functional con-
cepts that she already had. So the new
knowledge is just a new way of knowing old
facts. Before leaving the room, she knew
what it is like to see red in a third-person
way; after leaving the room, she acquires a
new way of knowing the same fact. If so,
what she acquires does not rule out any
POSSIBLE WORLDS that were not already
ruled out by the facts that she already
knew, and the thought-experiment poses no
danger to physicalistic doctrines. Incident-
ally, the recognitional disposition account
indicates how qualia could turn out to be
relational; perhaps the recognitional disposi-
tion picks out a relational physical state of
the brain or even a functional state. (But
see the criticism of Loar in FUNCTIONAL-
ISM(2).)
ABSENT QUALIA
Another familiar conundrum is the absent
qualia hypothesis. If human beings can be
described computationally, as is assumed by
the research programme of cognitive
science, a robot could in principle be built
that was computationally identical to a
QUALIA
human. But would there be anything it was
like to be that robot? Would it have qualia?
(See Shoemaker, 1975, 1981, and White,
1986.) Some thought experiments have
appealed to oddball realizations of our func-
tional organization, e.g. by the economy of
a country. If an economy can share our
functional organization, then our functional
organization cannot be sufficient for qualia.
Many critics simply bite the bullet at this
point, saying that the oddball realizations
do have qualia. Lycan (1987) responds by
making two additions to functionalism as
spelled out in FUNCTIONALISM (2). The
additions are designed to rule out oddball
realizations of our functional organization of
the ilk of the aforementioned economy. He
suggests thinking of the functional roles in
teleological terms and thinking of these
roles as involving the details of human phy-
siology (see TELEOLOGY). Economies don't
have the states with the right sort of evolu-
tionary 'purpose', and their states are not
physiological. On the first move, see FUNC-
TIONALISM (2). On the second, note that
including physiology in our functional
definitions of mental states will make them
so specific to humans that they won't apply
to other creatures that have mental states.
Further, this idea violates the spirit of the
functionalist proposal, which, being based
on the computer analogy, abstracts from
hardware realization. Functionalism with-
out multiple hardware realizations is func-
tionalism in name only.
THE INVERTED SPECTRUM
One familiar conundrum that uses a physic-
alistic idea of qualia against functionalist
and intentionalist ideas is the famous in-
verted spectrum hypothesis, the hypothesis
that things we both call 'red' look to you
the way things we both call 'green' look to
me, even though we are functionally (and
therefore behaviourally) identical. A first
step in motivating the inverted spectrum
hypothesis is the possibility that the brain
state that I have when I see red things is
the same as the brain state that you have
when you see green things, and conversely.
515
QUALIA
(Nida-Rumelin, forthcoming, presents evid-
ence that this is a naturally occurring phe-
nomenon.) Therefore, it might be said, our
experiences are inverted. What is assumed
here is a supervenience doctrine, that the
qualitative content of a state supervenes on
physiological properties of the brain.
There is a natural functionalist reply.
Notice that it is not possible that the brain
state that I get when I see things we both
call 'red' is exactly the same as the brain
state that you get when you see things we
both call 'green'. At least, the total brain
states can't be the same, since mine causes
me to say 'It's red', and to classify what I'm
seeing as the same colour as blood and fire
hydrants, whereas yours causes you to say
'It's green', and to classify what you are
seeing with grass and Granny Smith apples.
Suppose that the brain state that I get when
I see red and that you get when you see
green is X-oscillations in area V4, whereas
what I get when I see green and you get
when you see red are Y-oscillations in area
V4. The functionalist says that phenomenal
properties should not be identified with
brain states quite so 'localized' as X-oscilla-
tions or Y -oscillations, but rather with more
holistic brain states that include tendencies
to classify objects together as the same
colour (see HOLISM). Thus the functionalist
will want to say that my holistic brain state
that includes X-oscillations and your holis-
tic brain state that includes Y -oscillations
are just alternative realizations of the same
experiential state (Harman, 1990). So the
fact that red things give me X-oscillations
but they give you Y-oscillations doesn't
show that our experiences are inverted. The
defender of the claim that inverted spectra
are possible can point out that when some-
thing looks red to me, I get X-oscillations,
whereas when something looks green to
me, I get Y-oscillations, and so the differ-
ence in the phenomenal aspect of experi-
ence corresponds to a local brain state
difference. But the functionalist can parry
by pointing out that this difference has only
been demonstrated intra-personally, keeping
the larger brain state that specifies the roles
of X-oscillations in classifying things con-
516
stant. He can insist on typing brain states
for inter-personal comparisons holistically.
And most friends of the inverted spectrum
are in a poor position to insist on typing
experiential states locally rather than holist-
ically, given that they normally emphasize
the 'explanatory gap', the fact that there is
nothing known about the brain that can
adequately explain the facts of experience
(see CONSCIOUSNESS). SO the friend of the
inverted spectrum is in no position to insist
on local physiological individuation of
qualia. At this stage, the defender of the
inverted spectrum is stymied.
One move the defender of the possibility
of the inverted spectrum can make is to
move to an intra-personal inverted spec-
trum example. Think of this as a four-stage
process. (1) You have normal colour vision.
(2) You have colour inverting devices in-
serted in your retinas or in the lateral geni-
culate nucleus, the first way-station behind
the retina, and red things look the way
green things used to look, blue things look
the way yellow things used to look, etc. (3)
You have adapted, so that you naturally
and spontaneously call red things 'red', etc.,
but when reminded, you recall the days
long ago when ripe tomatoes looked to you,
colourwise, the way Granny Smith apples
do now. (4) You get amnesia about the days
before the lenses were inserted. Stage 1 is
functionally equivalent to stage 4 in the
relevant respects, but they are arguably
qualia-inverted. So we have an inverted
spectrum over time. The advantages of this
thought experiment are twofold. First, the
argument profits from the force of the sub-
ject's testimony at stages 2 and 3 for qualia
inversion. Second, the four-stage setup
forces the opponents to say what stage is
the one where my description goes wrong.
(See Shoemaker, 1981; Block, 1990.) Rey
(1993) attacks (3), Dennett (1991) attacks
(2) and (3), and White (1993) attacks (4).
In my view, the most vulnerable stage is (3)
because the functionalist can raise doubts
about whether what it's like to see red
things could remain the same during the
changes in responses that have to go on in
the process of adaptation.
Why, an opponent might ask, is the
inverted qualia argument against function-
alism any more powerful than the inverted
qualia argument against physicalism? After
all, it might be said, one can imagine par-
ticle-for-particle duplicates who have
spectra that are inverted with respect to one
another. But though physical duplicates
with inverted spectra may be imaginable,
they are ruled out by a highly plausible
principle that any materialist should accept:
that qualia supervene on physical constitu-
tion. The thought experiments that I have
been going through argue that even materi-
alists should accept the possibility of an
inverted spectrum, and further, that for all
we know, such cases are feasible via robot-
ics or genetic engineering, or even actual.
And in so doing, they make the case for
conceptual possibility stronger, for one is
surer that something is genuinely con-
ceptually possible if one can see how one
might go about making it actual.
INVERTED EARTH
An interesting variant of the inverted spec-
trum thought-experiment is Inverted Earth
(Block, 1990). Inverted Earth is a planet
that differs from Earth in two relevant ways.
First, everything is the complementary
colour of the corresponding earth object.
The sky is yellow, the grass-like stuff is red,
etc. (To avoid impossibility, we could
imagine, instead, two people raised in rooms
in which everything in one room is the com-
plementary colour of the corresponding item
in the other room.) Second, people on Inver-
ted Earth speak an inverted language. They
use 'red' to mean green, 'blue' to mean
yellow, etc. If you order paint from Inverted
Earth, and you want yellow paint, you fax
an order for 'Blue paint'. The effect of both
inversions is that if you are drugged and
kidnapped in the middle of the night, and
inverters are inserted behind your eyes (and
your body pigments are changed), you will
notice no difference if you are placed in the
bed of your counterpart on Inverted Earth.
(Let's assume that the victim does not know
anything about the science of colour.)
QUALIA
Now consider the comparison between
you and your counterpart on Inverted
Earth. The counterpart could be your iden-
tical twin who was fitted with inverting
lenses at birth and put up for adoption on
Inverted Earth. or the counterpart could be
you after you've been switched with your
twin and have been living there for a long
while. Looking at blue things gives you Z-
oscillations in the brain, yellow things give
you W -oscillations: your twin gets the
opposite. Now notice the interesting differ-
ence between this twin case and the one
mentioned earlier: there can be perfect
inversion in the holistic brain states as well
as the local ones. At this moment. you both
are looking at your respective skies. You get
Z-oscillations because your sky is blue, he
gets Z-oscillations because his sky is yellow.
Your Z-oscillations make you say 'How
bluel'. and his Z-oscillations make him say
'How bluel' too. Indeed, we can take your
brains to be molecular duplicates of one
another. Then the principle of the super-
venience of qualia on holistic brain state
dictates that experientially. at the moment
of looking at the skies, you and your twin
have the same qualia.
But though you and your twin have the
same qualia. you are functionally and
intentionally inverted. If you are asked to
match the colour of the sky with a Munsell
colour chip. you will pick a blue one. but if
your twin is shown the same (earth-made)
Munsell chips. he will pick a yellow one.
Further, when he says 'How blue!' he means
'How yellow!' Recall that the Inverted Earth
dialect. of which he is a loyal member. has
colour words whose meanings are inverted
with respect to ours. You and your twin are
at that moment functionally and intention-
ally inverted, but qualitatively identical. So
we have the converse of the inverted spec-
trum. And there is no problem about local
versus holistic brain states as in the inter-
subjective inverted spectrum; and no
problem about whether qualia could persist
unchanged through adaptation as in the
intra-subjective inverted spectrum.
The argument that you and your twin
are qualitatively the same can work either
517
QUALIA
of two ways. We can assume the principle
of supervenience of qualia on the brain,
building the brain-identity of the twins into
the story. Or we can run the story in terms
of your being kidnapped, drugged, and
placed in your twin's niche on Inverted
Earth. What justifies the idea that your
qualia are the same is that you notice no
difference when you wake up in your
Twin's bed after the switch; no appeal to
supervenience is required.
Notice that the functional differences
between these qualia-identical twins are
long-arm functional differences (see FUNC-
TIONALISM) and the intentional differences
are external intentional differences.
Perhaps, you might say, the twins are not
inverted in short-arm functional roles and
narrow intentional content. The cure for
this idea is to ask the question of what the
purely internal functional or intentional dif-
ferences could be that would define the dif-
ference between an experience as of red and
an experience as of green. The natural
answer would be to appeal to the internal
aspects of beliefs and desires. We believe, for
example, that blood is red but not that it is
green. However, someone could have colour
experience despite having no standing
beliefs or desires that differentiated colours.
Imagine a person raised in a room where
the colour of everything is controlled by a
computer, and nothing retains its colour for
more than 10 seconds. Or imagine a person
whose colour perception is normal but who
has forgotten all colour-facts.
There is no shortage of objections to these
lines of reasoning. I will very briefly
mention two closely related objections. It
has been objected (Hardin, 1988) that red is
intrinsically warm, whereas green is intrin-
sically cool, and thus inversion will either
violate functional identity or yield an inco-
herent cool-red state. (Note, incidentally,
that this isn't an objection to the inverted
earth thought experiment; since that is a
case of qualitative identity and functional
difference, no functional identity is involved.)
But the natural reply (Block, 1990) is that
warm and cool can be inverted too. So long
as there is no intrinsic connection between
518
colour qualia and behaviour, the inverted
spectrum is safe. But is there such an
intrinsic connection? Dennett (1991) says
there is. Blue calms, red excites. But perhaps
this is due to culture and experience;
perhaps people with very different cultures
and experiences would have colour experi-
ences without this asymmetry. The research
on this topic is equivocal; Dennett's sole
reference, Humphrey (1992), describes it as
'relatively second-rate', and that is also my
impression. The fact that we don't know is
itself interesting, however, for what it shows
is that this asymmetry is no part of our
colour concepts. As Shoemaker (1981)
points out, even if human colour experience
is genetically asymmetrical, there could
nonetheless be people much like us whose
colour experience is not asymmetrical. So
an inversion of the sort mentioned in the
thought experiments is conceptually pos-
sible, even if it is not possible for the human
species. But then colour inversion may be
possible for a closely related species whose
colour qualia are not in doubt, one which
could perhaps be produced by genetic
engineering. Functionalism would not be a
very palatable doctrine if it were said to
apply to some people's colour experiences
but not to others.
THE EXPLANATORY GAP AGAIN
At the outset, I mentioned the 'explanatory
gap', the idea that nothing now known
about the brain, nor anything anyone has
been able to imagine finding out would
explain qualia. We can distinguish infla-
tionary and deflationary attitudes towards
this gap among those who agree that the
gap is unclosable. McGinn (1991) argues
that the gap is unclosable because the fun-
damental nature of consciousness is inac-
cessible to us, though it might be accessible
to creatures with very different sorts of
minds. But a number of authors have
favoured a deflationary approach, arguing
that the unclosability of the explanatory
gap has to do with our concepts, not with
nature itself. Horgan (1984), Levine (1993),
Jackson (1993). Chalmers (1993) (and
interestingly, McGinn, 1991, too) have con-
tributed to working out the idea that reduc-
tive explanation in science depends on a
priori analyses of the phenomena to be
explained, usually in functional terms. (A
version of this point was made in Nagel,
1974.) Consider Chalmers' example of the
reductive explanation of life. Life can be
roughly analysed in terms of such general
notions as metabolism and adaptation, or
perhaps more specific notions such as diges-
tion, reproduction and locomotion, and
these concepts can themselves be given a
functional analysis. Once we have explained
these functions, suppose someone says 'Oh
yes, I see the explanation of those functions,
but what about explaining life?' We can
answer that, a priori, to explain these func-
tions is to explain life itself.
In some cases, the a priori analysis of the
item to be explained is more complicated.
Consider water. We can't give an a priori
analysis of water as the colourless, odour-
less liquid in rivers and lakes called 'water',
because water might not have been colour-
less, it might have been called 'glue', there
might not have been lakes, etc. But we can
formulate an a priori reference fixing defini-
tion of the sort that Kripke has emphasized:
water = R( the colourless, odourless liquid in
rivers and lakes called 'water'), where the
'R' is a rigidification operator that turns a
definite description into a rigid designator.
(A rigid designator picks out the same thing
in all possible worlds in which the thing
exists; for example, 'Aristotle' is rigid. To
rigidify a definite description is to treat it as
a name for whatever the definite description
actually picks out.) Thus, suppose we want
to explain the fact that water dissolves salt.
It suffices to explain that H
2
0 dissolves salt
and that H
2
0 is the colourless, odourless
liquid in rivers and lakes called 'water'. If
someone objects that we have only
explained how something colourless, odour-
less, etc. dissolves salt, not that water does,
we can point out that it is a priori that
water is the actual colourless, odourless,
etc., substance. And if someone objects that
we have only explained how H
2
0 dissolves
salt, not how water does, we can answer
QUALIA
that from the fact that H
2
0 is the colourless,
odourless, etc., stuff and that, a priori, water
is the (actual) colourless, odourless, etc.,
stuff, we can derive that water is H
2
0.
The upshot is that closing the explan-
atory gap requires an a priori functional
analysis of qualia. If Kripke (1980) is right
that we pick out qualia by their qualitative
character and not by their functional role,
then no a priori reference fixing definition
can be given for qualitative concepts of the
sort that can be given for 'water' and 'life'.
Of course, if there is a true functional
analysis that picks out a quale, it can be
rigidified, but it still won't be an a priori
characterization. Pain = R(Aunt Irma's
favourite sensation) can be true and neces-
sary without being a priori. And if the argu-
ments about qualia inversion just sketched
are right, there is no a priori conceptual
analysis of qualitative concepts either, and
so the explanatory gap is unclosable. As
Chalmers points out, with a physical or a
functional account, we can explain the func-
tions associated with qualia, the capacity to
classify things as red, for example. But once
we have explained these functions, there will
be a further question: why are these func-
tions accompanied by qualia? Such a further
question does not arise in the case of life and
water precisely because of the availability of
an a priori functional analysis.
It would be natural to suppose that the
explanatory gap derives from the fact that
neuroscientists have not yet come up with
the required concepts to explain qualia.
Nagel (1974) gives an analogy that sug-
gests this idea. We are in the situation, he
suggests, of a caveman who is told that
matter is energy. But he does not have the
concepts to appreciate how this could be so.
These concepts, however, are ones that
some of us do have now, and it is a natural
thought that a few hundred years from
now, the concepts might be available to
explain qualia physically. But the defla-
tionary account of reductive explanation
denies this, blaming the explanatory gap on
our ordinary concepts, not on science.
See also An Essay on Mind section 2.2.
519
QUALIA
BIBLIOGRAPHY
Block, N. 1990. Inverted earth. In Philo-
sophical Perspectives, Vol. 4, ed. J. Tomberlin.
Ridgeview.
Chalmers, D.J. 1993. Toward a Theory of Con-
sciousness. University of Indiana Ph.D.
thesis.
Davies, M., and Humphreys, G. 1993. Con-
sciousness. Oxford: Basil Blackwell.
Dennett, D. 1988. Quining qualia. In Con-
sciousness in Contemporary Society ed. A.
Marcel and E. Bisiach. Oxford University
Press.
Dennett, D. 1991. Consciousness Explained.
New York: Little, Brown.
Flanagan, O. 1992. Consciousness Reconsidered.
Hardin, C. 1988. Color for Philosophers. Indi-
anapolis: Hackett.
Harman, G. 1990. The intrinsic quality of
experience. In Philosophical Perspectives, Vol.
4, ed. J. Tomberlin. Atascadero, CA: Ridge-
view.
Horgan, T. 1984a. Supervenience and cosmic
hermeneutics. Southern Journal of Philosophy,
Supplement 22, 19-38.
Horgan, T. 1984b. Jackson on physical infor-
mation and qualia. Philosophical Quarterly,
34, 147-53.
Humphrey, N. 1992. A History of the Mind.
New York: Simon & Schuster.
Jackson, F. 1986. What Mary didn't know.
Journal of Philosophy, 83, 291-95
Jackson, F. 1993. Armchair metaphysics. In
Philosophy in Mind ed. J. O'Leary-Hawthorne
and M. Michael. Kluwer.
Kripke, S. 1980. Naming and Necessity. Cam-
Levine, J. 1993. On leaving out what it is like.
In Davies and Humphreys, 1993.
Loar, B. 1990. Phenomenal properties. In Phi-
losophical Perspectives, Vol. 4, ed. J. Tomber-
lin. Atascadero, CA: Ridgeview.
Lycan, W. 1987. Consciousness. Cambridge,
MA.: MIT Press.
McGinn, C. 1991. The Problem of Conscious-
Nagel. T. 1974. What is it like to be a bat?
Philosophical Review, 83, 435-50.
Nida-Rumelin, M. Forthcoming. Pseudonormal
vision. An actual case of qualia inversion?
Philosophical Studies.
Papineau, D. 1993. Physicalism, Conscious-
520
ness and the Antipathetic Fallacy. Austra-
lasian Journal of Philosophy, 71: 2,169-184
Peacocke, C. 1989. No resting place: a critical
notice of The View from Nowhere. Philosophi-
cal Review, 98, 65-82.
Rey, G. 1993. Sensational sentences switched.
Philosophical Studies, 70, 1.
Shoemaker, S. 1975. Functionalism and
qualia. Philosophical Studies, 27, 291-315.
Shoemaker, S. 1981. Absent qualia are impos-
sible - a reply to Block. Philosophical Review,
90: 4,581-99.
Van Gulick, R. 1993. Understanding the phe-
nomenal mind: Are we all just armadillos?
In Davies and Humphreys, 1993.
White, S.L. 1986. Curse of the qualia. Synth-
ese, 68, 333-68.
White, S.L. 1993. Color and the narrow con-
tents of experience. Paper delivered at the
Eastern Division of the American Philoso-
phical Association.
NED BLOCK
Quine, Willard Van Orman Although few
of his writings directly address traditional
issues in the philosophy of mind, Quine's
work has been a major influence upon the
area since the 1950s. His contributions to
the philosophy of language shook the
prejudices underlying much 'traditional'
philosophical analysis. forcing philosophers
to adopt new models and approaches in
thinking about the mind. By defending
'naturalism', the idea that philosophy is
continuous with the sciences and has no
privileged source of knowledge, he con-
tributed to the growth of cognitive science
and the idea that an adequate philosophical
understanding of mind should take into
account developments in psychology and
biology (see COGNITIVE PSYCHOLOGY). And
his uncompromising defence of physicalism
since the 1950s has shaped our under-
standing of what this position commits us
to.
Quine's earliest discussions of mind attack
'mentalism', the claim that there are
irreducibly mental entities or events. Since
his aim was normally to show that such
theories have no role in the explanation of
linguistic meaning, his underlying philoso-
phy of mind was rarely explicitly presented.
Fortunately, some of Quine's more recent
work remedies this lack (Quine, 1975, and
especially 1985, and 1990, ch. IV).
NATURALISM: MIND AND BODY
Quine's naturalism affirms that there can be
no 'philosophical' study of mind outside
psychology: progress in philosophical
understanding of the mind is inseparable
from progress in psychology. Furthermore,
psychology is a 'natural science' studying a
'natural phenomenon, viz. a physical
human subject' (1969, p. 82): Brentano
was wrong to advocate an autonomous dis-
cipline studying its own distinctively mental
objects. 'Unless a case can be made for dis-
embodied spirits', Quine argued, 'a dualism
of mind and body is an idle redundancy.'
(1985, p. 5)
'States of mind' (1985, p. 5) offers the
following brief argumerit:
Corresponding to every mental state,
however fleeting or however remotely
intellectual, the dualist is bound to admit
the existence of a bodily state that
obtains when and only when the mental
one obtains. The bodily state is trivially
specifiable in the dualist's own terms,
simply as the state accompanying a
mind that is in the mental state. Instead
of ascribing the one state to the mind,
then, we may equivalently ascribe the
other to the body. The mind goes by the
board, and will not be missed.
(The relevance of 'disembodied spirits',
whose states would not be correlated with
bodily ones, is clear.) This 'effortless physic-
alism' encourages us to retain the tradi-
tional mentalistic terms (we speak of pains,
beliefs, sudden thoughts and so on), but 'we
reckon mental states as states of the body
rather than as states of another substance,
the mind' (ibid). The target here appears to
be substance dualism, the idea that our
ontology should contain non-physical
mental objects (see DUALISM; PHYSICALISM).
But Quine may be making the stronger
QUINE. WILLARD VAN ORMAN
claim that all of our explanatory and
descriptive needs are met by talk of the
'bodily states' of physical objects. This
would exclude property dualism, the idea
that psychological terms express irreducibly
mental properties of physical objects. As we
shall see, his position is more complex than
this: there are irreducible psychological
properties, but all explanation is ultimately
physical.
His account of our mental concepts
emerges as he examines how we acquire
them, how we learn 'to call our anxieties
anxieties, and our dull aches dull aches, our
joys joys and our awareness awareness'. He
explains that:
such terms are applied in the light of
publicly observable symptoms: bodily
symptoms strictly of bodily states, and
the mind is as may be. Someone
observes my joyful or anxious expres-
sion, or perhaps observes my gratifying
or threatening situation itself, or hears
me tell about it. She then applies the
word 'joy' or 'anxiety'. After another
such lesson or two I find myself apply-
ing those words to some of my sub-
sequent states in cases where no
outward signs are to be observed
beyond my report itself. Without the
outward signs, to begin with, mentalis-
tic terms could not be learned at all.
(1985, p. 6)
He is no crude behaviourist: mental states
are inner states which may (but need not)
be manifested in behaviour (see BEHA-
VIOURISM). Nor is he a dualist: a mental
state is a bodily state, 'a state of nerves'.
Exploiting their links with behaviour, men-
talistic language enables us to refer to inner
neurological states and use them in expla-
nation, while ignorant of the neural
mechanisms involved. In effect, we identify
mental states as the bodily states (whatever
they may be) which occupy various specifi-
able causal roles in the determination of
behaviour: the position is close to FUNC-
TIONALISM. Patterns in behaviour enable us
to identify states which occupy distinctive
521
QUINE. WILLARD V AN ORMAN
causal roles; further research may help us
to construct neurophysiological descriptions
of them. This may accord with the apparent
rejection of property dualism: the apparent
ineliminability of psychological concepts is a
reflection of our ignorance of the neurophy-
siological underpinning of our thoughts and
actions. But the position is more complex
than that.
SOME COMPARISONS: BEHAVIOURISM.
ELIMIN A TIVISM. IDENTITY THEORIES
Quine is sometimes misdescribed as a beha-
viourist, and was certainly influenced by
writers such as B. F. Skinner. The passages
just cited show that he is closer to the lD E N -
TITY THEOR Y than to behaviourism: mental
states are physical states, usually of the
brain. But they are identified as the neuro-
physiological grounds of behavioural dis-
positions: behaviour is fundamental to the
development of our psychological concepts.
Moreover, although behaviourism is not
compulsory in psychology, it is 'mandatory'
in linguistics:
Each of us learns his language by obser-
ving other people's verbal and other
behaviour and having his own faltering
verbal behaviour observed and rein-
forced or corrected by others. We
depend strictly on overt behaviour in
observable situations. As long as our
command of our language fits all ex-
ternal checkpoints, where our utterance
or our reaction to someone's utterance
can be appraised in the light of some
shared situation, so long all is well. Our
mental life between checkpoints is indif-
ferent to our rating as a master of the
language. There is nothing in linguistic
meaning beyond what is to be gleaned
from overt behaviour in observable
circumstances. (1990, pp. 37-8)
Since many of Quine's remarks about men-
talism and the mind occur when his real
interest is in understanding and linguistic
representation, it is un surprising that his
writings often give an impression of an
522
adherence to behaviourism which reflects
his philosophy of language rather than his
philosophy of mind.
Materialist philosophers are often classi-
fied according to whether they hold that
ordinary psychological terms denote physi-
calor bodily states, or whether they hold
that traditional mentalistic language ('FOLK
PSYCHOLOGY') will eventually be repu-
diated, eliminated from the language as part
of a discredited theory. Quine occasionally
allies himself with each of these opposed
tendencies. In 'Mind and verbal dispositions'
(1975, p. 94) he emphasizes the advantages
of dispensing with mental states, doing our
theorizing in physiological terms. The more
recent 'States of mind' advocates 'an identi-
fication of mental states with bodily ones,
neural ones: a construing of the mental as
neural' (1985, p. 6).
In fact, as Quine sees, his philosophy of
language robs this debate of its point. A
philosopher who is realist about properties,
viewing the world as containing objectively
natural kinds of properties, and who
believes that the notion of reference is a
fundamental explanatory notion within an
account of how language works, might
address this question of whether two dis-
parate expressions generally refer to the
same property. But since Quine rejects
realism about properties and believes that
reference is radically inscrutable, no sense
attaches to the question whether 'pain'
expresses the property of being in a parti-
cular brain state. Quine consistently denies
that a distinction can be drawn between
'identifying the mental states with the states
of nerves ... and repudiating them in favor
of the states of nerves' (1985, p. 6). Hence
he is content to talk of identity and identifi-
cation; but one misunderstands his position
if one looks for a considered refutation of
elimination (see ELIMINATIVISM).
IDENTITY: TYPE. TOKEN. ANOMALOUS
MONISM
For Quine, expressions like 'Tom's tooth-
ache' or 'an itch behind the left ear' collect
together heterogeneous classes of neural
states and events. A 'type identity' theorist
expects that such classes will be unified by
some common neurophysiological trait: the
mental kind will correspond to a physical
kind. Those who expect only 'token iden-
tities' do not anticipate that such classes of
physical events will form physical kinds,
that they can be seen to 'belong together'
except by reference to the corresponding
mental properties. Quine has recently
declared that we should expect no more
than token identities: no neurophysiological
kinds correspond to our systems of psycho-
logical description (see An Essay on Mind
section 3.6; TYPE/TOKEN). Indeed, he has
declared his adherence to a position asso-
ciated with Donald DAVIDSON, 'ANOM-
ALOUS MONISM' (Quine, 1985, p. 7; 1990,
pp. 70-3; Davidson, 1980, pp. 214-25).
In Pursuit of Truth, he explains that Tom's
perceptions of rain can differ 'because there
are different indicators of rain'. Although
these perceptions may form a class that is
too complex and heterogeneous for us to
give it a satisfying general description in
neurophysiological terms, we can be con-
fident that there is a 'neural trait that
unifies these neural events as a class; for it
was by stimulus generalization, or sub-
jective similarity, that Tom eventually
learned to make the observation sentence
"It's raining" do for all of them' (1990, p.
62). Thus far, type identity. But if we move
to the more general mental property perceiv-
ing that it is raining (which can be shared by
many different people) the position is differ-
ent: 'a neurological rendering of "perceives
that it is raining", applicable to all comers,
would be out of the question' (1990, p. 70).
This is because people's 'nerve nets' differ in
consequence of their genetic endowments
and their educational histories. The expres-
sion 'cuts through all that hopeless neuro-
logical complexity' (1990, p. 62) unifying
such states by reference to a symptom
rather than by a common neural mechan-
ism: token identity seems to be the order of
the day.
Once we turn from perceptions to beliefs
and other propositional attitudes, the com-
plexities grow. The evidence upon which we
rely in ascribing beliefs to people is enor-
mously varied: we note what they say, the
wagers they are prepared to accept, such
behaviour as searching and fleeing, infor-
mation about past experience and training,
etc. (1985, pp. 6-7). Quine writes: 'The
empirical content of ascriptions of belief is
thus heterogeneous in the extreme, and the
physiological mechanisms involved are no
less so'. We group states together as cases of
'perceiving that p' and 'believing that p':
they are unified not by any common
physiological feature but rather by the 'that
clause' or CONTENT clause. Thus, for Quine,
although mental events are all physical
events, their occurrence being explicable by
reference to physical law, there are 'irre-
ducibly mental ... ways of grouping them:
grouping a lot of respectably physical per-
ceptions as perceptions that p, and grouping
a lot of respectably physical belief instances
as the belief that p' (p. 71). This is the view
that Quine calls 'anomalous monism': it
signals his acceptance of a form of property
dualism.
EMPATHY AND THE PROPOSITIONAL
ATTITUDES
How are we to understand our use of 'ir-
reducibly mental' ways of classifying physi-
cal events? The behavioural manifestations
of BELIEFS, DESIRES and INTENTIONS are
enormously varied, as we have suggested.
When we move away from perceptual
beliefs, the links with behaviour are intract-
able and indirect: the expectations I form on
the basis of a particular belief reflect the
influence of numerous other opinions; my
ACTIONS are formed by the totality of my
preferences and all those opinions which
have a bearing upon them. The causal pro-
cesses that produce my beliefs reflect my
opinions about those processes, about their
reliability and the interferences to which
they are subject. Thus behaviour justifies
the ascription of a particular belief only by
helping to warrant a more inclusive inter-
pretation of the overall cognitive position of
the individual in question. Psychological
description, like translation, is a HOLISTIC
523
business. And once this is taken into
account, it is all the less likely that a
common physical trait will be found which
grounds all instances of the same belief. The
ways in which all of our PROPOSITIONAL
A TTITUDES interact in the production of
behaviour reinforce the anomalous char-
acter of the mental and render any sort of
reduction of the mental to the physical
impossible (see RATIONALITY; REDUCTION).
Quine's hints concerning how we arrive
at an acceptable set of belief ascriptions are
vague. Clearly, we are to ascribe perceptual
beliefs only when the opinion in question is
appropriately linked to sensory input; and
we are required to ascribe sets of beliefs and
desires which are (broadly) logically coher-
ent. In Pursuit of Truth, Quine (1990) ges-
tures towards a role for 'empathy':
'Empathy is why we ascribe a propositional
attitude by a content clause' (p. 68).
Parents train children in their home lan-
guages by evaluating 'the appropriateness
of the child's observation sentences by
noting the child's orientation and how the
scene would look from there' (p. 42). This
'uncanny knack for empathizing another's
perceptual situation, however ignorant of
the physiological or optical mechanism of
his perception' (p. 42) underlies all transla-
tion and understanding of other people.
Quine's remark that it is comparable to 'our
ability to recognize faces while unable to
sketch or describe them' suggests that the
basis of psychological interpretation is not
reducible to a set of rules or principles.
'Practical psychology' employs the method
of 'empathy': we are to imagine ourselves in
our subject's position as well as we can (p.
46) - we expect our imagined responses to
provide a clue to how they have responded.
One of Quine's most famous theses
emerges here: feeling our way into an-
other's language and into their beliefs and
desires, we rely upon our empathetic under-
standing of their position and try to make
sense of their projects and their cognitive
position. Since the behavioural clues upon
which we rely reflect the complex inter-
actions of innumerable beliefs and desires,
there is no reason to expect that only one
524
set of ascriptions will enable us to under-
stand their behaviour and enter into dia-
logue with them. In talking of translation,
Quine remarks that 'What is utterly factual
is just the fluency of conversation and the
effectiveness of negotiation that one or
another manual of translation serves to
induce' (1990, p. 43). In parallel vein, we
might say of psychological ascription that
what is utterly factual is the fact that a set
of interpretations enables us to empathize
with someone, conversing, negotiating,
cooperating and so on. We have no reason
to suppose that only one set of interpreta-
tions will enable us to do this. Since
propositional attitudes do not identify
neurophysiological states which embody
any kind of physical unity, we should not
be surprised if a variety of sets of interpreta-
tions made psychological sense of this enor-
mous variety of physical events. Translation
and psychological description are both inde-
terminate: there is no fact of the matter
which translation or description is correct.
Fortunately this indeterminacy rarely
intrudes to block understanding in practice.
SCIENCE AND THE MIND
Quine employs an analogy between psycho-
logical terms and terms for diseases: each
identifies an initially unknown bodily state
by reference to its symptoms (1985, p. 6).
But there is a disanalogy: while we may
eventually hope to identify the kind of
bodily disorder associated with a particular
disease, Quine sees no reason to expect such
associations to be possible for propositional
attitudes. We cannot expect to show that
psychological generalizations are explicable
by reference to neurological laws; at best we
can hope for physical explanations of parti-
cular instances of psychological states. Men-
talistic psychology is not to be integrated
with natural science as chemistry is to be
integrated with physics.
As Quine sees it, this has given succour
to dualism: Brentano and his followers have
sought an autonomous science of content,
exploring further the properties of inten-
tional states or propositional attitudes
(1985, p. 7; 1960, p. 221). Agreeing with
Brentano about the irreducibility of mental
discourse to physical, he enjoins us not to
try to 'weave [intensional discourse] into
our scientific theory of the world to make a
more comprehensive system' (1990, p. 71).
The events referred to as beliefs and percep-
tions also receive physical descriptions, and
so described, they receive properly scientific
explanations. This is all the integration of
psychology and physical science that is
required. If, contrary to the indeterminacy
of translation, the evidence was sufficient to
determine one set of propositional attitudes,
the prospects for an autonomous science of
propositional attitudes may seem more
attractive. Since this is not the position,
there is no reason to follow Brentano (see
INTENTION ALITY).
Moreover, it is Quine's opinion that
natural science is extensional: its claims can
be regimented employing the familiar first-
order logic of quantifiers and relations.
Much work on the logic of propositional
attitudes has been prompted by Quine's
demonstration that such discourse is not
extensional. Quine's refusal to incorporate
mentalistic talk in science enables the latter
to enjoy 'the crystalline purity of extension-
ality' (1990, p. 71):
As long as extensional science can pro-
ceed autonomously and self-contained,
with no gaps of causality that inten-
sional intrusions could serve to close,
the sound strategy is the linguistic
dualism of anomalous monism. (p. 72)
So we find at least two reasons convincing
Quine that scientific explanation of mental
phenomena should be physiological and,
ultimately, physical. However valuable it
may be, talk of propositional attitudes has
no place in science.
This is not to suggest, as some have, that
intensional discourse could one day be
abandoned, replaced by a respectably scien-
tific vocabulary to characterize our mental
states: 'The stubborn idioms of propositional
attitude are as deeply rooted as the overtly
physical ones' (1985. p. 7). For example,
language learning is only possible because
our teachers empathize with us, estimating
what we perceive and correcting our utter-
ances accordingly. Rather propositional atti-
tude talk is a practically indispensable habit
(1960. pp. 216-221). irreducible to physi-
cal discourse, guided by empathy, wholly
rational and answerable to evidence; but it
is a way of talking about physical states and
events, grouping them in ways that do not
serve directly the needs of scientific.explana-
tion but which do serve other needs.
BIBLIOGRAPHY
Davidson. D. 1980. Essays on Actions and
Quine. W.V. 1960. Word and Object. Cam-
-- 1969. Ontological Relativity and Other
Essays. New York: Columbia University
Press.
-- 1975. Mind and verbal dispositions. In
Mind and Language, ed. S. Guttenplan.
-- 1976. The Ways of Paradox. Revised and
Enlarged Edition, Cambridge, MA.: Harvard
University Press. First published, 1966.
-- 1981. Theories and Things. Cambridge.
Quine, W.V. 1985. States of mind. Journal of
Philosophy. LXXXII, 5-8.
Quine, W.V. 1990. Pursuit of Truth. Cam-
CHRISTOPHER HOOKWA Y
525
R
radical interpretation Though not meant
as a practical procedure, it can help our
thinking about language and the mind if we
ask what would be involved in interpreting
someone's words and actions. Moreover, if
we imagine ourselves beginning this inter-
pretative process without any prior know-
ledge of what the person means by her
words or what PROPOSITIONAL ATTITUDES
she has, then we are engaged in what is
called 'radical interpretation'. QUINE origin-
ally discussed the idea of radical translation
in respect of another's language and
DA VIDSON, generalizing on this so that
interpretation and not merely translation is
at issue, has made this notion central to his
account of the mind. (See also RATION-
ALITY.)
SAMUEL GUTTENPLAN
rationality To be rational, a set of BELIEFS,
DESIRES, and ACTIONS (also PERCEPTIONS,
INTENTIONS, decisions) must fit together in
various ways. If they do not, in the extreme
case they fail to constitute a mind at all -
no rationality, no agent. This core notion of
rationality in philosophy of mind thus
concerns a cluster of personal identity
conditions, that is, holistic coherence
requirements upon the system of elements
comprising a person's mind (see HOLISM). A
person's putative beliefs must mesh with
the person's desires and decisions, or else
they cannot qualify as the individual's
beliefs; similarly, mutatis mutandis, for
desires, decisions, etc. This is 'agent-con-
stitutive rationality' - that agents possess it
is more than an empirical hypothesis. A
related conception is epistemic or 'norma-
tive rationality': To be rational (that is,
reasonable, well-founded, not subject to
526
epistemic criticism), a belief or decision at
least must cohere with the rest of a person's
cognitive system - for instance, in terms of
logical consistency and application of valid
inference procedures. Rationality con-
straints therefore are key linkages among
the cognitive, as distinct from qualitative,
mental states. The main issue is character-
izing these types of mental coherence. The
discussion below concentrates on agent-
constitutive rationality, proceeding through
a spectrum of accounts, from most to least
stringent.
STANDARD IDEALIZA TroNS
Current philosophical conceptions of ration-
ality inherit central features of models of the
rational agent propounded in micro-
economic, game, and decision theory of the
last half-century (e.g. Von Neumann and
Morgenstern, 1944; Hempel, 1965). These
standard rationality models tie an agent's
beliefs (or knowledge representation) and
desires (or goal structure) to its actions (or
decisions). The underlying idea is approxi-
mated by a necessary condition on the cog-
nitive system of an agent A: 'If A has belief-
set B and desire-set D, then A would decide
upon all and only actions that are (the
most) apparently appropriate.' Roughly, an
action is apparently appropriate for A if,
according to B, it would tend to satisfy D.
To guarantee unfailing choice of the best
of all possible acts, such rationality concep-
tions require an ideal agent of great infer-
ential insight. The agent must possess
deductive, perhaps also inductive, com-
petencies at least along the lines of: 'If A has
belief-set B, then A would make all and only
sound inferences from B that are apparently
appropriate.' Otherwise, A might fail to
identify some unobvious actions most desir-
able according to B. In addition, the belief-
set must perfectly cohere not only over
time, but also at any given time; such
agents must always maintain belief con-
sistency. That is, 'If A has belief-set B then if
any inconsistency arose in B, A would elim-
inate it.'
Some familiar philosophical positions not
explicitly under a rationality rubric are now
discernible. In standard epistemic logic
(Hintikka, 1962), the deductive closure
requirement on the agent's belief-set is even
stronger than the above ideal inference con-
dition: A actually believes (or infers) all
logical consequences of B. The principle of
charity of translation of QUINE entails the
above ideal consistency condition; in our
making sense of another agent's utterances,
the ruling maxim is, 'Pair translation pre-
serves logical laws', that is, the attributor is
obliged to construe the speaker's beliefs so
that the latter does not 'accept contra-
diction', explicit or tacit. (1960, p. 59)
However, one may then begin to worry
whether, from our present viewpoint, we
are implausibly supposed to 'translate'
away, for example, Prege's apparent affir-
mations of his set theory axiomatization in
The Basic Laws of Arithmetic - which is now
conventionally construed to be subject to
Russell's Paradox, i.e. inconsistent. Simi-
larly, the principles of charity of interpreta-
tion of DAVIDSON, broadening the holistic
Quinian paradigm to include, in making
behaviour intelligible, consideration of the
agent's desires as well as beliefs and mean-
ings, sometimes amount to ideal rationality
requirements. As an instance again of an
ideal consistency condition, in keeping with
the traditional decision-theoretic postulate
that the agent always maintains perfect pre-
ference transitivity, Davidson states, 'I do
not think we can clearly say what should
convince us that a man at a given time ...
preferred a to b, b to c, and c to a.' (1980,
p. 237) However, so perfectionistic an inter-
pretation methodology seems prima facie at
odds with a wide range of psychological
studies (e.g. Kahneman, Slovic, and
Tversky, 1982) of apparent preference
RA TION ALITY
transitivity failures in real-world human
decision-making.
More generally, from the perspective of
computational accounts of mind, ideal
rationality requires vast capacities indeed
(see COMPUTATIONAL MODELS). The ideal-
ization is committed to abstracting from a
fundamental fact of the human condition,
our finitude. If one assumes the ideal
agent's deductive ability is represented as
some finite algorithm for always deciding
whether or not any given sentence is a first-
order consequence of its belief-set, then the
creature must violate Church's Undecid-
ability Theorem for first-order predicate cal-
culus. Otherwise, there will have to be some
deductive questions that the creature's
inference algorithm cannot settle in any
finite runtime, yet that the creature could
believe to be high-priority (say, life or death)
matters; hence the creature would be para-
lysed on these and so not satisfy the earlier
ideal inference condition. Similarly, it
should be observed that for this ideal agent,
a significant portion of the deductive sci-
ences would be trivial. Por instance, such
an agent must have (e.g. for when it ranks
the task as highly desirable) a capacity to
determine whether Goldbach's Conjecture
or its negation follows from the axioms of
arithmetic, an important open question -
for human beings - in number theory for
over a century. In effect, the standard
rationality models deny that our inquiries
even have a history, where we prove a
theorem, then in turn use it as a lemma in
subsequent proofs of other theorems.
Indeed, characterizations of the agent's
deductive ability in standard epistemic logic
axiomatizations recall depictions of some of
the omniscience of God under conventional
conceptions of His perfection.
Debate about the role of idealizations thus
takes a central position in recent theorizing
about rationality. Certainly idealizations
have indispensable functions in mature sci-
entific theory in general (cf. the familiar
ideal gas law example, representation of the
complex structure of molecules as perfectly
elastic, dimensionless spheres). Witness also
the success stories of the agent-idealizations
527
RA TION ALITY
in economic, game, and decision theory;
such perfection may be profoundly impos-
sible, but it is simple, that is. manageable
for the purposes of useful formalization. Part
of the key to effective use of an idealization
consists in recognition - often only at the
level of unarticulated lore or practice - of
the degree of approximation and the limits
of applicability of the idealization to reality.
A sense that we tend uncritically to reify
the rationality idealizations, combined with
an awareness of how deeply they cannot in
fact apply to human beings, may in turn
motivate anti-realist or instrumentalist
accounts of all cognitive theory. The cogni-
tive system attributed to an individual then
attenuates to little more than a convenient
but impossible fiction to aid the attributor in
predicting the agent's behaviour (such a
tendency seems evident in some earlier
accounts of DENNETT (1978)).
The above conception of rationality con-
ditions that are sine qua non for agenthood
in the first place ought to be distinguished
from another type of even more stringent
norm of rationality in terms of which an
agent's reasoning is evaluated. A belief or
inference of mine may be criticized as irra-
tional while I still manage, by the rest of my
epistemic track record. to continue to
qualify as an agent or person. Failure to dis-
tinguish such normative rationality from
the above weaker agent-constitutive ration-
ality will tend to reinforce the high levels
demanded for the latter. In general, a sense
of the limited applicability of the ideal
rationality models has tended to fuel doubts
about the very possibility of a cognitive
science.
PSYCHOLOGICALLY REALISTIC ACCOUNTS
After recognizing that nothing could count
as an agent for which there are no ration-
ality constraints, one can stop to wonder
whether one has to jump to a conclusion
that the agent must therefore be ideally
rational. Is rationality all or nothing. or is
there some cognitive via media between
perfect, Cartesian unity of mind and utter,
anarchic disintegration of personhood?
528
Such an approach attempts to deal with the
aperru that human beings - indeed, all
physically realizable intelligences - must
have limited cognitive resources. A cogniti-
vist, post-behaviourist paradigm permits
modellers to abandon empty-organism
approaches and attempt to represent the
agent's internal psychological reality.
This latter type of option traces back at
least to the research programme of Simon
(1982) on normative and empirically
descriptive models of 'bounded rationality'.
Simon's account begins as a critique of tra-
ditional idealizations. Actual human beings
generally have neither perfect information
about their world of alternative choices nor
a capacity to use such knowledge. There-
fore, in place of the conventional postulate,
'A always acts so as to maximize its ex-
pected utility' (Le. A does, or ought to,
choose the apparently most appropriate
actions), Simon proposed a principle to the
effect that 'A "satisfices" its expected utility'.
Rational agents do, and ought only to. try
to 'satisfice' rather than maximize, in that
they should not be required to do the
impossible, to identify decisions that per-
fectly optimize their expected utility; they
need only choose options that in this respect
are 'good enough'.
An approach for real-world agent-
constitutive rationality similar to Simon's
for normative and empirically hypothesized
rationality can be found in Cherniak
(1986). Weakening the ideal rationality
requirement, while still recognizing that
nothing could qualify as a person without
'some' internal coherence, yields psycho-
logically more realistic rationality conditions
of the type, 'If A has belief-desire set BUD,
then A would decide upon some (but not
necessarily all) of the apparently appropriate
actions for that set.' In turn, such moderate
rationality only requires the agent to be a
fair logician in order to choose enough of
the right actions: 'If A has belief set B. then
A would make some (not necessarily all)
sound inferences from B that are apparently
appropriate.' - Namely, A must accomplish
enough of the inferences that are useful in
identifying apparently appropriate actions.
Unlike ideal inference capacity, such a mod-
erate capacity itself requires a non-trivial
ability to select apparently appropriate infer-
ential tasks to undertake so that the limited
cognitive resources are not wasted. For this
type of minimal agent, computation no
longer need be virtually costless. These
'some, but not all' conditions are conceived
as embedded in ancillary cognitive theory
that helps to fill in which inferences are most
likely to be made, including modules con-
cerning the reasoning psychology and the
memory structure of the agent.
Such a resource-realistic rationality
framework possesses systematicity in that it
links together recent independent research
programmes in COGNITIVE PSYCHOLOGY
and in computer science. The psychological
studies focused on the strikingly ubiquitous
human use of reasoning procedures that are
formally incorrect and/or incomplete 'heur-
istics'. An example studied by Kahneman
and Tversky (Kahneman, Slovic, and
Tversky, 1982) is the 'availability heuristic'
- the strategy, widespread in intuitive
assessment of probability of a type of event
- of judging how easy to recall or imagine
such an event is, instead of its actual object-
ive frequency. (Hence, e.g. rare but vividly
recollectable disaster scenarios are con-
sistently overestimated as risks.) Another
field that developed contemporaneously
with, but separately from, non-ideal ration-
ality accounts and experimental studies of
reasoning heuristics is computational com-
plexity theory in computer science (Cher-
niak, 1986, ch. 4); the basic insight is that
formally correct and complete inference
procedures (e.g. even for propositional cal-
culus) appear to be intrinsically intractable,
that is, with quite small problem instances
requiring literally cosmic-scale resources of
time and memory. From a minimal ration-
ality perspective, use of the heuristics need
not be viewed as mere uninteresting excep-
tions to a rule of perfect rationality, or even
as irrationality. For, if formally adequate
procedures mean practical paralysis, then
the heuristics may represent a trade-off of
perfection - formal correctness/complete-
ness - for speed and usability.
RATION ALITY
This approach in turn suggests an opti-
mistic reinterpretation of the semantic and
set-theoretic paradoxes, the response to
which has so shaped the rise of modern
logic. Instead of, for instance, the Liar or
Russell's Paradoxes being par excellence
symptoms of disease at the core of our con-
ceptual scheme, they may be manifestations
of our use of inconsistent procedures as
another type of 'quick but dirty heuristic' in
a rational speed-reliability trade-off. These
heuristics may work well in the ordinary
contexts where they developed, and break
down in the new situations into which
recent metamathematical enquiries have led
us.
ALTERNATIVE VIEWS
Ideal and more realistic rationality models
may well be able to coexist peacefully, where
simplicity of an approximation can be pro-
gressively exchanged for real-world applic-
ability as necessary. Another ideal-agent
type of response to account for the phenom-
enon of apparently widespread human use
of incorrect reasoning procedures employs a
'competence/performance' distinction along
lines promoted by CHOMSKY. In lingUistic
theory, a competence model of, say, a speak-
er's underlying knowledge of syntax is
represented as a set of formally adequate
rules the speaker accepts and uses; however,
the speaker's performance - actual linguistic
behaviour - often does not conform simply
to the postulated syntactic competence. FQr
example, limitations on the speaker's lan-
guage processing memory, time pressures,
other demands on attention, and even
fatigue will disrupt the speaker's observed
judgments of grammaticality. Cohen (1981)
similarly proposed to explain away people's
reasoning fallacies in terms of a perfect
underlying inferential competence that is to
be represented as a formally correct rule set
they accept. When a person's ideal com-
petence is manifested in actual inferential
performance, it is degraded just like linguis-
tic performance by the inevitable limitations
of memory, time, and so on, yielding the
observed errors.
529
RATIONALITY
The underlying inferential competence is
thus still very like the inference capacity of
the earlier standard ideally rational agent.
One concern about this competence
account focuses on how satisfying is an
explanation of the widespread heuristics as
so to speak mere 'noise' - is it too ad hoc,
does it acquire too many epicycles? The
concern can parallel questions about ex-
tremely 'abstract' competence accounts for
syntax, as well as for semantics (e.g. as in
FODOR). Formally adequate rules - for rea-
soning, syntax, or meaning-representation
- are supposed to be possessed and used by
agents to the degree that they approximate
'ideal speaker-hearers'. However, even for
moderate nativists, as eliciting conditions
for the tacit knowledge grow more elab-
orate, the genuineness of this acceptance
seems to fade (a problem already evident
for Plato's claims in the Meno that the
unlettered slave really 'knew' the Pythagor-
ean Theorem even before Socrates' brilliant
coaching). The psychological reality of
actual human beings' internal representa-
tions of the inference rules thereby seems to
dissipate; what does it mean to say that a
person really now is using so abstract a
rule, as opposed to, for instance, one of the
messy heuristics? Once again, it becomes
hard to tell whether the subject's accep-
tance of the ideal rules is a convenient
fiction of the theoretician or cognitive
reality of the subject.
There is a connection here in turn to
accounts of FOLK PSYCHOLOGY, the prac-
tical, pre-philosophical theory that human
beings possess (perhaps in part innately)
concerning their own and others' minds,
and use in everyday dealings. In particular,
a doctrinal symbiosis may operate between
a sense of the unreality of ideal competence
models of our psychology, and eliminativist
impulses against folk psychology (see ELIM-
INATIVISM). That is, to the extent that the
competence idealizations become of proble-
matic psychological reality, so also can the
entire nexus of intentional psychology
become a candidate for future extirpation as
unscientific lore. The concept of rationality
will of cuurse be embedded centrally within
530
folk psychology, and so a prime eliminativist
target.
Against this background appears the
most minimalist type of rationality account,
the pluralistic theory of Stich (1990). On
the spectrum of rationality constraints
ranging from maximal for the idealizations,
through moderate for so-called 'minimal'
theories, Stich's relativism most closely
approaches the true minimalist pole. The
relevant thesis here is that we can place no
a priori conditions on rationality for all pos-
sible agents. For, along pragmatist lines,
Stich views reasoning procedures <IS tools,
the effectiveness of which varies for differ-
ent purposes and circumstances, hence
there is no single inferential canon all
people ought to obey. Correspondingly, the
rationality constitutive of agenthood
becomes wide open; for example, possibil-
ities for radical cognitive diversity extend
without limit.
One 'No rationality, no agent' response
begins by insisting that rationality is a
metaphysical matter of part of the essential
nature of a particular type of object - a
mind, person, or agent. Hence, with the
identity of rationality itself also fixed (e.g.
including some logical consistency), some
rationality constraints are just true of any-
thing that can count as an agent. Latitude
for diversity still remains; interpretation
methodology can now be reconstructed in
terms of a principle of 'moderate charity',
where the agent's behaviour need only be
so construed that the attributed cognitive
system comes out moderately, instead of
perfectly, rational. However, one must
immediately, and uneasily, add a 'Neur-
ath's Boat' qualification; our conceptual
scheme, like a boat at sea which even as
we reconstruct we must still keep afloat,
cannot just be set aside wholesale. That is,
the required moderate rationality has to be
evaluated in terms of the only standards
available to us, our own current ones. And
so one may yet retain a feeling that any
constitutive rationality constraints,
however unavoidable, just amount to a
Procrustean Bed into which genuine cogni-
tive differences are unwarrantedly forced.
Perhaps such dialectical tug of war is a
sign that the nature of rationality remains
one of those pieces of unfinished business
so distinctive of the philosophical enter-
prise.
See also CONCEPTUAL ROLE SEMANTICS;
INN A TENESS; INTENTIONALITY ; LANGUAGE
OF THOUGHT; MEMORY; REPRESENTATION.
BIBLIOGRAPHY
Cherniak. C. 1986. Minimal Rationality. Cam-
Cohen. L. 1981. Can human irrationality be
experimentally demonstrated? Behavioral and
Brain Sciences. 4.317-31.
Davidson. D. 1980. Essays on Action and
Dennett. D. 1978. Brainstorms. Cambridge.
MA.: MIT Press.
Hempel. C. 1965. Aspects of scientific ex-
planation. In Aspects of Scientific Explanation.
1O. New York: Free Press.
Hintikka. J. 1962. Knowledge and Belief. Ithaca,
N.Y.: Cornell University Press.
Kahneman, D .. Slovic, P .. and Tversky. A ..
eds. 1982. Judgment Under Uncertainty:
Heuristics and Biases. Cambridge University
Press.
Quine, W. 1960. Word and Object. Cambridge,
MA.: MIT Press.
Simon, H. 1982. Models of Bounded Rationality.
2 vols. Vol. 2. Cambridge. MA.: MIT Press.
Stich. S. 1990. The Fragmentation of Reason:
Preface to a Pragmatic Theory of Cognitive
Evaluation. Cambridge. MA.: MIT Press.
Von Neumann. J .. and Morgenstern. O. 1944.
Theory of Games and Economic Behavior. New
York: Wiley.
CHRISTOPHER CHERNIAK
reasons and causes The psychological
level of description carries with it a mode of
explanation which 'has no echo in physical
theory' (Davidson. 1980, pp. 207-25);
explanation in terms of reasons. We regard
ourselves and each other as 'rational pur-
posive creatures, fitting our BELIEFS to the
world as we perceive it and seeking to
obtain what we desire in the light of them'.
(Hopkins. 1982, pp. vii-xix). Reason-giving
REASONS AND CAUSES
explanations can be offered not only for
ACTIONS and beliefs. which will gain most
attention in this entry; but also for
DESIRES, INTENTIONS, hopes, fears, angers
and affections, etc. Indeed. their positioning
within a network of rationalizing links is
part of the individuating characteristics of
this range of psychological states and the
intentional acts they explain (see INTEN-
TIONALITy).
THE REASON-GIVING RELATION
At the heart of the reason-giving relation is
a normative claim. An agent has a reason
for believing, acting, etc., if, given her other
psychological states, this belief/action is jus-
tified or appropriate. Displaying someone's
reasons consists in making clear this justifi-
catory link. Paradigmatically, the psycho-
logical states that provide an agent with
reasons are intentional states individuated
in terms of their propositional content (see
CONTENT; PROPOSITIONAL ATTITUDES).
There is a long tradition that emphasizes
that the reason-giving relation is a logical
or conceptual one. One way of bringing
out the nature of this conceptual link is by
the construction of reasoning, linking the
agent's reason-providing states with the
states for which they provide reasons. This
reasoning is easiest to reconstruct in the
case of reasons for belief where the contents
of the reason-providing beliefs inductively or
deductively support the content of the
rationalized belief. For example, I believe my
colleague is in her room now, and my
reasons are (1) she usually has a meeting in
her room at 9.30 on Mondays and (2) it is
9.30 on Monday. To believe a proposition is
to accept it as true: and it is relative to the
objective of reaching truth that the ration-
alizing relations between contents are set
for belief. They must be such that the truth
of the premises makes likely the truth of the
conclusion.
In the case of reasons for actions the pre-
mises of any reasoning are provided by
intentional states other than beliefs. Classic-
ally. an agent has a reason to perform a
certain kind of action when she has (a) a
531
REASONS AND CAUSES
pro-attitude towards some end or objective,
and (b) a belief that an action of that kind
will promote this end. The term pro-attitude
derives from Donald DAVIDSON. It includes
'desires, wantings, urges, promptings and a
great variety of moral views, aesthetic prin-
ciples .. .' (Davidson, 1980, pp. 3-19). It is
common to use 'desire' as a generic term for
such pro-attitudes. It is relative to the con-
stitutive objectives of desire that the ration-
alizing links are established in the practical
case. We might say that the objective of
desires is their own satisfaction. In the case
of reasons for acting therefore we are
looking for a relationship between the con-
tents of the agent's intentional states and
the description of the action which show
that performing an action of that kind has
some chance of promoting the desired goals.
I find it desirable to own a Burmese kitten. I
believe that putting an advertisement in the
paper will help bring it about that I own a
Burmese kitten; so putting an advertisement
in the paper is a thing to be done. (There
are a number of suggestions for formalizing
practical reasoning; see Von Wright. 1971;
Davidson, 1980; Kim, 1984; Lennon,
1990.)
The presence of a reason for believing or
acting doesn't necessarily make it rational
for an agent to believe or act in that way.
From the agent's point of view overall she
may have other beliefs which provide con-
flicting evidence, or conflicting desires. To
establish what is rational to believe or do
overall we would need to take into account
principles for weighing competing beliefs
and desires. Of course, we do not always
believe what is rational, or act in the light
of what we judge best (cases of SELF-
DECEPTION and WEAKNESS OF WILL show
this). However, a minimum of RATION-
ALITY must be present in the pattern of a
person's beliefs, desires, intentions and
actions before they can be regarded as an
agent with intentional states at all. (David-
son, 1980,pp. 207-25)
When the reason-glVlng relation is
articulated as a piece of reasoning, or calcu-
lation, it is not being suggested that when-
ever we have a reason we go through a
532
process of conscious or unconscious delib-
eration. Rather the reasoning is a device to
spell out the kind of conceptual links
between intentional contents which fre-
quently constitute the reason-giving rela-
tion. One of the things such a mode of
articulation serves to make clear is the
dependency of the reason-giving links on
the intentional descriptions of our psycholo-
gical states and our actions. This makes any
project of reducing such states to extension-
ally characterizable phenomena highly pro-
blematic. Such a reduction, it is argued,
cannot capture the rationalizing relations
constitutive of the intentional realm
(Davidson, 1980, pp. 207-25; Lennon,
1990, pp. 85-104).
Calculative presentation, however, does
not seem essential to all reason-giving
links. If providing reasons consists in dis-
playing the appropriateness of an agent's
response, given her other perceptions and
attitudes, then this may not always be cap-
turable in propositional form (e.g. what
reasons do you have for believing she took
the money? I saw her). This may be parti-
cularly the case when we are considering
reasons for emotions (for an illuminating
discussion of this point, see Taylor, 1985,
pp. 1-16). If we accept that reasons are
not always presentable in the form of
calculations, this also opens the door for
psychological states with phenomenal or
qualitative content to play a rationalizing
role (Gilbert, 1992).
For some writers the justificatory and
normative character of reason-giving rela-
tions renders such relations perspectival or
subjective in character (see SUBJECTIVITY).
Appreciating the rationality of a belief or an
action requires appreciating it from the per-
spective of an agent evaluating beliefs and
deciding how to act (McDowell, 1985). The
subjective or perspectival quality which is
here being claimed of reason-giving links
and consequently for the intentional kinds
which find their anchorage within them,
parallels that are commonly claimed to
attach to psychological states with phenom-
enal content (see PERCEPTUAL CONTENT;
QUALIA).
EXPLAINING WITH REASONS
What is the explanatory role of providing
reasons for our psychological states and
intentional acts? Clearly part of this role
comes from the justificatory nature of the
reason-giving relation: 'things are made
intelligible by being revealed to be, or to
approximate to being, as they rationally
ought to be' (McDowell, 1985). For some
writers the justificatory and explanatory
tasks of reason-giving simply coincide. The
manifestation of rationality is seen as suffi-
cient to explain states or acts quite indepen-
dently of questions regarding causal origin.
Within this model the greater the degree of
rationality we can detect, the more intelli-
gible the sequence will be. Where there is a
breakdown in rationality, as in cases of
weakness of will or self-deception, there is a
corresponding breakdown in our ability to
make the action/belief intelligible.
The equation of the justificatory and
explanatory role of rationalizing links can
be found within two quite distinct pictures.
One account views the attribute of ration-
ality from a third-person perspective. Attri-
buting intentional states to others, and by
analogy to ourselves, is a matter of applying
to them a certain pattern of interpretation.
We ascribe whatever states enable us to
make sense of their behaviour as conform-
ing to a rational pattern. Such a mode of
interpretation is commonly an ex post facto
affair, although such a mode of interpreta-
tion can also aid prediction. Our interpreta-
tions are never definitive or closed. They are
always open to revision and modification in
the light of future behaviour, if such revi-
sions enable the person as a whole to
appear more rational. Where we fail to
detect a rational pattern then we give up
the project of seeing a system as rational
and instead seek explanations of a mechan-
istic kind (DENNETT, 1979).
The other picture is resolutely first-
personal, linked to the claimed perspectivity
of rationalizing claims. When we provide
rationalizing explanations we make an
action, for example, intelligible by adopting
the agent's perspective on it. Understanding
REASONS AND CAUSES
is a reconstruction of actual or possible
decision making. It is from such a first-
person perspective that goals are detected as
desirable and the courses of action appro-
priate to the situation. The standpoint of an
agent deciding how to act is not that of an
observer predicting the next move. When I
find something desirable and judge an act
an appropriate route for achieving it, I con-
clude that a certain course of action should
be taken. This is different from my reflecting
on my past behaviour and concluding that I
will do X in the future (see Kim, 1984;
McDowell, 1985).
For many writers, however, the justifica-
tory and explanatory role of reasons cannot
simply be equated. To do so fails to distin-
guish cases where I have reasons from cases
where I believe or act because of these
reasons. I may have beliefs from which your
innocence would be deduced but none-
theless come to believe you are innocent
because you have blue eyes. I may have
intentional states that give me altruistic
reasons for giving to charity but none-
theless contribute out of a desire to earn
someone's good opinion. In both these
cases, although my belief could be shown to
be rational in the light of other beliefs, and
my actions in the light of my altruistic
states, neither of these rationalizing links
would form part of a valid explanation of
the phenomena concerned. Moreover, cases
of weakness of will show that I can have
sufficient reason for acting and yet fail to
act, e.g. I continue to smoke although I
judge it would be better to abstain. This
suggests that the mere availability of rea-
soning, however good, in favour of an
action cannot, in itself, be sufficient to
explain why it occurred.
If we resist the equation of the justifica-
tory and explanatory work of reason-giving,
we must look for a connection between
reasons and action/belief in cases where
these reasons genuinely explain, which is
absent in cases of mere rationalizations (a
connection that is present when I act on my
best judgment and not when I fail). Classi-
cally, of course, the connection that has
been suggested here is that of causality. In
533
REASONS AND CAUSES
cases of genuine explanation, the reason-
providing intentional states cause the belief!
actions for which they also provide reasons.
This position also seems to find support
from considering the conditionals and coun-
terfactuals that our reason-providing expla-
nations sustain, which parallel those in
cases of other causal explanations. Imagine
I am approaching the Arts Building looking
for the cafeteria. If I believe the cafe is to the
left, I turn to the left; if I believe it is to the
right, I will turn to the right. If my
approach to the building is explained simply
by my desire to find the cafe, then in the
absence of such a desire I would not have
walked in that direction. In general terms,
where my reasons explain my action, then
the presence of such reasons was, in those
circumstances, necessary for the action and
at least made probable its occurrence. These
conditional links can be explained if we
accept that the reason-giving link is also a
causal one. Any alternative account would
therefore also need to accommodate them
(Lennon, 1990, pp. 42-55).
The most famous defence of the view that
reasons are causes is found in Davidson's
article 'Actions, Reasons and Causes'
(Davidson, 1980, pp. 3-20). However,
within the Davidsonian picture reason-
giving explanations perform two quite dis-
connected tasks; namely showing an action!
belief to be rational and pointing to its
cause. For Davidson the fact that events are
of a kind to make an action rational is irre-
levant to their causal role in the production
of that action. This position is a result of his
claim that there can be no empirical causal
laws employing intentional vocabulary
(Davidson, 1980, pp. 207-24). The causal
generalizations required to support the sin-
gular causal links between reasons and
actions must therefore rest on the non-
intentional characteristics of these events.
Intentional states therefore cause actions in
virtue of their neurophysiological character-
istics.
For most causal theorists, however, the
radical separation of the causal and ration-
alizing role of reason-giving explanations is
unsatisfactory. For such theorists, where we
534
can legitimately point to an agent's reasons
to explain a certain belief or action, then
those features of the agent's intentional
states that render the belief or action
reasonable must be causally relevant in
explaining how the agent came to believe or
act in a way which they rationalize. One
way of putting this requirement is that
reason-giving states not only cause but also
causally explain their explananda.
On most accounts of causation an accept-
ance of the causal explanatory role of
reason-giving connections requires empiri-
cal causal laws employing intentional
vocabulary. It is arguments against the pos-
sibility of such laws that have, however,
been fundamental for those opposing a
causal explanatory view of reasons. What is
centrally at issue in these debates is the
status of the generalizations linking inten-
tional states to each other, and to ensuing
intentional acts. An example of such a
generalization would be 'If a person desires
X, believes A would be a way of promoting
X, is able to A and has no conflicting desires
then (ceteris paribus) she will do A'. For
many theorists such generalizations are
grounded in the constitutive relationships
between desire, belief and action. Grasping
the truth of such a generalization is
required to grasp the nature of the inten-
tional states concerned. For some theorists
the a priori elements within such general-
izations are sufficient to exclude them from
consideration as empirical laws. That,
however, seems too quick, for it would simi-
larly rule out any generalizations in the
physical sciences that contain a priori ele-
ments, as a consequence of the implicit defi-
nition of their theoretical kinds in a causal
explanatory theory. Causal theorists, includ-
ing functionalists in philosophy of mind (see
FUNCTIONALISM), can claim that it is just
such implicit definition that accounts for the
a priori status of our intentional general-
izations.
In the hands of Davidson, the a prlOfi
status of intentional generalizations is used
to a somewhat different end. Such general-
izations provide the constitutive principles of
rationality which govern our attribution of
intentional states to ourselves and others.
He argues that such attributions are open to
constant revision and retain a residual inde-
terminacy which render our intentional
notions quite unsuitable for inclusion in
strict causal laws (Davidson, 1980, pp.
207-225). The Davidsonian argument,
however, requires us to anchor our inten-
tional classifications in a necessarily inde-
terminate process of interpretation. For the
causal theorist, in contrast, they are akin to
natural kind terms, generating causal
explanatory generalizations, and subject to
no more than the epistemic indeterminancy
of other such terms (Lennon, 1990, pp. 99-
104).
The causal explanatory approach to
reason-giving explanations also requires an
account of the intentional content of our
psychological states, which makes it pos-
sible for such content to be doing such work
(Dretske 1989). It also provides a motiva-
tion for the reduction of intentional char-
acteristics to extensional ones, in an
attempt to fit such intentional causality into
a fundamentally materialist world picture.
The very nature of the reason-giving rela-
tion, however, can be seen to render such
reductive projects unrealizable (Davidson,
1980; Lennon, 1990). This, therefore,
leaves causal theorists with the task of
linking intentional and non-intentional
levels of description in such a way as to
accommodate intentional causality, without
either overdetermination or a miraculous
coincidence of prediction from within dis-
tinct causally explanatory frameworks.
CONTENT; DRETSKE; EPIPHENOMENALISM;
FODOR; IDENTITY THEORIES; PHYSICALISM.
BIBLIOGRAPHY
Events, Oxford University Press.
Dennett, D. 1979. Intentional systems. In
Brainstorms, Brighton; Harvester.
Dretske, F. 1991. Reasons and causes. In
Philosophical Perspectives, 3: Philosophy of
Mind and Action Theory, ed. J. Tomberlin.
California, Ridgeview.
REDUCTION
Gilbert, P. 1992. Immediate experience, Pro-
ceedings of the Aristotelian Society, XCII,
233-250.
Hopkins, J. 1982. Introduction. In Philosophi-
cal Essays on Freud, ed. J. Hopkins and R.
Wollheim. Cambridge University Press.
Kim, J. 1984. Self understanding and ration-
alising explanations. Philosophia Naturalis,
21.
Lennon, K. 1990. Explaining Human Action.
London: Duckworth.
McDowell, J. 1985. Functionalism and anom-
alous monism. In Action and Events: Perspec-
tives on the Philosophy of Donald Davidson, ed.
E. Lepore and B. McLaughlin. Oxford: Basil
Blackwell.
Taylor, G. 1985. Emotions and beliefs. In
Pride, Shame and Guilt. Oxford: Clarendon
Press.
Von Wright, G. 1971. Explanation and Under-
standing. Ithaca: Cornell University Press.
KATHLEEN LENNON
reduction This notion is best approached
in its most natural setting: the philosophical
understanding of science. The standard,
text-book example here is that of the rela-
tion between thermodynamics and mechan-
ics, Up until the nineteenth century, these
two theories played an important part in
our understanding of the physical world. On
the one hand, thermodynamics described
the behaviour of what was thought of as a
substance, heat; its principles accounted for
such facts as that heat moved from hotter to
colder bodies, and that the amount of heat
so transferred depended on the material
constitution of the relevant bodies. On the
other hand, the principles of mechanics -
largely due to the work of Newton -
accounted for the ways in which bodies,
either in motion or at rest, affected one
another.
Due to theoretical and experimental work
in the nineteenth century, it became pos-
sible to unify these two theories, and it is in
this unification that the notion of reduction
comes into play. Once we recognize that all
matter is made up of smaller 'particles'
(atoms or molecules) and that the motion of
these particles is governed by the principles
535
REPRESENT A TION
of mechanics, it becomes possible to identify
heat as the degree to which these particles
move or vibrate. But once this is done, then
the explanations offered in thermodynamics
can be subsumed under the broader expla-
nations of mechanics. For example, transfer
of heat is explained by the laws governing
the transmission of motion among atoms
and molecules. In this case, one can say
that thermodynamics has been reduced to
mechanics. However, in contrast to thermo-
dynamics, the account of combustion which
made central use of the notion of phlogiston
has proven to be irreducible to the laws of
chemistry and physics, and, being thus irre-
ducible, phlogiston is now treated as chi-
merical - as not being the name of any
genuine existent. So, reduction is often a
way of preserving the ontological security of
some item.
Outside of the scientific context, the
notion of reduction is used, but it is not
always obvious that the conditions for its
proper application are fulfilled. In philo-
sophy of mind, the type IDENTITY THEORY
is often thought of as reductionist. The
identification of various mental items with
types of physical phenomena seems to be
analogous to the above-mentioned identifi-
cation of heat with molecular motion, and
this then promises the reduction of our
understanding of the mind to our under-
standing of the brain and related physical
systems. Moreover, such reduction promises
to secure the ontological bona fides of the
mental, since, as was seen above, a reduced
realm is guaranteed a certain reality.
However, it is not clear that one can think
of our understanding of the mind as theo-
retical, as genuinely analogous to some-
thing like thermodynamics. And if our
everyday, 'folk psychological' conception of
the mind is not like an explanatory theory,
then the idea of reduction might well just be
inappropriate. Indeed, there are a number of
writers who insist, for various reasons, that
the mind is not reducible to anything phys-
ical, even though mental phenomena have
a material substrate. The notion that they
appeal to here is supervenience: the mental
is said to supervene on the physical without
536
being reducible to it. However, the relations
between supervenience and reduction are
controversial. (See BELIEF; DAVIDSON;
FODOR; FOLK PSYCHOLOGY; LEWIS; SUPER-
VENIENCE .)
SAMUEL GUTTENPLAN
representation Although linguistic and
pictorial representations are undoubtedly
the most prominent symbolic forms we
employ, the range of representational
systems humans understand and regularly
use is surprisingly large. Sculpture, maps,
diagrams, graphs, gestures, music notation,
traffic signs, gauges, scale models, and
tailors' swatches are but a few of the rep-
resentational systems that play a role in
communication, thought, and the guidance
of behaviour. Indeed, the importance and
prevalence of our symbolic activities has
been taken as a hallmark of being human.
What is it that distinguishes items that
serve as representations from other objects
or events? And what distinguishes the
various kinds of symbols from each other?
As for the first question, there has been
general agreement that the basic notion of a
representation involves one thing's 'standing
for', 'being about', 'referring to or denoting'
something else. The major debates here
have been over the nature of this connec-
tion between a representation and that
which it represents. (See below.) As for the
second question, perhaps the most famous
and extensive attempt to organize and dif-
ferentiate among alternative forms of rep-
resentation is found in the works of C. S.
Peirce (1931-1935). Peirce's theory of
signs is complex, involving a number of
concepts and distinctions that are no longer
paid much heed. The aspect of his theory
that remains influential and is widely cited
is his division of signs into Icons, Indices
and Symbols. Icons are signs that are said
to be like or resemble the things they repre-
sent (e.g. portrait paintings). Indices are
signs that are connected to their objects by
some causal dependency (e.g. smoke as a
sign of fire). Symbols are those signs that
are related to their object by virtue of use or
association; they are arbitrary labels (e.g.
the word 'table'). This tripartite division
among signs, or variants of this division, is
routinely put forth to explain differences in
the way representational systems are
thought to establish their links to the world.
Further, placing a representation in one of
the three divisions has been used to account
for the supposed differences between con-
ventional and non-conventional representa-
tions, between representations that do and
do not require learning to understand, and
between representations, like language, that
need to be read, and those which do not
require interpretation. Some theorists,
moreover, have maintained that it is only
the use of Symbols that exhibits or indicates
the presence of mind and mental states.
Over the years this tripartite division of
signs, although often challenged, has
retained its influence. More recently, an
alternative approach to representational
systems (or as he calls them 'symbolic
systems') has been put forth by N. Goodman
(1976). Goodman has proposed a set of syn-
tactic and semantic features for categorizing
representational systems. His theory pro-
vides for a finer discrimination among types
of systems than Peirce's, and the categoriza-
tions he elaborates cut across many of the
boundaries of the Peircian format. What
also emerges clearly is that many rich and
useful systems of representation lack a
number of features taken to be essential to
linguistic or sentential forms of representa-
tion (e.g. discrete alphabets and vocabul-
aries, syntax, logical structure, inference
rules, compositional semantics, and recur-
sive compounding devices). As a con-
sequence, although these representations
can be appraised for accuracy or correct-
ness, it does not seem possible to analyse
such evaluative notions along the lines of
standard truth theories, geared as they are
to the structures found in sentential
systems.
In light of this newer work, serious ques-
tions have been raised about the soundness
of the tripartite division and about whether
various of the psychological and philosophi-
cal claims concerning conventionality,
REPRESENTATION
learning, interpretation, etc., that have been
based on this traditional analysis, can be
sustained. It is of special significance that
Goodman has joined a number of theorists
in rejecting accounts of Iconic representa-
tion in terms of resemblance (similarity or
1-1 correspondence). (See Gombrich,
1960.) The rejection has been twofold. First,
as Peirce himself recognized, resemblance is
not sufficient to establish the appropriate
referential relations. The numerous prints of
a lithograph do not represent one another,
any more than an identical twin represents
his or her sibling. Something more than
resemblance is needed to establish the con-
nection between an Icon or picture and
what it represents. Second, since Iconic
representations lack as many properties as
they share with their referents, and certain
non-Iconic symbols can be put in 1-1 cor-
respondence with their referents, it is diffi-
cult to provide a non-circular account of
what the similarity is that distinguishes
Icons from other forms of representation.
What's more, even if these two difficulties
could be resolved, it would not show that
the representational function of pictures can
be understood independently of an asso-
ciated system of interpretation. The design,
!L;J, may be a picture of a mountain.
Alternatively, it could be a graph of the
economy, a circuit diagram, or a letter or
word in a foreign language. Or it may have
no representational significance at all.
Whether it is a representation and what
kind of representation it is, is relative to a
system of interpretation.
REPRESENT A TION AND INTENTION ALITY
Representations, along with mental states,
especially beliefs and thoughts, are said to
exhibit INTENTIONALITY in that they refer
to or stand for something else. The nature
of this special property, however, has
seemed puzzling. Not only is intentionality
often assumed to be limited to humans, and
possibly a few other species, but the prop-
erty itself appears to resist characterization
in physicalist terms. The problem is most
obvious in the case of 'arbitrary' signs, like
537
REPRESENT A TION
words, where it is clear that there is no
connection between the physical properties
of a word and what it denotes. (As noted
above, the problem also remains for Iconic
representations. )
Early attempts tried to establish the link
between sign and object via the mental
states of the sign user. A symbol # stands
for * for S, if it triggers a * -idea in S. On one
account, the reference of # is the * -idea
itself. On the other major account, the
denotation of # is whatever the * -idea
denotes. The first account is problematic in
that it fails to explain the link between
symbols and the world. The second is prob-
lematic in that it just shifts the puzzle
inward. For example, if the word 'table'
triggers the image 'n' or 'TABLE' what
gives this mental picture or word any refer-
ence at all, let alone the denotation nor-
mally associated with the word 'table'?
An alternative to these mentalistic the-
ories has been to adopt a behaviouristic
analysis (see BEHAVIOURISM). On this
account, # denotes * for S is explained along
the lines of either: (i) S is disposed to behave
to # as to *; or (ii) S is disposed to behave in
ways appropriate to * when presented #.
Both versions prove faulty in that the very
notions of the behaviour associated with or
appropriate to * are obscure. In addition,
once one gets beyond a few trivial cases
there seems to be no reasonable correlations
between behaviour towards signs and beha-
vior towards their objects that is capable of
accounting for the referential relations.
A currently influential attempt to 'natur-
alize' the representation relation takes its
cue from Indices. The crucial link between
sign and object is established by some
causal connection between * and #. (See
Dretske, 1986.) It is allowed, however, that
such a causal relation is not sufficient for
full-blown intentional representation. An
increase in temperature causes the mercury
to rise in the thermometer, but the mercury
level is not a representation for the thermo-
meter. In order for # to represent * to S, #, it
is said, must play an appropriate role in the
functional economy of S's activities. The
notion of 'function', in turn, is to be spelled
538
out along biological or other lines so as to
remain within 'naturalistic' constraints (see
NATURALISM). This approach runs into pro-
blems in specifying a suitable notion of
'function' and in accounting for the pos-
sibility of misrepresentation. Also, it is not
obvious how to extend the analysis to
encompass the seman tical force of more
abstract or theoretical symbols. These diffi-
culties are further compounded when one
takes into account the social factors that
seem to play a role in determining the
denotative properties of our symbols.
The problems faced in providing a reduc-
tive naturalistic analysis of representation
has led many to doubt that this task is
achievable or necessary. Although a story
can be told about how some words or signs
are learned via association or other causal
connections with their referents, there is no
reason to believe that the 'stand-for' rela-
tion, or semantic notions in general, can be
reduced to or eliminated in favour of non-
semantic terms.
DIGITAL AND ANALOGUE
Spurred by the work of CHOMSKY and
advances in computer science, the 1960s
saw a rebirth of 'mentalistic' or cognitivist
approachs to psychology and the study of
mind (see COGNITIVE PSYCHOLOGY). Two
features have loomed large in these develop-
ments. One is the extensive use of the
notion of a representation in characterizing
mental states and activities. (See Fodor,
1975; Mandler, 1983. See also CON-
NECTIONISM for an alternative approach.)
The other is the adoption of computer
models of mind and cognitive processing
(see COMPUTATIONAL MODELS). In turn, two
questions have become the centre of much
debate and discussion: (i) What kinds of
representational systems are employed in
cognition? (ii) What type of computer, if
any, is a cognitive being? Not surprisingly,
in philosophy, with its 20th-century focus
on language, logic and formal systems, the
answers to these two questions tended to be
respectively (i) language-like systems of
representation and (ii) a digital computer.
The two answers fit together nicely in that
language is assumed to be a digital form of
representation and, therefore, readily suited
for use by a digital computer. To give these
assumptions some bite and generality,
though, it is important to provide a clear
specification of the difference between
digital and non-digital (analogue) repre-
sentations and processing.
There appears to be wide agreement that
digitalness has to do with things that are
discrete and analogue involves the con-
tinuous. Beyond this the issues get fuzzy.
Not only is there no consensus about the
correct definitions of 'digital' and 'analogue',
but there is much confusion over whether
the distinction applies to representational
systems, to the person or machine using the
system, to a combination of system and
user, or to the nature of the physical laws
and physical stuff the symbols or the symbol
user are made of. Those following Goodman
propose that a distinction be drawn between
digital and analogue symbol systems based
on the syntactic and semantic density of the
elements of the system. Others have taken
the crucial distinction between digital and
analogue symbols to depend on whether
the representation is an analogue of that
which it represents. (See Pylyshyn, 1986.)
The notion of an 'analogue' being appealed
to here varies. Some stress a 1-1 correspon-
dence between sign and object, assuming
pictures and maps (Le. Icons) to be para-
digm cases of analogue representation.
Others think that representation is analog-
ical if there is some law-like connection
between the referents and signs. For them,
an example of an analogue system would be
the representation of temperature by the
level of mercury in a thermometer. This last
analysis allies analogue representation with
Indices that reflect (continuously valued)
law-like relations between sign and object.
The digital/analogue distinction has taken
on metaphysical importance in that many
theorists assume that analogue signs are not
'real' representations. They are not the sorts
of representations that exhibit intentionality
or implicate mind. Only Symbols, free as
they are from physical constraints, do this. If
REPRESENT A TION
the digital/analogue distinction is treated in
terms of the syntactic and semantic features
of the system, however, metaphysical claims
of this sort would seem to be unfounded.
They result from running together claims
about the structural properties of the rep-
resentational system with claims about how
the system is used. We tend to reject attri-
butions of intentionality when production
and/or response to a sign is 'triggered' by
present stimuli, hence our reluctance to
consider the doings of the thermometer or
the bee dances as 'truly' symbolic activities,
indicating the presence of mind. (See
Bennett, 1964.) But the question whether
use of a representational system is, in
Chomsky's (1968) terms, creative and free
from the control of stimuli is separate from
whether the system is syntactically or
semantically dense. We can program a
bank-teller machine to communicate and
respond in a totally fixed or unfree manner,
while using a digital symbol system, like
English. Alternatively, we humans could
come to understand the symbolically dense
bee language without the sight of a dance or
the sighting of nectar triggering any beha-
viour on our part. We could, that is, employ
the bee language in the same stimulus-free
ways we use English.
The idea that analogue processing is
somehow non-mental or non-cognitive also
seems peculiar from the syntactic/semantic
density perspective taken above. The pocket
calulator that computes on discrete rep-
resentations for each number is usually
considered to operate digitally, while com-
putation by slide-rule seems to involve con-
tinuous or analogue processing. Human
mathematical competence would appear to
be a paradigm case of cognitive activity. It is
hard to see, however, why it should lose
this status if it turned out that our compu-
tations are carried out by a neuronal device
that operates like a slide-rule rather than
like a calculator.
IMAGERY
The ascendancy of cognitive approaches to
mind has brought with it a renewed interest
539
REPRESENT ATION
in imagery. (See IMAGERY. See also Block,
1981.) Two problems concerning repre-
sentation have held centre stage in these
discussions. The first problem is of a piece
with older ontological worries over the
status of so-called 'pictures in the mind'.
Proponents of imagistic theories often talk
in ways that seem to presuppose that
images are objects, like physical objects,
that can be rotated, scanned, approached,
enlarged, etc. Yet it is hard to make sense of
such reification, given that mental images
have no mass, physical size, shape, or
location. The second problem concerning
imagery has close ties to debates over the
adequacy of the (digital) computer model of
mind. The reason for this is that images are
typically identified with pictures and thus
allied with analogue representation. So it is
held that if we employ images in cognition,
it shows that claims that all mental repre-
sentation is propositional or sentential (i.e.
digital) is false. In turn, if mental processing
involves the use of non-digital, pictorial
representations, our minds and cognitive
activities cannot be understood within the
constraints of the standard computer model.
Although seemingly separate matters, the
issue of ontological reification and the issue
of analogue representation come together
for those who assume that analogue repre-
sentations function via their sharing or
having features analogous to those they
represent. Most proponents of imagistic
explanations allow that their theories would
be unsustainable if they did require that
there literally be items in the mind that pos-
sessed spatial dimensions and other physical
properties. They have offered various pro-
posals attempting to show how it is possible
to cash in on talk of using or manipulating
images without falling into the trap of reifi-
cation. In any case, it should be clear that
questions of reification also pose a problem
for proponents of sentential models of mind,
who claim that we think in words. For the
ontological quandary of giving a satisfac-
tory account of how there can be pictures
or maps in the head is at root no different
than the problem of how there can be
words and sentences in the head. And if a
540
satisfactory answer is available to the latter,
it should be adaptable to the former.
A good deal of the debate over imagery
has been obscured by problematic accounts
of the basis of the 'stand for' relation and by
unsupported assumptions about the nature,
function and distinction between and
among linguistic and non-linguistic forms of
representation. For example, it is common
for both proponents and critics of imagery
to identify images with pictures or picture-
like items, and then take it for granted that
pictorial representation can be explained in
terms of resemblance or some other notion
of 1-1 correspondence, or assume that
since pictures are like their referents they
require no interpretation. But it is highly
questionable whether such accounts are
adequate for dealing with our everyday use
of pictures (maps, diagrams, etc.) in cogni-
tion. The difficulties involved with this older
understanding of Iconic representation
become more acute when applied to imagis-
tic or mental pictures.
EXP ANDING THE REPRESENT A TION AL
DOMAIN
There is something problematic in the very
way the imagery controversy, along with
other debates over mind and cognition,
have been set up as a choice between
whether humans employ one or two kinds
of representational systems. As mentioned
at the start, we know that humans make
use of an enormous number of different
types of (external) representational systems.
These systems differ in form and structure
along a variety of syntactic, seman tical and
other dimensions. It would appear there is
no useful sense in which these various and
diverse systems can be divided into two
well-specified kinds. Nor does it seem pos-
sible to reduce, decode, or capture the cog-
nitive content of all of these forms of
representation into sentential symbols. Any
adequate theory of mind is going to have to
deal with the fact that many more than two
types of representation are employed in our
cognitive activities. It would seem all the
more premature, then, to assume that yet-
to-be discovered modes of internal repre-
sentation must fit neatly into one or two
pre-ordained categories.
Appeals to representations play a promi-
nent role in contemporary work in the study
of mind. With some justification. most atten-
tion has been focused on language or lan-
guage-like symbol systems. Even when some
non-linguistic systems are countenanced.
they tend to be given second-class status.
This practice. however. has had a rather
constricting affect on our understanding of
human cognitive activities. It has. for
example. resulted in a lack of serious exam-
ination of the function of the arts in organiz-
ing and reorganizing our world. And the
cognitive uses of metaphor. expression.
exemplification. and the like are typically
ignored. Moreover. recognizing that a much
broader range of representational systems
plays a role in our cognitive activities should
throw a number of philosophical presupposi-
tions and doctrines in the study of mind into
question. Among these are: (1) claims about
the unique significance of language or syn-
tactic systems of representation as the mark
of the mental; (2) the identification of con-
tentful or informational states with the sen-
tential or PROPOSITIONAL ATTITUDES; (3)
the idea that all thought can be expressed in
language; (4) the assumption that composi-
tional accounts of the structure of language
provide the only model we have for the crea-
tive or productive nature of representational
systems in general; and (5) the tendency to
construe all cognitive transitions among
representations as cases of INFERENCE
(based on syntactic or logical form.)
See also CONTENT; DEVELOPMENTAL PSY-
cHoLoGy; DRETSKE; FODOR; LANGUAGE OF
THOUGHT; PERCEPTION; RA TION ALITY.
BIBLIOGRAPHY
Bennett. J. 1964. Rationality. London: Rout-
ledge & Kegan Paul.
Block. N .. ed. 1981. Imagery. Cambridge. MA.:
MIT Press.
Chomsky. N. 1968. Language and Mind. New
York: Harcourt Brace Jovanovich.
R YLE. GILBERT
Dretske. F. 1986. Misrepresentation. In Belief.
ed. R. Bogdan. Oxford University Press.
Fodor. J. 1975. Language of Thought. New
York: T. Crowell.
Gombrich. E. 1960. Art and Illusion. New
York: Pantheon Books.
Goodman. N. 1976. Languages of Art. 2nd edn.
Indianapolis: Hackett Publishing.
Mandler. J.M. 1983. Representation. In Hand-
book of Child Psychology. Vol. III: Cognitive
Development. ed. J. Flavell and E. Markman.
New York: John Wiley and Sons.
Peirce. C.S. 1931-5. Collected Papers. vol. II.
Cambridge. MA.: Harvard University Press.
Pylyshyn. Z. 1986. Computation and Cognition.
ROBERT SCHW AR TZ
Ryle, Gilbert (1900-1982) Gilbert Ryle's
The Concept of Mind (1949) had for a con-
siderable time an enormous influence. but.
despite its open and direct style. it was
never easy to characterize. One reason for
this can be found in its origins. Ryle tells us
in his Autobiographical Sketch (1970) that
his main interest was in the metaphilo-
sophical question: 'what constitutes a philo-
sophical problem and what is the way to
solve it?' Questions specifically to do with
the mind were not in the forefront of his
thinking when he embarked on The Concept
of Mind. Instead. he was looking around for.
as he wrote. 'some notorious and large-size
Gordian Knot' upon which to 'exhibit a sus-
tained piece of analytical hatchet-work'.
Leaving aside the fact that Ryle was more
interested in philosophical method than in
the philosophy of mind itself. there is a
second reason for the difficulty one has in
interpreting the book. The Gordian knot
Ryle chose was the product of a conception
of mind that is generally attributed to
Descartes. On that conception. there is. in
addition to material substance. a kind of
mind substance. and we have a more direct
- indeed privileged - access to what passes
in the mind than to what takes place in the
material world. Thus. for example. I can
know for certain and immediately that I
have a pain. though I could doubt that
541
RYLE. GILBERT
I have a damaged hand which I take to be
the cause of my pain. The notorious Carte-
sian possibility of there being an evil demon
who could deceive me about my body and
about the external world does not extend to
what I am currently thinking or experien-
cing, and this makes the mental substance
knowable in a way nothing else can be.
Now one may feel that there is a kind of
authority we each have about the contents
of our minds - that we know these contents
in a way unavailable to anyone else - but it
is possible to say things about this kind of
authority which do not go so far as to
commit one to the Cartesian conception (see
CONSCIOUSNESS; FIRST-PERSON AUTHOR-
ITY; INTROSPECTION). One can separate
certain problematic features of the mind
from the particular dualistic Cartesian
account of those features. However, in Ryle,
it is not easy to tell whether the target is the
Cartesian conception or, more radically, the
intuitions it seeks to explain. Officially, it is
the 'Ghost in the Machine' - a Cartesian
ghost if ever there was one - which is to be
exorcized. Yet, when it comes to giving
detailed arguments, Ryle can sound as if he
is denying that we have any kind of intro-
spective knowledge of our own mental
states. And this leads to the third elusive
aspect of the Concept of Mind - its degree of
commitment to BEHA VIOURISM.
Some behaviourists, most notably psy-
chologists such as Skinner and Watson,
seem to have flirted with the idea that there
simply are no phenomena of mind, that the
mind is a kind of fiction superimposed on
the complex movements of human bodies.
Such an extreme eliminativist behaviourism
was certainly never part of Ryle's project.
However, less radically, there is a kind of
behaviourism which treats the mind, not as
a fiction, but as itself consisting in, or defin-
able in terms of, behaviour. And there is a
real question just how far Ryle can be
understood as advocating this sort of view.
On the one hand, it is not easy to see
what Ryle is getting at when he speaks of
the 'category mistakes' we make in speak-
ing about mental phenomena unless there
is some kind of behaviourism lurking in the
542
background. According to Ryle, a category
mistake consists in taking one kind of thing
for another as when, to use his example,
the confused tourist says that he has seen
all the college buildings in Oxford, but has
yet to find the University. What the tourist
has failed to appreciate is that the Uni-
versity is a different kind of thing from an
assortment of colleges and buildings.
Applied to the mind, the idea seems to be
this: it is easy to be misled into thinking it is
a special kind of thing, different from, but
belonging to, the same' general category as
the matter that makes up the physical
world. (This would be like thinking that the
University, being different from, but in the
same category as, the colleges and build-
ings, was itself something one could see by
being in the right place in Oxford.) But Ryle
insists: 'the hallowed contrast between Mind
and Matter will be dissipated. but dissipated
not by either of the equally hallowed
absorptions of Mind by Matter or of Matter
by Mind (p. 23).' So, for example, he sug-
gests that when one tries to find the differ-
ence between intelligence and lack of it, one
should not look for some special mind-stuff,
the operation of which makes someone
intelligent. Instead, one should be 'asking
by what criteria intelligent behaviour is
actually distinguished from non-intelligent
behaviour.' In this and in numerous other
examples Ryle suggests that the mind con-
sists in patterns of behaviour - that it can
be reductively identified with behaviour -
and that to think otherwise about it is pre-
cisely to categorize it wrongly.
Yet, on the other hand, Ryle never gave
the kind of analysis of behaviour that would
be necessary to support a fully reductive
account. In his examples, behaviour is
always treated as fully intentional; there is
no attempt to characterize it in non-mental.
physicalistic terms. So, it remains simply
unclear how thoroughgoing a behaviourism
one ought to find in The Concept of Mind.
Some of the things he says might even
encourage one to think of Ryle as suggest-
ing an early form of FUNCTIONALISM. On
this view (roughly) each mental phenom-
enon is thought of as a state disposition ally
located in a network of behaviour and other
mental states. But here again Ryle's text
would seem to confound such an inter-
pretation, for he offers an account of dis-
positions that is both highly idiosyncratic
and inconsistent with the functional
account.
In spite of the difficulties of interpretation
and the fact that The Concept of Mind is less
often cited in present work in the philo-
sophy of mind, it contains a wealth of
highly specific, important insights into
mental concepts. And no account of Ryle's
masterwork should neglect to mention the
R YLE. GILBERT
exquisite style and subtle humour it con-
tains, which was characteristic of the man
himself.
See also DENNETT; WITTGENSTEIN.
BIBLIOGRAPHY
Ryle, G. 1949. The Concept of Mind. London:
Hutchinson.
Woods, 0., and Pitcher, G. eds. 1970. Ryle:
Critical Essays. New York: Doubleday.
SAMUEL GUTTENPLAN
543
s
Searle, John R. My work in the philosophy
of mind developed out of my early work in
the philosophy of language, especially the
theory of speech acts. Most of my work in
the philosophy of mind has been concerned
with the topics of INTENTION ALITY and its
structure, particularly the intentionality of
PERCEPTION and ACTION and the relation of
the intentionality of the mind to the inten-
tionality of language. I have also written
extensively on cognitive science (see COGNI-
TIVE PSYCHOLOGY), especially on the limita-
tions of the COMPUTATIONAL MODEL of the
mind. Other work has concerned the MIND-
BODY PROBLEM, the nature and structure of
CONSCIOUSNESS, the relation of conscious-
ness to UNCONSCIOUSNESS, the proper form
of explanation in the social sciences and in
the explanation of human behaviour gen-
erally, and the background of intentionality.
I can only discuss a small number of these
topics in this article, so I will confine myself
to four of the most controversial areas.
THE MIND-BODY PROBLEM
The traditional mind-body problem arises
out of the Cartesian assumption that
'mental' and 'physical' name two different
metaphysical categories of phenomena. Given
that assumption, the problem is what are
the relations between the two, and, specifi-
cally, how can there be causal relations?
Many contemporary philosophers of mind,
though they are not always aware of it, still
accept the basic assumption that 'mental',
construed naively in terms of consciousness,
subjectivity, privacy, qualia, etc., implies
'non-physical'; and 'physical' implies non-
mental. Those who think of themselves as
materialists typically deny the existence of
any special, irreducible mental phenomena;
544
and those who think of themselves as prop-
erty dualists or dual aspect theorists think
that we must recognize the existence of
some non-physical phenomena in the
world. I reject these assumptions and I
think both sides are mistaken. I think that
this is one of those rare questions in philo-
sophy where you can have your cake and
eat it too. The materialists are right to think
that the world consists entirely of physical
phenomena; the dualists are right to think
that it contains irreducible mental phenom-
ena, such as my present state of conscious-
ness. My claim is that these are not
inconsistent. Given that they are both true,
we need to redefine the notions of the
'mental' and the 'physical'.
The solution to the apparent paradox is
what I call biological naturalism. Mental
phenomena are part of our natural bio-
logical history, as much a part of biology as
growth, digestion, enzyme secretion or
reproduction. So construed, the general
form of the relations between the mental
and the physical can be stated as follows.
All mental states, from the profoundest
philosophical thoughts to the most trivial
itches and tickles are caused by neurobio-
logical processes in the brain. As far as we
know anything at all about how the world
works, this point is well established. Neur-
onal processes cause mental states and
events. But what, then, are these mental
states and events? Does the causal relation
between the two not commit us to some
kind of dualism? No! Mental phenomena,
such as my present state of conscious
awareness of the table in front of me, are
higher-level features of the brain. But how
can that be? How can there be causal rela-
tions without two distinct events, hence two
different phenomena, the mental and the
physical? Actually this sort of relation is
quite common in nature. Consider the
liquidity of the water in the glass in front of
me or the solidity of the table on which I
am working. In both cases a higher-level
feature of a system is caused by the beha-
viour of lower-level elements, molecules,
even though the entire system is composed
of those lower-level elements. Similarly in
the brain a higher-level feature of a system,
consciousness, is caused by the behaviour of
lower-level elements - synapses, neurons,
modules, etc., even though the entire
system is composed of those (and other)
lower-level elements. And, just as the
liquidity and solidity of systems of molecules
is not a property of any molecule, so the
consciousness in the brain is not a feature
of any single neuron. And, just as liquidity
and solidity are not separate substances, so
to speak, squirted out by the molecules, so
consciousness and intentionality are not
separate juices squirted out by the neurons,
they are simply states that the whole neur-
onal system is in at certain times and under
certain conditions.
We can summarize biological naturalism
rather crudely in two propositions:
(1) Brains cause minds.
(2) Minds are higher-level features of brains.
Two caveats. First, like all analogies, the
ones I have used only go so far. There are
many dis analogies between the relations of
solidity and liquidity to molecular behaviour
on the one hand, and that of consciousness
and intentionality to neuronal behaviour,
on the other. Most importantly, conscious-
ness is by its very nature subjective and, in
some sense at least, 'private' and 'inner',
not equally accessible to all competent
observers.
Second, I am not suggesting that all our
problems about mind-body relations are
now solved. On the contrary, neurobio-
logically speaking we are just getting
started. Once we see, however, that there is
no philosophical or metaphysical mystery to
mind-brain relations we can kick the
problem of specifying the exact details of
SEARLE, JOHN R.
mind-body relations out of philosophy,
where it does not belong, and into neuro-
biology, where it does. The questions of how
exactly which brain processes cause which
mental phenomena and how they are real-
ized in neuroanatomy are empirical scien-
tific issues, not problems for philosophical
analysis.
The 'solution' to the mind-body problem
enables us to deal with several other tradi-
tional problems:
The Other Minds Problem
The standard view is that:
( 1) we know of the existence of conscious-
ness in other people by inference;
(2) the inference is based on observation of
their behaviour, especially their verbal
behaviour;
(3) the principle of the inference is analogy.
Because their behaviour is similar to
ours in similar circumstances we infer
that their mental states are like ours.
I believe that (1)-(3) are all false. Our
relations to other people, and to the world
generally, are not epistemic except in odd
circumstances. I do not infer that other
people are conscious, except in very special
circumstances. My confidence that other
people are conscious is not based on their
behaviour. If it were a matter of behaviour I
would have to infer that my radio is con-
scious because it engages in more varied
and more coherent verbal behaviour than
any human I ever met. The basis of my
total confidence that other people are con-
scious is that I can see that their causal
structure is like mine. They have eyes, nose,
skin, mouth and all the rest, and the beha-
viour is relevant to the question of their
conscious states only because we see it as
situated in an overall causal order. The
principle that warrants my complete con-
fidence in the existence of other minds is
not: similar-behaviour-ergo-similar-mental-
states. Rather the principle is: similar-
causal-structures-ergo-similar-cause-and-
effect-relations.
545
SEARLE. JOHN R.
Epiphenomenalism
It is obvious to any conscious agent that
consciousness enters crucially into his or
her causal relations with the world. I decide
to raise my arm, and 10 and behold it goes
up. No philosopher or neurobiologist is
going to convince us that there is no causal
relation. But how is such a thing possible?
Does consciousness secrete acetylcholene
(the neurotransmitter most responsible for
muscle movements), does it shake the axons
and dendrites? On the account given by bio-
logical naturalism, consciousness is causally
efficacious in a way that is typical of higher-
level system features. For example the solid-
ity of the piston in the car engine is entirely
caused by and realized in the system made
up of the microelements, but the solidity
does not thereby become epiphenomenal.
The solidity of the piston is causally essen-
tial to the functioning of the engine.
Similarly consciousness is entirely caused
by and realized in the system made up of
neurons, but consciousness does not
thereby become epiphenomenal. It is caus-
ally essential to the functioning of the
organism. The fact that it is grounded in the
system of neurons, far from showing it to be
epiphenomenal, precisely explains how it
can, like other higher-level causal features
of systems, be causally efficacious. (See EPI-
PHENOMENALISM)
THE CRITIQUE OF COGNITIVE SCIENCE
Cognitive Science is currently one of the
most exciting areas in the study of the
mind, but it unfortunately was founded on
a mistake. This need not be fatal, many
other successful sciences have been founded
on mistakes. In cognitive science the
mistake is: the mind is a computer program;
and the mind is to the brain as the program
is to the hardware. Minds, in short, are
computer programs implemented in brains.
This view, which I have baptized 'Strong
AR TIFICIAL INTELLIGENCE' (Strong AI for
short), can be refuted in one sentence.
Minds cannot be identical with computer
programs, because programs are defined
syntactically in terms of the manipulation of
546
formal symbols, such as as and Is, whereas
minds have mental or semantic CONTENTS;
they have more than a syntax, they have a
semantics. This refutation came to be
known as The Chinese Room Argument,
because I originally illustrated it with the
following parable.
Imagine that I, a non-Chinese speaker,
am locked in a room with a lot of Chinese
symbols in boxes. I am given an instruction
book in English for matching Chinese
symbols with other Chinese symbols and for
giving back bunches of Chinese symbols in
response to bunches of Chinese symbols put
into the room through a small window.
Unknown to me, the symbols put in
through the window are called questions.
The symbols I give back are called answers
to the questions. The boxes of symbols I
have are called a database, and the instruc-
tion book in English is called a program.
The people who give me questions and
designed the instruction book are called the
programmers, and I am called the com-
puter. We imagine that I get so good at
shuffling the symbols, and the programmers
get so good at writing the program, that
eventually my 'answers' to the 'questions'
are indistinguishable from those of a native
Chinese speaker. I pass the T URI N G test for
understanding Chinese. But all the same, I
don't understand a word of Chinese and -
this is the point of the parable - if I don't
understand Chinese on the basis of imple-
menting the program for understanding
Chinese, then neither does any digital com-
puter solely on that basis because no digital
computer has anything that I do not have.
This is a simple refutation of Strong AI.
Like all arguments it has a logical structure.
It is a derivation of a conclusion from three
premises:
premise 1: programs are formal (syn-
tactical),
premise 2: minds have contents (seman-
tics),
premise 3: syntax is not sufficient for
semantics.
The story about The Chinese Room illus-
trates the truth of premise 3. (See SYNTAX/
SEMANTICS.) From these three propositions
the conclusion logically follows:
programs are not minds.
And this conclusion refutes Strong AI.
Of all the arguments that I have ever
published I regard this one as among the
most obviously sound and valid, but I have
to say that nothing has aroused more
debate than this. I am constantly amazed to
see the implausibility of the various replies
that for over a decade have been given to
the Chinese Room Argument.
There are several common mis-
understandings of the Chinese Room Argu-
ment and of its significance. Many people
suppose that it proves that 'computers
cannot think'. But that is a misstatement.
The original meaning of 'computer' was
'person who computes'. On this definition,
we are all computers whenever we compute
anything and we can certainly think. The
definition of 'computer' and 'computation'
has since evolved to mean anything to
which we can attach a computational inter-
pretation. This has the consequence that
everything is a computer because you can
always attach zeros and ones to anything.
For example, consider the door: let door
open = 0, door closed = 1. So the door is a
primitive computer. Neither the original
narrow nor the new expanded definition of
computation has the consequence that
computers cannot think. On the contrary,
on the narrow definition all computers are
thinkers. On the wide definition, everything
is a computer, so a fortiori, all thinkers are
computers.
Another misunderstanding of the Chinese
Room Argument is to suppose that I am
arguing that as a matter of logic, as an a
priori necessity, only brains can have con-
sciousness and intentionality. But I make no
such claim. The point is that we know in
fact that brains do it causally. And from this
it follows as a logical consequence that any
other system that does it causally, i.e. that
produces consciousness and intentionality,
must have causal powers to do it at least
equal to those of human and animal brains.
SEARLE, JOHN R.
But it does not follow that other systems
have to have neurons to do it. (Compare:
airplanes do not have to have feathers in
order to fly, but they do have to share with
birds the causal powers to overcome the
force of gravity in the earth's atmosphere.)
The question of which systems are causally
capable of producing consciousness and
intentionality is an empirical factual issue,
not to be settled by a priori theorizing. Since
we do not know exactly how brains do it
we are in a poor position to figure out how
other sorts of systems, natural or artificial
might do it. But there is no logical or meta-
physical obstacle to consciousness and
intentionality being caused in some other
sorts of system, whether natural or arti-
ficial.
There is an even more powerful argu-
ment against the computational model of
the mind which I did not publish until after
1990. A basic distinction fundamental to all
the sciences is between those intrinsic fea-
tures of the world whose existence is inde-
pendent of us and those features of the
world whose existence is relative to obser-
vers, users, makers and to intentionality in
general. That an object is made of cellulose
fibres is intrinsic, that it is a chair is observer-
relative, even though the same object is both
made of cellulose fibres and is a chair. In
general the natural sciences deal with
intrinsic features of nature; the social
sciences deal with observer-relative features.
I am using the expression 'observer-relative'
to include the notions of user/maker/
designer/intentionality-relative, etc.
Now what is the status of computation? Is
it intrinsic or observer-relative? Well, on the
narrow definition considered above, where
conscious agents actually do, e.g. arith-
metic, it is intrinsic. It is not a matter of
what any outside observer says. If I am
adding 2 + 2 to get 4, that is an intrinsic fact
about me. But what about the electronic
circuit on which I am now writing this?
Such cases of computation are observer-
relative. Intrinsically the object is an electro-
nic circuit with state transitions between
voltage levels. But that it is a 'computer',
like the fact that the illuminated shapes on
547
SEARLE. JOHN R.
the screen in front of me are words and sen-
tences, is observer-relative. The simple fact
that computation is observer-relative is
devastating to the claim that the brain is a
digital computer. The question, 'Is the brain
a digital computer?' lacks a clear sense. If it
asks, 'Is the brain intrinsically a digital
computer?', the answer is: Nothing is intrin-
sically a digital computer. A process is com-
putational only relative to some observer or
user who assigns a computational inter-
pretation to it. If it asks, 'Could we assign a
computational interpretation to the brain?',
the answer is: We can assign a computa-
tional interpretation to anything.
This is a separate argument from the
Chinese Room. That argument showed that
semantics is not intrinsic to syntax, this one
shows that syntax is not intrinsic to
physics. The upshot is that the computa-
tional theory of the mind is incoherent. It is
untestable because it lacks a clear sense.
There is no answer to the question: what
intrinsic physical facts about the system
make it computational?
THE CENTRALITY OF CONSCIOUSNESS
Until recently there were few discussions of
consciousness in philosophy, psychology or
cognitive science. There are historical
reasons for this neglect, but it has had
unfortunate consequences because con-
sciousness is the central mental notion. All
of the other crucial mental notions, such as
subjectivity, intentionality and mental cau-
sation can only be understood in terms of
conscious mental states and processes.
It may seem strange to say that con-
sciousness is the central mental notion
when at any given moment most of our
mental states, our beliefs, desires, memories,
etc., are unconscious. But the crucial con-
nection between consciousness and the
unconscious can be stated as follows: There
is a logical connection between the notion of
consciousness and the notion of the uncon-
scious such that in order for a state to be an
unconscious mental state it must be the sort
of thing that could be conscious in principle.
I call this 'The Connection Principle'.
548
There are many arguments for the Con-
nection Principle but perhaps the simplest is
this. There is no way to make intelligible
the notion of an aspectual shape of an
intentional state except in terms of con-
sciousness or accessibility to consciousness.
Every intentional state has what I call an
'aspectual shape'. This just means that it
represents its conditions of satisfaction
under some aspects and not others. Thus.
for example, the desire for water is a differ-
ent desire from the desire for H
2
0 even
though water and H
2
0 are identical. If I
represent what I desire under the aspect
'water', that is a different aspectual shape
from representing the same substance under
the aspect 'H
2
0'. What is true of this
example is true generally. All intentional
states represent their conditions of satisfac-
tion under some aspects and not others; and
this has the consequence that every inten-
tional state, conscious or unconscious, has
an aspectual shape.
For conscious intentional states, such as
the conscious desire for water, there is no
problem about specifying the aspectual
shape. It is determined by how one thinks
about whatever it is that the state is about.
But what about unconscious intentional
states? When a mental state is entirely
unconscious, the only occurrent reality of
that state is in the form of neurophysiologi-
cal states and processes. Therefore, the only
sense that we can give to the claim that an
unconscious intentional state has a determi-
nate aspectual shape, is that it is the sort of
thing that could be brought to conscious-
ness in the form of conscious thoughts.
actions, etc.
What then, is the ontology of the uncon-
scious when unconscious? Strictly speak-
ing, the only occurrent ontology of the
unconscious is that of neurobiological
states and processes. As far as the occur-
rent reality is concerned, all of my mental
life consists of two and only two features:
consciousness and neurobiological pro-
cesses. When we talk of unconscious inten-
tional states, we are talking about the
capacity of the brain to produce conscious
thoughts. actions, etc.
Attributions of unconscious mental
states. then. are in a sense 'dispositional'.
To say of the man who is sound asleep that
he believes that Clinton is president is like
saying of a bottle of fluid on the shelf that
it is poison or bleach. It does not imply that
the substance is poisoning or bleaching
anyone or anything right then and there.
but it describes the substance in terms of its
causal capacity. not in terms of its occur-
rent realization of that capacity. Now.
similarly with attributions of unconscious
mental states. when we say of someone
that he has an unconscious mental state.
we are describing his brain not in terms of
its structural features. but in terms of its
causal capacities. such as the causal capa-
city to think the appropriate thoughts on
waking.
The Connection Principle gives us a
further basis for the criticism of cognitive
science. Much of the cognitive science
literature postulates mental states that are
not only unconscious in fact. but uncon-
scious in principle. They are not the sort of
thing that could be brought to conscious-
ness. If my account is correct. this view is
incoherent. There are no such states. The
only unconscious mental states are those
features of the brain that are capable of
causing the state in a conscious form.
THE BACKGROUND OF INTENTIONALITY
A large part of my work in the philosophy
of mind has dealt with intentionality. Since
I present some of my views on this subject
in another article in this volume. INTEN-
TIONALITY. I will confine this part of the
discussion to an aspect where my views are
different from those which are common in
the field.
On my view. all intentional states can
function. that is they only determine their
conditions of satisfaction. such as truth con-
ditions in the case of beliefs. or fulfilment in
the case of desires and intentions. against a
Background of capacities. abilities. tenden-
cies. dispositions and other causal structures
that are not and could not be analysed in
terms of other intentional states. This point
SEARLE. JOHN R.
differs from the current orthodoxy in the
following respect: the currently accepted
view recognizes that intentional states are
not atomistic but function within holistic
networks of other intentional states. For
example. in order to believe that dinner is
ready. I have to have a network of other
beliefs. But the point I am making is that in
addition to all of these other beliefs. the
beliefs themselves only function against a
Background of abilities. capacities. know-
how. etc .. which are not and could not be
analysable into further sets of beliefs.
The simplest argument for the thesis of
the Background is that if you try to follow
out the threads in the network. the task is
endless. This is not simply because you do
not know where to stop. but because each
further intentional state allows for an inde-
finitely large range of interpretations unless
the correct interpretation is fixed by some-
thing not itself an intentional state. Thus. if
I understand the sentence 'dinner is ready',
I have to understand that dinner is some-
thing you eat. Further. I have to know that
it is eaten through the mouth and not
through the ear or the toe. I have to know
that dinner is made of edible material
objects as opposed to prime numbers or car
factories. This list of beliefs goes on indefin-
itely. but - and this is the crucial point -
each of these other beliefs is itself subject to
an indefinitely large range of different inter-
pretations. and that has the consequence
that there has to be some point at which
the series comes to a stop. It comes to a stop
with my simple ability to cope with the
world. As WITTGENSTEIN, who presents a
similar argument. maintains, understanding
is fixed by an ungrounded way of acting -
we just know what to do - and on my view,
what to do includes how to understand.
interpret and apply intentional states.
The thesis of the Background together
with the critique of cognitive science has
significant implications for the under-
standing of human cognition. Instead of
saying that human beings in the exercise of
their cognitive capacities are everywhere
following rules that are inaccessible in prin-
ciple to consciousness. we should say,
549
THE SELF
rather, that they have a Background struc-
ture which enables them to cope in certain
ways with their environment, and that
structure is what it is because the environ-
ment is what it is and their relations to that
environment are what they are.
CONCLUSION
My work in the philosophy of mind is by
no means complete. I am currently
working on the social character of many
mental phenomena and the role of the
mind in constructing a social reality, a
reality of money, property, marriage, pro-
fessions, government and other institutions.
In this article I have only discussed a few
of the topics that interest me. But even in
this narrow scope I hope to have sketched
the main features of my overall vision.
Mental phenomena are above all biological
phenomena and they are as real as any
other biological phenomena, such as
growth, digestion, photosynthesis or the
secretion of bile. They cannot be reduced to
something else, such as behaviour or com-
puter programs. There are exactly two
sorts of processes that go on in the brain,
first-person subjective consciousness and
third-person neurobiological phenomena.
Talk of unconscious mental states and pro-
cesses is always dispositional. Talk of
unconscious mental phenomena that are in
principle inaccessible to consciousness is
incoherent. There is no intermediate level
of computation between the level of the
mental and the level of the neurobiological.
And the problem is not that we lack evi-
dence for such a level, rather the claim is
untestable because it is incoherent. A
genuine science of cognition would allow
for at least three levels of explanation - a
neurobiological level, a level of intention-
ality, and a functional level where we iden-
tify the operation of Background capacities
in terms of their functional role in the life
of the organism.
See also BELIEF; FUNCTIONALISM; IDENTITY
THEORIES; THOUGHT; THOUGHT AND LAN-
GUAGE.
550
BIBLIOGRAPHY
Lepore, E., and van Gulick, R., eds. 1991. John
Searle and His Critics. Oxford: Basil Black-
well.
Searle, J.R. 1979. Expression and Meaning.
-- 1983. Intentionality: An Essay in the Phi-
losophy of Mind. Cambridge University Press.
-- 1980. Minds, brains and programs.
Behavioral and Brain Sciences, 3, 417-24.
-- 1984. Minds, Brains and Science: The
1984 Reith Lectures. Cambridge, MA.:
-- 1992. The Rediscovery of the Mind. Cam-
JOHN R. SEARLE
the self The individuals to which we attri-
bute thoughts, beliefs, desires, emotions,
intentions, memories, sensations, and inten-
tional actions - together with responsibility
for those actions, are persons. Human
beings are the obvious bearers of these attri-
butes, sharing with other physical objects
those properties that accrue to them by
virtue of their location in space and time.
But two particularly problematic features of
the attribution of the mental properties to
persons may be isolated that give rise to the
idea that a person is something 'over and
above' a material body.
The first feature is that of self-conscious-
ness: experiences require a subject, at least
potentially aware of itself as the subject of
those experiences. But how can the thing
that has the experiences itself be an item
within these experiences? As will emerge in
the next section, when we consider the dif-
ferences between self-awareness and the
awareness of the objects of experience that
enters into the content of experience, it is
tempting to posit a 'pure' experiencing con-
sciousness that is identifiable only from the
first-person perspective - an entity of a quite
different sort from the physical body by
which, as things are, third-person identifica-
tion is available of the experiencing subject.
The second feature is the capacity of
persons for intentional action (see THE
WILL). Although much of what happens to
persons is accountable for in the same way
as that which happens to material bodies -
for instance the ways in which the world
impacts on them are conditioned by where
they happen to be located at particular
times - many of their interactions with the
world are conditioned not by that world
impersonally conceived. but by how they
themselves represent the world to be.
Persons are self-motivated beings with a
considerable degree of autonomy over what
happens to them and the world they
inhabit. Again it is tempting to view the
author of action as something over and
above a material body. An apparently
unbridgeable gulf opens between the sub-
jective and the objective. with subjectivity
here conceived as an extra-worldy source of
a unique perspective on what is available
from the third-person view.
Such considerations can make it seem
that a person is an individual that. in addi-
tion to a body. has a self. Perhaps a person
is a composite of a self - in some sense
requiring explication - and a body. There
has been much debate in the literature
about the role the conception of the self
should play in arriving at the identity con-
ditions for persons over time.
SELF-CON SCIOUSNESS
Philosophical problems about self-conscious-
ness are best approached by consideration
of the reference of the T of the '1 think'
that, as Kant insisted, is in principle
appendable to all my conscious experiences
(see CONSCIOUSNESS). A peculiarity of self-
reference is what Evans (1982) calls its
'immunity to error through misidentifica-
tion' (but also see Evans, 1982, for argu-
ments that self-reference is not unique in
this respect). Thus I employ no criterion of
identity in order to know which entity 1 am
thinking about in thoughts about myself as
experiencing subject, nor is there is any
question of my needing to keep track of
some entity - at least in the way that 1
must in the case of reindentification of
material objects - in order for my thoughts
about myself at different times to latch on to
THE SELF
the correct object of such thoughts. If as
noted above the T considered merely with
respect to its role as the owner of experi-
ences cannot itself be encountered in
experience, it cannot be mistakenly encoun-
tered in experience. Famously, such con-
siderations led Descartes to reason that the
only thing that was immune to doubt was
his own existence as a 'thinking thing' - a
substantial, but non-material, self or soul,
that had infallible access to an interior
mental realm. however fallible that realm
might be in its purported representation of
an external world. (See HISTORY; INTRO-
SPECTION .)
Hume by contrast took the unencounter-
ability of the self in experience to impugn its
very existence. He argued:
For my part, when I enter most intim-
ately into what 1 call myself. 1 always
stumble on some perception or other
. . . 1 never catch myself at anytime
without a perception, and never can
observe anything but the perception
. . . nor do I conceive what is further
requisite to make me a perfect non-
entity. (1978, p. 254)
He concluded that the idea of the self as an
entity that owns experiences should be
replaced with the idea of the sum of those
experiences themselves - T am nothing but
a 'bundle of perceptions' - a conclusion that
left him highly dissatisfied. for his own phil-
osophical framework lacked the resources
to individuate such bundles. Hume's insis-
tence that no a priori conclusions about the
self are derivable from an idea that lacks
experiential content marked an important
development in thought about self-
consciousness, which was expanded upon
by Kant (see the next section). But Hume's
conclusions were flawed by his inheritance
of the Cartesian view of mind according to
which it is mental items, or 'perceptions',
that are themselves the objects of thought
and perception in an independent interior
realm - a view that invites the postulation
of an elusive inner observer to witness such
items (see, further, McDowell, 1986).
551
THE SELF
A more recent 'no-ownership' view of the
self, or at least the view that a certain class
of 'I' statements fail to refer, may be attrib-
uted to Wittgenstein (1958) who contrasted
statements such as 'I have grown six
inches' with those such as 'I have a tooth-
ache'. In the latter case he writes 'It is as
impossible that in making the statement ...
I should have mistaken another person for
myself, as it is to moan with pain by
mistake, having mistaken someone else for
me. To say "I have pain" is no more a
statement about a particular person than
moaning is.' And Anscombe (1975) con-
cludes, from considerations about the
feature of immunity to reference-failure of
'I' -thoughts when 'I' is taken to be a refer-
ring expression, that self-conscious thought
is not about an object at all.
An alternative approach to establishing
what selfhood consists in is to abandon the
quest for an entity answering to the pecu-
liarities of the logical grammar of 'I'-
thought sentences, and focus instead on the
criteria of identity for persons overtime. This
is in effect Locke's strategy, who invokes as
crucial a memory criterion:
in this alone consists personal identity,
i.e., the sameness of a rational being:
. . . as far as . . . consciousness can be
extended backwards to any past action
or thought, so far reaches the identity
of that person. (1959, p. 333)
Many objections have been raised against
Locke's criterion, and many refinements
that meet them proposed. Indeed Locke's
own elaborations of his criterion suggest
that it is a more general criterion of con-
tinuity of consciousness that he has in
mind. I shall be considering some accounts
that might be called Lockean in flavour in
the next section but one. But for now the
important point to note is that Locke's
criterion is, essentially, detachable from a
bodily criterion. He is himself explicit
about this, arguing that were a prince to
exchange his soul, 'carrying with it the
consciousness of the prince's past life' with
a cobbler, then 'everyone would see' that
552
the cobbler would then be the same
person as the prince. (But see Williams,
1956, for doubts about the coherence of
such thought-experiments.) As such,
Locke's account of the self shares with
that of Descartes and Hume an underlying
assumption that considerations about what
is constitutive of the subjective are detach-
able from considerations about that which
is available to the objective view. It is time
to examine that assumption.
SUBJECTIVITY AND THE OBJECTIVE
WORLD
How, then, is the subject of experience to be
conceived as situated in the objective world?
A first step is to look to Kant, who argued
for a crucial interdependence between the
notions of subjectivity and objectivity.
In his famous Transcendental Deduction
of the Categories Kant attempts to prove
that there are certain concepts - the cat-
egories - whose application within experi-
ence is presupposed by the very possibility
of experience. Kant uncontroversially
assumes that judgments are the vehicles of
knowledge claims, and so of experience
itself. He begins by arguing that all judg-
ments - requiring as they do a judge, and
so their (at least potential) accompaniment
by 'I think' - must conform to the require-
ments of what he calls the 'necessary unity
of apperception [or self-consciousness],. This
requirement is expressed by the analytic
principle 'All my representations in any
given intuition must be subject to that con-
dition under which alone I can ascribe them
to the identical self as my representations'
[Bl38]. Although no substantive a priori
truths about the self are derivable from the
analytic principle that is Kant's starting
point, he nevertheless derives some impor-
tant 'synthetic a priori' claims about the
nature experience must have in order to
conform to the condition of being the
experience of a self-conscious subject. He
proceeds to 'deduce' the categories by estab-
lishing a reciprocal relation between judg-
ments, and the judge's representation of the
content judged as belonging to a unitary
consciousness. It emerges that it is only if
the data presented to sensibility are con-
ceived of - via the categories employed in
judgments - as constituting relatively per-
manent items, enjoying law-governed
causal relations, that the experiencing
subject would be able to distinguish itself as
judge distinct from them, the order of whose
subjective representational states is distin-
guishable from the order of the represented
items, thus allowing room for the thought
that they are my representations. And this,
famously, amounts to the thought of my
series of representations jointly constituting
a course of experience through an objective
realm. As Strawson (1966) puts it: 'its
members collectively build up or yield,
though not all of them contribute to, a
picture of a unified objective world through
which the experiences themselves constitute
a single, subjective, experiential route, one
among other possible subjective routes
through the same objective world.' The pro-
vision for, at the very least, what Strawson
terms the 'self-reflexivity' of experience - the
room within experience for the thought of
experience - must be made if a requirement
on making judgments at all is to be met: the
ability to discriminate between 'this is how
things are', and 'this is how things seem'.
But once that minimal self-conception
within experience has been provided for, so
have the conditions for the conception of
the world as independent from the judging
subject's experience of it. Subjectivity and
objectivity from this perspective are not
opposed but interdependent.
In the Paralogisms, which deal explicitly
with the problem of the self, Kant stresses
that the necessity of the unity of appercep-
tion is a necessity of thought. Although we
must conceive of ourselves as unitary con-
sciousnesses, it does not follow from this
that we know ourselves as unitary con-
sciousnesses; nor even may we conclude
that we know ourselves to be unitary con-
sciousnesses. The identity of the T of the 'I
think' that is a 'permanent element' accom-
panying my judgments as a mere logical
form is a purely logical identity, and no
'permanent element in appearance' should
THE SELF
be postulated to answer to it. That we do
conceive ourselves as unitary conscious-
nesses is indeed inevitable, given that we
must so conceive ourselves in order to con-
ceive - in judgment - anything at all. Hence
the fallacious reasoning engaged in deriving
a priori truths about the self - its simplicity,
immortality, spirituality and so forth - is
'paralogistic', 'without intent to deceive'.
But the reflective, second-order thought
about the nature of experience, that we
must so conceive ourselves, is not a thought
that we ought to try to render 'objectively
valid'. That is to say, in recognizing a
necessity of experience we should not take
ourselves to have supplied ourselves with
an intuition - an item within experience - to
which the idea of a unitary consciousness
corresponds. Rather, all we have is an
explanation of why, groundlessly. we con-
sider ourselves to know ourselves as unitary
consciousnesses.
We are still some way from a conception
of the self as itself an element of the objec-
tive world, and so a candidate for knowl-
edge. But if we could show that the subjects
of experience must further conceive of
themselves as physical entities in order to
instantiate the attributes about whose own-
ership they have such direct awareness,
then we could reinstate the common-sense
view that the proper bearers of those attri-
butes are ordinary, physical, human beings.
And the Kantian conception of the T as a
pure logical form may be considered to be
an abstraction from the common-sense con-
ception rather than a component of it,
requiring an entity or some sort of con-
struction to answer to it.
This is roughly Strawson's (1959)
strategy. He starts with two questions: '(I)
why are states of consciousness ascribed to
anything at all? and (2) why are they
ascribed to the very same thing as certain
corporeal characteristics?' Strawson ack-
nowledges that experiential attributes and
states owe their existence to the identity of
the individual whose states they are: they
are 'logically non-transferable'. He argues
that a necessary condition of ascribing
states of consciousness to oneself is that one
553
THE SELF
be able to make sense of their ascription to
others. And a necessary condition of ascrib-
ing states of consciousness to others is that
one be able to identify them as individuals
of the same type as oneself. If the owners of
states of consciousness were Cartesian egos,
then there would be no means of identifying
them, and so no means of ascribing states of
consciousness at all. On the other hand, a
no-ownership theorist who is prepared to
allow that all my experiences contingently
depend on a certain body either illegitim-
ately presupposes a referent for the 'my' of a
sort that he has disallowed, or ends up
stating a necessary truth. Rather, Stawson
concludes, 'person' is a logically primitive
concept of an entity to which both states of
consciousness and corporeal characteristics
are ascribable. Furthermore the logical
character of predicates denoting states of
consciousness requires that they have third-
personal - constitutive behavioural - and
first-personal attributional conditions.
Mastery of such predicates requires an
understanding of both such conditions of
attribution. Thus questions (1) and (2) are
answered by noting that (2) is a necessary
condition for (1).
Where Strawson stresses the importance
of the ascribability of different types of pre-
dicate to the same type of individual. Evans
(1980) emphasizes the other aspect of
what he calls the 'Generality Constraint'
on knowledge of which object one is think-
ing about. Namely, that self-knowledge
requires an understanding of what it would
be to ascribe predicates of the same type to
different individuals of the same type as
oneself. Indeed self-knowledge requires a
sensitivity to certain ways of gaining infor-
mation (for instance knowledge of what
must be the case for 'I am in pain' to be
true rests upon my ability to decide
whether or not I am in pain merely on the
basis of how I feel), as well as a capacity to
manifest T - thoughts in action. But
equally self-knowledge requires the capa-
city to understand propositions about our-
selves whose truth is not so grounded on
these distinctive ways of gaining informa-
tion, or manifesting our grasp of them (for
554
instance propositions about our infancy, or
future events involving ourselves). And this
type of self-knowledge requires that one
know what it would be for an identity
statement of the form 'I = x' to be true,
where 'x' is 'an identification of a person
which - unlike one's "I" - identification -
is of a kind which could be available to
someone else'.
This, crucially, requires a conception of
persons as physical things, locatable in
space and time, and so the practical ability
to know how to locate oneself in space and
time (Evans, 1980, p. 209).
Evans shows that there is a mistaken
tendency to assume that the immunity to
error through misidentification of T-
thoughts arises only with respect to
thoughts about oneself as the subject of
mental self-ascriptions. But this phenom-
enon attaches equally to bodily self-ascrip-
tions such as 'my legs are crossed', 'I am
moving', 'I am in front of a house', where -
at least in ordinary circumstances - no
knowledge of anything would be gained if
the ways in which such thoughts were
arrived at involved the application of some
identity criterion that left room for error
about who such thoughts concerned.
In the case of mental self-ascriptions it
cannot be overemphasized that it is the
world, and not some inner informational
state, to which the subject looks in thinking
those thoughts that grant him an infallible
awareness of himself as their owner. Indeed
an 'I think', as Kant claims, accompanies all
my 'representations'; but as Kant insists, as
such it is purely formal. Self-knowledge
requires in addition the range of capacities
that Evans describes. But such knowledge
is, on Evans account, achievable, for '[t]he
very idea of a perceivable, objective, spatial
world brings with it the idea of the subject
as being in the world, with the course of his
perceptions due to his changing position in
the world and the more or less stable way
the world is' (Evans, 1980, p. 222). On this
view the concept of the self involved in self-
reflection is the concept of the same entity
as a person, objectively conceived, and not
a component of it.
PERSONAL IDENTITY
Even if it can be argued that selves are
necessarily embodied, it does not follow that
bodily identity is a necessary condition of
personal identity. Considerations about dead
bodies would seem to show that it is clearly
not a sufficient condition. But various
thought experiments have been devised to
show that it is not a necessary condition
either; and others devised that seem to
allow the concepts of self and person once
again to diverge.
The underlying intuition driving such
thought experiments is an elaboration of
the Lockean thought that what matters in
questions about personal identity is the
fulfilment of some sort of psychological-
continuity criterion (underwritten by a
plausible causal condition). Thus, Shoe-
maker (1963) asks us to consider a case
where Brown's brain is successfully trans-
planted into Robinson's head. Intuition
seems to argue in favour of allowing that
Brown survives as the resulting man,
'Brownson'. After all, he will, it is to be
assumed, retain Brown's memories, disposi-
tions, character traits and so forth (see
again, Williams, 1956b, for doubts about
this). Such a case could, of course, be con-
sidered a case of bodily continuity, with the
brain taken to be a vital enough bodily
component for brain-identity to be suffi-
cient for bodily identity. But more tenden-
tious examples, involving 'brain state
transfer devices', that are intended to elicit
the same intuitions suggest that it is
merely the brain's (contingent) casual role
as realizer of a particular set of psychologi-
cal states that grants it the privileged
status as identity criterion. 'Psychological
continuity' is essentially abstractable from
it, and the self may be viewed not as an
entity at all, but as a construction of
casually related states, or perhaps (see
Lewis, 1976) as a sum of person-stages.
(But see Williams (1970), for an argument
to show that the intuitions elicited by
thought-experiments of this latter sort are
merely the result of the way such examples
are described.)
THE SELF
Perhaps the most perplexing case is of the
sort discussed by Wiggins (1967) and Parfit
(1971), where the brain of a man, let us
call him S, is divided, and each half -
retaining S's memories, character traits,
etc., - transplanted in a new body. Both
resulting people, Sl and S2, would seem to
be psychologically continuous with S in the
way that it is proposed 'matters' for perso-
nal identity. But we surely do not have here
a case of identity, for identity is a transitive
relation. And while the appropriate condi-
tions seem to exist for S's identity with both
Sl and S2, it seems clear that Sl and S2 are
not identical: they have different spatial
locations, for instance, and proceed to enjoy
entirely separate lives. It seems clear that S
'survives' in some sense - after all, were
half his brain to be destroyed in the process
and the other half to leave him with his
memories and character traits intact, his
survival would be indisputable. Why should
the existence of a second person resulting
from the success of the operation cast doubt
upon his survival?
Parfit's own conclusion is that what
matters about survival is not identity at all,
but psychological continuity, which only
contingently coincides with identity. Psy-
chological continuity differs from identity
not only in its intransitivity, but also in
being a matter of degree. Sl and S2 may
themselves divide, and subdivide, and the
more remote 'branches' will eventually fail
to retain S's memories, still less character
traits. Still, a relation Parfit characterizes as
'connectedness' will be retained between S
and such branches, by virtue of the stron-
ger, direct relation of continuity that
obtains between the intervening stages. And
a vagueness attending the issue of at which
stage we should allow that S himself sur-
vives is inevitable.
The example has been much discussed
(see especially Lewis, 1976; Perry, 1972,
1976; Parfit, 1976; Wiggins, 1976, 1980).
But for now let us note that the chief diffi-
culty it gives rise to is how a person -
individuated in any of the ways the ima-
gined circumstances allow as a matter of
arbitration - is to conceive of himself. If it
555
THE SELF
were I who were to be divided in this way
tomorrow, and I were to learn that some
piece of good fortune I now covet were to
befall my left-half survivor rather than the
right-half survivor, I am unable to avoid
hoping that I will turn out to be left-half
rather than right. But in so doing I am
adopting the Cartesian conception, derided
by Kant, assuming that 'I' refers to a pure
consciousness that I will, at some future
date, know to be identical with that that
expresses the hope now. Left-half, I hope,
will somehow house this consciousness,
and right-half merely mimic it. But it is
unclear what alternative conception of my
future self - the fulfiller of my hopes - is
available to me. Further difficulties attend
thoughts about the past on this concep-
tion. For instance, as Wiggins (1980) asks,
how is S later to represent to himself some
past achievement (to use Wiggins's
example, sailing single-handed between
Scylla and Charybdis)? Let us say there are
three paths psychologically continuous
with S and so three equally good clai-
mants for remembering the feat. But such
claimants will have differing attitudes
towards the feat, attitudes that, on the
conception driving the claim that psycho-
logical continuity is what matters, are
partly constitutive of their own self-concep-
tions. But unless we are to claim that all
three at one time were in the same place,
the only plausible notion of a person avail-
able to each is that of being a fusion of
person-stages. But as Wiggins notes, the
properties of fusions are derivative from
those attributable to its constituent stages.
And too many of the predicates ascribable
to persons are not so derivable: Wiggins
offers, amongst others, 'weak, strong,
clever, stupid, a good goalkeeper, a fair
weather friend'. Wiggins argues that the
concept of a person is the concept of a
member of a NATURAL KIND - an organ-
ism, with clear principles of individuation.
An illusion of technical omnipotence
encourages the mistaken idea that
thought-experiments of the sort under con-
sideration, which violate those principles,
are genuine conceptual possibilities.
556
THE SELF AND ACTION
I omitted from the discussion of Kant's
account of the self (above) a feature of his
framework that ultimately renders that
account untenable. It was noted that the
conception of an objective world is inter-
dependent with one's own self-conception -
a conclusion also reached by Evans. Kant's
insight was to show that the ordinary world
of experience is, necessarily, a world con-
ceived in certain ways. There is no uncon-
ceptualized experience: that which is
represented - what experiential states are
about - is the world as experienced, which
is the empirical world. But Kant also warns
against our claim to knowledge of 'things in
themselves', unnervingly characterized as
inhabiting a 'supersensible' realm. A relat-
ively innocuous construal of 'things in
themselves' might simply be that they are
unconceptualized things, or things con-
sidered in abstraction from the necessary
conditions for our thinking about them,
where the implication is that such things
may not be considered as 'things' - in any
sense that carries ontological commitments
- at all. Individuation of objects requires
concepts, and so a judging subject. And the
supersensible is merely the empirical realm
considered in abstraction from the condi-
tions for experiencing it. But Kant unam-
biguously consigns the self to the realm of
the supersensible: there are no empirical
conditions of application for selfhood. But
we are the conceptualizers, and if it is 'we'
as we are in ourselves, from a supersensible
realm, that do the conceptualizing, then it
seems there is after all a perspective from
which the ordinary world of experience is a
mere appearance of things in themselves.
And it surely undermines Kant's insight if
the conceptualized world - the world of
experience - turns out to be only qualifiedly
real.
Why should Kant have refused to allow
that as conceptualizing beings we are ele-
ments of the empirical world? The most
likely answer is the link Kant insisted upon
between rationality and freedom. And he
held that it is only in our capacity as moral
agents that we are genuinely free. Kant
argued for causality as a crucial objective
feature of the world. but he equated caus-
ality with the causally determined - the
province of the events of the natural sci-
ences as opposed to exercises of rationality.
In order to exercise our moral capacities we
must escape the deterministic net.
But if we reinstate the self in the empirical
world. then taking Kant's point that the
world of experience is one represented by
the subject presents no obvious obstacle to
our allowing that our causal interactions
with the world that involve our believing it
to be a certain way. or desiring it to be some
other way - in other words our intentions -
are exercises of our rational capacities. and
so manifestations of freedom. But if this is
the conception of the self we want to
vindicate. it is surely an important project in
the philosophy of mind to show how such
causation is possible within a world
governed by physical law (see IDENTITY
THEORIES; PHYSICALISM; SUPERVENIENCE).
For let us assume that intentional concepts
and the common-sense. lawlike general-
izations in which they figure are in some
sense incommensurate with physical con-
cepts and physical laws (see DAVIDSON).
Then without some account of the relation
instantiations of intentional properties by
persons bear to instantiations of physical
properties. it will seem mysterious. if not
miraculous. how instances of the laws at
the physical level harmonize so happily
with the generalizations of the other. If we
just insist on the disparate nature of the
two explanatory schemes. and take their
explanatory success to underwrite their
status as causally explanatory schemes
without saying more. then the picture of
the mental realm that we are likely to be
left with is of a superimposed pattern that -
from a perspective that is coloured by our
own concerns - we discern in what is
independently there. But how then are 'we'
as pattern-discerners to conceive of our-
selves? What are we doing in imposing our
patterns if there is a perspective from which
particular exercises of our pattern-discern-
ing capacities are abstractable from the
THE SELF
pattern discerned? It is difficult to see how
such a picture allows for intentional prop-
erties making a genuine difference to what
happens - to what is there on which to
impose our interpretations.
The problem of establishing a plausible
conception of the subject of experience. or
the self. as an element of the objective
world. turns out to be inseparable from a
major problem in the philosophy of mind:
how to vindicate the causal efficacy of the
mental properties distinctive of subjects of
experience (see EPIPHENOMENALISM).
See also An Essay on Mind section 1; CON-
scIOusNEss; REASONS AND CAUSES; SUB-
JECTIVITY.
BIBLIOGRAPHY
Anscombe. E. 1975. The First Person. In Mind
and Language. ed. S. Guttenplan. Oxford:
Basil Blackwell.
Evans. G. 1982. The Varieties of Reference.
Hume. D. 1978. A Treatise of Human Nature.
ed. L. A. Selby-Bigge. 2nd edn rev. P. H.
Nidditch. Oxford: Clarendon Press.
Kant. I. 1929. Critique of Pure Reason. trans. N.
Kemp Smith. London: Macmillan & Co. Ltd.
Lewis. D. 1976. Survival and Identity. In
Rorty. ed .. 1976.
Locke. J. 1975. An Essay Concerning Human
Understanding. ed. P. H. Nidditch. Oxford
University Press.
McDowell. J. 1986. Singular thought and the
extent of inner space. In Subject. Thought.
and Context. ed. P. Pettit and J. McDowell.
Nagel. T. 1970. The Possibility of Altruism.
Parfit. D. 1971. Personal identity. Philosophical
Review. LXXX: 1. 3-27.
Parfit. D. 1976. Lewis. Perry. and what
matters. In Rorty. ed .. 1976.
Perry. J. 1972. Can the self divide? Journal of
PhilBsophy. LXIX. 463-88.
Perry. J. 1976. The importance of being iden-
tical. In Rorty. ed .. 1976.
Rorty. A.. ed. 1976. The Identities of Persons.
University of California Press.
Shoemaker. S. 1963. Self-Knowledge and Self-
Identity. Ithaca: Cornell University Press.
557
SELF-DECEPTION
Strawson, P.F. 1959, Individuals: An Essay in
Descriptive Metaphysics. London: Methuen.
Strawson, P.F. 1966. The Bounds of Sense: An
Essay on Kant's Critique of Pure Reason.
London: Methuen.
Wiggins, D. 1967. Identity and Spatio-Temporal
Continuity. Oxford: Basil Blackwell.
Wiggins, D. 1976. Locke, Butler and the
stream of consciousness: And men as a
natural kind. In Rorty, ed., 1976.
Wiggins, D. 1980. Sameness and Substance.
Williams, B. 1956. Personal identity and indi-
viduation. Proceedings of the Aristotelian
Society, LVII.
Williams, B. 1970. The self and the future.
Philosophical Review, LXXIX.
KIRSTIE MORRISON
self-deception The idea of self-deception is
problematic on two levels. First, there seems
to be something superficially paradoxical
about it from an analytical point of view.
But, secondly, given that it seems to be a
fundamental and pervasive feature of
human life, there are much deeper ques-
tions about its rationality and its connection
with agency and action.
The paradoxical nature of self-deception
shows itself as soon as one thinks about the
conditions necessary for deception in
general. If I set about to deceive Smith -
say, by lying - then I would typically try to
get him to believe something true, which I
know (or believe) to be false. Thus, if I know
there to be a train to Oxford at 11 p.m.,
then I deceive Smith if I can get him to
believe that there is no such train. How or
why I would do this is not of immediate
relevance, but what is crucial to deception
is that there is a certain deliberateness
about it. If I myself believe there to be no
such train - if I have made a mistake - then
telling Smith this does not count as attemp-
ted deception. To be sure, Smith may come
to believe something false, but, insofar as
my assertion was sincere, I have not
deceived him; my assertion was not a lying
one. Now, as one presumes from the label.
what one must do to be 'self-deceived' is to
558
fulfil the conditions of deception in respect
of oneself. For example, lying to oneself
would count as a central case. But how
could one possibly try to convince oneself of
something believing it false? If I do believe
that some proposition p is false, then it
would seem to be impossible for me to also
believe that p is true. So, it would seem that
self-deception consists in bringing about
something which is impossible.
Of course in less schematically character-
ized cases, there seems to be a bit more
room to manreuvre. Suppose that Jones's
husband is having an affair, and suppose
also that he is not very good at hiding this.
All Jones's friends are aware of what is
going on and they see in her behaviour
evidence that she too knows. For example,
though she does not normally like to spend
a night away at her sister's, she does so
often enough to make it seem as if she is
somehow getting out of the way. This and
other examples add up to a pattern of beha-
viour that only makes sense if she knows
what is happening, but, when confronted
by the inevitable busybody, she seems
wholly surprised that anyone would have
thought such a thing. Additionally, she
makes it clear how incapable she would be
of tolerating any infidelity on the part of her
husband. Her surprise and revulsion at the
idea seem so genuine that it would be very
difficult to think she was feigning ignor-
ance. Yet she takes no steps to find out the
truth, whilst all the while continuing to do
things consistent with her really knowing it.
In the end, her friends come to think that
she is deceiving herself about her husband's
affair.
What this sort of example introduces is
the important connection between how one
acts and what one believes. This makes it
possible to think that self-deception is not
the paradoxical believing that p and not-p,
so much as acting as if one believed that p,
whilst also acting as if p was false. Much of
jones's non-verbal behaviour showed her to
believe that the affair was going on. But her
verbal behaviour, when confronted with the
unpleasant fact, was unequivocally on the
other side.
Whilst the complex interaction between
action and belief might seem to get us out of
logical trouble, its ultimate success is doubt-
ful. For it seems reasonable, given the
nature of propositional attitudes, to ask
whether someone does or does not believe
something: acting as if one believed p is not
yet believing. But, as the general conditions
for deception revealed, you are only really
deceived when you are intentionally led to
believe something true which is believed by
your informant to be false. So, to be self-
deceived, Jones, believing the affair was
going on, would have had to convince
herself quite intentionally that it wasn't. If
in some way she merely acted as if she
believed it - perhaps in some unconscious
way - this wouldn't be enough: her denial
would not have been the result of any
intention on her part.
One way it might be tempting to take
matters further is by appealing to the notion
mentioned above in passing: THE UNCON-
SCIOUS. The thought might be that our
minds are composed of at least two ele-
ments: a conscious and an unconscious self.
In cases of self-deception, the one would be
the engineer of the deception of the other.
Or, less psycholanalytically, one might go in
for talk of the mind as inherently compart-
mental. Given the right circumstances, it
can happen that a single agent is able to
harbour the beliefs and propensities to act of
two or more distinct individuals. In the case
of Jones, as described above, her revulsion
at the thought of her husband's infidelity
was strong enough to push the belief that
he was indeed having an affair into a com-
partment of her mind insulated from her
otherwise positive picture of him.
Of course, talk of unconscious and mul-
tiple selves needs much more filling out
before it can be taken seriously. And in the
case of self-deception there is a particularly
important constraint on any such back-
ground story. For as we have seen, the self-
deceiver seems to go in for the deception
deliberately - it is motivated - and the very
idea of motivation presupposes that there is
an agent at work. Now consider how we
would view a compartmentalized mind if
SELF-DECEPTION
what we were discussing were the overt
actions of some agent. Suppose that Peter
were to do two things which appeared to be
in tension, say buying a particularly extra-
vagant car and yet going miles out of his
way to save a small amount on washing
powder. Our feeling that these are in tension
comes from the fact that we have trouble
imagining the motivational background in a
single agent that would explain two such
different ways of dispensing money. Of
course, it may be that Peter, loving expen-
sive cars, believes that any little savings on
other things will make ownership of such a
car possible. But, supposing that we could
dismiss this sort of circumstance, what
could be said about Peter's behaviour?
Surely, we would say that it is in some
sense irrational. However, if we went in for
the compartment view of Peter's mind, then
we seem free to say that one part of Peter
did the one thing and another part, the
other. And this would seem to rule out the
judgment that Peter's actions were irra-
tional; indeed it would remove the basis for
saying that they were even in tension with
one another. For if we treat Peter as
somehow harbouring two different agents,
then we can do no more than remark on
the very different kinds of money-directed
behaviour these two agents display. But of
course that is not how we would treat such
a case: for agency is not so easily divisible
when it comes to things people do.
This is not to say that we would never
compartmentalize agency. In cases of mul-
tiple personality and schizophrenia we may
well do just that. But this makes rather
than challenges the point. For cases of
behaviour like Peter's are not usually taken
as evidence of multiple personality, and yet
that is what one form of the compartment
theory requires of us. Of course, nothing
said here about how we understand agency
is meant to tell finally against the compart-
ment view. It is simply that such a view
must be spelled out in a way that makes it
clear how one and the same agent can have
what we might call a divided mind.
Trying to soften the paradoxical hard
edges of the idea of self-deception has led us
559
SELF-DECEPTION
to consider the intricate connection of atti-
tude and action, as well as the possibility of
less-than-integrated selves. Still another
way to complicate matters so as to avoid
paradox is to bring in the notion of time.
Unlike ordinary cases of deception which
can, as it were, happen in an instant, when
the deceived victim takes in and accepts the
falsehood deliberately provided, cases of self-
deception are typically more spread out in
time. Jones's friends were happy enough to
describe her as self-deceived about the faith-
fulness of her husband, but they came to
this conclusion only after a fairly protracted
examination of Jones's behaviour and only
after dismissing a number of initially more
plausible alternatives. Moreover, accepting
for the argument's sake the idea that Jones
deceived herself because she could find no
other way to reconcile her love of her
husband with her detestation of infidelity, it
should be clear that deceiving oneself, if it
can happen, is more in the nature of a
project. It takes time and a degree of careful,
if devious, planning to get oneself to ignore
what would otherwise be plainly admitted.
And perhaps it is in the temporal extended-
ness of the feat that one can defuse the
paradoxical air of self-deception. None-
theless, however the feat is managed, it is
undeniable that self-deception is something
ordinarily accepted as fairly commonplace,
as well as often beneficial. For there is no
denying that, however irrational the state of
mind that comes with self-deception, there
are circumstances when, in a broader sense,
it is rational to go in for it. When what one
believes is just too painful to acknowledge,
the alchemical transformation of belief
characteristic of self-deception can come to
the rescue. In such cases, the deceiving of
oneself may well be a project whose goal is
in one's own best interests, and is thus
wholly intelligible as well as rational.
BIBLIOGRAPHY
Pears, D. 1975. The paradoxes of self-
deception. In his collection Questions in the
Philosophy of Mind. London: Duckworth.
560
Fingarette, H. 1969. Self-Deception. London:
Davidson, D. 1982. Paradoxes of irrationality.
In Philosophical Essays on Freud, ed. R. Woll-
heim and J. Hopkins. Cambridge University
Press.
SAMUEL GUTTENPLAN
self-intimation see FIRST-PERSON
AUTHORITY.
sensation What one experiences when
touching a radiator or diving into a cold
mountain stream are sensations of heat and
cold. Thus understood, a sensation is some
experience or feeling that arises from the
way the world affects certain of our senses.
However, it is not obvious either that sensa-
tions require a specific sensory modality or
that all sensory modalities have associated
sensations. For example, we are said to
have sensations arising from our awareness
of the internal states of our bodies and these
are not the result of touch or any other
'external' sense. Of course, one might
simply maintain that we have a non-
external bodily sensory system which gives
rise to sensations of pain, upsets in the
stomach, etc. In this way, one would be
able to say that a sensation was an experi-
ence brought about either by some mechan-
ism for sensing the world or by one which
tells us about the states of our bodies.
However, the most difficult problem in
this area comes from considering the ques-
tion of whether sensations -- as just de-
scribed - are typical of all the sensory
modalities. Many feel that touch, smell and
taste are unproblematically accompanied by
sensations, but hearing, and especially
sight, give rise to lots of problems. For
example, when we see something blue, do
we have a sensation, and, if so, how can we
characterize it? Moreover, what, if any, role
does such a sensation play in the accounts
we might give of PERCEPTION? Answers to
these questions range surprisingly widely.
At one extreme is the view that the very
idea of seeing something blue is to be ana-
lysed as the having of a particular kind of
sensation (sometimes called a 'sense-
datum') which itself has properties (qualia)
that resemble the seen object. Thus, we see
something as blue because we have a 'blue-
ish' sensation and infer from that that the
object of our seeing is blue. (Of course, it
would be odd to describe the sensation itself
as blue, so one is forced to talk about these
sensational properties in quotation marks,
or in some other way.)
At the other extreme, is the view that
visual perception does not involve sensation
- that it is simply a mistake to assimilate
the case of seeing something blue to the
kind of thing we experience when we touch
something hot. This kind of view is often
called the 'intentional' theory of perception,
because what is attempted is an account of
perception as wholly consisting in content-
ful intentional states on a par with the pro-
positional attitudes (see QUALIA; PERCEP-
TU AL CONTENT).
See also CONSCIOUSNESS; FUNCTIONALISM;
IDENTITY THEORIES.
SAMUEL GUTTENPLAN
simulation theory and theory theory These
are two, many think competing, views of
the nature of our commonsense, proposi-
tional attitude explanations of action. For
example, when we say that our neighbour
cut down his mulberry tree because he
believed that it was ruining his patio and
didn't want it ruined, we are offering a typi-
cally commonsense explanation of his
action in terms of his beliefs and desires.
But, even though wholly familiar, it is not
clear what kind of explanation is at issue.
On one view, the attribution of beliefs and
desires is taken as the application to actions
of a theory which, in its informal way, func-
tions very much like theoretical explanation
in science. This is known as the 'theory
theory' of everyday psychological explana-
tion. In contrast, it has been argued that
our propositional attributions are not theor-
etical claims so much as reports of a kind
of simulation. On such a 'simulation theory'
STALNAKER. ROBERT
of the matter, we decide what our neigh-
bour will do (and thereby why he did it) by
imagining ourselves in his position and
deciding what we would do. For fuller
accounts of these two views and further
examples, see BELIEF (2) and FOLK PSY-
CHOLOGY.
SAMUEL GUTTENPLAN
Stalnaker. Robert In this sketch of my
own philosophical views and approach I
will focus on the problem of INTENTION-
ALITY, a problem that has held my atten-
tion since I first started thinking about the
philosophy of mind. I will say what I take
the problem to be, what strategies for
responding to it I advocate, and what some
of the remaining puzzles about intention-
alit yare.
Like most fat philosophical problems with
such labels (the problem of free will, the
MIND-BODY PROBLEM) the problem of
intentionality does not pose a single well-
defined question, but is a cluster of inter-
related questions and puzzles that seem to
derive from what is, in some sense, a
common underlying philosophical problem.
As with those other problems, there is
philosophical work to be done to turn the
puzzles and vague feelings of conceptual dis-
comfort into well-defined questions and pro-
jects - to articulate what needs to be done
to solve the problem. As with the other
problems, one cannot even begin the task of
formulating a well-focused project without
making controversial philosophical commit-
ments. A large part of the task is to defend
claims about what questions should be
asked, and about what would count as an
answer. I will begin by trying to say, in
general and impressionistic terms, what the
intentional phenomena are, what is puz-
zling and problematic about them, and
what needs to be done to clarify them.
Some things - for example, names and
utterances, memories, mental images and
feelings of anger, pictures, charts and
graphs - are said to represent, or be about,
or be directed at other things (see REPRE-
SENTATION). Intentional relations are rela-
561
STALNAKER. ROBERT
tions that hold between such things and
what they represent. or are directed at.
Intentional relations seem to be distinctively
mental; to the extent that inanimate objects
such as names or maps have intentional
properties. these properties seem to derive
from the intentional properties of the thin-
kers and agents who create or use the
objects. And intentional relations seem to
have peculiarities not shared by relations
found to hold more widely between things
in the natural order - causal relations.
spatiotemporal relations. and relations of
similarity and difference - and these pecu-
liarities seem not only to make it difficult to
understand how rational agents could be a
part of the natural order. they also are puz-
zling in themselves. How is it possible to
stand in a relation of any kind to a thing.
such as Zeus or the planet Vulcan. that does
not exist? How can one admire the man on
the beach while having a low opinion of the
man in a brown hat when they are one and
the same man? A satisfactory clarification of
intentional relations must begin by respond-
ing to these traditional puzzles (see An Essay
on Mind section 2.1).
The first move in this direction that I and
many others make is to propose that inten-
tional relations generally are to be explained
in terms of relations to propOSitions. Inten-
tional relations between agents and other
persons or physical object (x wants. loves.
admires. worships. or fears y. for example)
are to be explained in terms of PROPOSI-
TION AL ATTITUDES that the agent x has
toward propositions that involve or are
defined in terms of the object y. There will
be work to be done saying how individual
objects and concepts of them are involved in
defining or identifying propositions. but the
prior tasks. according to this way of
approaching the problem. are to clarify.
first. what propositions are. and second.
what it is for speakers. thinkers and agents
to be in intentional states or to perform
intentional acts such as asserting. believing.
hoping and fearing that take propositional
objects. One of my central concerns has
been to connect the two questions: what is
a proposition? and what is it to stand in an
562
intentional relation to one? The strategy for
connecting and clarifying these questions is
to use assumptions about the answer to the
second in order to motivate an answer to
the first. That is. one begins with a certain
kind of account of how we are related to
propositions. and asks what propositions
must be in order for us to be related to them
in that way.
One cannot begin without saying some-
thing about what propositions are. but the
idea is to make only tentative and relatively
neutral assumptions about these abstract
objects: they are what is said or thought.
whatever is the content of a BELIEF or a
DESIRE. whatever is the referent of a clause
of the form that P when it is the direct
object of an attribution of speech or
thought. Then the first question is this:
what is it to stand in a relation such as
belief or desire to such an object? But more
needs to be said about what kind of answer
to this rather vague question we are looking
for. and this requires that we say more
about what is problematic about intentional
relations: what puzzlement about them we
expect an explanation of intentional rela-
tions to resolve.
Even before committing ourselves to any
specific account of propositions. it does seem
reasonable to assume that they are abstract
objects of some kind. and one may be
puzzled about how human agents. if they
are a part of the natural order. manage to
get themselves into relationships with
abstract objects. The metaphors with which
we characterize our relations to propositions
are metaphors of causal interaction: we
'grasp' the propositions we understand; in
communication. we receive and convey
information. But how is it possible to stand
in something like a causal relation to an
abstract object? And more specifically. how
is it possible for intentional relations to the
abstract objects. propositions. to exhibit the
peculiar and distinctive features such as
referential opacity. and the apparent capa-
city to be related. by way of propositions
'about' them. to things that do not exist?
The strategy for solving this problem that
I adopted (see 1984. ch. 1) was a modest
one, but one that I would argue suffices to
solve the problem, at least in its initial form.
Rather than attempting to provide analyses
or definitions of the problematic intentional
relations - propositional attitudes such as
belief, desire and INTENTION - I aimed only
to exhibit some simple relations between
persons or physical objects and propositions,
relations that were manifestly unmyster-
ious, intelligible in themselves, and
obviously compatible with a naturalistic
understanding of the things that stood in
the relation, but that had the features that
were said to be problematic about inten-
tional relations (see NATURALISM). The
simple and artificial relations that I used to
make this point were relations defined in
terms of causal and counterfactual con-
structions. Let me give a purely abstract
example, just to make the strategy clear:
suppose we define a relation R between a
person and a proposition in terms of a
counterfactual in something like the follow-
ing way: R(x, that P) = df if it were the case
that p, then x would be F. ('x' and 'that P'
are here variables for persons or physical
objects and propositions, respectively, while
'F' stands in for some particular predicate of
things of the kind that the 'x' ranges over.)
Such a definition might, of course, be much
more complex. All that is required to show
how persons or other things can be related
to propositions is that the relation be
defined in terms of some open expression
that contains two variables, one taking an
individual as value, and the other a proposi-
tion. Now if such examples succeed in
making their point - that is, if the relations
defined are unmysterious, but also manifest
the features that were said to be what was
puzzling about intentional relations - then
they show that there is nothing mysterious
or problematic about an empirical relation
between a person and a proposition, or
about the features that were said to distin-
guish intentional relations from relations
that hold between things in the natural
order. This may not solve the underlying
problem, but it at least sharpens the issue
by challenging the poser of the problem of
intentionality to say what features distin-
STALNAKER. ROBERT
guish real intentional relations from the
simple relations exhibited.
The particular relations that I defined to
make this point - specifically a 'tendency to
bring about' relation, and a relation of
'indication' - were intended also to suggest
a slightly more ambitious claim: that inten-
tional relations might actually be explained
in terms of relations of the kind I defined.
'Indication' as I defined it was a kind of ten-
dency to carry information, where carrying
information is understood in terms of rela-
tions of counterfactual dependency between
the thing carrying the information and its
environment. It does not seem implausible
to suggest that belief is a kind of indication,
one that plays a certain role, in interaction
with motivational states, in the determina-
tion of rational action. Even if some such
claim is right, this is a long way from an
analysis of belief. At best, it puts belief into a
category that includes many kinds of states
of persons and systems that are clearly not
belief states, or intentional states in any
ordinary sense, for example states of the
immune system, of hard disks, of thermo-
stats. But even if the work of distinguishing
full-blooded rational agents and their inten-
tional mental states from simpler systems
and simpler representational states remains
to be done, it may be that much of the
puzzlement about representation is removed
by getting clear about how simple repre-
sentation is possible. And it may be that the
puzzlement that remains is more easily
articulated in the context of a more general
account of representation.
Now if we assume that the way
intentional states such as belief, desire and
intention relate us to propositions is to be
explained in terms of the kinds of causal
and counterfactual constructions used to
define the simple example relations, then we
can draw some conclusions about what
propositions must be. Specifically, we can
conclude that propositions should be indi-
viduated by truth conditions; necessarily
equivalent propositions - propositions true
under the same conditions - will behave
identically in causal and counterfactual
constructions, and so should behave
563
STALNAKER. ROBERT
identically in contexts explained in terms of
those constructions.
The original version that I gave of this
argument (in Stalnaker, 1976) made more
specific assumptions about the way in
which the intentional states, belief and
desire, should be analysed - assumptions
that I later came to think left out an essen-
tial dimension of intentionality (a historical,
or backward-looking dimension that is
exhibited by the relation of indication). But
the argument depended only on very
abstract and general features of the kind of
analysis proposed, features that are shared
by a wide range of different kinds of
hypotheses about the nature of the relations
that give intentional states their content.
Intentional states of agents may have the
content they have partly because of facts
about the way they came to be, and partly
because of the way they dispose the agent
to behave, and one need assume nothing
about the particular way in which content
depends on such facts about the way agents
are situated and disposed.
The conception of CONTENT motivated by
this argument is one that individuates pro-
positions by their truth conditions. The pos-
SIBLE WORLDS analysis of proposition -
propositions are sets of possible worlds, or
functions from possible worlds into truth
values - is no more than a way of repre-
senting a conception of propositional
content that is individuated in this way. In
saying this, I am advocating a metaphysi-
cally deflationary conception of possible
worlds: possible worlds should be under-
stood simply as the elements of the con-
ceptual space used to characterize
intentional states and acts. The intuitive
idea is this: to form a conception of the
world - to have beliefs, or to make supposi-
tions about the world - is to locate it in a
space of alternative possibilities; to under-
stand a thought or a statement is to see
how that thought or statement locates the
world in the relevant space of possibilities.
The possible worlds representation of
content provides a conceptual distinction
between truth conditions themselves and
the statements we use to say what the truth
564
conditions are. The advantage of this con-
ceptual separation is that it gives us the
resources to make sense of intentional states
that are not expressed in speech, or repre-
sented in linguistic form, and to clarify
questions about the relation between
language and thought (see THOUGHT AND
LANGUAGE). Many philosophers in this
century have argued or assumed that the
order of explanation of intentional phenom-
ena should go from language to thought:
first solve the problem of intentionality for
language and then explain the intention-
ality of thought in terms of the intention-
ality of the language with which thoughts
are expressed. This seems to me to get
things backwards, but even if one thinks
that in some sense speech precedes and is
presupposed by thought - it is useful, and
begs no questions, to begin with a con-
ceptual distinction between content and the
means by which it is expressed, and to have
a neutral characterization of content. An
explanation of content in terms of speech
could take the form of an explanation of the
possibilities used to define content in terms
of patterns of linguistic behaviour.
While I think the truth-conditional con-
ception of propositional content has con-
siderable intuitive appeal, it was recognized
from the beginning that the conclusion of
the argument sketched above - that the
contents of ordinary mental states are pro-
positions of this kind - has consequences
that are strongly counterintuitive, even
paradoxical. It seems that if we really believe
all necessary equivalents of our beliefs, then
we must be essentially logically omniscient,
but obviously we are not. Some take the
argument as a reductio ad absurdum of its
assumption. I prefer to take it as the next
step in the dialectical discussion of the
problem of intentionality: the initial puzzles
about intentionality are dissolved by exhibit-
ing unmysterious relations between persons
and propositions, including simple relations
that are plausibly described as relations of
representation. The challenge then is to say
what problems about the full-blooded inten-
tional states are not solved by seeing them
as states of the same kind as simple repre-
sentational states that can be shown to be
unproblematic. The argument that if inten-
tional content is explained in the way pro-
posed, then propositions must be coarse-
grained contents individuated by their truth
conditions can be interpreted as an attempt
to sharpen the problem in this way.
Whatever further things need to be said
about intentional states in order to solve the
further problems about them, I think it
should not be controversial to conclude that
one thing about us that is essentially con-
nected with the fact that we are intentional
agents is that we absorb and use informa-
tion, that intentional states are states
involved in such processes, and that seeing
them as such is an important part of under-
standing intentionality. If this is right, then
it is reasonable to conclude that the concep-
tion of informational content - content indi-
viduated by truth conditions - is at least
one important aspect of content. one that
must playa role in a full story about inten-
tionality. And it is reasonable to conclude
that the causal or information-theoretic
conception of representation is at least a
step on the way to an adequate account of
intentionality, a step that provides a context
for formulation of further problems.
Whether this step is progress or a dead-end
depends on whether the further problems
that this step leaves us with are problems
that it is fruitful to try to solve, and whether
the problems that we all share are clarified
by being seen in this context. (Some may be
inclined to think of the further problems
raised by a philosophical account as costs in
the cost-benefit analysis used to evaluate
such an account. But sometimes the prob-
lems faced by an account should be placed
on the benefit side. It is the job of a philo-
sophical project to raise problems as well as
to solve them, and it may be a limitation of,
for example, an analysis of propositions,
that it fails to raise certain problems. Of
course there is no neutral or easy way to
distinguish cases where a real problem has
been swept under the rug from cases where
a pseudo-problem has been successfully dis-
solved. To make this distinction, we need to
look at the details.)
STALNAKER. ROBERT
In conclusion, I will sketch very briefly
four clusters of problems about intentional
phenomena that arise in the context of the
strategy I have advocated for solving the
problem of intentionality, and that I think
the information theoretic story about
intentional relations and the possible worlds
conception of propositions that it motivates
help to clarify. Then I will comment in a
little more detail about the last and most
difficult of the problems.
First, there are problems raised by extern-
alism: how to come to terms with the fact,
or apparent fact, that meanings, and even
beliefs, 'ain't in the head'. (See EXTERN AL-
ISM/INTERNALISM; PUTNAM.) The thesis
that the contents of speech and thought
are determined in part by facts about the
physical and social environment of the
speaker or thinker was given a compelling
defence with intuitive examples and
thought experiments, but the thesis itself
was thought to be paradoxical (see TWIN
EARTH). How can the contents of my
thoughts - the way the world seems to me
- depend on facts about my environment
that are, in some sense, inaccessible to me?
The possible worlds conception of proposi-
tions and the information-theoretic concep-
tion of intentionality helps to provide the
externalist thesis with a theoretical context,
to clarify what is paradoxical about it, and
to respond at least to some of the reasons
for resisting it. (See 1989, 1990, 1993.)
Second, there are problems about self-
locating, or indexical attitudes. Can my
belief that today is Tuesday, my realization
that I am late for the meeting, or my wish
that my dentist appointment were not
tomorrow be understood as attitudes
toward impersonal and timeless proposi-
tions? I have argued that we can reconcile
this phenomenon with the assumption that
the contents of belief are propositions, and
clarify the status of indexical attitudes and
expressions of them, if we define proposi-
tions in terms of possible worlds (or at least
in terms of relevant alternative possibilities).
(See 1981.)
Third, there are problems about the rela-
tion between intentional states themselves
565
STALNAKER. ROBERT
and the semantic devices we use to attribute
and describe intentional states. It is widely
recognized that attitude ascriptions are
highly context dependent, and that some of
the problems and puzzles about proposi-
tional attitudes arise from the complexities
of the way we talk about attitudes rather
from complexities in the attitudes being
talked about, but it is controversial and
unclear exactly how these should be dis-
tinguished. If we have a representation of a
belief state that is conceptually distinguish-
able from the language used to describe it,
then we can more easily separate questions
such as, 'what is the world like according to
Pierre?' from questions about the truth or
falsity of specific belief attributions. ('Does
Pierre believe that London is pretty or
not?'). Separating the questions does not by
itself resolve the puzzles, but it does help to
bring them into focus. (See 1987, 1988.)
Fourth, there is the recalcitrant problem
alluded to above, the problem of deduction.
or of logical omniscience. If the contents of
intentional states are individuated by their
truth conditions, and if the contents of belief
states are explained in terms of the informa-
tion they tend to carry, then how is it pos-
sible to explain the fact that we don't
believe all the consequences of our beliefs?
To try to reconcile the facts about belief
attribution with the thesis that propositions
are individuated by truth conditions, I have
argued, first, that one should recognize
more complexity in the relation between
sentences and the propositions they express:
even if P and Q, as usually construed, are
necessarily equivalent, perhaps they should
be construed differently in the relevant
belief attribution contexts. Second, it must
be recognized that beliefs may be frag-
mented: one may be in separate belief states
that are not integrated into a single concep-
tion of the world, and so one may fail to
believe conjunctions of propositions that
one believes. (See Stalnaker, 1984. ch. 5.) I
think these strategies show some promise of
throwing light on deductive ignorance and
enquiry, but it is obvious that there is no
quick fix; solving this problem in the terms
that are set by the assumptions about inten-
566
tionality that it makes is a large and daunt-
ing project. One might ask, why bother
trying to jump over the hurdles that this
conception of intentionality and content
puts in our way when there are alternative
accounts of the contents of intentional
states that bypass these problems? Since it is
obvious that we take different attitudes
toward necessarily equivalent propositions
and that we are often ignorant of deductive
relationships, why accept an account of
intentionality that seems to imply that we
are not, or that at least raises the question
how it is possible for us to do something
that we obviously do? Even if such man-
(Euvres succeed, for the moment, in evading
decisive refutation of the hypothesis that the
objects of attitudes are coarse-grained pro-
positions, wouldn't it be better to respond
more directly to the problem by trying to
find a fine-grained conception of propOSition
that gets the examples right? I concede that
some kind of fine-grained object may playa
role in the solution to the problem, but
simply accepting the thesis that such an
object is the content of attitudes is not
responsive to the problem. It is clear that
the problem of deduction is a genuine
problem, and not just an artefact of certain
philosophical assumptions; it is a problem
that raises its head in theoretical contexts
that have no stake in any philosophical
account of intentionality, or of the content
of intentional states. In accounts of reason-
ing and knowledge representation studied
by theoretical computer scientists, and in
accounts of deliberative reasoning in eco-
nomics, decision theory and game theory,
all of the well-understood models of repre-
sentation, reasoning and decision-making
involve the kind of idealization that
abstracts away from deductive ignorance,
and the fact that they do so is regarded as a
major conceptual problem (see RATION-
ALITY). It is no help with the problem as it
arises in those non-philosophical contexts
simply to choose a more fine-grained object
- for example a structured meaning or a
Russellian proposition - as the object of
knowledge. belief and partial belief, pre-
ference and desire.
The assumption that I started with - that
an account of what the objects of belief are
must be motivated by an account of what
belief is - provides a challenge to the defen-
der of the thesis that the object of belief is
some kind of fine-grained object. Suppose
we grant that the object denoted by a clause
of the form that P in an attribution of atti-
tude is a structured meaning containing
individuals or individual concepts. truth
functions. properties and relations. and so
forth. According to this conception. the
object of belief is a kind of recipe for deter-
mmmg a coarse-grained informational
content; the object is more fine-grained
because the same informational content
may be determined in different ways - as
the values of different functions applied to
different things. or to things in a different
order. Suppose we grant further that where
M and N are two such recipes that deter-
mine the same truth-conditional content.
an agent might stand in the belief relation
to M. but not to N. The challenge is to
explain the role of the structure that distin-
guishes M from N in distinguishing a state
of believing M from a state of believing N. Is
it. for example. that a claim about what one
believes makes a claim. not only about the
informational content. but also about the
form in which that information is stored in
the brain? Or alternatively. is it that a belief
attribution makes a claim about the struc-
ture of the sentences that a believer is dis-
posed to use to express his belief in speech? I
have argued (in Stalnaker. 1990) that
neither of these answers is defensible. but
even if I am wrong about this. I think that
the move to structured propositions is not
responsive to the real problem of deduction.
or more generally to the problem of inten-
tionality.
One reason that the problem of intention-
ality is of central philosophical concern is
that our conception of a proposition con-
strains our conception of the distinctions
there are to be made in describing the
world. The propositions we express in
saying how the world is - in doing physics.
sociology and metaphysics - must be the
propositions that our account of intention-
STALNAKER. ROBERT
ality says are the objects of speech and
thought. If. in our metaphysics. we claim
that there is a distinction to be made
between the world being such that P and
being such that Q. and a fact of the matter
about which way it is. then our account of
intentionality must allow that there is a dis-
tinction between believing or supposing or
expressing the thought that P and believing.
supposing. or expressing the thought that
Q. This is an obvious point. but in the light
of it. at least some of the disputes between
proponents of coarse-grained informational
contents and fine-grained structured mean-
ings or Russellian propositions can be seen
as a side issue for the following reason.
Whatever view of propositions one takes.
one may distinguish two questions about
anything (for example. an utterance or an
intentional mental state) that expresses a
proposition: (1) what information does it
convey (where this is understood to mean.
what is the informational content - the
coarse-grained proposition expressed). and
(2) by what means does it convey that
information? A fine-grained account of the
object of belief is an account that says that
the object referred to in the that-clause of
an attitude attribution is an object that
answers a part of question (2) as well as
question (1). whereas the coarse-grained
account says that the referent of a that-
clause answers question (1). but says
nothing about question (2). But whatever
view one takes about this issue. if one of the
things at stake in an account of intention-
ality is the range of possibilities that we
have the capacity to represent. then it is
important to distinguish the question of
informational content from the question of
the means by which that content is ex-
pressed. and it is only the first that is rel-
evant to this concern.
If we rely on an intuitive distinction
between content and form - between what
is said or thought and how it is said or
thought - then everyone will of course
agree that to say what a person says or
thinks is to say only what that person says
or thinks. and not to say how it was said or
thought. The problem is that distinctions
567
SUBJECTIVITY
between fine-grained propositions seem,
from an intuitive point of view, to be dis-
tinctions of content. If I am ignorant of the
fact that 343 is 7 cubed, then it seems
intuitively that I lack a piece of information,
and not just that I fail to use a certain
means to represent the necessary proposi-
tion. This is indeed a problem - it is, I think,
the most pressing remaining part of the
problem of intentionality. But the problem is
evaded rather than solved by a conception
of content that simply builds a part of the
means by which content is represented into
the object that is called 'content'. The real
problem is to understand the nature of the
information that is conveyed, expressed or
stored when one conveys, expresses or
stores information that does not seem, on
the face of it, to be explainable in terms of
distinctions between possibilities. The
account of intentionality that I have been
promoting does not offer an easy solution to
this problem, but I think it does help to
bring it into focus.
See also DENNETT; DRETSKE; FODOR;
THOUGHT.
BIBLIOGRAPHY
Stalnaker, R. 1976. Propositions. In Issues in
the Philosophy of Language, ed. A. MacKay
and D. Merrill. New Haven and London:
Yale University Press.
-- 1981. Indexical belief. Synthese, 49,
129-51.
-- 1984. Inquiry. Cambridge, MA.: MIT
-- 1987. Semantics for belief. Philosophical
Topics, 15, 177-90.
-- 1988. Belief attribution and context. In
Contents of Thought, ed. R. Grimm and
D. Merrill. Tucson: University of Arizona
Press.
-- 1989. On what's in the head. In Philo-
sophical Perspectives, 3: Philosophy of Mind
and Action Theory; reprinted in The Nature of
Mind, ed. D. M. Rosenthal. Oxford Uni-
versity Press.
-1990a. Mental content and linguistic form.
Philosophical Studies, 58, 129-46.
-1990b. Narrow content. In Propositional
568
Attitudes: The Role of Content in Logie, Lan-
guage and Mind, ed. C. A. Anderson and J.
Owens. Stanford: CSU.
-- 1991. The problem of logical omnis-
cience, I. Synthese, 89, 425-40.
-- 1993. Twin Earth revisited. Proceedings of
the Aristotelian Society, XCIII, 297-311.
ROBERT STALNAKER
subjectivity With a fair amount of irony,
and a goodly amount of truth, one may say
that 'subjectivity' means different things to
different people. This entry will discuss three
notions of subjectivity. There is, first, the
subjectivity of phenomenal experience. By
'phenomenality', I mean the 'feel' of certain
of our experiences. Because a great deal has
been said about phenomenality elsewhere in
this volume (see CONSCIOUSNESS; QUALIA),
only a little will be said here. Second, there
is a notion of subjectivity associated with
THE WILL. This notion is central to develop-
mental psychologists, like Piaget; and the
greatest part of the present entry will focus
on it. This notion of subjectivity underlies
one's concept of oneself as a subject of
experience, distinguished, in the first place,
from the objects of experience, and, latterly,
from other subjects of experience as well.
The third idea of subjectivity, tied in a deep
way to a notion of a point of view, is the
realization that one is not only a different
subject of experience from other subjects of
experience but also that the world is experi-
enced differently by different subjects of
experience. This third idea of subjectivity is
also prominent among DEVELOPMENTAL
PSYCHOLOGISTS and is closely tied to the
previous one: that of a subject of experience.
PHENOMENALITY
People experience phenomenal states. Per-
ceptual and other sensory experiences, espe-
cially, seem to possess phenomenal qualities
(qualia). It is very difficult to say exactly
what phenomenal states or properties are;
but, in experiencing them, we somehow dis-
tinguish them from other sorts of states, and
from each other. An ice-cube just feels dif-
ferent from a warm compress; a blue cube
just looks different from a red square; a
mental image of a blue circle just 'looks' dif-
ferent from an image of a red square; and a
toothache just feels different from the SEN-
SA TION of relief that occurs when the tooth
stops hurting.
Although there is no way to capture it
completely, the idea of phenomenality is
perhaps best expressed in Nagel's (1974,
1979a) well-known claim (I will call it
'Nagel's slogan') that when an organism
experiences phenomenality, there is some-
thing it is like to be that organism. Nagel
points out that we could know the physiol-
ogy of a bat's 'sonar' sense and still not
know what it is like to be a bat. We would
not know what the bat's sonar experience
feels like for the bat. In a similar way, a con-
genitally blind person is unable to grasp
what the quality of a sighted person's
colour experience is like.
Closely related to Nagel's slogan is the
privacy of phenomenal states. Phenomenal
states are individuated by the person whose
states they are, as are broken arms; but
while both Sarah and Benjamin can observe
Sarah's broken arm, only Sarah can feel the
pain caused by it. If Benjamin feels pain on
observing Sarah's broken arm, that is Ben-
jamin's pain, not Sarah's. Benjamin cannot
observe or experience Sarah's pain - at least
not in the direct way Sarah can. The same
is true for all other phenomenal states.
Besides this ontological privacy of phe-
nomenal states, various forms of epistemo-
logical privacy are claimed for them. (i)
Only the person experiencing a phenomenal
state can know he or she is experiencing it
and what it is like; others can have, at best,
only probable beliefs about it. (ii) One
knows one is experiencing some phenom-
enal state - pain, for instance - simply by
experiencing it; no other criteria are
required. (iii) If one is in a given phenom-
enal state, then one knows oneself to be
in that state. (iv) If one believes oneself to
be in that state, then one knows oneself
to be in that state.
While ontological privacy and epistemo-
logical privacy have often been thought to
SUBJECTIVITY
go hand in hand, they are distinct (as are
the epistemological forms from each other).
Furthermore, the privacy claims might be
questioned. (1) How do we know that only
a single person can experience a given phe-
nomenal state? Is this a deep fact about
phenomenal states, or only a convention of
our use of words like 'pain', 'after-image',
and so forth? (2) If only the person under-
going a phenomenal experience - say, pain
- can know it, and knows it only on the
basis of experiencing it, how did that person
ever come to have a concept of that sort of
experience? Call the person P. Surely, one
may argue, the person would at best have a
concept of P-pain, not of pain itself. If so,
one could never ascribe this state to others,
for others cannot have P-pains. So, if the
privacy claims were correct, solipsism
would follow. (3) Worse, could the person
even acquire the concept of P-pain? What
would be the criteria for whether instances
fell under the concept? What would be the
criteria of relevant similarity among in-
stances? Wouldn't it be a case of whatever
one says goes? If so, such a 'criterion' is
inadequate for concept possession. (See An
Essay on Mind section 2.2.)
Such questions and objections, only men-
tioned here, have at times been backed by
extensive use of argument and example.
The most famous of these discussions is
WITTGENSTEIN'S (1953) 'Private-language
Argument'. But exactly what the argument
is, and what its intended target is, are in
dispute. Wittgenstein has been read var-
iously as saying that there are no phenom-
enal states that are private, even
onto logically; as saying that phenomenal
states are not sufficient determiners of pain
(and other sensations) - an interpretation
supported by the first part of the 'beetle in
the box' paragraph: 'Here it would be quite
possible for everyone to have something
different in his box' (1953, 293, lOOe); or
as saying that phenomenal states are not
even necessary to pains (and other sensa-
tions) - as suggested by the second claim of
the 'beetle in the box' paragraph: 'The box
might even be empty'. No modern discus-
sion of phenomenal subjectivity can avoid
569
SUBJECTIVITY
considering the issues raised by Wittgen-
stein.
One last set of remarks on phenomen-
ality. It is often claimed that Nagel's slogan
applies to another sort of experience, which
is not quite like phenomenal experience. It
might be agreed, for instance, that there is
no phenomenal experience required for
thinking that a lOOO-sided figure is larger
than a 999-sided one - phenomenal experi-
ences (mental images and the like) at most
accompany such thoughts. But one might go
on to claim that such thoughts are 'felt',
even if not felt, and that we are aware of
these 'feelings' when we introspect our
thinking. 'How else could we be conscious of
our thoughts?' seems to be the motivating
question. For we usually are conscious that
we are thinking such thoughts. If so, there
must be a basis for our awareness. Surely,
the claim runs, it just 'feels' different when
we are thinking this thought from when we
are not. And these 'feelings' are the basis of
our consciousness of the thought. Thoughts
(and other experiences like them) have
phenomenological properties, even if not
phenomenal ones; and it is these properties
that constitute the experience of those
thoughts.
My own view is to be sceptical about the
existence of such properties. Like Hume,
when he looked inside himself but could not
find his self, when I look inside myself, I find
no such states. I don't mean to deny that I
am able to introspect my thoughts; but
thoughts, and the introspecting of them,
have no phenomenal, or phenomenal-like,
properties. And Nagel's slogan does not
apply to mental states like thought, belief,
hope, and so forth (see Nelkin, 1993, for a
fuller argument).
For similar reasons, it is a mistake to treat
all notions of subjectivity on the paradigm
of phenomenal states. Other notions of sub-
jectivity can be usefully discussed without
even mentioning phenomenal states.
SUBJECTIVITY AND THE WILL
Nagel (1979a, 1986) has spent a good deal
of time discussing two ways of viewing the
570
world: the objective and the subjective. An
objective point of view involves viewing the
world from an 'impartial', third-person
aspect, while the subjective point of view
involves viewing the world as it appears to
oneself from a 'partial', first-person aspect.
Nagel conflates this 'first-personness' with
his slogan that there is something it is like to
be that person, but these are different
notions of subjectivity. As we will see, the
notion of a subject of experience, the quintes-
sential notion of subjectivity. need make no
reference to any sort of phenomenal feeling
(or phenomenological 'feeling', for that
matter).
Nagel argues that, on many questions of
life, whether descriptive or normative, the
subjective and objective views are incom-
patible and that there is no way to resolve
disputes between them. Before investigating
this claim, we will consider a prior question.
How do we separate the subjective and the
objective in the first place? And for answer-
ing this question, the most interesting set of
views to turn to are those that might
broadly be called Piagetian (see Piaget,
1954; Piaget and Inhelder, 1969).
We have good reason to believe that
human beings have no simple relation of
apprehension and belief to the world they
inhabit. PERCEPTION involves large amounts
of processing; and whether one believes per-
ception is representational (as do most psy-
chologists) or non-representational (as do
followers of J. J. Gibson), all would agree
that the perceptual story is a highly complex
one (see REPRESENTATION, IMAGERY). And
if concepts and beliefs based on perception
are prior to any of our other concepts and
beliefs, then certain difficult questions need
answers. For instance, if we are in a sense
being bombarded by information from the
external world, but in a fashion such that
the information needs to be filtered, pro-
cessed, subtracted from, added to, attended
to, recognized, and so on in order to be
useful, how does our perceptual experience
ever lead us to conceive our SELF, as a
subject of that experience, and the world of
spatiotemporal things that make up the not-
self, as the object of that experience?
Piagetians believe that we are able to
make these distinctions because there is,
given as a primitive in our experience, a recog-
nition of a distinction between doing things
and having things happen to us. Because
we can sometimes manipulate the world,
we come to distinguish our self (the sub-
jective), from the out-there (the objective).
The story to follow is brief and over-
simplified (and not every Piagetian would
agree to every part of the story), but it gives
the flavour of a Piagetian view (a view with
which I am broadly in sympathy).
As newborn infants (though there is evi-
dence that the process begins before we are
born - fetuses seem to initiate movements in
the womb), we are presented with an
unbroken stream of experience (where
'experience' includes more than phenom-
enal states - see Nelkin, 1989). This stream,
as such, displays neither its types nor its
tokens on its surface. The neonate's experi-
ence is a 'buzzing, blooming confusion'.
However, two factors allow the infant to
begin sorting this undifferentiated stream
into tokens and types.
The first factor is INTROSPECTION: the
infant is aware of its awareness of at least
some of this unbroken stream. The second
factor is that in some cases the infant is able
to affect and effect its own experiences,
while in others it is not. Putting these two
factors together, Piagetians maintain that
the infant finds itself apparently in control
of some of its' experiences, while finding that
other experiences are apparently not in its
control. There is evidence that infants are
already aware of this distinction as early as
eight weeks of age. Piagetians claim that
infants are aware of this distinction right
from the beginning. The words, 'finds itself'
are somewhat misleading, because at this
point the infant has no awareness of itself as
itself, i.e. no concept of a self.
It may be objected that infants cannot do
anything, so they cannot be in control of
any of their experiences. But experiments
show that newborn infants can do quite a
bit. They can turn their heads, move their
eyes, fix them on an object, track an object
as it moves, kick their legs, and wave their
SUBJECTIVITY
arms - among other things. When they do
these things, their experiences alter.
The basic distinction neonates are able to
make, then, is that of in-control/not-in-
control. The dichotomy is an either-or one:
either this (kind of) cause of experience (an
in-control experience - an action) or that
(kind of) cause of experience (a not-in-
control experience - not an action). This
primitive distinction allows the infant to
begin dividing the stream at first into
tokens. (These parts of the stream were in-
control; those parts, not-in-control.) Equally
important, the in-control/not-in-control dis-
tinction is also the basis for the first type
distinction, which I will call the me/not-me
(or self/not-self) distinction, though the
name is somewhat misleading, because at
this point the 'me' is not exactly conceived
of as an individual thing, a self. We can
think of it as a kind of proto-self (and there
is evidence in the developmental literature
for this proto-self). The not-me is also not
yet individualized, but it can be thought of
as the proto-external-world (the proto-out-
there). It is crucial to emphasize that the in-
control/not-in-control distinction underlies
both one's primitive concepts of one's very
self and of an external world. It is only
because one takes oneself to successfully
will certain behaviours, but not others, that
this primitive (and primary) distinction of
self and other can be made. All concept for-
mation begins with this in-control/not-in-
control distinction and our introspective
awareness of it.
To continue the story, we need to focus
on the not-self. The not-self, as cause of
those experiences that are not-in-control,
gets broken up, by whatever mechanisms,
into objects: bodies. And those bodies are
categorized into types of objects. Only after
bodies are differentiated do we come to
recognize our self as an individual thing: a
body inhabiting the same world as all other
bodies, yet different, as a token, from all
other bodies. Perception itself also leads to
this same notion of a bodily self (see Gibson,
1979). We first identify our individual self
as a body, one that not only has shape, size,
and so forth, but also is a subject of experi-
571
SUBJECTIVITY
ence: one that thinks - especially one that
wills. We then come to perceive that our
self, this particular body, is a member of a
type: human being. So we come thereby to
ascribe subjectivity to all members of the
type, to all human beings (and also to many
other animals, though to human beings first
and foremost).
There is much evidence for this account
in the development literature. For instance,
infants seem to grasp at least a crude idea of
bodies very early in the first year of life, but
the idea of the mental is not evidenced until
the middle of the second year. When an
infant does grasp the mental as divisible
into types, it ascribes those states to itself
first, only later to others (and, then, first to
human beings; later, to other animals).
Although we ascribe the mental features of
human beings to our self first, we appar-
ently learn human body parts first of others,
and only later as ascribable to our self. That
is, only when we conceive of our self as the
same type as other human beings do we
begin to ascribe subjectivity to others.
On this account, underlying all concept
formation is an essential subjectivity: that
experiential distinction of the in-control from
the not-in-control, and an introspective
capacity that makes one aware of this dis-
tinction in one's own experience. But the
underlying, essential subjectivity is one of
will and introspection, not of phenomenal
states.
With this brief, Piagetian account as
background, there is a road that appears to
be open, one which leads swiftly to the
fundamental tension that Nagel (1979a)
finds between the subjective and the objec-
tive (though the 'road' represents only one
way of interpreting the Piagetian account).
An important feature of the above account
of concept formation is that we first under-
stand subjects of experience (thinking,
willing beings) - our own self and others -
as physical objects: objects in space and
time. The mind/body division comes only
later. As we come to realize that the body is
often outside our control - as in disease, or
when a bodily part is asleep, or when we
are frozen with fear, and so on - we are
572
tempted to identify the self only with our
mental activities, especially with our will.
Inherent in these loss-of-control experiences
is an increased recognition of how much is
out of our control, of how even what we
believed to be in our control is instead at
the mercy of external forces.
Finally, we come to realize that our very
mind is often out of our control. Despite our
best efforts, we cannot stop thinking about
someone or something; despite our strong
wishes not to, we cannot help eating a
second pastry; and so forth. At this point we
begin to suspect that our very self, however
we identify it, is not in our control, that we
are always in subjection to the not-self, that
our sense of being in control is only an illu-
sion. When we view ourselves from the
objective, impartial world of science, we
seem to be influenced, both in action and
thought, by external causes as much as is
anything we label an object.
In initially identifying our self, we relied
on a sense of control that made it appear
that our actions originate entirely within
ourselves. This sense was not one of omni-
potence (as Freud thought): without a
recognition of the not-in-control, we would
not have had a concept of self at all. Rather
it was a sense of total freedom in those
actions over which we took ourselves to
have control. But at some point we discover
that we are not, and never have been, the
sole and unfettered origin of our own beha-
viours. The external world hinders/helps
our actions, it affects our very thoughts,
and it influences or thwarts our very will
itself. The sense of control, of autonomy,
looks to be mere illusion.
But if the key ingredient - the in-control
- is illusory, what can be said of all the
crucial distinctions based on it? If we can be
so wrong about that very foundation for
making all distinctions, if our sense of
control is not itself based in fact, then how
can we be certain - how can we be any-
thing but uncertain - about the resulting
distinctions themselves? Why shouldn't illu-
sion just give rise to further illusion? It all
begins to slide away. All the distinctions
dependent on our sense of control - all our
distinctions - seem threatened: that there is
a community we are part of; our very self;
our very world. No wonder the problem of
free will is so disturbing.
This essential subjectivity - the sense of
control - underlies our 'discovery' not only
of self but also of the objective. Moreover,
and ironically, objective science - in this
case, developmental psychology - provides a
basis for believing that this essential sub-
jectivity itself exists. Yet, at the same time,
objective science seems to tell us that such a
subjectivity is an illusion. Science tells us
that no causes can be autonomous in the
sense required for the in-control/not-in-
control distinction to be genuine. But if the
subjectivity that underlies all our distinc-
tions is itself called into question, then our
hold on both the subjective and the objective
is loosened, for the path to the objective is
by way of the subjective. It is only because
of the in-control (the subjective) that we
ever come to recognize the not-in-control
(the objective world), but then 'objectively'
we come to doubt the subjective altogether.
There appears to be no resting place. Our
only basis for believing in the objective is
the subjective, and the objective tells us the
subjective is illusory. It is this tension that
Nagel. perhaps more than anyone else since
Kant, has realized and called to our atten-
tion.
If there is no solution to these problems,
then it appears that there really is a funda-
mental tension in our lives: the subjective
leading inevitably to the objective, and the
objective casting doubt on the autonomy of
the essential subjectivity that made distin-
guishing the objective possible.
Is there any solution to this problem?
Certainly attempts have been made to solve
it (Nagel just recognizes it, but makes no
real attempt to solve it). Kant tried to solve
the problem by putting the real subject of
experience in a different world altogether:
the transcendent, non-empirical world. For
Kant, the empirical self is an object like all
other objects. It is not the subject whose
subjectivity underlies all distinctions, the
one that makes experience of things possible.
Buddhism may be seen as another sort of
SUBJECTIVITY
attempt to resolve the problem. The aim of
meditation is to 'escape' into that pure sub-
jectivity that precedes all distinction, that
precedes any notion of the objective, but
also precedes any notion of the subjective,
of the self. The aim of Buddhist meditation
is to eliminate the self, to eliminate the sub-
jective/objective distinction itself.
Another, more familiar, solution to the
problem is to defend some sort of soft deter-
mmism (compatibilism). Compatibilists
argue that the problem is a spurious one,
the result of identifying the crucial notion of
'in-control' with a not-understandable
notion of 'autonomy'. Compatibilists claim
that a more realistic notion of 'in-control'
will do the Piagetian job of making concept
formation explainable, without getting us
into all the apparent metaphysical entangle-
ments that Nagel. Kant, and Buddhism each
claims to see.
Whether Kant, Buddhists, or compati-
bilists actually resolve the paradox seen by
Nagel (who doesn't put the paradox in quite
the way I have), is beyond the scope of this
entry to evaluate. But the notion of sub-
jectivity discussed in this section, the gist of
which remains whether Nagel or one of the
others is right, puts much of the history of
philosophy into clear relief - in addition to
providing an apparently coherent and
evidence-sensitive theory of concept forma-
tion.
One's realization of oneself as a subject of
experience, based on the idea of the in-
control. also makes .understandable why
many philosophers have identified persons
with their subjectivity. Persons, whatever
else they are, seem to be enduring things. If
the essential subjectivity that makes a dis-
tinction of self possible is rooted in the will,
then that subject of experience must endure
in time. To be in-control is to have one's
plans (and intentions) for the future result
as one planned (or intended), for one to
remember - when the actions do result as
planned (or intended) - that one had
planned (or intended) these results, and so
on. The notion of 'in-control' is a time-
encompassing notion. The person in-control
must be a thing which persists through the
573
SUBJECTIVITY
relevant time-span (or be a construction of
things which participate in this process - as
suggested by Locke, and defended by Parfit
(1984), who claims that persons are con-
structions out of person-stages). Since one's
subjectivity persists, it is natural to make
the identification of self and subjectivity.
Yet, even if subjectivity makes possible
our conceiving of our self in the first place,
it may still be a mistake to identify persons
with that subjectivity rather than as bodily
things (I think it is a mistake). While this
topic has been widely investigated, like most
philosophical questions, it bears con-
siderably more investigation.
POINTS OF VIEW
The third notion of subjectivity, as Nagel
again sees, is the notion of a point of view,
what psychologists call a constructivist
theory of mind (Chandler, 1988). Undoubt-
edly, this notion is closely tied to the notion
of essential subjectivity, but it is worth dis-
cussing in its own right. This kind of sub-
jectivity is constituted by an awareness of
the world's being experienced differently by
different subjects of experience. (It is thus
possible to see how the privacy of phenom-
enal experience might be easily confused
with the kind of privacy inherent in a point
of view.)
Point-of-view subjectivity (PVS) seems to
take time to develop. The developmental
evidence suggests that even toddlers are
able to understand others as being subjects
of experience. For instance, at a very early
age, we begin ascribing mental states to
other things - generally to those same
things to which we ascribe 'eating' (Carey,
1985). And at quite an early age we can
say what others would see from where they
are standing, Le. we early on demonstrate
an understanding that the information
available is different from different percep-
tual pOints of view, and so to different per-
ceivers. It is in these perceptual senses that
we first ascribe PVS to others.
However, some experiments seem to show
that the PVS then ascribed to others is
limited. A well-known and influential series
574
of experiments by Wimmer and Perner
(1983) is usually taken to illustrate these
limitations (though there are disagreements
about the interpretation of these and the
other experiments to be mentioned). Two
children - Dick and Jane, say - watch as an
experimenter puts a box of candy some-
where, such as in a cookie jar, which is
opaque. Jane leaves the room. Dick is asked
where Jane will look for the candies, and he
correctly answers, 'In the cookie jar'. The
experimenter, in Dick's view, then takes the
candy out of the cookie jar and puts it in
another opaque place, a drawer, say. When
Dick is asked where to look for the candy, he
says quite correctly, 'In the drawer'. When
asked where Jane will look for the candy
when she returns, Dick answers, 'In the
drawer'. Dick ascribes to Jane, not the PVS
she is likely to have, but the one that fits the
facts. Dick is unable to ascribe to Jane false
beliefs - his ascription is 'reality driven' -
and his inability demonstrates that Dick
does not yet have a fully developed PVS.
At around the age of four, children in
Dick's position do ascribe the likely PVS to
children in Jane's position (Le. 'Jane will
look in the cookie jar'); but, even so, a fully
developed notion of a PVS is not yet
attained. Suppose that Dick and Jane are
shown a dog under a tree, but only Dick is
shown the dog's arriving there by chasing a
boy up the tree. If Dick is asked to describe
what Jane, who he knows not to have seen
the chase, will describe when seeing the dog
under the tree, Dick will display a more
fully developed PVS only if his description
will not entail the preliminaries that only he
witnessed. It turns out that four-year-olds,
despite passing the Wimmer-Perner test, are
unable to pass role-playing tests such as this
one. Only when children are six to seven do
they succeed.
Yet, even when successful in these cases,
children's PVS is reality-driven. Ascribing a
PVS to others is still in terms relative to
information available. Only in our teens do
we seem capable of understanding that
others can view the world differently from
ourselves, even when given access to the
same information. Only then do we seem to
become aware of the subjectivity of the
knowing procedure itself: interpreting the
'facts' can be coloured by one's knowing
procedure and history. There are no
'merely' objective facts.
Thus, there is evidence that we ascribe a
more and more subjective point of view to
others: from the PVSs we ascribe being
completely reality-driven, to the possibility
that others have insufficient information, to
their having merely different information,
and, finally, to their understanding the
same information differently. This develop-
mental picture seems insufficiently familiar
to philosophers - and yet well worth our
thinking about and critically evaluating.
The following questions all need answer-
ing. Does the apparent fact that our PVS
ascriptions to others develop over time,
becoming more and more of a 'private'
notion, shed any light on the sort of sub-
jectivity we ascribe even to our own self?
Do our self-ascriptions of subjectivity them-
selves become more and more 'private',
more and more removed both from the sub-
jectivity of others and from the objective
world? If so, what is the philosophical
importance of these facts? At the least, this
developmental history shows that disen-
tangling our self from the world we live in is
a complicated matter.
CONCLUSION
Even this brief survey makes clear that
'subjectivity' is not the name of a single
concept, but names a set of intertwining
and overlapping concepts. I have discussed
three notions: phenomenality, will, and
point of view. And three claims are worth
re-emphasizing. (1) Nagel's slogan - that
there is something it is like to be in a sub-
jective state - applies only to phenomen-
ality. (2) The subjectivity associated with a
sense of control is the deepest sort of sub-
jectivity, underlying all concept formation,
including one's concept of one's very self.
(3) Finally, experimentation seems to show
that the PVS we ascribe to others becomes
more and more sophisticated, more and
more subjective.
SUPER VENIENCE
BIBLIOGRAPHY
Carey, S. 1985. Conceptual Change in Childhood.
Cambridge, MA: MIT Press/Bradford Books.
Chandler, M. 1988. Doubt and developing
theories of mind. Developing Theories of Mind,
ed. J. W. Astington, P. 1. Harris, and D. R.
Olson. Cambridge University Press.
Gibson, J.J. 1979. The Ecological Approach to
Vision. Boston: Houghton-Mifflin; reprinted
1986, Hillsdale, N.J.: Lawrence Erlbaum.
Nagel, T. 1974. What is it like to be a bat?
Philosophical Review, 83, 435-50.
Nagel, T. 1979a. Mortal Questions. Cambridge
University Press.
Nagel, T. 1979b. Moral luck. In Nagel, 1979,
pp.24-38.
Nagel, T. 1986. The View from Nowhere.
Nelkin, N. 1989. Propositional attitudes and
consciousness. Philosophy and Phenomen-
ological Research, 49, 413-30.
Neikin, N. 1993. What is consciousness? Phi-
losophy of Science, 60, 419-34.
Parfit, D. 1984. Reasons and Persons. Oxford:
Clarendon Press.
Piaget, J. 1954. The Construction of Reality in
the Child, trans. Margaret Cook. New York:
Basic Books.
Piaget, J., and Inhelder, B. 1969. The Psychol-
ogy of the Child, trans. H. Weaver. New
York: Basic Books.
Wimmer, H., and Perner, J. 1983. Beliefs
about beliefs: representation and constrain-
ing the function of wrong beliefs in young
children's understanding of deception. Cog-
nition, 13, 103-28.
would like to thank Edward Johnson,
Carolyn Morillo, and Samuel Guttenplan for
comments made on earlier drafts of this entry.
NORTON NELKIN
supervenience During the past two
decades or so, the concept of supervenience
has seen increasing service in philosophy of
mind. The thesis that the mental is super-
venient on the physical- roughly, the claim
that the mental character of a thing is
wholly determined by its physical nature -
575
SUPER VENIENCE
has played a key role in the formulation of
some influential positions on the MIND-
BODY PROBLEM, in particular versions of
non-reductive PHYSICALISM. Mind-body
supervenience has also been invoked in
arguments for or against certain specific
claims about the mental, and has been used
to devise solutions to some central problems
about the mind - for example, the problem
of mental causation (see PROPOSITIONAL
ATTITUDES; REASONS AND CAUSES). There
are also questions about mind-body super-
venience itself, such as exactly how it ought
to be formulated, what its relationship is to
mind-body reduction, whether certain
types of mental states, such as 'QUALIA' and
intentional states, are plausibly regarded as
physically supervenient, and what the
implications are for physicalism if they fail
to supervene.
The idea of supervenience is usually
thought to have originated in moral theory,
in the works of such philosophers as G. E.
Moore and R. M. Hare (see below, however,
on emergentism). Hare, for example,
claimed that ethical predicates are 'super-
venient predicates' in the sense that no two
things (persons, acts, states of affairs) could
be exactly alike in all descriptive or natur-
alistic respects but unlike in that some
ethical predicate ('good', 'right', etc.) truly
applies to one but not to the other. That is,
there could be no difference in a moral
respect without a difference in some descrip-
tive, or non-moral, respect. Evidently the
idea is generalizable so as to apply to any
two sets of properties (to secure greater gen-
erality it is more convenient to speak of
PROPERTIES than predicates). Donald
DAVIDSON (1970) was perhaps first to
introduce supervenience into discussions of
the mind-body problem, when he wrote:
' ... mental characteristics are in some
sense dependent, or supervenient, on phys-
ical characteristics. Such supervenience
might be taken to mean that there cannot
be two events alike in all physical respects
but differing in some mental respect, or that
an object cannot alter in some mental
respect without altering in some physical
respect.' Following Moore and Hare, from
576
whom he avowedly borrowed the idea of
supervenience, Davidson went on to assert
that supervenience in this sense is con-
sistent with the irreducibility of the super-
venient to their 'subvenient', or 'base',
properties: 'Dependence or supervenience of
this kind does not entail reducibility
through law or definition ... '
Thus, three ideas have come to be closely
associated with supervenience: (1) property
covariation (if two things are indiscernible in
base properties, they must be indiscernible
in supervenient. properties); (2) Dependence
(supervenient properties are dependent on,
or determined by, their subvenient bases);
and (3) non-reducibility (property covariation
and dependence involved in supervenience
can obtain even if supervenient properties
are not reducible to their base properties).
Hellman and Thompson (1975), appar-
ently independently of earlier writers, pro-
posed a model-theoretic analysis of
determination, in terms of which they for-
mulated a version of physicalism free of
reductionist commitments. The central
thesis of their physicalism is the claim that
the physical facts of the world 'determine'
all the facts, including facts concerning the
mental. Interestingly, their concept of deter-
mination turns out to be a form of what is
now known as 'global supervenience' (see
below), although they did not use the termi-
nology of 'supervenience' and seem not to
have been aware of the historical ante-
cedents of their approach.
These writers, however, were not the first
to use the idea of supervenience, or the
term 'supervenience', in connection with
the mind-body problem. The emergentists,
who were active during the first half of this
century, especially in Britain, sometimes
used 'supervenient' as a stylistic variant of
'emergent' (especially Lloyd Morgan, 1923),
and there is a striking Similarity between
emergence and supervenience. According to
emergentism, higher-level properties,
notably CONSCIOUSNESS and other mental
properties, emerge when, and only when,
an appropriate set of lower-level 'basal con-
ditions' are present, and this means that the
occurrence of the higher properties is deter-
mined by. and dependent on. the instantia-
tion of appropriate lower-level properties
and relations. In spite of this. emergent
properties were held to be 'genuinely novel'
characteristics irreducible to the lower-level
processes from which they emerge. Clearly.
then. the concept of emergence combines
the three components of supervenience de-
lineated above. namely property covariance.
dependence. and non-reducibility. In fact.
emergentism can be regarded as the first
systematic formulation of non-reductive
physicalism.
TYPES OF PROPER TY COV ARIATION
The idea that mentality supervenes on
physical-biological nature arises quite
intuitively. Suppose that we could build an
exact physical replica of you. a creature
who is molecule-for-molecule. cell-for-cell.
indistinguishable from you. (It would make
no difference if your replica had been natu-
rally created.) Would this creature. who is
physically indiscernible from you. be also
psychologically indiscernible? To put it
another way: are physical duplicates neces-
sarily psychological duplicates as well? If you
are inclined to answer yes. that means that
you are disposed to accept mind-body
supervenience: no two things (organisms.
events. etc.) could differ in a mental respect
unless they differed in some physical respect
- that is. indiscernibility with respect to
physical properties entails indiscernibility
with respect to mental properties. That is
the core idea of mind-body supervenience.
Recent work has shown. however. that
this core idea can be explicated in various
non-equivalent. but interestingly related.
ways. yielding supervenience claims of
varying strengths (Kim. 1984. 1987). In
what follows. we will assume that our
domain consists of organisms and other
structures of interest from the psychological
point of view (which could include electro-
mechanical systems. parts of whole organ-
isms. etc.). and that two sets of properties
are defined over this domain. one. M. con-
sisting of mental properties and the other. P.
consisting of physical properties (in the
SUPER VENIENCE
broad sense that includes biological as well
as phYSicochemical properties). If we wish.
we could think of 'states' and 'events' as an
organism's (or structure's) instantiating. or
changing in respect of. one or more of these
properties at a time. Thus, the super-
venience of events and states can be
explained in terms of property super-
venience. In any case. what is it for M to
supervene on P over our domain (henceforth
we will often delete reference to the
domain)?
We may begin with the following:
Weak supervenience: Necessarily (that is.
in every possible world). if any x and y
(in the domain) are indiscernible in P
(,P-indiscernible' for short). x and yare
M-indiscernible.
On weak mind-body supervenience. then.
no possible world contains two individuals
which are alike in respect of properties in P
but unlike in some mental property in M. It
is clear that if P were so comprehensive
(e.g. it includes spatiotemporal locations as
physical properties) that no two individuals
in the domain could have exactly the same
P-properties. weak supervenience would
hold trivially. no matter what the super-
venient properties are. To generate a mean-
ingful thesis of weak supervenience.
therefore. P may be appropriately circum-
scribed (e.g. to biological properties. compu-
tational properties. etc.).
Suppose you are creating worlds: weak
supervenience prohibits you from placing in
the same world physical duplicates that are
not mental duplicates. But it does not prohi-
bit you from creating two physical dupli-
cates that are not mental duplicates as long
as you put them in different worlds. The dis-
tinctive feature of weak supervenience is
that its constraint applies only intra-world.
not cross-world; that is. the way in which
mental and physical properties are dis-
tributed in one world places no restriction
whatever on how they may be distributed
in another world. In fact. weak mind-body
supervenience permits: (1) a world which is
exactly like the actual world in all physical
577
SUPERVENIENCE
respects but which is totally devoid of men-
tality; (2) worlds that are physically just like
our world but in which everything is con-
scious in exactly the same way; and (3)
worlds that, again. are physically indis-
tinguishable from our world but in which
unicellular organisms. but no humans or
other higher animals. are conscious. The
conclusion seems unavoidable. then. that
weak supervenience is not strong enough to
support the kind of mind-body dependence
demanded by robust physicalism. Physical-
ism must require at least this much: phys-
ical facts of a world determine all the facts of
that world. This means that once physical
properties have been distributed over the
individuals in a certain way. that leaves but
one way to distribute psychological proper-
ties. But that is precisely what is not
required by weak supervenience.
Consider then the following two theses of
mind-body supervenience:
Global supervenience: Any two worlds
that are indiscernible with respect to P
(i.e. worlds in which physical properties
are distributed over the individuals in
the same way) are indiscernible with
respect to M (that is. they cannot differ
in how mental properties are dis-
tributed).
Strong supervenience: For any individuals
x and y, and any worlds Wi and Wk. if x
in Wi is P-indiscernible from y in Wk
(that is. x has in Wi exactly the same
properties in P that y has in wd. then x
in Wi is M-indiscernible from y in Wk.
Global supervenience applies indiscernibility
considerations globally. to whole worlds
taken as units rather than to individuals
within worlds. and requires that worlds that
are indistinguishable from the physical
point of view do not differ from the mental
point of view. although of course worlds
that are alike mentally could differ in physi-
cal respects. This evidently explicates one
clear sense in which the mental character of
a world can be said to depend on. or be
determined by. its physical character.
578
Strong supervenience. unlike global
supervenience. applies indiscernibility con-
siderations 'locally'. to individuals, not
directly to whole worlds. It says of any indi-
viduals from anywhere in the panoply of all
possible worlds that they cannot differ psy-
chologically without differing physically.
Strong supervenience. therefore. differs from
weak supervenience in that individuals com-
pared for indiscernibility may be recruited
from different worlds. whereas on weak
supervenience individuals are compared
only as they are within the same world. This
is why the constraint of strong super-
venience. unlike that of weak supervenience.
applies cross-world as well as intra-world.
It is easily seen that strong supervenience
entails both global and weak supervenience.
and that weak supervenience entails neither
strong nor global supervenience (Kim.
1984. 1987). Does global supervenience
entail strong or weak supervenience? It for-
mally implies neither (Kim. 1987; Paull and
Sider. 1992), but this does not fully settle
the issue. since there may be plausible
metaphysical premises which together with
global supervenience yield strong super-
venience (Paull and Sider. 1992; Kim.
1993). Moreover, it has been shown (Paull
and Sider. 1992) that when restricted to
'intrinsic properties'. strong and global
supervenience are fully equivalent.
SUPERVENIENCE AND NON-REDUCTIVE
PHYSICALISM
For various reasons. psychophysical reduc-
tionism. or type physicalism. had lost favour
with philosophers by the early 1970s. and
many physicalists looked for a way of char-
acterizing the primacy and priority of the
physical that is free from reductionist impli-
cations. As we saw with Davidson and
Hellman and Thompson. the key attraction
of supervenience to physicalists has been its
promise to deliver dependence without reduc-
tion. Here. the example of moral theory has
seemed encouraging: Moore and Hare, who
made much of the supervenience of the
moral on the naturalistic. were. at the same
time. strong critics of ethical naturalism. the
principal reductionist position in ethical
theory. And there has been a broad con-
sensus among ethical theorists that Moore
and Hare were right, that the moral. or
more broadly the normative, is supervenient
on the non-moral without being reducible
to it.
The issue is complicated, however. For
one thing, it is possible that Moore and Hare
were just wrong about the relationship
between supervenience and reducibility. For
another, the issue surely depends on the
kind of supervenience involved and what
one understands by 'reduction'. It may well
be that the sense of reduction appropriate
for the mind-body case differs crucially from
the kind of a priori definitional reduction
Moore and other ethical theorists apparently
had in mind; and the supervenience rela-
tions involved in the two cases may be dis-
similar in some respect that makes a
difference to the question of reduction.
Most anti-reductionist arguments con-
cerning the mental focus on the availability
of type-type correlations between the mental
and the physical - that is, correlations
between mental and physical properties. The
heart of these arguments is that psycho-
physical reduction requires such correla-
tions, but that they are unavailable. But
what sorts of correlations are required for
reduction? Logical BERA VIOURISM tried to
provide each mental predicate with an
analytically equivalent definition framed
exclusively in terms of physical-behavioural
expressions; so the behaviouristic reduction
was to be implemented by providing every
mental property with an analytically
equivalent physical property, a project aban-
doned as entirely hopeless several decades
ago. The demise of logical behaviourism has
led would-be reductionists to look to the
model of nomological reduction, which was
thought to characterize inter-theoretic
reductions in empirical science (e.g. thermo-
dynamics to statistical mechanics, optics to
electromagnetic theory). Mind-body reduc-
tion, on this model. would consist in the
derivation of psychological laws from those
of some underlying physical theory (pre-
sumably, neurobiology) taken together with
SUPER VENIENCE
'bridge laws', empirical laws correlating
mental kinds with physical-neural kinds.
The idea was that reduction could be carried
out by finding for each mental property a
nomologically coextensive physical property,
rather than an analytically coextensive one.
In any case, the idea that mind-body reduc-
tion must be underwritten by a pervasive
system of type-type correlations between the
mental and the physical has been a tacit but
widely shared presupposition of the recent
debate on reductionism. This requirement
can be stated as follows:
The requirement of strong connectibility:
For each mental property M there is a
physical property P such that neces-
sarily, M is instantiated by a system at t
if and only if P is instantiated by it at t.
Definitional reduction and nomological
reduction differ from each other only in that
different senses of 'necessarily' are involved,
analytic or logical necessity in the former
and nomological necessity in the latter.
This means that the question whether
supervenience is consistent with the irredu-
cibility of the mental amounts to the ques-
tion whether it is consistent with the failure
of strong connectibility between the mental
and the physical. But which supervenience
relation do we have in mind? It is clear that
weak supervenience is not a candidate, for
it is unable to generate psychophysical type-
type correlations with an appropriate modal
force of necessity. This leaves strong and
global supervenience. It will be convenient
to make use of the following alternative for-
mulation of strong supervenience:
Strong supervenience (alternative
version): Necessarily, for each property
M in M, if anything x has M, then
there is a property P in P such that x
has P, and necessarily if anything has P
it has M.
When a thing has a mental property, M,
strong supervenience calls for a subvenient
physical base, P, for M, and the relationship
between P and M carries over into other
579
SUPERVENIENCE
possible worlds, as indicated by the italicized
inner modal operator 'necessarily'. This
stability of psychophysical type-type correla-
tions across possible worlds is, as may be
recalled, precisely what was lacking in weak
supervenience. The exact force of 'neces-
sarily' is open to further specification: some
might want only nomological necessity, but
others might insist on metaphysical neces-
sity or even analytic necessity. (This version
of strong supervenience is sometimes called
'the operator formulation', to contrast it
with the earlier 'possible-world formula-
tion'. The two formulations are provably
equivalent under certain assumptions con-
cerning property composition (Kim, 1987).
Under strong psychophysical super-
venience, therefore, every mental property
M has a (possibly infinite) series of physical
properties P
lo
P
20
, each of which is suf-
ficient for M. Consider then the union (dis-
junction) of these Ps, UP
i
. It is easily seen
that uP
i
is a necessary coextension of M.
Does this mean that strong supervenience
entails the satisfaction of the strong con-
nectibility requirement? Could uP
i
serve as
a reduction base for M?
It can scarcely be denied that strong super-
venience, at least in its alternate version,
entails the strong connectibility ofthe mental
with the physical. The anti-reductionist
physicalist can, however, respond in one of
the following two ways: first, she might
argue that the strong connectibility is not
sufficient for reduction, and, in particular,
that properties like uP
i
, because oftheir com-
plexity and artificiality, are unsuited as a
reduction base for mental properties; second,
she might look to global supervenience to
secure psychophysical dependence without
the threat of reductionism.
Neither of these moves is entirely unprob-
lematic, however. Consider the first move: it
has never been convincingly argued why
'complexity' and 'artificiality' of uP
i
should
count against it as a reduction base. The
point is often made to the effect that the
complexity and open-ended heterogeneity of
UP
i
makes it entirely unusable in scientific
theory reduction; it may not even be pos-
sible to state, in a finite way, bridge laws
580
involving such properties. It may be replied,
though, that the alleged complexity and
heterogeneity are primarily characteristic of
how UP is represented or specified, not
necessarily of the property uP
i
itself, and
that as long as this property is there, there
is always the possibility of its being repre-
sented by a perspicuous description in an
appropriate scientific theory. Even if this
never happens and, indeed, cannot happen
given our cognitive capacities or inclina-
tions, why isn't the demonstrated existence
of UP
i
for M, and similar physical coexten-
sions for all other mental properties, suffi-
cient to show the metaphysical reducibility of
mental properties? Obviously, it is not the
business of a philosophical argument to
generate actual reductions of scientific the-
ories or properties, but only to show their
metaphysical possibility.
Could global supervenience then help the
non-reductivist? Some non-reductive phys-
icalists find this form of supervenience parti-
cularly attractive because, unlike strong
supervenience, it does not posit, at least not
directly, type-type relationships between the
mental and the physical (Post, 1987). But,
as we saw, the question whether global
supervenience entails strong supervenience
has not been fully settled. When restricted
to intrinsic properties, global and strong
supervenience turn out to be equivalent.
and it appears that global supervenience
fails to entail strong supervenience only if
mental properties, but not the subvenient
physical properties, are allowed to include
non-intrinsic properties. Moreover, it should
be noted that in the sense in which global
supervenience does not imply strong super-
venience, it does not imply weak super-
venience either (Kim, 1987). And we may
very well doubt whether a supervenience
relation which permits the violation of weak
supervenience could yield a robust enough
dependence relation adequate for physical-
ism.
DOES MIND-BODY SUPER VENIENCE HOLD?
Much of our evidence for mind-body super-
venience seems to consist in our knowledge
of specific correlations between mental
states and physical (in particular, neural)
processes in humans and other organisms.
Such knowledge, although extensive and in
some ways impressive, is still quite rudi-
mentary and far from complete (what do we
know, or can we expect to know, about the
exact neural substrate for, say, the sudden
thought that you are late with your rent
payment this month?). It may well be that
our willingness to accept mind-body super-
venience, although based in part on specific
psychophysical dependencies, has to be sup-
ported by a deeper metaphysical commit-
ment to the primacy of the physical; it may
in fact be an expression of such a commit-
ment.
But there are kinds of mental states that
raise special issues for mind-body super-
venience. One such kind is 'wide content'
states, i.e. contentful mental states that
seem to be individuated essentially by refer-
ence to objects and events outside the
subject (see CONTENT; PROPOSITIONAL
ATTITUDES; TWIN EARTH). You believe that
water is wet, but your physical duplicate on
Twin Earth (where a substance, XYZ, obser-
vationally indistinguishable from H
2
0, fills
the lakes and oceans, comes out of the tap,
etc.) believes that XYZ is wet, not that
water is wet. It would seem then that the
belief that water is wet does not supervene
on the physical states of organisms. The
same considerations evidently apply to all
wide-content states.
However, what this shows is only the
failure of 'local supervenience', i.e. the
supervenience of wide-content states on the
local. intrinsic physical-neural properties of
a given subject. It does not show that these
states fail to supervene physically tout court;
in order to secure supervenience we need to
broaden the supervenience base, to include
the relational. or extrinsic, physical proper-
ties of the subject, including historical-
causal properties (e.g. having matured in an
environment including XYZ but no water).
What this implies about the causal powers
of wide-content states is a question that has
been intensely debated in recent years.
Another type of mental states that has
SUPER VENIENCE
raised doubts about mind-body super-
venience are 'phenomenological' or 'quali-
tative' states or events ('QU ALIA' for short).
These are mental states with sensory quali-
ties, like pains and itches, seeing of green,
etc. Many philosophers believe that 'qualia
inversion' is perfectly conceivable - that is,
it is conceivable and perfectly intelligible
that when you look at a ripe tomato, the
colour you sense is the colour your physical
duplicate senses when she looks at a fresh
cucumber, and vice versa (see CON-
SCIOUSNESS). Now, if such a possibility is
coherently conceivable, as it seems to be,
that could defeat supervenience, both strong
and global. (The details here will depend on
the specific type of supervenience claim
involved - in particular, its modal force.) If
you take the view that not only is qualia
inversion coherently conceivable, but there
is no way to know that it doesn't occur all
around us (even if we knew all about
human neurophysiology), you are taking
the position that there is no good evidence
for weak supervenience either.
Many physicalists take mental properties
to be 'physically realized' - in fact, realized
by a multiple set of physical properties. The
idea that mental properties are multiply rea-
lized physically is very closely related to the
idea that they are physically supervenient.
There seems to be no single concept of
'realization'; however, those who use the
idiom of 'realization' will agree that physical
duplicates realize exactly the same set of
mental properties - except possibly wide-
content properties noted above. The idea
that mentality is physically realized is inte-
gral to the FUNCTIONALIST conception of
mentality, and this commits most function-
alists to mind-body supervenience in one
form or another.
PSYCHOPHYSICAL SUPERVENIENCE AS A
THEOR Y OF MIND
Supervenience of the mental - in the form
of strong supervenience, or at least global
supervenience - is arguably a minimum
commitment of physicalism. But can we
think of the thesis of mind-body super-
581
SUPER VENIENCE
venience itself as a theory of the mind-body
relation - that is, as a solution to the mind-
body problem?
As we saw, a supervenience claim con-
sists of a claim of covariance and a claim of
dependence (leaving aside the controversial
claim of non-reducibility). This means that
the thesis that the mental supervenes on the
physical amounts to the conjunction of
the two claims: (1) the mental covaries with
the physical (a la strong or global super-
venience), and (2) the mental depends on
the physical. Notice, however, the fact that
the thesis says nothing about just what kind of
dependence is involved in mind-body super-
venience. When you compare the super-
venience thesis with the standard positions
on the mind-body problem, you are struck
by what the supervenience thesis doesn't
say. For each of the classic mind-body the-
ories has something to say, not necessarily
anything very plausible, about the kind of
dependence that characterizes the mind-
body relationship. According to epipheno-
menalism, for example, the dependence is
one of causal dependence; on logical beha-
viourism, dependence is rooted in meaning
dependence, or definability; on the standard
type physicalism, the dependence is one
that is involved in the dependence of macro-
properties on micro-structural properties;
etc. Even Leibniz and Malebranche had
something to say about this: the observed
property covariation is due not to a direct
dependency relation between mind and
body but rather to divine plans and inter-
ventions. That is, mind-body covariation
was explained in terms of their dependence
on a third factor - a sort of 'common cause'
explanation.
It would seem that any serious theory
addressing the mind-body problem must
say something illuminating about the
nature of psychophysical dependence, or
why, contrary to common belief, there is no
dependence here either way. However,
there is reason to think that 'supervenient
dependence' does not signify a special type
of dependence relation. This is evident when
we reflect on the varieties of ways in which
we could explain why the supervenience
582
relation holds in a given case. For example,
consider the supervenience of the moral on
the descriptive: the ethical naturalist will
explain this on the basis of definability; the
ethical intuitionist will say that the super-
venience, and also the dependence, here is a
brute fact you discern through moral intui-
tion; and the prescriptivist will attribute the
supervenience to some form of consistency
requirement on the language of evaluation
and prescription. And distinct from all of
these is mereological supervenience, namely
the supervenience of properties of a whole
on properties and relations of its parts.
What all this shows is that there is no
single type of dependence relation common
to all cases of supervenience; supervenience
holds in different cases for different reasons,
and does not represent a type of dependence
that can be put alongside causal depen-
dence, meaning dependence, mereological
dependence, etc.
If this is right, the supervenience thesis
concerning the mental does not constitute
an explanatory account of the mind-body
relation, on a par with the classic alter-
natives on the mind-body problem. It is
merely the claim that the mental covaries in
a systematic way with the physical, and
that this is due to a certain dependence
relation yet to be specified and explained. In
this sense, the supervenience thesis states
the mind-body problem rather than offers a
solution to it.
There seems to be a promising strategy
for turning the supervenience thesis into a
more substantive theory of mind, and it is
this: to explicate mind-body supervenience
as a special case of mereological super-
venience - that is, the dependence of the
properties of a whole on the properties and
relations characterizing its proper parts.
Mereological dependence does seem to be a
special form of dependence that is meta-
physically sui generis and highly important.
If one takes this approach, one would/have
to explain psychological properties as
macroproperties of a whole organism that
covary, in appropriate ways, with its micro-
properties, i.e. the way its constituent
organs, tissues, etc. are organized and func-
tion. This more specific supervenience thesis
may well be a serious theory of the mind-
body relation that can vie with the classic
options in the field.
See also; CONTENT; IDENTITY THEORIES;
PHYSICALISM.
BIBLIOGRAPHY
Davidson. D. 1970. Mental events. In Experi-
ence and Theory. ed. 1. Foster and J. W.
setts Press.
Hellman. G .. and Thompson. F. 1975. Physic-
alism: ontology. determination. and reduc-
tion. Journal of Philosophy. 72. 551-64.
Horgan. T. 1982. Supervenience and micro-
physics. Pacific Philosophical Quarterly. 63.
29-43.
Kim. J. 1984. Concepts of supervenience. Phi-
losophy and Phenomenological Research. 65.
153-76.
Kim. J. 1987. 'Strong' and 'global' super-
venience revisited. Philosophy and Phenomen-
ological Research. 68. 315-26.
Kim. 1990. Supervenience as a philosophical
concept. Metaphilosophy. 21. 1-27.
Kim. J. 1993. Supervenience and Mind. Cam-
Lloyd Morgan. C. 1923. Emergent Evolution.
London: Williams and Norgate.
Paull. R.C .. and Sider. T.R. 1992. In defense of
global supervenience. Philosophy and Phe-
nomenological Research. 52. 833-54.
Post. J. 1987. The Faces of Existence. Ithaca:
Cornell University Press.
Teller. P. 1984. Poor man's guide to super-
venience and determination. Southern
Journal of Philosophy. 22 (The Spindel Con-
ference Supplement). 137-62.
JAEGWON KIM
syntax/semantics Consider the sentence.
'The cat is on the mat.' It is constructed
from six words. it has a subject and a pre-
dicate. two of its words are nouns. there is a
preposition and two definite articles as well
as the copula 'is'. All these are remarks
about the syntax of this sentence - its form.
One could know all of them and still not
know what the sentence expresses. what it
SYNTAX/SEMANTICS
means. The latter sort of knowledge is
seman tical.
The syntax/semantics distinction seems
straightforward. but there are deep issues in
linguistics and the philosophy of language
lying in wait to make things more complex.
First of all. though syntax is a matter of
form. there are many possible levels of such
form. Thus. knowing that a sentence is of
the subject-predicate sort is a fairly sophisti-
cated level of formal description; one must
know something about grammatical cat-
egories to appreciate it. Whereas saying of
the original sentence that it contains 16
letters and 5 spaces. or that it is composed
of certain kinds of black-an-white shapes
are descriptively no less formal. though they
can be appreciated without any background
grammatical knowledge.
However. the complications really mul-
tiply in respect of semantics. It is one thing
to say that the semantics of a sentence is its
meaning. it is another to say what meaning
is. or even to say how one would go about
describing the meaning of words or sen-
tences. Is it enough to say that the sentence
'The cat is on the mat' expresses the fact
that the cat is on the mat? On the one
hand. this seems uninformative - imagine it
was the sole explanation of the meaning of
this sentence. On the other hand. it is not
clear how to understand 'expresses the fact
that'.
Exactly what form a theory of meaning
should take. and what level of syntactical
description is most appropriate to under-
standing language. are problems for lin-
guists and philosophers of language. But the
notions of syntax and semantics also play
an important part in philosophy of mind.
This arises because it is widely maintained
that words and sentences are not the only
kinds of thing that have syntax and seman-
tics; in one way or another these features
have been claimed for mental phenomena
such as beliefs and other propositional atti-
tudes (see REPRESENT A TIONS). Thus. there
is a view known as the 'LANGUAGE OF
THOUGHT' theory which maintains that
beliefs are syntactically characterizable
items in the mind/brain and that they are
583
SYNTAX/SEMANTICS
semantically evaluable. According to this
account, we can best explain, for example,
Smith's belief that snow is white as his
having in his mind/brain a token of a lan-
guage of thought sentence - a sentence
with some kind of syntax - which has as a
semantical value the appropriate relation to
snow and whiteness. Also, many not com-
mitted to the idea of a language of thought
would still believe there to be a semantics of
attitude states. So, the very difficult issue of
584
how to describe the seman tical relation
carries over from the philosophy of lan-
guage to the philosophy of mind. It is often
called the 'problem of intentionality',
though this label covers other issues as well.
(See CONTENT; DRETSKE; FODOR; INTEN-
TIONALITY; PROPOSITIONAL ATTITUDES;
THOUGHTS.)
SAMUEL GUTTENPLAN
T
teleology Derived from the Greek word
'telos' meaning purpose or goal. 'teleology'.
as it is most often used in the philosophy of
mind. is thought of as the study of the pur-
poses. goals or. more broadly. biological
functions of various elements of the mental
realm. For example. it has been suggested
that we can better understand the PROPOSI-
TIONAL ATTITUDES when we have dis-
cerned their evolutionary function. It has
even been suggested that one can begin to
understand specific propositional attitude
contents in this way. (See CONTENT.)
SAMUEL GUTTENPLAN
thought The most significant feature of
thought is its INTENTIONALITY or
CONTENT: in thinking. one thinks about
certain things. and one thinks certain
things of those things - one entertains pro-
positions that stand for states of affairs.
Nearly all the interesting properties of
thoughts depend upon their content: their
being coherent or incoherent. disturbing or
reassuring. revolutionary or banal. con-
nected logically or illogically to other
thoughts. It is thus hard to see why we
would bother to talk of thought at all unless
we were also prepared to recognize the
intentionality of thought. So we are natur-
ally curious about the nature of content: we
want to understand what makes it possible.
what constitutes it. what it stems from. To
have a theory of thought is to have a theory
of its content.
Four issues have dominated recent think-
ing about the content of thought; each may
be construed as a question about what
thought depends on. and about the con-
sequences of its so depending (or not
depending). These potential dependencies
concern: (i) the world outside of the thinker
himself. (ii) language. (iii) logical truth. (iv)
CONSCIOUSNESS. In each case the question
is whether intentionality is essentially or
accidentally related to the items mentioned:
does it exist. that is. only by courtesy of the
dependence of thought on the said items?
And this question determines what the
intrinsic nature of thought is. Let us con-
sider each question in turn.
THOUGHT AND THE WORLD
Thoughts are obviously about things in the
world. but it is a further question whether
they could exist and have the content they
do whether or not their putative objects
themselves exist. Is what I think intrinSically
dependent upon the world in which I
happen to think it? This question was given
impetus and definition by a thought experi-
ment due to Hilary PUTNAM. concerning a
planet called TWIN EARTH. On Twin Earth
there live thinkers who are duplicates of us
in all internal respects but whose surround-
ing environment contains different kinds of
natural object. The suggestion then is that
what these thinkers refer to and think about
is individuatively dependent upon their
actual environment. so that where we think
about cats when we say 'cat' they think
about some other species when they use
that word - the different species that actu-
ally sits on their mats and so on. The key
point is that since it is not possible to indi-
viduate natural kinds like cats solely by
reference to the way they strike the people
who think about them. thinking about them
cannot be a function simply of internal
properties of the thinker. Thought content.
here. is relational in nature; it is fixed by
external facts as they bear upon the thinker.
585
THOUGHT
Much the same point can be made by con-
sidering repeated demonstrative reference to
distinct particular objects: what I refer to
when I say 'that bomb', of different bombs,
depends upon the particular bomb in front
of me and cannot be deduced from what is
going on inside me. Context contributes to
content.
Inspired by such examples, many philo-
sophers have adopted an 'externalist' view
of thought content: thoughts are not
autonomous states of the individual,
capable of transcending the contingent facts
of the surrounding world. One is therefore
not free to think whatever one likes, as it
were, whether or not the world beyond
cooperates in containing suitable referents
for those thoughts. And this conclusion has
generated a number of consequential ques-
tions. Can we know our thoughts with
special authority, given that they are thus
hostage to external circumstances? How do
thoughts cause other thoughts and beha-
viour, given that they are not identical
with any internal states we are in (see PRO-
POSITIONAL ATTITUDES; REASONS AND
CAUSES)? What kind of explanation are we
giving when we cite thoughts? Can there
be a science of thought if content does not
generalize across environments? These
questions have received many different
answers, and of course not everyone agrees
that thought has the kind of world-depen-
dence claimed. What has not been con-
sidered carefully enough, however, is the
scope of the externalist thesis - whether it
applies to all forms of thought, all concepts.
For unless this question can be answered
affirmatively we cannot rule out the possi-
bility that thought in general depends on
there being some thought that is purely
internally determined, so that the exter-
nally fixed thoughts are a secondary phe-
nomenon. What about thoughts concern-
ing one's present sensory experience, or
logical thoughts, or ethical thoughts? Could
there, indeed, be a thinker for whom
internalism was generally correct? Is ex-
ternal individuation the rule or the excep-
tion? And might it take different forms in
different cases?
586
Since words are also about things, it is
natural to ask how their intentionality is
connected to that of thoughts. Two views
have been advocated: one view takes
thought content to be self-subsistent relative
to linguistic content, with the latter depen-
dent upon the former; the other view takes
thought content to be derivative upon lin-
guistiC content, so that there can be no
thought without a bedrock of language (see
THOUGHT AND LANGUAGE). Thus arise con-
troversies about whether animals really
think, being non-speakers, or computers
really use language, being non-thinkers. All
such questions depend critically upon what
one is to mean by 'language'. Some hold
that spoken language is unnecessary for
thought but that there must be an inner
language in order for thought to be possible;
while others reject the very idea of an inner
language, preferring to suspend thought
from outer speech. However. it is not
entirely clear what it amounts to to assert
(or deny) that there is an inner LANGUAGE
OF THOUGHT. If it means merely that con-
cepts (thought-constituents) are structured
in such a way as to be isomorphic with
spoken language, then the claim is trivially
true, given some natural assumptions. But if
it means that concepts just are 'syntactic'
items orchestrated into strings of the same,
then the claim is acceptable only in so far as
syntax is an adequate basis for meaning -
which, on the face of it, it is not. Concepts
no doubt have combinatorial powers com-
parable to those of words, but the question
is whether anything else can plausibly be
meant by the hypothesis of an inner lan-
guage.
On the other hand. it appears undeniable
that spoken language does not have auton-
omous intentionality, but instead derives its
meaning from the thoughts of speakers -
though language may augment one's con-
ceptual capacities. So thought cannot post-
date spoken language. The truth seems to
be that in human psychology speech and
thought are interdependent. in many ways,
but that there is no conceptual necessity
about this. The only 'language' on which
thought essentially depends is that of the
structured system of concepts itself: thought
indeed depends upon there being isolable
concepts that can join with others to
produce complete propositions. But this is
merely to draw attention to a property any
system of concepts must have; it is not to
say what concepts are or how they succeed
in moving between thoughts as they do.
Appeals to language at this point are apt to
founder on circularity, since words take on
the powers of concepts only insofar as they
express them. Thus there seems little philo-
sophical illumination to be got from making
thought depend upon language.
THOUGHT AND LOGIC
This third dependency question is prompted
by the reflection that, while people are no
doubt often irrational, woefully so, there
seems to be some kind of intrinsic limit to
their unreason. Even the sloppiest thinker
will not infer anything from anything. To
do so is a sign of madness. The question
then is what grounds this apparent conces-
sion to logical prescription. Whence the
hold of logic over thought? For the depen-
dence here can seem puzzling: why should
the natural causal processes of thought
mirror the normative abstract relations of
logic? I am free to flout the moral law to
any degree I desire, but my freedom to think
unreasonably appears to encounter an
obstacle in the requirements of logic. My
thoughts are sensitive to logical truth in
somewhat the way they are sensitive to the
world surrounding me; they have not the
independence of what lies outside my will or
self that I fondly imagined. I may try to
reason contrary to modus ponens, but my
efforts will be systematically frustrated. Pure
logic takes possession of my reasoning pro-
cesses and steers them according to its own
dictates; not invariably, of course, but in a
systematic way that seems perplexing (see
RATIONALITY).
One view of this is that ascriptions of
thought are not attempts to map a realm of
independent causal relations, which might
THOUGHT
then conceivably come apart from logical
relations, but are rather just a useful
method of summing up people's behaviour.
Another view insists that we must acknow-
ledge that thought is not a natural phenom-
enon in the way merely physical facts are:
thoughts are inherently normative in their
nature, so that logical relations constitute
their inner essence (see DAVIDSON).
Thought incorporates logic in somewhat the
way externalists say it incorporates the
world. Accordingly, the study of thought
cannot be a natural science in the way the
study of (say) chemical compounds is.
Whether this view is acceptable depends
upon whether we can make sense of the
idea that transitions in nature, such as
reasonings appear to be, can also be transi-
tions in logical space, Le. be confined by the
structure of that space. What must thought
be such that this combination of features is
possible? Put differently, what is it for
logical truth to be self-evident?
THOUGHT AND CONSCIOUSNESS
This dependency question has been studied
less intensively than the previous three. The
question is whether intentionality is depen-
dent upon CONSCIOUSNESS for its very exis-
tence, and if so why. Could our thoughts
have the very content they now have if we
were not to be conscious beings at all?
Unfortunately, it is difficult to see how to
mount an argument in either direction. On
the one hand, it can hardly be an accident
that our thoughts are conscious and that
their content is reflected in the intrinsic
condition of our state of consciousness; it is
not as if consciousness leaves off where
thought content begins - as it does with,
say, the neural basis of thought. Yet, on the
other hand, it is by no means clear what it
is about consciousness that links it to inten-
tionality in this way. Much of the trouble
here stems from our exceedingly poor
understanding of the nature of conscious-
ness in general. Just as we cannot see how
consciousness could arise from brain tissue
(the mind-body problem), so we fail to
grasp the manner in which conscious states
587
THOUGHT
bear meaning. Perhaps content is fixed by
extra-conscious properties and relations and
only subsequently shows up in conscious-
ness, as various naturalistic reductive
accounts would suggest; or perhaps con-
sciousness itself plays a more enabling role,
allowing meaning to come into the world,
hard as this may be to penetrate. In some
ways the question is analogous to, say, the
properties of PAIN: is the aversive property
of pain, causing avoidance behaviour and
so on, essentially independent of the con-
scious state of feeling pain, being possibly
present without the feeling; or is it that pain
could only have its aversive function in
virtue of the conscious feeling? This is part
of the more general question of the epiphe-
nomenal character of consciousness (see
EPIPHENOMENALISM): is conscious aware-
ness just a dispensable accompaniment of
some mental feature - such as content or
causal power - or is it that consciousness is
structurally involved in the very determina-
tion of the feature? It is only too easy to feel
pulled in both directions on this question,
neither alternative being utterly felicitous.
Some theorists, indeed, suspect that our
uncertainty over such questions stems from
a constitutional limitation to human under-
standing. We just cannot develop the neces-
sary theoretical tools with which to provide
answers to these questions; so we may not
in principle be able to make any progress
with the issue of whether thought depends
upon consciousness and why. Certainly our
present understanding falls far short of pro-
viding us with any clear route into the
question.
It is extremely tempting to picture thought
as some kind of inscription in a mental
medium, and of reasoning as a temporal
sequence of such inscriptions. The model
here is language and its spoken and written
expression (hence the appeal of the second
dependency thesis above). On this picture
all that a particular thought requires in
order to exist is that the medium in ques-
tion should be impressed with the right
inscription. This makes thought indepen-
dent of anything else. On some views the
588
medium is conceived as consciousness itself,
so that thought depends on consciousness
as writing depends on paper and ink. But
ever since WITTGENSTEIN wrote, we have
seen that this conception of thought has to
be mistaken; in particular, it cannot provide
an acceptable account of intentionality. The
definitive characteristics of thoughts cannot
be captured within this model. Thus, it
cannot make room for the idea of intrinsic
world-dependence, since any inner inscrip-
tion would be individuatively independent
of items outside the putative medium of
thought. Nor can it be made to square with
the dependence of thought on logical pat-
terns, since the medium could be configured
in any way permitted by its intrinsic nature,
without regard for logical truth - as sen-
tences can be written down in any old order
one likes. And it misconstrues the relation
between thought and consciousness, since
content cannot consist in marks on the
surface of consciousness, so to speak. States
of consciousness do contain particular
meanings but not as a page contains sen-
tences: the medium conception of the rela-
tion between content and consciousness is
thus deeply mistaken. The only way to
make meaning enter internally into con-
sciousness is to deny that it acts as a
medium for meaning to be expressed.
However, as remarked above, it is devilishly
difficult to form an adequate conception of
how consciousness does carry content - one
puzzle being how the external determinants
of content find their way into the fabric of
consciousness.
Only the alleged dependence of thought
upon language fits the naIve tempting
inscriptional picture, but as we have seen
this idea tends to crumble under examina-
tion. The indicated conclusion seems to be
that we simply do not possess a conception
of thought that makes its real nature theor-
etically comprehensible; which is to say that
we have no adequate conception of mind.
Once we form a conception of thought that
makes it seem unmysterious, as with the
inscriptional picture, it turns out to have no
room for content as it presents itself; while
building in content as it is leaves us with no
clear picture of what could have such
content. Thought is real. then, if and only if
it is mysterious.
See also BELIEF; FODOR; IDENTITY THEORIES;
PHYSICALISM.
BIBLIOGRAPHY
York: Crowell.
McGinn, C. 1989. Mental Content. Oxford:
Basil Blackwell.
McGinn, C. 1991. The Problem of Conscious-
Putnam, H. 1981. The meaning of 'Meaning'.
In Mind, Language and Reality. Cambridge
University Press.
COLIN McGINN
thought and language The relation
between language and thought is philoso-
phy's chicken-or-egg problem. Language
and thought are evidently importantly
related, but how exactly are they related?
Does language come first and make thought
possible, or vice versa? Or are they on a par,
each making the other possible?
When the question is stated this gen-
erally, however, no unqualified answer is
possible. In some respects language is prior,
in other respects thought is prior, and in
still other respects neither is prior. For
example, it is arguable that a language is an
abstract pairing of expressions and mean-
ings, a function, in the set-theoretic sense,
from expressions onto meanings (see LEWIS,
1983). This makes sense of the fact that
Esperanto is a language no one speaks, and
it explains why it is that, while it is a con-
tingent fact that 'La neige est blanche'
means that snow is white among the French,
it is a necessary truth that it means that in
French. But if natural languages such as
French and English are abstract objects in
this sense, then they exist whether or not
anyone speaks them; they even exist in pos-
SIBLE WORLDS in which there are no thin-
kers. In this respect, then, language, as well
as such notions as meaning and truth in a
language, is prior to thought.
THE DEPENDENCE OF LANGUAGE ON
THOUGHT
But even if languages are construed as
abstract expression-meaning pairings, they
are construed that way as abstractions
from actual linguistic practice - from the
use of language in communicative beha-
viour - and there remains a clear sense in
which language is dependent on thought.
The sequence of marks 'Naples is south of
Rome' means among us that Naples is
south of Rome. This is a contingent fact,
dependent on the way we use 'Naples',
'Rome' and the other parts of that sen-
tence. Had our linguistic practices been dif-
ferent, 'Naples is south of Rome' might
have meant something entirely different or
nothing at all among us. Plainly, the fact
that 'Naples is south of Rome' means
among us that Naples is south of Rome has
something to do with the BELIEFS and
INTENTIONS underlying our use of the
words and structures that compose the sen-
tence. More generally, it is a platitude that
the semantic features that marks and
sounds have in a population of speakers are
at least partly determined by the PROPOSI-
TIONAL ATTITUDES those speakers have in
using those marks and sounds, or in using
the parts and structures that compose
them. This is the same platitude, of course,
which says that meaning depends at least
partly on use; for the use in question is
intentional use in communicative beha-
viour. So here is one clear sense in which
language is dependent on thought: thought
is required to imbue marks and sounds
with the semantic features they have in
populations of speakers.
The sense in which language does depend
on thought can be wedded to the sense in
which language does not depend on
thought in the following way. We can say
that a sequence of marks or sounds (or
whatever) cr means q in a language L, con-
strued as a function from expressions onto
meanings, iff L(cr)=q. This notion of
589
meaning-in-a-language. like the notion of a
language. is a mere set-theoretic notion that
is independent of thought in that it pre-
supposes nothing about the propositional
attitudes of language users: (J can mean q
in L even if L has never been used. But then
we can say that (J also means q in a popu-
lation P just in case members of P use some
language in which (J means q; that is. just
in case some such language is a language of
P. The question of moment then becomes:
What relation must a population P bear to
a language L in order for it to be the case
that L is a language of P. a language
members of P actually speak? (see Lewis.
1983. 1992). Whatever the answer to this
question is. this much seems right: in order
for a language to be a language of a popu-
lation of speakers. those speakers must
produce sentences of the language in their
communicative behaviour. Since such beha-
viour is intentional. we know that the
notion of a language's being the language
of a population of speakers presupposes the
notion of thought. And since that notion
presupposes the notion of thought. we also
know that the same is true of the correct
account of the semantic features expressions
have in populations of speakers.
This is a pretty thin result. not one likely
to be disputed. and the difficult questions
remain. We know that there is some rela-
tion R such that a language L is used by a
population P iff L bears R to P. Let us call
this relation. whatever it turns out to be. the
actual-language relation. We know that to
explain the actual-language relation is to
explain the semantic features expressions
have among those who are apt to produce
those expressions. and we know that any
account of the relation must require lan-
guage users to have certain propositional
attitudes. But how exactly is the actual-
language relation to be explained in terms
of the propositional attitudes of language
users? And what sort of dependence might
those propositional attitudes in turn have
on language or on the semantic features
that are fixed by the actual-language rela-
tion? Let us continue for a while with the
first question. about the relation of lan-
590
guage to thought. before turning to the
relation of thought to language.
All must agree that the actual-language
relation. and with it the semantic features
linguistic items have among speakers. is at
least partly determined by the propositional
attitudes of language users. This still leaves
plenty of room for philosophers to disagree
both about the extent of the determination
and the nature of the determining proposi-
tional attitudes. At one end of the determi-
nation spectrum. we have those who hold
that the actual-language relation is wholly
definable in terms of non-semantic proposi-
tional attitudes. This position in logical
space is most famously occupied by the pro-
gramme. sometimes called intention-based
semantics (IBS). of the late Paul Grice and
others. The foundational notion in this
enterprise is a certain notion of speaker
meaning. It is the species of communicative
behaviour reported when we say. for
example. that in uttering 'Il pleut'. Pierre
meant that it was raining, or that in
waving her hand. the Queen meant that
you were to leave the room. IBS seeks to
define this notion of speaker meaning
wholly in terms of communicators'
audience-directed intentions and without
recourse to any semantic notions. Then it
seeks to define the actual-language relation
in terms of the now-defined notion of
speaker meaning. together with certain
ancillary notions such as that of a conven-
tional regularity or practice, themselves
defined wholly in terms of non-semantic
propositional attitudes. The definition of the
actual-language relation in terms of speaker
meaning will require the prior definition in
terms of speaker meaning of other agent-
semantic notions. such as the notions of
speaker reference and Austin's notion of an
illocutionary act. and this. too. is part of the
IBS programme.
Some philosophers object to IBS because
they think it precludes a dependence of
thought on the communicative use of lan-
guage. This is a mistake. Even if IBS defini-
tions are given a strong reductionist
reading. as saying that public-language
semantic properties (Le. those semantic
properties that supervene on use in com-
municative behaviour) just are psychologi-
cal properties. it might still be that one
could not have propositional attitudes
unless one had mastery of a public lan-
guage. (See REDUCTION: SUPERVENIENCE.)
Whether or not this is plausible (that is a
separate question). it would be no more
logically puzzling than the idea that one
could not have any propositional attitudes
unless one had ones with certain sorts of
contents. Tyler Burge's insight, to be dis-
cussed later, that the contents of one's
thoughts is partly determined by the mean-
ings of one's words in one's linguistic com-
munity (see Burge, 1979), is perfectly
consistent with any IBS reduction of the
semantic to the psychological. Nevertheless,
there is reason to be sceptical of the IBS
programme. First, no IBS theorist has suc-
ceeded in stating a sufficient condition for
speaker meaning, let alone succeeded at the
much more difficult task of stating a neces-
sary-and-sufficient condition. And a plau-
sible explanation of this failure is that what
typically makes an utterance an act of
speaker meaning is the speaker's intention
to be meaning or saying something, where
the concept of meaning or saying used in
the content of the intention is irreducibly
semantic. Second, whether or not an IBS
account of speaker meaning can be
achieved, there are difficulties with the IBS
way of accounting for the actual-language
relation in terms of speaker meaning. The
essence of the IBS approach is that sen-
tences are used as conventional devices for
making known a speaker's communicative
intentions, and on this view language
understanding is an inferential process
wherein a hearer perceives an utterance
and, thanks to being party to relevant con-
ventions or practices, infers the speaker's
communicative intentions. Yet it appears
that this inferential model is subject to in-
superable epistemological difficulties (see
Schiffer. 1987, ch. 7). Third, there is no
pressing reason to think that the semantic
needs to be definable in terms of the psycho-
logical. Many IBS theorists have been moti-
vated by a strong version of PHYSICALISM,
which requires the reduction of all inten-
tional properties (Le. all semantic and pro-
positional-attitude properties) to physical or
at least topic-neutral, or functional, proper-
ties: for it is plausible that there could be no
reduction of the semantic and the psycho-
logical to the physical without a prior
reduction of the semantic to the psychologi-
cal (see Loar, 1981: Schiffer, 1982). But it
is arguable that such a strong version of
physicalism is not what is required in order
to fit the intentional into the natural order.
So the most reasonable view about the
actual-language relation is that it requires
language users to have certain propositional
attitudes, but there is no prospect of defin-
ing the relation wholly in terms of non-
semantic propositional attitudes. It is further
plausible, I submit, that any account of the
actual-language relation must appeal to
speech acts such as speaker meaning, where
the correct account of these speech acts is
irreducibly semantic (they will fail to super-
vene on the non-semantic propositional
attitudes of speakers in the way that inten-
tions fail to supervene on an agent's beliefs
and desires). If this is right, it would still
leave a further issue about the definability of
the actual-language relation. Is it possible to
define the actual-language relation, and if
so, will any irreducibly semantic notions
enter into that definition other than the
sorts of speech act notions already alluded
to? These questions have not been much
discussed in the literature: there is neither
an established answer nor competing
schools of thought. My own view is that the
actual-language relation is one of the few
things in philosophy that can be defined,
and that speech act notions are the only
irreducibly semantic notions the definition
must appeal to (see Schiffer, 1993).
THE DEPENDENCE OF THOUGHT ON
LANGUAGE
This brings us to the dependence of thought
on language. A useful starting point is a
claimed dependence I think does not obtain.
This is the claim that propositional attitudes
are relations to linguistic items which
591
obtain at least partly by virtue of the
CONTENT those items have among language
users. This position does not imply that
believers have to be language users, but it
does make language an essential ingredient
in the concept of belief. The position is moti-
vated by two considerations: (a) the suppo-
sition that believing is a relation to things
believed, which things have truth values
and stand in logical relations to one
another; and (b) the desire not to take
things believed to be propositions - abstract,
mind- and language-independent objects
that have essentially the truth conditions
they have. Now (a) is well motivated: the
relational construal of propositional atti-
tudes is probably the best way to account
for the quantification in 'Harvey believes
something nasty about you'. But there are
problems with taking linguistic items, rather
than propositions, as the objects of belief. In
the first place, if 'Harvey believes that floun-
ders snore' is represented along the lines of
'B(Harvey, "flounders snore")', then one
could know the truth expressed by the sen-
tence about Harvey without knowing the
content of his belief; for one could know
that he stands in the belief relation to
'flounders snore' without knowing its
content. This is unacceptable (see Schiffer,
1987, ch. 5). In the second place, if Harvey
believes that flounders snore, then what he
believes - the reference of 'that flounders
snore' - is that flounders snore. But what is
this thing, that flounders snore? Well, it is
abstract, in that it has no spatial location; it
is mind and language independent, in that
it exists in possible worlds in which there
are neither thinkers nor speakers; and,
necessarily, it is true iff flounders snore. In
short, it is a proposition - an abstract,
mind- and language-independent thing that
has a truth condition and has essentially
the truth condition it has.
A more plausible way that thought
depends on language is suggested by the
topical thesis that we think in a 'LANGUAGE
OF THOUGHT'. On one reading, this is
nothing more than the vague idea that the
neural states that realize our thoughts 'have
elements and structure in a way that is
592
analogous to the way in which sentences
have elements and structure' (Harman,
1978, p. 58). But we can get a more literal
rendering by relating it to the abstract con-
ception of languages already recommended.
On this conception, a language is a function
from 'expressions' - sequences of marks or
sounds or neural states or whatever - onto
meanings, which meanings will include the
propositions our propositional-attitude rela-
tions relate us to. We could then read the
language of thought hypothesis as the claim
that having propositional attitudes requires
standing in a certain relation to a language
whose expressions are neural states. There
would now be more than one 'actual-
language relation'. The one discussed earlier
might be better called the public-language
relation, since it makes a language the
instrument of communication of a popula-
tion of speakers. Another relation might
be called the language-oj-thought relation,
because standing in that relation to a lan-
guage makes it one's lingua mentis. Since
the abstract notion of a language has been
so weakly construed, it is hard to see how
the minimal language-of-thought proposal
just sketched could fail to be true. At the
same time, it has been given no interesting
work to do. In trying to give it more inter-
esting work, further dependencies of
thought on language might come into play.
For example, it has been claimed that the
language of thought of a public-language
user is the public language she uses: her
neural sentences are related to her spoken
and written sentences in something like the
way her written sentences are related to her
spoken sentences. For another example, it
might be claimed that even if one's lan-
guage of thought is distinct from one's
public language, the language-of-thought
relation makes presuppositions about the
public-language relation in ways that make
the content of one's thoughts dependent on
the meanings of one's words in one's public-
language community.
Tyler Burge has in fact shown that there
is a sense in which thought content is
dependent on the meanings of words in
one's linguistic community (Burge, 1979).
Alfred's use of 'arthritis' is fairly standard,
except that he is under the misconception
that arthritis is not confined to the joints; he
also applies the word to rheumatoid ail-
ments not in the joints. Noticing an ailment
in his thigh that is symptomatically like the
disease in his hands and ankles, he says to
his doctor, 'I have arthritis in the thigh'.
Here Alfred is expressing his false belief that
he has arthritis in the thigh. But now con-
sider a counterfactual situation that differs
in just one respect (and whatever it entails):
Alfred's use of 'arthritis' is the correct use in
his linguistic community. In this situation,
Alfred would be expressing a true belief
when he says 'I have arthritis in the thigh'.
Since the proposition he believes is true
while the proposition that he has arthritis
in the thigh is false, he believes some other
proposition. This shows that standing in the
belief relation to a proposition can be partly
determined by the meanings of words in
one's public language. The Burge phenom-
enon seems real, but it would be nice to
have a deep explanation of why thought
content should be dependent on language
in this way.
Finally, there is the old question of
whether, or to what extent, a creature who
does not understand a natural language
can have thoughts. Now it seems pretty
compelling that higher mammals and
humans raised without language have their
behaviour controlled by mental states that
are sufficiently like our beliefs, desires and
intentions to share those labels. It also
seems easy to imagine non-communicating
creatures who have sophisticated mental
lives (they build weapons, dams, bridges,
have clever hunting devices, etc.). At the
same time, ascriptions of particular contents
to non-language-using creatures typically
seem exercises in loose speaking (does the
dog really believe that there is a bone in the
yard?), and it is no accident that, as a
matter of fact, creatures who do not under-
stand a natural language have at best prim-
itive mental lives. There is no accepted
explanation of these facts. It is possible that
the primitive mental lives of animals
account for their failure to master natural
languages, but the better explanation may
be CHOMSKY'S, that animals lack a special
language faculty unique to our species (see
Chomsky 1968, 1975). As regards the
inevitably primitive mental life of an other-
wise normal human raised without lan-
guage, this might simply be due to the
ignorance and lack of intellectual stimula-
tion such a person would be doomed to. On
the other hand, it might also be that higher
thought requires a neural language with a
structure comparable to that of a natural
language, and that such neural languages
are somehow acquired pari passu as the
child learns its native language. Finally, the
ascription of content to the propositional-
attitude states of languageless creatures is a
difficult topic that needs more attention. It is
possible that as we learn more about the
logic of our ascriptions of propositional
content, we will realize that these ascrip-
tions are egocentrically based on a similar-
ity to the language in which we express our
beliefs. We might then learn that we have
no principled basis for ascribing proposi-
tional content to a creature who does not
speak something a lot like one of our
natural languages, or who does not have
internal states with natural-language-like
structure. It is somewhat surprising how
little we know about thought's dependence
on language.
See also FODOR; PUTNAM; STALNAKER.
BIBLIOGRAPHY
Austin, J.L. 1962. How to Do Things with
Words. Oxford University Press.
Avramides, A. 1989. Meaning and Mind, Cam-
Bennett, J. 1976. Linguistic Behaviour. Cam-
Studies in Philosophy, 4,73-121.
Chomsky, N. 1968. Language and Mind. New
York: Harcourt Brace.
Chomsky, N. 1975. Reflections on Language.
New York: Pantheon Books.
Curtiss, S. 1977. Genie: A Psycholinguistic
Study of a Modern-Day 'Wild Child'. New
York: Academic Press.
593
TURING. ALAN
Davidson, D. 1984a. Truth and Interpretation.
Davidson, D. 1984b. On saying that. In David-
son 1984a.
Davidson, D. 1984c. Thought and talk. In
Davidson, 1984a.
Dummett, M. 1989. Language and commu-
nication. In Reflections on Chomsky, ed. A.
George. Oxford: Basil Blackwell.
Field, H. 1978. Mental representation.
Erkenntnis, 13, 9-61.
York: Crowell.
Grice, P.1989. Studies in the Way of Words.
Harman, G. 1973. Thought. Princeton Uni-
versity Press.
Harman, G. 1978. Is there mental representa-
tion? In Perception and Cognition: Issues in the
Foundations of Psychology, ed. C. Savage.
Minneapolis: University of Minnesota Press.
Higgenbotham, J. 1986. Linguistic theory and
Davidson's program in semantics. In Truth
and Interpretation: Perspectives on the Philoso-
phy of Donald Davidson, ed. E. LePore.
Lewis, D. 1969. Convention. Cambridge, MA.:
Lewis, D. 1983. Languages and language. In
Philosophical Papers, vol. 1. Oxford Uni-
versity Press.
Lewis, D. 1992. Meaning without use: Reply
to Hawthorne. Australasian Journal of Philo-
sophy, 70, 106-110.
Loar, B. 1976. Two theories of meaning. In
Truth and Meaning, ed. G. Evans and J.
McDowell. Oxford University Press.
Loar, B. 1981. Mind and Meaning. Cambridge
University Press.
Sacks, O. 1989. Seeing Voices: A Journey. into
the World of the Deaf. Berkeley: University of
California Press.
Segal, G. 1989. A preference for sense and
reference. Journal of Philosophy, 86, 73-89.
Schiffer, S. 1972. Meaning. Oxford University
Press.
Schiffer, S. 1981. Indexicals and the theory of
reference. Synthese, 49, 43-100.
Schiffer, S. 1982. Intention-based semantics.
Notre Dame Journal of Formal Logic, 23,
119-56.
Schiffer, S. 1987. Remnants of Meaning. Cam-
594
Schiffer, S. 1992. Belief ascription. The Journal
of Philosophy, 89,499-521.
Schiffer, S. 1993. Actual-language relations.
Philosophical Perspectives, 7,231-58.
Strawson, P. 1964. Intention and convention
in speech acts. Philosophical Review, 73,
439-60.
Strawson, P. 1969. Meaning and Truth. Oxford
University Press.
STEPHEN SCHIFFER
Turing, Alan (1912-54) Alan Turing was
a mathematical logician who played a
crucial role in the development of the
theory of computation. His most well-
known contribution to this theory was cast
in terms of what has come to be called a
'Turing machine', though for reasons that
will be obvious, this was not a machine in
any concrete sense. What is relevant in the
present context is that the notion of a
Turing machine has had a decisive influ-
ence on certain views about the nature of
the mind, and, fittingly, Turing himself
published a celebrated article in the philo-
sophical journal Mind outlining some of
the philosophical consequences of his
'machine'.
Ironically, given the fact that Turing's
work underpins the ideas that gave birth to
the digital computer, Turing used the term
'computer' - long before there were such
devices - of a human being engaged in cal-
culation. In fact, it was partly by thinking
about such human 'computers' that he
came to develop the idea of Turing
machines. What he suggested was that one
could simplify and mechanize the process by
which a human computer did calculation,
and then, by generalizing on this mechan-
ized process, one could use it to define a
special class of numbers. Turing called these
'computable' numbers. Finally, by reflecting
on the way in which these numbers were
mechanically generated, he showed that
there were bound to be numbers that were
not computable, though there was no
mechanically describable way by which you
could demonstrate this of a particular
number. In essence, Turing had used his
imagined device - the generalized mechan-
ical calculator - to prove that there were
what mathematicians call 'undecidable'
problems.
Important as it has been in the sophisti-
cated reaches of computational and logical
theory, the idea of a Turing machine is very
simple. Imagine that you have a type-
writing device which can do a restricted
range of things: it can type a symbol on a
paper tape, it can remove such a symbol
and it can move left or right one unit along
the tape. Figure 1 shows the device poised
over a section of an infinitely long paper
tape which is divided into squares, some of
which are empty and some of which
contain an's' (for 'symbol').
Is I s Is I
Figure 1
The whole of what is pictured in figure 1 is
a Turing machine. By specifying more pre-
cisely what symbols the typewriter can use,
and how it is disposed to react to them as it
passes along the tape, we can get the
Turing machine to transform one set of
symbols and spaces (the 'inpuf) into
another (the 'output'). And despite its sim-
plicity, it is enormously powerful. Turing
showed that his device could take any input
and transform it into any output so long as
there is some computable relation between
them. Of course, if the function relating
input to output is very complex, the
machine will take a long time to do the cal-
culation. So no one would seriously con-
template building one as a practical way of
doing computation, nor is it even possible to
build a true Turing machine given the fact
that it is essential to the machine's opera-
tion that the tape be infinite. As mentioned
earlier, a Turing machine is essentially a
machine in thought.
Despite the fact that a Turing machine is
not intended to be a real device, and that its
original field of application was the theory
TURING, ALAN
of mathematical functions, it has had a
major influence on thinking about the
mind. To see why this is so, one must first
accept - as most are prepared to do - that
mental phenomena are dependent on the
workings of the brain. Then one is invited
to recognize that, at some level of descrip-
tion more general than the neurophysio-
logical, the brain is a device that receives
complex inputs from sensory systems and
effects equally complex outputs to the motor
systems. Moreover, as seems plausible, these
relations between input and output are func-
tionally well-behaved enough to be describ-
able by various - albeit mind-bendingly
complicated - mathematical relationships.
This is plausible for many reasons, not least
because our mental lives have a certain
orderliness, to them: we avoid obstacles we
can see, WI.e tend to pursue goals based on
our n e e ~ s and desires, we come to believe
new things on the basis of those we already
accept, etc. Finally, it should be noted that
we do not have actually to know what the
mathematical relationships are for the story
to be interesting. For, so long as they exist,
we know that some specific version of a
Turing machine will be able to mimic them.
There will be a Turing machine that per-
fectly simulates input-output structure of
any specific brain.
What the above reasoning comes to is
this: the brain is a computational device
and when we talk about the mind, we are
in effect, describing this device. Of course,
when we speak about what persons want,
believe, need, intend, etc. we are speaking
at a much higher level of description than
would be appropriate to a Turing machine.
But the fact that we can depend on there
being a Turing machine that captures the
functional relations of the brain supports
the idea that our talk about the mind is also
a functional characterization. Or so many
currently think. Turing (1950) himself was
quite clear about this. Using his celebrated
'imitation' game, he suggested that one
could envisage a computational device
that would be indistinguishable from a
human 'computer' (to revert to Turing's
original use of this word). The game
595
TYPE/TOKEN
essentially consists of three 'players': a com-
puter which responds uses a teletype to
respond to various inputs it receives, a
human being who also responds via a tele-
type, and a second human being who pro-
vides the input to the other two - who
interrogates them. The object of the game is
to see if the second human being can, by
asking questions and exammmg the
answers, tell which of the unseen 'devices'
is human and which a machine. Turing
was in no doubt that, if a computational
device succeeded in fooling the human
questioner, then that device would have all
that is required for having a mind. And, as
mentioned earlier, he thought that some
such device would one day manage the
task.
Much has been written about Turing's
own interpretation of the imitation game,
and not all those otherwise committed to
COMPUTA TION AL MODELS OF MIND would
accept Turing's conception of the mind. But,
in the development of his 'machine', Turing
managed to make a crucial contribution to
logical and mathematical theory as well as
to the philosophy of mind. The functionalist
theory of the mind owes him a great debt,
even if in some of its forms, it uses the idea
of a Turing machine only in the back-
ground (see FUNCTIONALISM).
See also ARTIFICIAL INTELLIGENCE.
BIBLIOGRAPHY
Turing, A.M. 1950. Computing machinery
and intelligence. Mind, 59, 433-60.
Hodges, A. 1983. Alan Turing: The Enigma of
Intelligence. London: Hutchinson.
SAMUEL GUTTENPLAN
twin earth In a paper published in 1975,
Hilary Putnam described a thought experi-
ment in which one imagined a duplicate of
of our planet. down to almost the last detail.
However, what is crucial to the thought
experiment is that there is one small differ-
ence between earth and twin earth: on twin
earth the substance which looks and
596
behaves just like water is not in fact water.
Its chemical composition is imagined to be
something called 'XYZ' rather than H
2
0.
The original purpose of this thought experi-
ment was to show that the meanings of
words in language couldn't be, as Putnam
put it, 'in the head'. This consequence was
held to follow from the fact that a thinker
on earth and a thinker on twin earth - both
of whom were ignorant of the chemical
composition of things on their planets -
would be exact duplicates as far as brain
and behavioural organization went, and
they would produce exactly the same
sounds in the same circumstances. But a
speaker on earth would be referring to
water (that is, H
2
0) when he spoke,
whereas the speaker on twin earth would
be referring to XYZ (twin water, or 'twater'
as it is called when one tells the story).
The original thought experiment, and
variations on it, have been used extensively
in the philosophy of mind, though often
with aims that go beyond any in Putnam's
original paper. In this volume, the thought
experiment is either described in detail or
mentioned by many of the authors. (See
CONCEPTUAL ROLE SEMANTICS; EPIPHENO-
MEN ALISM; EXTERNALISM/INTERN ALISM;
THOUGHTS.)
BIBLIOGRAPHY
Putnam, H. 1975 The meaning of 'meaning'.
In Philosophical Papers, Vol. 2: Mind, Lan-
guage and Reality. Cambridge University
Press.
SAMUEL GUTTENPLAN
type/token How many words are there in
the sentence: 'The cat is on the mat'? There
are of course at least two answers to this
question, precisely because one can either
count word types, of which there are five, or
individual occurrences - known as tokens -
of which there are six. Moreover, depending
on how one chooses to think of word types,
another answer is possible. Since the sen-
tence contains definite articles, nouns, a
preposition and a verb, there are four gram-
matically different types of word in the sen-
tence.
The type/token distinction, understood as
a distinction between sorts of thing and
instances, is commonly applied to mental
phenomena. For example, one can think of
pain in the type way as when we say that
we have experienced burning pain many
times; or, in the token way, as when we
speak of the burning pain currently being
suffered. The type/token distinction for
TYPE/TOKEN
mental states and events becomes important
in the context of attempts to describe the
relationship between mental and physical
phenomena. In particular, the IDENTITY
THEOR Y asserts that mental states are phys-
ical states, and this raises the question
whether the identity in question is of types
or tokens. (See also DAVIDSON; LEWIS;
SUPERVENIENCE.)
SAMUEL GUTTENPLAN
597
u
the unconscious Psychoanalytic theory
describes a range of motives, mental states,
and processes of which persons are ordi-
narily unaware, and which they can
acknowledge, avow, and alter only with dif-
ficulty. Freud's collective term for these, and
for the functional division of the mind to
which he assigned them, was the uncon-
scious. (For references and further discussion
of italicized terms see Laplanche and Pointa-
lais, 1973). The term has also been used
to describe other mental states, such as
hypothesized beliefs about language, taken
to playa comparable role (Fodor, 1991,
p. 278). In what follows, however, we shall
concentrate on the psychoanalytic use.
Freud sometimes illustrated unconscious
motivation by examples from hypnosis.
Someone may, for example, comply with a
post-hypnotic suggestion, while seeming to
remember nothing about it, and citing some
implausible motive of his or her own. Here,
it seems, we do not accept the subject's own
account, but rather suppose that the action
is caused by a motive (e.g. a desire to do
what the hypnotist said) of which the
subject is unaware. Further hypnotic
research has produced a variety of examples
apparently fitting Freud's descriptions of the
unconscious and its working (Erickson,
1939; Luria, 1976, ch. 4). While such phe-
nomena seem genuinely illustrative, it
remains unclear how far they should be
assimilated to those encountered in psycho-
analytic practice. So let us begin with the
clinical work of Freud and his successors,
and then turn to the more abstract meta-
psychology based on work of this kind.
CLINICAL
Early in his career Freud discovered that
dreams and symptoms could be seen as
598
related, causally and in their contents, to
motives. In particular, both could be seen as
wishfulfilments, that is, as representing the
satisfaction of DESIRES or wishes, which had
not been subjected to the rational thought
requisite for INTENTIONAL action. This
emerged when the dream or symptom was
considered in the context of the patient's
full and uncensored account of related
thoughts and feelings, as obtained through
the process of free association.
This can be illustrated by the example
with which Freud begins The Interpretation
of Dreams (Freud, 1974 vol. 4, ch. 2), his
own dream of Irma's injection. In this
dream Freud met Irma, a family friend and
patient, whom he had diagnosed as hysteri-
cal, and treated by analysis. He told her that
if she still felt pains, this was her own fault,
for not accepting his solution. He became
alarmed, however, that she was suffering
from an organic illness which he had failed
to diagnose, and this turned out to be true.
His senior colleague M examined Irma, and
confirmed that she was indeed organically
ill; and it became manifest that her illness
was caused by a toxic injection given by
another of Freud's colleagues, his family
doctor Otto. The dream ended with Freud
censuring Otto's practice, saying 'Injections
of that kind ought not to be made so
thoughtlessly' and adding 'probably the
syringe had not been clean'.
On the surface this dream dealt with
topics that were not pleasant to Freud, such
as the continued suffering of a friend and
patient, and the possibility that he had mis-
diagnosed an organic illness, which he
described as 'a constant anxiety' to someone
offering psychological treatment. Freud's
associations, however, enable us to see that
the treatment of these topics in the dream
was in fact thoroughly wishful. The day
before the dream Otto - who had recently
visited Irma and her family - had briefly
discussed Irma with Freud. Otto had said
that Irma was looking 'better, but not yet
well'; Freud had thought he detected a
reproof in this, and was vaguely annoyed.
That night, in order to justify himself, Freud
had started to write up Irma's case to show
to M, who was respected by both himself
and Otto, and who appeared in the dream
as diagnosing Irma's illness and becoming
aware that it was Otto's fault. Also, as it
happened, Otto had been called on to give
someone an injection while at Irma's (cf.
the topic of the dream), and Freud had just
had news indicating that another of his
female patients had been given a careless
injection by some other doctor, and had
been contemplating his own careful practice
in this respect.
In considering the dream Freud noted
that his desire to justify himself in respect of
Irma's case, and in particular not to be
responsible for her suffering, was apparent
from the beginning, in which he told Irma
that her pains were now her own fault.
Also, he felt that his alarm at her illness in
the dream was not entirely genuine. So,
Freud realized, it seemed that he was actu-
ally wishing that Irma be organically ill: for
as he undertook to treat only psychological
complaints, this also would mean that he
could not be held responsible for her condi-
tion. This theme, indeed, seemed carried
further in the rest of the dream, in which M
found that Otto, not Freud, bore responsi-
bility for Irma's illness. The whole dream, in
fact, could be seen as a wishful response to
Otto's remark. According to the dream, and
contrary to what Freud had taken Otto to
imply, Freud bore no responsibility what-
ever for Irma's condition. Rather, Otto was
the sole cause of her suffering, and this was
a result of Otto's bad practice with injec-
tions, a matter about which Freud himself
was particularly careful.
To see the role of wishfulfilment here
more clearly, let us consider Freud's desire
that he be cleared of responsibility for
Irma's suffering, as this operated, on the
THE UNCONSCIOUS
one hand, in his intentional action, and on
the other, in his dream. Very schematically,
we hold that in rational action the causal
role of a desire that P is to bring about
(cause) a situation that p, which both satis-
fies the desire and pacifies it, that is, causes
the desire to cease to operate. Acting on a
desire that P (that one be cleared of culp-
able responsibility) should ideally bring it
about that P (that one is cleared of culpable
responsibility), that is, should bring about a
situation which constitutes the satisfaction
of the desire. This, in turn, should cause
the belief that P (that one has been
cleared ... ), and this, perhaps acting to-
gether with the satisfying situation, should
pacify the desire that P, so that it ceases to
govern action. This is approximately the
sequence of results that Freud was seeking
to produce, in accord with standard
medical practice, in writing up Irma's case
history on the night of the dream to discuss
with M, his respected senior colleague. M
would be able to offer an independent,
authoritative opinion on Freud's treatment
of Irma; so his judgment could partly serve
to clear Freud, and, we may presume,
Freud's conscience.
In Freud's dream the same motive was
apparently also at work, but in a different
way. There it produced no rational action,
but rather gave rise directly to a (dreamt)
representation of a situation in which Freud
was cleared of responsibility, and by M. This
representation, moreover, was extrava-
gantly wishful - Irma was made physically
ill, Freud was cleared in a great number of
ways, Otto was elaborately blamed, and so
on. Taking this example as typical, we can
contrast the causal role of desire, as
between action and wishfulfilment. In
rational action a desire that P serves to
bring about a situation that p, and this to
cause a (justified and true) belief that P, so
that the desire is pacified. In wishfulfilment,
by contrast, this process is short-circuited,
so as to leave a satisfying or justifying
reality out. Here the desire that P causes a
wishful and belief-like representation that P
directly, and this serves to pacify the desire,
regardless of reality, at least temporarily. In
599
THE UNCONSCIOUS
rational action we find both the real satis-
faction and also the pacification of desire,
with the latter a causal and rational con-
sequence of the former. In wishfulfilment
we find only pacification, via a version of
wishful imagining or make-believe: that is,
imaginary pacification without real satisfac-
tion. Although Freud did not describe
matters in these terms, he took this feature
to be characteristic of wishfulfilment
generally.
We can thus put part of Freud's concep-
tion by saying that wishfulfilment seems to
be the mind's (or brain's) way of pacifying
desires - and thus stabilizing or redirecting
its own functioning in a certain way -
without actually satisfying them. Still the
mode of pacification seems analogous in
both cases. In rational action pacification is
consequent on satisfaction and veridical
BELIEF, and in wishfulfilment on belief-like
representation. Belief itself, however, can be
regarded as the limiting case of belief-like
representation. So we can say that in
general pacification proceeds via representa-
tion of this kind.
Freud also found that a given dream,
symptom, or other wishfulfilment character-
istically involved a range of wishes, con-
nected in their contents. We have seen that
Freud's wish to avoid responsibility for Irma
went with one to blame Otto. But also his
associations make clear that the dream was
wishfulfilling on levels deeper than his
present concern with Irma. For example in
analysing his dream Freud realized that
Irma was linked in his mind with two
persons who had previously died as a result
of his medical interventions. One of his
friends had suffered from incurable nerve
pain, and was addicted to the morphia he
used for relief from it. Freud had suggested
that his friend use cocaine instead, not
grasping that it too was addictive. The
friend later died from injections of cocaine.
Also, Freud had himself once repeatedly
prescribed a woman patient a standard
medication, which, unpredictably, had
killed her; and he had consulted with M
about this case also.
These memories were integral to Freud's
600
associations, and connected with many
other details of the dream; so they can be
seen to have influenced the dream as well.
Hence the remark with which Freud ended
the dream - 'Injections of that kind ought
not to be made so thoughtlessly' - was
actually one with which he might well have
reproached himself, in respect of treatments
he associated with Irma's. But in the dream
this deeper reproach - regarding thought-
lessness, the misuse of toxic substances, and
damaging injections - was also wishfully
deflected on to Otto. In representing Otto
but not himself as responsible for Irma's suf-
fering, Freud also represented Otto but not
himself as bearing precisely the kinds of
responsibility involved in the deaths of his
other friend and other patient. Hence this
dream can also be regarded as representing
the fulfilment of a wish on Freud's part not
to be responsible in these cases also. But
this wish, and indeed the whole topic of his
own responsibility for death, was entirely
kept from Freud's CONSCIOUSNESS in the
dream, and came to light only via his
associations.
This also illustrates further mechanisms
that Freud found to be common in dreams,
and characteristic of unconscious function-
ing generally. In the dream the figure of
Freud's friend and patient Irma also repre-
sented, or stood for, Freud's other friend and
other patient who had died as a result of his
therapeutic interventions. So this example
shows what Freud called the condensation of
several significant figures and topics from
the latent content of the dream - the
thoughts and feelings uncovered by associ-
ation as related to the dream, which in this
case included the links between Irma and
these dead others - into one composite
figure and topic appearing in the manifest
content of which the dreamer was aware.
This went also with a displacement of Freud's
guilt, again in a way connected with all
three cases, on to the figure of Otto. These
processes contributed to the distortion of the
manifest mental CONTENT effected by what
Freud called the dream work. As noted
below, Freud later found MEMORY and con-
scious belief generally to be liable to similar
distortion; and some of this may be visible
in the material connected with this analysis,
for example in Freud's own conscious in-
clination to regard Otto as 'thoughtless' or
'jumping to a conclusion' about Irma's case.
In light of the above we can give the
following preliminary and schematic char-
acterization of Freud's clinical method and
project. In commonsense psychology we
interpret actions in accord with a basic gen-
eralization about desire: the role of a desire
that P is to produce a situation that P,
which in turn should produce a belief that P
serving, together with the situation, to
pacify the desire, and so to redirect action
(see FOLK PSYCHOLOGY). In our everyday
understanding of persons we both tacitly
use this generalization, and also sustain it
inductively, bearing it out through the suc-
cessful interpretation of desire in action in
case after case. This generalization includes
the idea that a representation (belief) that P
plays a role in pacifying a desire that P.
Hence we also take it as an intelligible, and
indeed common, phenomenon that a desire
that P should playa role in causing a belief-
like imaginative representation that p,
which tends to pacify the desire. This is
another generalization which we both use
and sustain, in understanding many forms
of wishful imagining, make-believe, and so
forth, with which we are familiar.
Freud's work on dreams and symptoms
uses, extends, and supports this latter gen-
eralization, by finding instance after
instance, and in previously unsuspected
cases, such as the dream above. Such inter-
pretative work, as Freud claimed, enables
one to see dreams and symptoms as pacify-
ing deeper desires with the same sort of
regularity as actions can be seen as satisfy-
ing them. This in turn serves also to extend
and support the basic generalization about
desires (above): for each interpretation of a
wishfulfilment adds new values for P to the
contents of the probable desires of an agent,
and so gives rise to further and better inter-
pretations of other of the agent's thoughts
and actions as well. (Thus in the course of
understanding Freud's interpretation of the
Irma dream, we naturally frame further and
THE UNCONSCIOUS
deeper explanations of his annoyance at
Otto's remark, his desire to justify himself,
his contemplation of his own conscientious-
ness about injections, etc.). Such further
ascriptions of desires, in turn, may make it
possible to detect further wishfulfilments;
and so on.
We can thus say that Freud sought to
extend commonsense psychology by means
internal to it: namely, the supportive exten-
sion of basic causal generalizations concern-
ing the satisfaction and pacification of desire
already employed in commonsense inter-
pretative practice. His extension is therefore
potentially sound, cumulative, and radical.
Sound, because the extending interpreta-
tions can gain support from the basic gen-
eralizations, and can also support them in
turn, as in commonsense psychology itself.
Cumulative, because each addition to the
contents of probable wishes or desires can
facilitate the discovery of others. And
radical. because the extension offers sig-
nificantly deeper and fuller explanations of
actions and wishfulfilments generally, and
by reference to motives which, in the main,
had not previously been contemplated.
Freud found that the unconscious
motives characteristically pacified in adult
dreams and symptoms could be traced back
into childhood, and included sensual love
for one parent combined with rivalry and
jealous hatred for the other, a constellation
he called the Oedipus Complex. This, as it
emerged, had one version in which the
child's love was for the parent of the oppo-
site sex, and another in which the love was
for the parent of the same sex (and vice
versa for the concomitant rivalry). It thus
appeared that the feelings and phantasies of
very young children showed remarkable
plasticity, and, in particular, a degree of
bisexuality. Little children were liable to
intense psychic conflict, as between desires
to harm or displace each parent, envied and
hated as a rival for the love of the other,
and desires to preserve and protect that
same parent, loved sensually and also as a
caretaker, helper, and model. In con-
sequence, Freud thought, these conflicting
motives were subjected to a process of
601
THE UNCONSCIOUS
repression, which removed them from think-
ing and planning of which the agent was
aware; and concomitantly, in the course of
normal development, they were both organ-
ized and modified by the child's formative
identification with the parent of the same
sex, that is, the child's taking that parent as
a basic model for agency and the satisfac-
tion of desire. Still, the repressed motives
continued to exist in the unconscious, and
to exercise their causal role in the produc-
tion of dreams, symptoms, and parapraxes;
and, in those cases in which conflict
remained particularly extreme, in forms of
neurotic or psychotic illness.
Following Freud's description of the role
of belief-like representation in the pacifica-
tion of desire, psychoanalysts now com-
monly describe the kind of representation
which serves to pacify unconscious desire as
phantasy. Particular phantasies, moreover,
can be seen as constituting or implementing
many of the unconscious mental processes,
including those of both development and
defence, which are described in psycho-
analytic theory. Thus persons form lasting
and life-shaping phantasies of themselves on
the model of others, thereby establishing
identifications with those others, as men-
tioned above. Again, persons represent
others as having, and themselves as
lacking, certain of their own impulses,
aspects of mind, or traits of character, and
thus accomplish the projection of these items
onto or into others. The projection, or
phantasied location, of parts of oneself in
another may create a particular kind of
mirror-image identification with that other,
now often called projective identification. (See
also Hinshelwood, 1991.) Such a process
can also effect the splitting of the self, for
example into good and bad, with the bad
located elsewhere; and likewise (the rep-
resentation of) the other may by the same
means be split into good and bad, as with
the image of the good mother and evil
(step)mother in a fairy tale. (These processes
are similar to those observed in dreams and
symptoms; for example, the dream above
might be taken as exemplifying a phantasy
in which Freud represented motives con-
602
nected with lack of professional care as in
Otto rather than himself, and hence as an
instance of splitting and projection on
Freud's part.)
Although Freud's hypotheses about child-
hood were mainly based on data from
adults, later analysts, and in particular
Anna Freud (1946, 1974) and Melanie
Klein (1932) were able to extend his
techniques to children. Even very small
children often have symptoms and diffi-
culties analogous to those of adults; but
they characteristically cannot produce such
articulate thoughts and feelings connected
with these, as Freud used in analysing their
elders. They do, however, spontaneously
and constantly represent things in play,
with, e.g. dolls, toys, clay, paints, and
games of make-believe. Child analysts have
been able to understand these representa-
tions as Freud understood dreams, that is,
as systematically reflecting motive and
mental state, and in particular as embody-
ing wishfulfilling phantasy. This has made it
possible to analyse disturbed children, and
hence to learn more about their mental life.
Such work is now taken both as confirming
and extending hypotheses based on the
analysis of adults. So let us consider some
material from the treatment of a little boy
(Loeb, 1992), in order to illustrate some of
the ideas sketched above.
This little boy suffered from nightmares -
for example about 'red crayfish monsters' -
and also behaved in an exaggeratedly fem-
inine way. From the age of two he had
wished to grow up to be a 'mommy', and as
a toddler he would cover his chest with a
towel after his bath, as if he had breasts.
When he began therapy at four and a half
he liked to pretend that he had breasts, and
to dress as a 'fancy lady' in women's
clothes, and to walk and talk accordingly.
He took female parts in his play with other
children, and by himself played with Barbie
dolls; and in his daydreams he imagined
himself to be Wonder Woman.
In part this behaviour showed an identifi-
cation with his father's attractive and fash-
ionable mother, his 'fancy grandmother',
with whom he had spent a lot of time as a
baby. This woman both behaved seductively
towards the little boy and fostered his femi-
nine ways. Thus she took off her clothes in
front of him, and also would, for example,
ask him to feel the soft leather pants she
was wearing, as a result of which he got an
erection and felt anxious. But also she let
him wear her own high-heeled shoes, and
dressed him in the make-up, jewellery, and
other female finery he had come to make
his own. We can see that from this seem-
ingly contradictory behaviour one could
extract a single coherent message, as to the
overriding power and desirability of the
grandmother's own feminine glamour; and
it seems that the little boy had done so. In
his first session of therapy he played with
two Barbie dolls, one of which he dressed in
plain clothing, the other in a 'fancy' low-cut
gown. The plain doll he called 'mother', and
the fancy doll 'queen grandmother'.
The little boy was thus able to express
feelings about his parental figures - includ-
ing here, perhaps, a sense of rivalry
between his mother and grandmother, and
also a division in his representation of
women as between plain and 'fancy' - in
terms of his play with dolls. At the same
time he began the transference of these feel-
ings on to his (female) analyst. He asked
her, for example, to undress for him as his
grandmother did; and when he was upset
he would attack her, saying that it was the
monsters that came in his nightmares who
were doing it. In one such nightmare a
'half-lady, half-pinching lobster' chased
him, and ran in and out of his mother's
nose. It could thus be seen that in his mind
an important sort of aggression was repre-
sented by pinching, and through phantasies
involving pinching figures or creatures.
Such aggression could be expressed in a
dream, as related to himself and his mother,
or in his behaviour, as related to the
analyst; and he was liable to imagine the
analyst as a fearful pinching figure as well.
Later in his analysis, as the little boy
began to play out the marriage of the dolls
Ken and Barbie, the role of such figures
emerged more vividly. After the wedding, as
the boy represented things, Ken would put
THE UNCONSCIOUS
his penis in Barbie's vagina; and then
Barbie would take the penis, leaving Ken
with a vagina. The little boy would scream
'Ken lost his penis.' Often he said 'If you
dress and act like a girl, nobody will think
you have a penis. Then you don't have to
worry that anyone will take it.' In time he
was able to make one basis of these fears
more clear. He talked about his (female)
analyst having a 'hidden penis', and said it
was 'the one that was taken from Ken - the
one women get back.' Women, he said
'steal penises because they are jealous of
men ... Women come to the men at night
and steal their penises. They have pinchers
. . . the press-on nails are their pinchers . . .
But no woman will ever get mine.' It thus
appeared that he likened women in general.
and his grandmother in particular, to the
pinching monsters of his dreams, and also
likened such women's pinching to castra-
tion, aimed at taking away men's penises so
as to keep them for themselves.
This material can be seen both in light of
Freud's general method, and also a number
of particular claims about the unconscious,
as sketched above. We can see, for example,
how it might be that the little boy's wish to
be a 'mommy' who had breasts (itself
perhaps an indication of a natural bisexu-
ality) was reflected even from the age of two
in wishfulfilling identification with female
behaviour, such as hiding his chest after a
bath. Apparently such desires were later
organized and represented as satisfied
through identification with his 'fancy
grandmother', and were expressed, elab-
orated, and pacified in a variety of rep-
resentational activities, ranging from day-
dreams through play to dress, posture, and
behaviour. Also we can see some of what
Freud described as the sexual phantasies of
children: e.g. that of the phallic woman, who
has a hidden penis; or of the primal scene of
parental intercourse, as one of violence and,
in this particular case, danger to men.
The little boy's phantasy life thus seemed
dominated by imagos - perhaps formed
partly by projection - of fearful pinching
figures, salient both in his nightmares and
the underlying phantasies about women
603
THE UNCONSCIOUS
which emerged in his analysis. The material
suggests that he was liable to identify
himself with these phantasy figures, and
that this served two connected functions, as
specified by psychoanalytic theory. First, it
enabled him wishfully to represent himself
as the kind of powerful, glamorous, and
castrating female figure he unconsciously
imagined his grandmother, or again his
analyst, to be. Secondly, it served to protect
the masculinity that was threatened by
figures of this same kind - if he represented
himself as such a woman, he might escape
the castration which such women dealt to
men. Thus, it would seem, through project-
ive identification, or again, identification with
the (phantasied) aggressor, this little boy
sought both to enjoy, and to escape, a form
of aggression with which he was pre-
occupied. It would seem that such deep pro-
jective and identificatory phantasies were
constitutive of his unconscious mental life,
and hence both of his character and the
conflicts he suffered, until understood and
thereby altered through analysis.
METAPSYCHOLOGICAL
In clinical work Freud described the uncon-
scious in commonsense terms, as including
wishes, beliefs, memories, and so forth. But
he also sought to integrate his clinical find-
ings with more abstract and theoretical
concepts, as well as with physiological
research, which was beginning to focus on
the neurons composing the brain.
In his early Project for a Scientific Psycho-
logy Freud hypothesized that the working of
the brain could be understood as the
passage among neurons of some form of
excitation, or cathexis, via connections
which he called 'contact-barriers'. Informa-
tion, on this hypothesis, would be stored in
the brain in the form of alterations - facil-
itations or inhibitions - of these connec-
tions, and would be processed by the
passage through the interconnected
neurons themselves. Hence, as Freud put it,
'psychic acquisition generally', including
memory, would be 'represented by the differ-
ences in the facilitations' of neural connec-
604
tions (Freud, 1974, vol. I, p. 300). Freud
thus anticipated the contemporary claim
that the brain can be understood as a com-
putational device whose 'knowledge is in the
connections' among neuronal processing
units (Rumelhart et aI., 1988, p. 75), and
also the associated view of mental processes
as forms of neural activation, and mental
states as dispositions to these, or structures
determining them (see Glymour 1992; and
also CONNECTIONISM). He sketched a model
representing his early clinical findings in
these terms, and seems to have framed his
later discussions to be consistent with this.
On Freud's early physiological model the
signalling of a bodily need, instinct, or drive
- say for nutrition in an infant - causes a
disequilbrium in neural excitation. This at
first results in crying and uncoordinated
bodily movements, which have at best a
fleeting tendency to stabilize it. Better and
more lasting equilibration requires satisfac-
tion, e.g. by feeding; and this causes the
facilitation of the neural connections
involved in the satisfying events. The brain
thus lays down neural records, or proto-
types, of the sequences of perceptions, inter-
nal changes, bodily movements, and so on,
involved in the restoration of equilibrium by
satisfaction. Then when disequilibrium
again occurs - e.g. when the infant is again
hungry - the input signals engage pre-
viously facilitated pathways, so that the
records of relevant past satisfactions are
naturally reactivated. This, Freud hypo-
thesized, constitutes early wishfulfilment.
Freud thus identified the wishfulfilling
pacification of infantile proto-desire with
what can be regarded as a form of neural
prototype activation. (For a recent general
account of this notion see Churchland,
1989, chs 9 and 10.) He took it that this
provided more stability in disequilibrium
than the random ennervations it replaced,
and also that it served to organize the
infant's responses, e.g. to hunger, by repro-
ducing those previously associated with
satisfaction. Then as the infant continued to
lay down prototype upon prototype, the
original wishful stabilizations evolved
towards a system of thought, while also
coming to govern a growing range of
behaviour, increasingly coordinated to the
securing of satisfaction. This, however,
required the brain to learn to delay the
wishfulfilment-governed neural behaviour
associated with past satisfaction until
present circumstances were perceptibly
appropriate - that is, to come increasingly
under the sway of what Freud called the
reality principle.
This capacity for delay depended upon a
tolerance of frustration, and of the absence
of the satisfying object, which permitted
reality testing, and hence the binding of the
neural connections involved in the securing
of satisfaction to perceptual information
about the object, and later to rational
thought. By this means what Freud re-
garded as a primary process leading to pre-
cipitate wishfulfilment was progressively
overlaid and inhibited by a secondary one,
which provided for the securing of satisfac-
tion in realistic conditions. This benign
development could, however, be blighted, if
frustration (or intolerance of it) too much
led to the overactivation of inappropriate
prototypes, and this to greater frustration.
Such a process could render the mind/brain
increasingly vulnerable to disequilibrium
and delusion, and hence increasingly reliant
on earlier and more wishfulfilling modes of
stabilization, in a vicious circle constitutive
of mental disturbance and illness.
Freud allocated the task of fostering the
sense of reality, and so providing for the
satisfaction and reality-based pacification of
desire, to a hypothetical neural structure, or
functional part of the mind, which he called
the ego. In later work Freud extended his
account of the ego to include, among other
things, the way it developed through identi-
fication with other persons. As noted above,
the child's ego was partly formed through
its identification with the parents in their
role as agents, or satisfiers of their own
desires. But the child also achieved self-
regulation by laying down images of the
parents as others in relation to the self, that
is, in their role as satisfiers, or again frus-
trators or controllers, of its own bodily
impulses and desires, and particularly the
THE UNCONSCIOUS
early impulses connected with feeding, defe-
cation, and the like. The child thus intro-
jected helpful or controlling figures, and
internalized its relations with them, as these
were registered in the perspective of early
experience, distorted both by projection and
by the extremes of infantile emotion. The
resulting distorted and controlling imagos
formed the basis of a distinct, self-critical
part of the ego, which Freud called the
super-ego. This faculty tended to be far more
aggressive, threatening, and punitive than
the actual parents, and so could be a source
of great anxiety or guilt, and even, in the
extreme, suicide.
Freud also related the development of the
ego and of conscious thought to language.
The earliest prototypes, he assumed, were
concerned with needs and actions bearing
on objects in the immediate environment,
and so with what he called thing-presenta-
tions. A limited relation of symbolism might
obtain among thing-presentations, in the
sense that one such presentation could
become activated by, or in place of, another.
In learning language, however, the brain
laid down a further set of facilitations, con-
stituting a network of word-presentations,
including 'sound-images', 'word-images',
and a system of 'speech associations' which
linked these linguistic prototypes both with
one another and with (those of) the things
and situations associated with words and
sentences. This system, Freud hypothesized,
was responsible for 'cognition' and for
'conscious observing thought'. Cognition
could partly be understood in terms of the
activation of connections which were medi-
ated by linguistic prototypes, and which,
therefore, might be logical or rational. The
consciousness of thought could be seen as
resulting from the interactivation of linguis-
tic and objectual representations; and in
consequence the unconscious could be
understood as that which was not properly
linked with, or was somehow cut off from,
the system of thought-facilitating connec-
tions laid down with the acquisition of lan-
guage.
Freud elaborated these ideas on symbol-
ism, language, and the unconscious in his
605
THE UNCONSCIOUS
later work (see, e.g. Freud, 1974, vol. r.
p. 365, and vol. XIV, p. 209ft'), He also
attempted to describe how infantile sexual
and aggressive motives could undergo sub-
limation, and thus be redirected towards
ends which were benign, or socially valued.
Subsequent psychoanalytic research, parti-
cularly with schizophrenic patients, has
suggested that both the capacity for such
emotional development, and that for ration-
ally integrated thought and feeling, depend
upon certain abilities to form and use
symbols; and that these in turn depend
upon a capacity to tolerate frustration, and
in particular to bear the absence, distinct-
ness, and separateness of the satisfying
object, in ways related to Freud's original
suppositions. (See Segal, 1986, ch. 4; Bion,
1967, ch. 4; Hinshelwood, 1991.)
CONCLUSION
As sketched above, Freud's early clinical
work began a systematic and potentially
cogent extension of commonsense psychol-
ogy, providing deeper explanations for
dreams, symptoms, and also many aspects
of everyday thought, feeling, and action, by
reference to unconscious motives. This
provided the basis for a more general and
theoretical account of normal and patho-
logical functioning and development, which
has been revised and extended by relation
to data gained from the analysis both of
children and psychotic patients. Much of
this account can be cast in terms of the
concept of unconscious phantasy, and asso-
ciated processes such as projection and
identification: and many of the constituent
hypotheses were framed to accord with a
conception of the working of the brain that
has recently become an independent focus
of research. Psychoanalytic hypotheses
about the unconscious thus provide an
explanatory and unifying account of a great
range of mental and behavioural phenom-
ena, many of which are commonsensically
or clinically observable, and which are
addressed by no other theory. Since these
hypotheses are arguably cogent, and based
on data gathered by many researchers over
606
years of systematic observation, they
deserve serious philosophical attention.
See also CONSCIOUSNESS; PSYCHOANALYTIC
EXPLANATION; PSYCHOLOGY AND PHILO-
SOPHY; RATIONALITY; SEARLE; SUB-
JECTIVITY.
BIBLIOGRAPHY
Bion, W. 1967. A theory of thinking. In
Second Thoughts. London: Heinemann.
Churchland, P. 1989. A Neurocomputational
Perspective. Cambridge, MA.: MIT Press.
Erickson, M. 1939. Experimental demon-
strations of the psychopathology of every-
day life. Psychoanalytic Quarterly, vol. 8,
1939; reprinted in Freud and Psychology:
Selected Readings, ed. S. G. M. Lee and M.
Herbert. Harmondsworth: Penguin Books,
1970.
Fodor J. 1991. Replies. In Meaning in Mind:
Fodor and his Critics, ed. B. Loewer and G.
Rey. Oxford: Basil Blackwell.
Freud, A. 1946. Introduction to the technique
of the analysis of children. In The Psycho-
analytical Treatment of Children. New York:
International Universities Press.
Freud, A. 1974. Four lectures on child ana-
lysis. In The Writings of Anna Freud, vol. 1.
New York: International Universities Press.
Freud, S. 1974. The Standard Edition of the
Complete Psychological Works of Sigmund
Freud. London: Hogarth.
Gardner, S. 1992. The unconscious. In The
Cambridge Companion to Freud, ed. J. Neu.
Glymour, C. 1992. Freud's androids. In The
Cambridge Companion to Freud, ed. J. Neu.
Hinshelwood, R., 1991. A Dictionary of Klein-
ian Thought. London: Free Associations
Books.
Klein, M. 1932. The Psychoanalysis of Children.
London: Hogarth; reprinted in The Collected
Works of Melanie Klein, vol. 2. London:
Hogarth,1975.
Laplanche, J., and Pontalis, J.-B., eds. 1973.
The Language of Psycho-Analysis. London:
Hogarth.
Loeb, L. 1992. Transsexual symptoms in a
child. Journal of the American Psychoanalytic
Association, 40: 2.
Luria, A. 1976. The investigation of com-
plexes produced during hypnosis by sugges-
tion. In The Nature of Human Conflicts. New
York: Liveright.
Segal, H. 1986. Notes on symbol formation. In
The Work of Hanna Segal. London: Free Asso-
ciations Press.
THE UNCONSCIOUS
Rumelhart, D., Hinton, G., and McClelland, J.
1988. A general framework for parallel dis-
tributed processing. In Parallel Distributed
Processing, Vol. 1, ed. D. Rumelhart et. al.
Wollheim, R. 1991. Freud. London: Fontana.
JIM HOPKINS
607
w
weakness of will The notion of weakness
of will or 'akrasia' (to use its Greek name)
figures importantly in moral philosophy.
Agents are said to be weak-willed when
they have reached conclusions about their
moral duties, but then fail to act on these
conclusions. Since it is often difficult to be
moral - to live up to one's moral principles
- there would seem to be nothing particu-
larly surprising or troubling about this
notion, and certainly nothing especially
pressing for the philosophy of mind. But this
appearance is wrong on both counts.
First, there are certain conceptions of
morality that make akratic action very
puzzling indeed. Suppose, as did Plato and
Aristotle, that an agent's opinion about
what is morally required of him in some
specific situation is an expression of what
that agent thinks true about the situation -
it is a cognitive state. One might here say
that the agent sees the situation as morally
requiring him to do some particular act.
And suppose further that what leads an
agent to act is precisely this view of how
things are in respect of the relevant situ-
ation. That is, the complete motive or source
of the action is the state of mind described
above as 'seeing the situation as requiring a
particular act'. Given these two things -
that a moral opinion is a cognitive state and
that this state provides the motivation for
action - the very idea of an agent having
such an opinion and not acting on it is
problematic. For if an agent does genuinely
have the moral view that, say action A, is
morally required then there would seem to
be no source of motivation that could
explain why the agent does something else.
Yet failing to do A, i.e. doing not-A, is
plausibly describable as doing something
else.
608
Of course, someone may form an opinion
and then have a change of heart. Or the
opinion may not be held in a fully sincere
way. But in neither of these cases do we
have the right conditions for akrasia. What
is required is that the moral opinion be sin-
cerely held, and continue to hold, whilst the
agent does something other than what is
dictated by that opinion. The usual view is
that such weak-willed action is perfectly
possible, but, given the cognitive account of
morality described above, it would seem
impossible.
This is not the place to consider the kinds
of solution that Aristotle and others have
offered to this problem; that investigation
belongs to moral philosophy. And of course
it might seem that the best thing to do in
this case is simply to reject the moral
account that makes akratic action so prob-
lematic. For example, one might insist that
moral opinions are only part of our motiva-
tion - that various non-cognitive elements
such as desires are equally necessary. If you
regard moral opinions as requiring moral
desires to be translated into action, then
any particular case of weak-willed action
might be ascribable to the effect of some
non-moral desire lurking in the back-
ground. However, though there may be
good reasons to reject the cognitive picture
of morality, the problem of weakness of will
by itself should not count as one. For, even
in the brief sketch of this problem in moral
philosophy, one can recognize the in-
gredients of a more generalized version of
akrasia which is, if anything, more puzzling,
and which cannot be solved by tinkering
with one's conception of morality. Indeed,
in its most general form, the problem of
weakness of will is a problem for the very
idea of rational action.
One can see the problem by juxtaposing
certain apparently obvious principles of
rational action with a description of akrasia
that leaves behind all talk of morality and
any wrangles about cognitive or non-
cognitive motivation. First, we shall con-
sider the principles of rational action,
though it should be noted that 'rational
action' does not mean action that deserves
praise for its intelligence or logicality. It just
means actions intelligible in the light of an
agent's beliefs, desires and intentions -
ingredients in what are thought of as an
agent's reasons.
Suppose you think one course of action is
much better for you than any other you
can think of, where 'better' is to be under-
stood as neutral in respect of morality. It
would seem just obvious that in this state of
mind you would want to undertake that
course of action in preference to any other.
Moreover, given that you have such a pre-
ference - that you want to do something
more than you want to do anything else - it
seems equally obvious that you will inten-
tionally do it, so long of course as you count
yourself able and free to do that thing. For
example, suppose that I think it better for
me to go to the cinema this evening - better
than anything else I could do, and in abso-
lutely no conflict with anything that I
ought morally to do or not to do. From this
it would seem to follow that I want to go to
the cinema more than I want to do any-
thing else. (There is a sense of 'better for
me' which might not carry that implication,
as when one speaks of something not all
that pleasant but that would be better for
me if I did it. A trip to the dentist might be
described this way. However, in the prin-
ciple above, one should not take the judg-
ment that way: 'better for me' does not
carry any sense of something undesirable
but necessary. What you judge best for you
is just that thing you put top of the list.)
Finally, given that I want to go to the
cinema more than I want to do anything
else, it would seem equally to follow that I
will intentionally go to the cinema when
the time comes, assuming of course that I
feel free to do so and have not changed my
WEAKNESS OF WILL
mind. These claims are straightforward
considerations governing our conception of
what it is to be an agent who makes judg-
ments, has wants or preferences, and acts
intentionally after deliberation. In short,
they are claims about rational agency.
We turn next to a description of akratic
action - one borrowed from Davidson
(1980). To a first approximation, an agent
is said to act in a weak-willed way when he
or she does something at the same time as
thinking that what is done is not the best
thing. Thus, I may decide that it would be
best if I did not stay up late, given that I
have a lot to do the next day - things that I
have been looking forward to doing and
that I know can only be fully enjoyed if I
am not exhausted by lack of sleep. But
being weak-willed, I nonetheless do stay up.
Generalizing on this example, and being
careful about certain details, we can sayan
agent acts in a weak-willed way when: (i)
the agent does something intentionally
whilst (ii) believing there to be another
available course of action which, all things
considered, that agent regards as genuinely
better. The word 'intentionally' is important
here because we wouldn't count a piece of
behaviour as weak-willed unless it was an
action in the full sense. And it is equally
important that the other course of action be
both available and judged better by the
agent's own lights.
Combining the discussions of the last two
paragraphs, it is not too difficult to appreci-
ate that we have trouble. For, according to
the principles of rational action, insofar as
an agent regards one of two courses of
action as better, then that agent wants most
to do that which is better and will in fact do
it, unless for some reason neither of them is
undertaken. But, as in the case of going to
bed early, it seems possible for an agent to
be akratic - to do intentionally something
which is judged to be less good than some-
thing else, though both are thought pos-
sible. The principles of rational action seem
to rule out akrasia, but not many examples
are needed to show us that there is in fact
such a thing. The combination of things we
accept about action and weakness entails
609
THE WILL
the contradictory and thus paradoxical con-
clusion that there both can and cannot be
such a thing as akrasia.
As characterized, weakness of will creates
difficulties that go right to the heart of our
general conception of what it is to act for
reasons. Morality and moral theories just
don't come into it. Moreover, there is no
universally accepted way to deal with the
paradox. Yet, since the work expended on
akrasia, has also served to sharpen our
understanding of such things as desire,
intention, action, and judgment, there is
every reason to think that the problem of
weakness of will is genuine and quite deeply
embedded in our concept of action.
See also DESIRE; INTENTION; RATIONALITY;
REASONS AND CAUSES.
BIBLIOGRAPHY
Davidson, D. 1980. How is weakness of will
possible? In his Essays on Actions and Events.
McDowell, J. 1981. The role of eudaimonia in
Aristotle's ethics. In Essays on Aristotle's
Ethics, ed. A. O. Rorty. University of Cali-
fornia Press.
Mortimore, G., ed. 1972. Weakness of Will: A
Collection of Papers. London: Macmillan &
Co.
Wiggins, D. 1987. Weakness of will, commen-
surability, and the objects of deliberation
and desire. In his Needs, Values, Truth.
Oxford: Basil Blackwell).
SAMUEL GUTTENPLAN
the will The word 'will' has been used in
at least two different ways by philosophers.
At the time of the German Romantics there
was a tendency to use it to designate a psy-
chological phenomenon, force-like in char-
acter and linked in ways to traits like
determination, which was broadly of the
nature of desire or instinct - the main pre-
cursor of Freudian 'Libido' and 'Id'. The
connection with intentional action was
close, but no more than that: thus, we find
Schopenhauer speaking of 'acts of will', but
also of 'the will' itself with a broader sig-
610
nification. According to him the will con-
stituted one great half of the mind, indeed
constituted its reality or essence. 'Will' here
signifies something to be opposed to the
cognitive and intellectual part of the mind,
close to 'heart' and remote from 'head', in
which our very being lay.
This usage is to be distinguished from a
different and more widespread usage in
which 'will' is exclusively linked with acting
and trying to act. Here 'will' stands for a
supposed psychological event which is
present of necessity and uniquely on such
active occasions. It is this supposed phenom-
enon that is the subject-matter of the
ensuing discussion. In short, I shall be
discussing 'the will' as the term is under-
stood by Descartes, Locke, Hume, (later)
WITTGENSTEIN and RYLE; and not as
understood by Fichte, Schelling, and Scho-
penhauer.
Before I begin this discussion I would like
to make three preliminary points, (ex), ( ~ ) ,
(y). The first point ((X) is that questions of
language are of particular relevance when
we debate the existence of contentious phe-
nomena like 'willing' and 'sense-data'. This
is because we embark upon such discus-
sions in ignorance of the necessary and suf-
ficient marks of the phenomena in question.
In my opinion we are right to do so. Science
has taught us that this is a genuine possi-
bility - in science: that it is possible to single
out a particular something in nature (e.g.
gold) by means of a designating expression,
which accordingly acquires the sense of
standing for that something, without our
knowing what the something is. A similar
situation can occur in philosophy; and in
particular in the case of those debatable
entities 'willing' and 'sense-data'. Thus, it
seems to me to be perfectly possible that
Descartes and others had truly divined the
presence of a psychological something
which they described as 'willing', even
though the characterization which they
gave of 'willing' was incorrect. We should
not judge the validity of their claim that
'willing' is a reality, by the veridicality of
their theories as to its nature. To be sure,
we can tolerate so much and no more
discrepancy between the actual nature of
that item and what they take that nature to
be; for we need reason in what they say to
know what it is that they are talking of.
Nonetheless the discerning of, and the
correct characterizing of the truly discerned,
are distinguishable phenomena. We should
therefore note that we can approach the
question of the reality of the will, without as
yet being able to state criterial necessary
and sufficient conditions for the use of 'will'.
Designation can precede identification.
The second preliminary point ( ~ ) is, that
the problem(s) of the Will is not the same as
the problem(s) of ACTION. The first problem
is, whether or not the will exists, and what
is its nature. More exactly, whether or not
there exists anything that is endowed with
a set of properties that sufficiently resemble
the set that philosophers (traditionally using
the word 'will') took to be the identifying
marks of 'willing', to justify crediting them
with its detection; and the nature of that
phenomenon. The problem of Action is dif-
ferent. Now the actual existence of such a
type as 'action' - which is allowed to range
over INTENTIONAL and unintentional
examples, and which excludes mere causing
'actions' such as storms 'do' - is I think a
matter concerning which there is wide-
spread agreement. It is a consensus that
extends far beyond Philosophy, and exists in
most legal and moral systems. While non-
rational animals act (e.g. walk) in precisely
the sense we do, most crime and wicked-
ness takes the form either of action or its
neglect. Then the problems concerning
action are at least two. What are the neces-
sary and sufficient conditions for the occur-
rence of an act of given type; and what are
the phenomenal constituents of such an
act? Now it may well be the case that the
necessary conditions of any action include
the occurrence of an event we are entitled
to describe as a 'willing'. If so, it may well
be the case that such an event is either
identical with or part of or necessary cause
of the act in question. Evidently the pro-
blems arising over the phenomenon of
action are intimately connected with those
appertaining to the will. Indeed, a resolu-
THE WILL
tion of the latter problems appears to be a
pre-condition of a resolution of the former.
Then I would like to emphasize that my
concern here is exclusively with the
problem of the will. It is not that of the con-
ditions of, and the constitution of, actions.
The third and last preliminary point (y)
concerns the close relation between theories
as to the existence and nature of the will,
and mind-body theory. The identification-
conditions employed by DUALISTS like
Descartes and Locke show a tendency to
include elements that betray the presence of
that dualist standpoint. Are such beliefs
essential to their belief in the reality of the
will? I do not think so. After all, Hume and
Berkeley cannot be described as dualists, yet
unquestionably believed in the selfsame
phenomenon mentioned by Locke. Indeed,
Hume in the Enquiries (section vii, part I)
impressively refutes Locke's attempt to trace
the origin of our idea of Power to the will,
by appeal to the mysteriousness of the rela-
tion between willing and its bodily effect,
and could scarcely have done so had he and
Locke had different phenomena in mind.
Theories of the will's existence and nature
almost wholly transcend theories of the
mind-body relation. There is nothing to
prevent a latter-day PHYSICALIST of all but
the most destructively reductionist variety
from believing in a theory of the will which
is more or less in total agreement with
seventeenth- and eighteenth-century dual-
ists, eighteenth-century idealists, and the
twentieth-century heirs of the theory.
In sum, we must distinguish (cx) theories
concerning the existence of the will, from
theories as to its nature, ( ~ ) theories con-
cerning the existence and nature of the will,
from theories concerning the conditions of
and constitution of actions, (y) theories con-
cerning the nature of the will, from mind-
body theory.
IDENTIFYING CONDITIONS OF THE WILL
I pass now to a consideration of the identi-
fying conditions of this supposed phenom-
enon. The identifying conditions adopted by
such traditional believers in the phenom-
611
THE WILL
enon as Descartes, Locke, Berkeley, Hume
(et al.), include the following at least:
(i) the item in question is a mental event
W,
(ii) which we 'do' in the very sense in
which we do actings and tryings,
(iii) which occurs of necessity whenever
we either act or try to act,
(iv) W is a distinct event from, and gen-
erally a regular causally sufficient con-
dition of, the bodily event x which
occurs when we engage in a bodily act
x' of x-production (e.g. if x is arm-rise,
x' is arm-raising).
A Cartesian might very well embellish
this (existential) theory with the following
additional features:
(v) W occurs in a purely mental sub-
stance,
(vi) W is non-physical in nature.
Since (v) and (vi) derive from more
general metaphysical views on the person/
mind/body relation, I think we have a right
to disregard them, and say that those philo-
sophers who endorsed (i)-(iv) and those
who endorsed (i)-(vi) shared a belief in the
existence of the will. More, they shared a
belief as to its nature: namely, that it is a
distinctive mental event that is 'done' and
necessary to and unique to active situations
(whether of success or failure, intentional-
ness or unintentionalness). And they shared
in addition a belief concerning its causal
explanatory power. These three beliefs - in
the will's existence, nature, and explanatory
power - will be held by anyone endorsing
(i)-(iv); and might very well be held by a
physicalist. They are not identical beliefs,
but they are nonetheless very closely linked.
Thus it is not possible to entertain a belief in
the will's existence without holding at least
some view as to its nature; though it may
well be possible to hold beliefs on these
latter two matters without endorsing any
particular theory as to the will's causal
explanatory properties.
The distinguishing of these three beliefs -
612
in the existence, nature, and power of will -
as separate from one another, is a little
more difficult when we come to consider
philosophers like Schopenhauer (or Spinoza)
who conceive of the mind-body relation in
radically unified terms. Schopenhauer
described 'acts of the body' as 'acts of the
will objectified'; and what he meant was
that physical actions like (say) walking
were at once physical phenomena and iden-
tical with willings. In short, not only would
Schopenhauer not endorse (v) and (vi), he
would not endorse (iv) either! Now the
causal explanatory properties of the will
have generally been taken to be of the first
importance - rightly, in my view. Ought we
therefore to say that Schopenhauer cannot
have believed in the existence of the phe-
nomenon postulated by Descartes and Locke
(etc.)? Or ought we instead to say that he
believed in the existence of the selfsame
phenomenon, yet held a different theory
concerning its pivotal causal-explanatory
properties, and conceivably also a different
theory as to its nature? I have no doubt the
latter is the correct interpretation of the
situation. I have no doubt that Schopen-
hauer believed in 'acts of the will' in the
received traditional sense. The considera-
tions advanced earlier wherein we dis-
tinguished grounds for saying a
phenomenon had been discerned, and
grounds for attributing a particular theory
concerning either the nature or pivotal
causal properties of that phenomenon, are
surely decisive. There can I think be no
doubt that Descartes, Hume, Schopenhauer,
and (say) Pritchard, are in agreement as to
the existence of willing; and in disagree-
ment with (say) R YLE and Anglo-American
philosophers in the Wittgensteinian era.
So much for what it is to entertain a
theory concerning the existence and char-
acter of the will. A word now about disbelief
in this phenomenon. Belief in the existence
of the will (and sense-data) became very
unpopular between (say) 1935 and 1970.
The main cause of this disbelief was I think
verificationist and neo-behaviourist styles of
thought, and in particular the influence of
the earlier and the later philosophies of
WITTGENSTEIN. Both the will and sense-
data smelled of 'bad metaphysics'. To many
they seemed philosophical inventions, scar-
cely on a par with The Absolute or The
Transcendental Ego, but suffering like them
from the unforgivable sin of being postu-
lated for reasons of an entirely philosophical
kind. The very idea of phenomenal items,
whose existence was a matter for philo-
sophy to decide, ran against the entire ver-
ificationist spirit of the era (which extends
into our own!) Now I think it is no exag-
geration to say that the general attitude of
the philosophical community to will (and
sense-data) was (and to a degree still is)
near-phobic in character and marked by a
special form of intolerance. Lodged at the
strategic points of interaction between mind
and extra-mental physical nature, these
entities were seen as threatening us with a
domain of 'private objects' and a mind-
body divide which would plunge us all back
in 'reactionary' Cartesian dualistic individu-
alism. I think this is superstitious. And if
anything the boot is on the other foot! For
such a view of will and sense-data evidences
a somewhat flimsy commitment to a natur-
alistic-physicalistic account of the mind-
body relation. Thus, it vastly overestimates
the extent to which the admission of such
entities could weaken such an account.
Provided those entities are embedded in a
law-governed psychophysical framework,
they constitute no more of a threat to a
unified account of mind and body than does
the admission of mental phenomena into
one's ontology. Indeed, I personally am per-
suaded that a particular account of willing
manages to accomplish a vitally significant
form of unification of mind and body which
cannot be expressed in other terms. Inside
many opponents of will and sense-data -
indeed, inside some physicalists! - a closet
or unconscious Cartesian is wildly signal-
ling to be let out!
These are no more than speculations.
And it has to be admitted that it is some-
thing of a paradox, if acts of the will and
sense-data do indeed exist, that Philosophy
should be called upon to demonstrate their
existence. Surely it is an empirical issue
THE WILL
whether or not these phenomena are reali-
ties? Surely observation, in these particular
cases introspective observation, must be the
ultimate court of appeal? And it has here
and now to be acknowledged that in either
case the findings of INTROSPECTIVE obser-
vation are precisely nil. Stare within as you
act or visually perceive and you come
across nothing but acting or seeing! And yet
many philosophers have posited the exist-
ence of these entities. Why? What is the
way out of this seeming impasse? It is, to
begin with, at least facilitated with the
assistance of a very little philosophy of lan-
guage. The first thing to note is, that 'will'
and 'sense-datum' are philosophical terms
of art, which stand in need of explanation,
an explanation which must take the form
either of a paraphrase or a definite descrip-
tion. In short, we need to be able to say
what it is that we are looking for if we are to
come across these entities. Here we have
one non-observational issue. And we need
in addition to know what is to count as an
introspective awareness of the phenomenon
in question. And so on. It is through loop-
holes like these that the question as to
whether or not these entities exist manages
to escape the rigid ruling that the existence
of phenomenal entities is as such purely and
simply an observational issue. Whether or
not it is a matter for observation, it is
without doubt a matter for philosophy.
THE REALITY OF THE WILL:
PRELIM IN AR Y CONSIDERA TIONS
Then what philosophical considerations
support belief in the reality of the will?
Before I advance an argument in the follow-
ing section, a few preliminary considera-
tions. One important preliminary to the
argument is linguistic, the discovery of a
name or description of the phenomenon
that is cast in terms of ordinary established
usage and eschews philosophical terminol-
ogy. Once this has been settled, we should
have a clearer idea of what is to count as
evidence pro or can its existence. Now one
such descriptive term is 'try' ('strive',
'attempt'), which in my opinion precisely
613
THE WILL
singles out the phenomenon traditionally
singled out by the word 'will'. However, this
expression suffers from the disadvantage
that pragmatic linguistic presumptions
restrict its application in certain situations
in which truth considerations alone would
license its application. A tendency to
confuse truthfulness of saying with pertin-
ence of saying obscures the vital issue of
truth. While there is absolutely nothing the
matter with expressing the argument in
terms of trying or attempting, it involves
one in argumentative antics one could do
without. Por these reasons alone I shall
instead appeal to the word 'do'. There can
be no doubt that there is a sense of the
word 'do' that is uniquely reserved for
actings (whether intentional of uninten-
tional) and tryings (whether successful or
failing). 'What are you doing?' 'Pixing the
lock', 'Daydreaming', 'Trying to fix the
lock', are acceptable answers, as 'Reflecting
the light', 'Imprinting my image on your
retina', 'Generating knowledge of my pres-
ence in you', are not. These last are unac-
ceptable responses to the above question,
and acceptable answers instead to a differ-
ent question, 'What are you doing?', which
merely amounts to 'What are you causing?'
However intimately linked, these are differ-
ent varieties of 'doing': a fact that is simply
demonstrated by their diverse space-time
properties: the time and place of the action-
doing being that generally of a bodily
action, of the merely-causal-doing that of
an effect. Thus, the act of killing someone
by intentionally starting an avalanche
occurred when and where the murderer
(stood, and with unnerving slowness) did
his repellent deed; but the killing by the
avalanche, which the murderer intention-
ally engineered, occurred precisely when
and where the victim expired. The agent
act-killed in one place and cause-killed in
another. 'Killed' is ambiguous.
I am trying to assemble an argument in
favour of the reality of the will, expressed in
words whose use is agreed, and I am
appealing to the familiar active use of the
word 'do' (as expounded above). The argu-
ment - which I shall set out in the next
614
section - seeks to demonstrate a particular
complex proposition, which is in my opinion
equivalent to an affirmation of the reality of
the will. Namely, that whenever we perform
an action, whether it be intentional or
unintentional, or a trying, whether it be
successful or unsuccessful. then an item W
comes into existence which is endowed with
the following properties:
(a) W is an event,
(b) W is psychological in status,
(c) we 'do' W in a distinctive sense which
is the sense in which we 'do' actings
and tryings - as distinct from that in
which we 'do' mere causings,
(d) the psychological type of W is a 'doing'
in the above sense,
(e) the content of W is the doing of the
intentional deed or trying.
To demonstrate the reality of the will, I
think we need do no more than prove this
complex claim. It will be noted that (a)-(e)
include no mention of the causal powers of
W - which accords with the position
adopted earlier when I urged that Schopen-
hauer's unusual account of the will's causal
properties did not disqualify him from
believing in its existence. However, I can
well believe that to some people such a the-
oretical position must verge upon absurdity.
Is not the production of phenomena pre-
cisely what the will is all about? And does
not the above theory threaten to trivialize
will-theory? Thus, if one omits a causal dif-
ferentia, and endorses (a)-(e) as necessary
and sufficient conditions of the phenomenon
of willing, it looks as if all one need believe
to believe in the reality of the will is that
actings and tryings exist and are psycholo-
gical events. But this seems relatively
un controversial - unlike the theory of the
will. Many people might be ready to endorse
this seeming truism, and at the same time
firmly disavow belief in the existence of the
'volitional' phenomenon posited by Des-
cartes and Hume (etc.). Surely we need to
add a causal clause to (a)-(e) to block this
trivialization of the claim, and to ensure
that the W in question really is the willI
I do not find this objection cogent. It
encapsulates two errors. In the first place it
objects to the claim that (a)-(e) amounts to
an affirmation of the will's existence, on the
grounds that one must know the will's
causal powers to know of its existence. But
this is false. No doubt we must know that
'willing' in some sense explains 'willed' bodily
movements; but one can be wholly ignorant
as to whether the 'willing' encompasses or is
distinct from these movements, wholly
ignorant therefore as to its causal powers,
and know nonetheless that whenever we
act or try, a distinctive immediately-given
active psychological event occurs (which
seems enough to know the will exists). The
second flaw in the above objection is this.
The objection rests on the supposition that
the psychologicality of actions is a non-con-
troversial matter. But how could it be?
Suppose that one believed - as many have
believed - that the physical act (x') of
moving a limb was identical with the limb
movement (x). In that case the claim that
actions were psychological in status would
imply that suitably originated finger-move-
ments (say) could be psychological events.
But could they? Could any causal history
make of such a simple mechanical event
something psychological? It would need to
be a psychological event whose immediate
origin is muscular, whose existence is dis-
covered proprioceptively, which is wholly
distinct from cerebral events! The sugges-
tion is surely absurd. These considerations
make it clear that a belief that bodily
actions are psychological in status commits
one to a position on the constitution of
bodily actions - which is a highly con-
troversial matter.
I think this disposes of the objection that
the complex proposition (a)-(e) is trivially
true, and cannot therefore amount to an
affirmation of the existence of the will.
Nevertheless the question remains: are (a)-(e)
sufficient to demonstrate the reality of the
phenomenon putatively designated by the
traditional term 'will'? I am strongly
inclined to say that they are, bearing in
mind that (a)-(e) ensure the existence of an
active psychological event of type 'doing'
THE WILL
whenever we act or try - which looks like
enough to guarantee coincidence of refer-
ence. However, it must be emphasized that
the issue, at least as stated above, is not in
itself a matter of prime philosophical import-
ance. What in effect we are asking is: what
propositions would prove the traditionalists
right concerning the existence of the will?
This is undoubtedly of philosophical-histor-
ical importance, which is not a negligible
consideration, but not I think of great philo-
sophical moment in its own right. Now
some may detect the unfolding of a sort of
dialectical process in the preceding discus-
sion. With the advent of Constitution-
Theory, and theories as to the nature of the
mental and/or psychological, it might seem
that our concept of will must develop, closing
off some possibilities and opening up others
as it gathers in complexity. In a word, that it
follows a line of development which it need
not have followed. Then how can we be
sure that we are contesting or endorsing the
existence of what earlier simpler conceptual
systems purported to identify? This counsels
a relativistic reading of the situation - which
I resist. Here I have to be dogmatic. I can
merely reiterate that if (a) - (e) is demon-
strable, then the reality of what Descartes
and others had in mind when they spoke of
'the will' will have been demonstrated.
What is at issue is the existence, and omni-
presence in active situations, of an immedi-
ately given psychological event of purely
active ('doing'-) type: a question that surely
transcends philosophical-historical develop-
ment. How can such a question be relative
to era?
THE REALITY OF THE WILL
What is certainly not historical, and is a
matter of prime philosophical importance, is
the truth of (a)-(e). If this can be estab-
lished, we will have established the exist-
ence in animal life of a purely active
psychological event matching the above
specifications. It is at that point that the
causal properties of such a psychological
event become a fit subject for investigation.
Now I am myself of the opinion that good
615
THE WILL
arguments exist in favour of (a)-(e). They
are for the most part variants on the much
despised - but absolutely invaluable - argu-
ment from illusion (error, failure, etc.).
Certain intelligible situations in which (say)
it erroneously seems to a wide-awake
subject in his right mind, who is in fact busy
obeying an order to move a finger (say),
that he has just been and even now still is
the victim of several perceptual illusions,
can reveal that whenever we intentionally
and confidently perform a bodily action we
are mentalistically-immediately aware of a
mentalistically-immediately-caused 'doing'
of something or another. One has merely to
imagine that the above person is brought
perfectly rationally (but erroneously) to
doubt that he is or has been moving his
finger, let us say by his being convinced
that scientifically induced neurological
proprioceptor and ocular events are and
have been generating proprioceptive and
visual illusions. Such a person would know,
and with the authority typical of first-
person present-tense experience, that an
immediately-mentally-caused event of an
active kind (a 'doing') had occurred. He
would know, not merely that he had inten-
ded to obey the order, but that he had in
addition done all that he could to obey: he
would know that an event had expressed his
intention. These situations of apparent
active failure combined with apparent epis-
temological error act as filters, whereby the
common presence across a whole series of
cases of a 'doing' endowed with all the
appurtenances of psycho logicality is exposed
to view. Such a 'doing' is to be found in
normal situations of confident and inten-
tional success, as well as in situations of
total and timorous failure. Along these lines
we can I think establish the existence of a
phenomenon that is the same as that tradi-
tionally singled out under the word 'will'
and disbelieved in by so many twentieth-
century philosophers.
The residual and very important question
remains as to the causal properties of the
aforementioned psychological event. Space
will not permit me to discuss this question
satisfactorily. It is undoubtedly true that
616
seventeenth- and eighteenth-century philo-
sophers understood the will to be both
cause and causal explanation of the 'willed'
bodily movement; and true in addition that
their beliefs concerning the mind-body rela-
tion were a partial determinant of this
theory. I myself do not in any way wish to
play down the significance of the causal
properties of the will, and am convinced
that if the will is a reality it must in some
sense constitute the causal explanation of
the 'willed' bodily movement. Thus, if I had
not willed it, the finger would not have
moved - this, surely, is true. However,
whether or not the will is a distinct cause of
that movement is an additional and difficult
question. The view that recommends itself
to me, for a series of rather complicated
reasons which I cannot here consider, is
that willing is generally a causally sufficient
condition of, but not a distinct cause of,
such 'willed' bodily movement. The proof of
this theory depends upon a contentious
conjunctive proposition: namely, that bodily
actions (like walking) are non-distinct from
the 'willed' bodily movement (like leg
motion) their occurrence necessitates; and
that all succeeded-in attempts-to-do-act-X
are identical with X-acts. The first half of this
conjunction seems relatively obvious: con-
siderations of visibility, locality, etc.,
strongly support the view that (say)
walking does not occur in the head alone! It
is the second half of the conjunction that is
difficult to establish - though there seems
little doubt that the claim or principle is
true of absolutely all instrumental action. I
shall not here try to demonstrate that 'the
attempt to perform basic act X that is suc-
cessful is the basic act X', though I am
myself persuaded that it can be done. Once
the conjunctive proposition is established, a
particular theory of bodily actions seems
natural. This theory involves the supposi-
tion that the bodily action is a willing,
which is at once a psychological event and a
physical event constituted out of nothing
but motor-events, a phenomenon that
spreads developmentally out from mind/
brain to the bodily extremity involved.
When we conjoin this theory with the
theory affirming (a)-(e), we have a theory -
now enriched with an account of the vitally
important causal properties of the phenom-
enon in question - which continues to suffi-
ciently well match the traditional
specifications of willing, to count as a
theory of the will. Whether or not the phe-
nomenon has this latter historical property,
pales in significance beside the truly philo-
sophical question as to whether or not any-
thing has the former complex array of
features. This last is of the first importance
in the understanding of the relation of mind
and body.
See also ACTION; An Essay on Mind section
2.3.; HISTORY; INTENTION; REASONS AND
CAUSES.
BIBLIOGRAPHY
O'Shaughnessy, B. 1980. The Will. (2 vols.)
Cambridge: Cambridge University Press.
BRIAN O'SHAUGHNESSY
Wittgenstein, Ludwig Wittgenstein's philo-
sophy of mind must be seen in the light of
his conception of the nature of philosophical
problems and what is necessary to resolve
them. For he believed that philosophical
problems are puzzles induced by mis-
interpretations of the use of words in our
language and they are solved by paying
attention to the ways in which the prob-
lematic words really are used. Hence it is
wrong to think that the aim of philosophy is
to provide an explanation of various phe-
nomena by means of a theory. Rather,
philosophy must eschew explanation and be
purely descriptive. What it describes are the
ways in which words are used in our lan-
guage, or, as Wittgenstein often expresses it,
the 'grammar' of words. Accordingly, Witt-
genstein's work in the philosophy of mind
does not issue in philosophical theses about
the mind, but is directed towards dissipating
philosophical puzzlement about the mind by
the identification of misleading images and
superficial similarities and by a description
of the actual use of those words that lie at
WITTGENSTEIN. LUDWIG
the heart of the problem. This makes his
philosophy of mind unusually resistant to
summary presentation, its force often deriv-
ing from its particularity. This should be
borne in mind in reading what follows.
Three main themes in Wittgenstein's
philosophy of mind are (i) the diversity or
heterogeneity of mental concepts, (ii) the
illusion of the essential privacy of states of
consciousness, and (iii) the nature and basis
of mental REPRESENTATION or INTEN-
TIONALITY.
THE HETEROGENEITY OF MENTAL
CONCEPTS
A fundamental feature of Wittgenstein's
later philosophy is the insistence that super-
ficial linguistic uniformity conceals 'gram-
matical' diversity, so that the variety of
reality is hidden from our reflective under-
standing by similarities in our means of
representation of that reality in language.
Wittgenstein applies this insight to words in
our mental vocabulary in two ways. First,
there is the simple point that the grammars
of words for SENSATIONS, EMOTIONS,
THOUGHTS, IMAGES, INTENTIONS, and so
on, are very -different from one another,
despite more or less superficial resem-
blances. Accordingly, Wittgenstein is at
pains to spell out these differences and warn
against the assimilation of one kind of
mental state or occurrence to a different
kind (for example, the feeling of how your
arm is moving to the feelings you experi-
ence in your arm when it is so moving
(Wittgenstein, 1953, II, viii)). Second,
there is the more subtle point that even an
apparently uniform mental category can
exhibit considerable diversity in the kinds of
state that fall within it, so that a failure to
recognize this diversity results in a mistaken
assimilation of mental states. Perhaps Witt-
genstein's best illustration of this is the
variety of visual experiences. We can see
colour, shape, a likeness between one face
and another, what a picture depicts, a per-
son's melancholy expression or hesitant
posture, a sign as the mirror-image of an F,
a person's glance, and endless other sorts of
617
WITTGENSTEIN, LUDWIG
thing, All of these experiences fall under our
concept of seeing (see PERCEPTION;
IMAGERY; IMAGINATION). Yet, as Wittgen-
stein shows, there are subtle differences
between the concept of one kind of visual
experience and the concept of another. Con-
sider the experience of suddenly seeing mel-
ancholy in someone's face - a face that does
not change in any way at the moment you
see the melancholy. This is one instance of
what Wittgenstein calls 'seeing an aspect'.
Wittgenstein rejects two opposed positions
as false alternatives. One insists that this
visual experience is on all fours with the
paradigm of a genuine visual experience,
namely the experience of seeing a colour or
a shape. The other insists that it is not
really a case of seeing, but instead some-
thing that is not a visual experience at all,
namely an interpretation of what is seen.
Both positions are mistaken, Wittgenstein
argues, for the experience of seeing melan-
choly - like any experience of seeing an
aspect - shares some, but only some, gram-
matical features with the alleged paradigm,
and some, but not all, features with the
suggested alternative. Whereas seeing an
aspect resembles seeing a colour with
respect to 'genuine duration' - that is, each
is a continuous state the onset and end of
which take place at precise moments -
seeing an aspect, unlike seeing a colour, is
'subject to the WILL', that is, it is the kind of
state which it makes sense to try to induce
in oneself when seeing an object without
distorting the object's visual appearance.
This means not that the experience of
seeing melancholy is in any way proble-
matic, but that the concept of that kind of
experience is not reducible to either of the
proposed kinds: the concept of seeing mel-
ancholy exhibits a significant difference
from that of seeing a colour or a shape and
yet it is distinguishable from the concept of
a mere adjunct of thought to what is
allowed to be a genuine visual experience.
To adapt the well-known remark of Bishop
Butler that impressed Wittgenstein: the
concept of seeing something as we interpret
it 'is what it is and not another thing'. This
is the moral that Wittgenstein emphasizes
618
time and time again in his examination of
mental concepts.
THE ILLUSION OF ESSENTIAL PRIV ACY
Wittgenstein's target in his investigation of
the concept of a state of consciousness, such
as the experience of pain, is the seductive
idea that a state of CONSCIOUSNESS is
essentially private, in the sense that, in
virtue of the nature of a state of conscious-
ness, it is such that only the subject of that
state can know whether he is in that state.
This conception is often thought to underlie
the notable first-person/third-person asym-
metry in the use of psychological words:
whereas the third-person use is based on
observation, the first-person singular
present indicative is not based on observa-
tion. This way of thinking is usually supple-
mented by the claim that the nature of a
state of consciousness guarantees not only
essential privacy, but also the subject's
immediate and infallible awareness of what
his present state of consciousness is.
Accordingly, a person's consciousness is
thought of as an essentially private realm,
and his assertions about his states of con-
sciousness as infallible descriptions of what
happens in this realm - descriptions that
cannot be checked against the reality by
any other person and about which the
person himself cannot be mistaken. (See
INTROSPECTION .)
Wittgenstein argues on a number of
grounds that this is a misrepresentation of
the concept of pain. First, it is false that, in
the ordinary sense of the expression, one
person cannot know that another person is
in pain: on the contrary, one person often
does know that another is in pain. (I return
to this blunt rejection of the essential
privacy of pain later.) Second, the idea that,
in virtue of the nature of pain, the subject of
pain knows ('with certainty') that he is in
pain is a misconstruction of the grammar of
the word 'pain'. It is true that it is senseless
for the subject of pain, but not another
person, to express doubt about whether he
is in pain; but this is not a matter of the
subject's being endowed by the nature of
pain with a uniquely privileged epistemic
authority about the current state of his con-
sciousness. The grammatical rules govern-
ing the word 'pain' are no more derivable
from the nature of pain than are any other
grammatical rules derivable from the reality
with which they deal; and the senselessness
of uncertainty should not be represented as
a peculiarly well-founded guarantee that
whatever seems to the subject right about
his state of consciousness is right. Rather,
the senselessness of uncertainty is a feature
of the 'language-game' we play with the
word 'pain', and a person's assertion that
he is in pain is an 'utterance' (Ausserung) of
the pain, not a description of an essentially
private and self-intimating event. Third, if
pain were essentially private, then, in the
first place, each person would know only
what he calls 'pain', not what anyone else
does. But then what each of us calls 'pain'
would be irrelevant to the use of the word
'pain' in our language: the use of a word in
a public language cannot be constrained by
considerations about what is essentially
private, and whether someone understands
the word 'pain' and uses it correctly is
determined by publicly accessible criteria.
Hence it would not matter whether the sup-
posed private reference of the word 'pain'
were the same or different across people or
in any individual's case: there would be
neither an intersubjective nor an intra-
subjective requirement of common reference
to something essentially private - a require-
ment that would need to be met for the
word 'pain' to have a constant meaning in
the language. This is the force of Wittgen-
stein's remark that 'if we construe the
grammar of the expression of sensation on
the model of "object and name" the object
drops out of consideration as irrelevant' -
the moral of his famous 'beetle-in-the-box'
analogy (Wittgenstein, 1953, 293).
These considerations are buttressed by
Wittgenstein's critique of so-called 'private
ostensive definition'. If there were to be
essentially private events or states that a
subject of consciousness refers to in his own
language, he would himself have to intro-
duce the names of these events or states
into his language, rather than acquire an
understanding of them from others. This
would require him to introduce a word as a
name of a kind of state - an essentially
private state - solely on the basis of his
apparent awareness of the occurrence of the
state. But how would it be possible for him
to do so? It would seem that all he could do
would be to concentrate his attention on
the state he is in on some occasion, as it
were pointing to it in the privacy of his
consciousness, and give himself a quasi-
ostensive definition of some sign'S' by
saying to himself 'This is called "S" '. But, as
Wittgenstein demonstrates, this would be a
pointless exercise: it could not serve as an
introduction of the word'S' into the per-
son's language as a name of some kind of
thing that happens to the person. The
reason it could not do this is that a defini-
tion provides a rule for the use of a word
and this act of quasi-pointing and 'stipula-
tion' fails to specify a rule. But without a
rule for the application of'S' no sense has
been given to the idea that some applica-
tions of the sign will be correct (Le. the
named item will be of the same kind as
the original (this)) and some incorrect (Le.
the named item will not be of the same
kind).
To appreciate the force of this considera-
tion, ask yourself the question, What could
the rule be? It could not be 'Call something
"S" if it is the same as this', for since this in
no way constrains the use of'S', Le. it fails
to make some uses of'S' correct and some
incorrect, it is not a rule. The reason it does
not provide a standard of correct application
is that it is entirely indefinite, failing to indi-
cate the kind that'S' is supposed to be the
name of. To be a rule it would need to be
completed by an answer to the question,
'Same what as this?'. But what completion
would be possible, given that'S' is intended
to be the name of something that is essen-
tially private? The rule could not be 'Call
something "S" if it is the same kind of sen-
sation as this'. Apart from any other con-
siderations, 'sensation' is a word of our
common language, and so its use in the
rule would be legitimate only if'S' were to
619
be used in accordance with the grammar of
words of sensations. which is precluded by
its being intended as the name of something
essentially private. There are two questions
that can be asked about a sign that is
claimed to be the name of a sensation: Is
the sign a word for a sensation? and Is the
sign a word for a sensation (i.e. a sensation
of some kind S, rather than T)? Wittgen-
stein rightly faults the attempt to give sense
to the sign'S' by 'private ostensive defini-
tion' on both grounds: 's' is not used as a
name of a sensation (Wittgenstein, 1953,
258), nor as a name of a sensation (Witt-
genstein, 1953, 261). But he also rightly
goes further. For it would be equally illegit-
imate to claim that'S' is being used as a
name of something the subject has, or as a
name of ... , where the gap is filled by any
words of our common language.
So we reach the conclusion that the
alleged essential privacy of states of con-
sciousness implies not only that one person
cannot impart his private 'knowledge' to
others but that he cannot express it to
himself. He will believe that he can express
it to himself only if he illegitimately allows
himself to use words of our common lan-
guage as the vehicle of his soliloquy. But he
has no right to use any of our words to refer
to what he forever denies us access to.
Wittgenstein's blunt rejection of the thesis
of the essential privacy of pain and his cri-
tique of philosophical scepticism about the
minds of other people is closely connected
with the failure of private ostensive defini-
tion. Both the other minds sceptic and the
believer in private ostensive definition help
themselves to a notion of sameness without
being prepared to accept the requirement
the use of the notion implies. Whereas the
believer in private ostensive definition needs
an intrasubjective notion of sameness but
cannot specify a criterion of identity that
would give content to the idea, the other
minds sceptic needs an intersubjective
notion of sameness, i.e. a notion of same-
ness to apply across people's sensations (or
other mental states), but can maintain his
scepticism - which is thereby rendered inco-
herent - only by failing to specify a criterion
620
of identity. These failings are, of course, two
sides of the same coin: the idea of a kind of
sensation has content only if it has both an
intrasubjective and intersubjective applica-
tion; and the criterion of sameness across
persons is all-important, since your applica-
tion of a name you understand to your
present sensation is criterionless (Wittgen-
stein, 1953, 374-82).
MENTAL REPRESENTATION
A major theme of Wittgenstein's later philo-
sophy is his opposition to a 'mentalistic'
conception of someone's meaning, under-
standing, or intending something by some-
thing he does or encounters; and the target
he aims at is not restricted to these central
instances but includes many other cases in
which a person's state is directed towards
an object or has an intentional content, as
in propositional remembering, imagining,
thinking, wishing or expecting. (See
INTENTION ALITY.)
There are two principal propositions that
Wittgenstein advances against the mental-
istic conception, and these define the men-
talistic view he opposes. First, the
intentional CONTENT of a state is not an
'experience': meaning, understanding,
intending, propositional remembering,
thinking, and so on, do not have an experi-
ential content (Wittgenstein, 1953, II, pp.
216-17, 231). The prime reason that this
is so is that the intentional content of these
mental states lacks the kind of duration
characteristic of an experience: unlike a
pain, which you can begin to experience at
a certain moment, endure for a time, and
which might suddenly cease, there is no
question of your thought that p beginning
at a moment, continuing and then coming
to an end. An intentional state of one of
these kinds does not consist in the con-
tinuous presence to consciousness of its
intentional content, as a pain can remain
present to consciousness. In fact, 'meaning',
'understanding' and 'thinking a thought'
are not the names of processes of any kind,
and so not the names of conscious pro-
cesses. A thought is neither 'articulated'
nor 'non-articulated': it is neither a series of
distinguishable phases of a process occur-
ring in a segment of time nor a uniform
state lasting from one point of time to
another. It follows that an individual's lin-
guistic expressions of his thoughts, mem-
ories, intentions, or expectations are not
infallible reports of essentially private pro-
cesses that take place in him when he
thinks, remembers, intends or expects.
Second, whether a subject is in a state with
a certain intentional content is not deter-
mined by the intrinsic, non-relational
nature of anything that happens to the
subject when in the state but is dependent
upon what else is true of the subject, and
the context in which the subject is located.
The significance of any occurrence, even an
occurrence in the individual's conscious-
ness, is a matter of how it is embedded in
the individual's life and the interpretation
he gives it. If we consider in isolation any-
thing that happens in someone when he
thinks the thought that p - if we abstract it
from the individual's past, his present capa-
cities and dispositions, and the prevailing
circumstances - we will fail to find any-
thing to ground the attribution of that
thought to the person. Even if the occur-
rence is one the subject is aware of, how he
understands or intends it is not a state of
consciousness but a matter of the use he
makes of it. Moreover, it is not a condition
of someone's thinking the thought that p
that he should be aware of anything hap-
pening in him when he thinks the thought.
We must therefore change our focus and
look elsewhere, in particular at how the
individual would represent his thought if he
were to give it expression. For the subject's
sincere expression of his state in words -
what memory, understanding, intention or
thought the subject would attribute to
himself - is criterial of the state's content.
The common conception of a person's
accounts of his memories, intentions or
thoughts needs to be turned upside down.
Rather than looking at a subject's sincere
assertion of what he means, remembers,
intends, understands or expects as a report,
somehow automatically guaranteed to be
WITTGENSTEIN, LUDWIG
correct, of an occurrence the nature of
which he monitors at the moment it
happens, his verbal expression should be
seen as being definitive of the content of his
mental state. His mental state does not
have a nature that is logically independent
of how he would express it: there is no state
with an intrinsic nature specifiable
independently of the subject's dispOSition to
express it by an utterance of 'p' (or some
synonymous sentence) which the subject
reports when he confesses his thought that
p. This might seem to render problematic
an individual's ability to attribute thoughts
to himself, a mystery that can be removed
only by an explanation of this ability. But
remember that Wittgenstein's aim is not to
explain anything, in this case an indi-
vidual's ability to attribute thoughts to
himself. Rather, he wants to encourage a
clear understanding of the language-game
played with the word 'thought'. Hence he
supplements the identification of a person's
utterance of his thought as a criterion of
the thought by drawing attention to the
circumstances in which it would be appro-
priate to attribute a thought to someone
and to the consequences of such an attribu-
tion; for at the level of a description of the
language-game, there is nothing more to
do. As he says, the significance of a truthful
confession of a thought resides, not in its
being a true description of a process hidden
from all but the thinker, but in 'the special
consequences which can be drawn from a
confession whose truth is guaranteed by
the special criteria of truthfulness' (Witt-
genstein, 1953, II, p. 222).
Wittgenstein's approach to the philo-
sophy of mind is rooted in and dictated by
his philosophy of language (and hence his
philosophy of philosophy). The meaning of
a word is its use in the language, and
understanding a word is mastery of the
language-games in which it figures - not
something that underlies and explains that
mastery (such as an awareness of the
nature of what the word refers to). Further-
more, language-games are heterogeneous
and do not need to conform to or be reduc-
ible to some paradigm of language use (as
621
was required by the so-called 'picture
theory' of the Tractatus). Hence when we
have described, compared and contrasted
the various language-games played with
psychological words, which often ramify in
ways that preclude a simple description,
there is nothing for philosophy to do except
to expose false accounts and to diagnose the
insidious seductiveness of such misconcep-
tions.
Wittgenstein's philosophy of mind repres-
ents a radical break with the past. For his
investigation of psychological concepts
shows that the traditional philosophical
picture of the mind is a fantasy. We have
seen that the expression of our mental
events in words is wrongly thought of as
the report of essentially private occurrences
that we observe within ourselves and that
we can only conjecture in others. Further-
more, mental words are not such that we
are able to understand them only in virtue
of being directly acquainted in our own case
with what they refer to - thus possessing
exemplars that guide our use of the terms.
Accordingly, the mind is not an internal
microcosm, the contents of which are inevi-
622
tably hidden from everyone but the subject,
who, in contradistinction from others, has
direct and infallible access to whatever takes
place in his inner world.
See also BEHAVIOURISM; SUBJECTIVITY.
BIBLIOGRAPHY
Budd, M. 1989. Wittgenstein's Philosophy of
Psychology. London: Routledge & Kegan
Pau!'
Geach, P.T., ed. 1988. Wittgenstein's Lectures
on Philosophical Psychology 1946-47
London: Harvester Wheatsheaf.
Wittgenstein, 1. 1953. Philosophical Investiga-
tions, trans. G. E. M. Anscombe. Oxford:
Basil Blackwell.
Wittgenstein, 1. 1960. The Blue and Brown
Books. Oxford: Basil Blackwell.
Wittgenstein, L. 1980. Remarks on the Philoso-
phy of Psychology, 2 vols, trans. G. E. M.
Anscombe. Oxford: Basil Blackwell.
Wittgenstein, L. 1982. I.ast Writings on the
Philosophy of Psychology, vo!' 1. trans. C. G.
Luckhardt and M. A. E. Aue. Oxford: Basil
Blackwell.
MALCOLM BUDD
Index
Note: Page references in bold type indicate chief discussions of major topics or persons. References in
italics indicate figures. Where names of contributors to the Companion are indexed, the references are to
citations in articles other than their own or to specific articles describing their own philosophy.
accessibility 17-19,24,27,
157,396-9
and attitudes 18-19,23
and experience 18, 24,
46-7, 51, 56, 60, 165
and memory 435-6
see also authority, first-person;
consciousness;
introspection; knowledge
acting and action 8, 13-15
action 111-21
and accessibility 18 -1 9
and 'act-beginner' 115-16
and acting 13-15
action control system
118-19,301
akratic 120-1, 607
automatic 117-19
beginning 69-72,114-16
as category of mind 7,
13-14,26,28,61-75
commitment to 376
and directionality 21-2,
112-14
at a distance 156-7, 376
and effect 69-74, 112
and emotion 16, 168
and experience 15, 80
and expressibility ] 9-20
and folk psychology 301,
307
and happening 62-9, 74,
78,112-17,118-20,
121-2
individuation 66-7, 68,
111-12,113,114,115,
220-1
mentalism 112-13
and observability 16-17
as output 266
physical 14-15
and rationality 245, 526,
534,609
selection and control
118-21, 215, 302
and self 15, 550-1.
556-7
and supervisory attentional
system (SAS) 119, 121
and temporality 68-74,
112, 114-16, 614
and theoreticity 23-4
volitional theory 120,
610-17
see also agent; attitudes,
propositional; behaviour;
causation; Davidson, D.;
desire; intentionality;
reasons and causes
activation
and connections 178-9,
206-7
trigger schema 118-19
vectors 129, 312, 314-15,
409
adequacy
computational/empirical
176-9
descriptive/explanatory
159-60
adverbialism 398, 403-4
agency 121-2
awarenessof 116-17,119
child's perception of 252-3,
256
and self-deception 558-9
see also agent; belief; desire;
intention; self
agent
and 'act-beginner' 115-16
and action 15, 26-7, 63-4,
68, 111, 133, 608-9
and behaviour 132-3
causation 115-16
and intentionality 64-5,
113,277
see also belief; desire;
rationality; self
Akins, K. 215-16
akrasia see will, weakness of
Albert the Great 335
Alexander of Aphrodisias 337
Alexandrism 337
Alfarabi, Abu Nasr 334
algorithm
cognitive processing via
181, 182-4
definition 124-5
lookup 126
practicable 125-6, 129,
130
and quantifier logic 12 7-8
Altmann, G. T. 443
analogy, argument from 54,
58-9, 309, 313, 545
Anderson, J. A. 200
Anscombe, G. E. M.
and desire 246
and intention 237, 379
and self-consciousness 552
Aquinas, St Thomas 334, 335,
336-7, 341
Arbib, M. A. 237,442,444
Aristotle
and desire 248
and imagery 355
and imagination 362
and medieval philosophy
333-4, 335-7, 341
and morality and action
608
and perception 460
and pluralism 122
and practical reasoning 485
and reasons for action 80
and substance 343
Armstrong, D. M.
and analytic functionalism
135-6, 318, 323
623
INDEX
Armstrong, D. M. (cont'd)
and behaviourism 1, 132,
135, 418
and distinct existences
argument 396-7
and identity theory 349,
351
and materialism 412
and perception 468-9
Arnauld, Antoine 335
Asch, S. E. 301-2
Ashby, Ross 237
aspect, dualism 92, 198, 342,
513,544
association, and memory
434-5
associationism 173, 296
atomism
mechanistic 341, 343
representational 189,
406-7,409-10,411
attitudes and posture 11
attitudes, propositional 12 -13,
29,31 n.1, 175,488-92
and accessibility 18-19,23
and action 80, 488
attribution 36-46, 232,
450, 492, 498, -561
and behaviourism 135,
137-8
as category of mind 11-13,
15-16, 28-46
causal-explanatory approach
225, 226
and child's theory of mind
254
complement 12, 13, 20, 29,
32; see also content
and directionality 11-12,
20-2, 29-36, 40, 44-5,
95-6, 102-3, 321,
387-8, 487
and eliminative materialism
491-2
and emotions 272
and expressibility 19-20
207, 301, 306, 310,
315, 491
and functionalism 328-9
and holism 231-2, 347
and intension/extension
374-5,488
intentionality 208, 349-50,
375, 387-94, 562
and language 36-8, 39-40,
44-5, 139, 589-93
as local mental phenomena
387-8, 394
and naturalism 488-90
and observability 16-17,
23-4
624
and pro-attitudes 113, 532
problems 29-32
as relational 30-45, 328-9,
388, 403-4, 491
scientific status 491
and Stalnaker 565-8
and teleology 585
and theoreticity 23
verbs of 185, 254
see also behaviour; belief;
consciousness; content;
desire; externalism;
intentionality;
internalism; thought
attitudinizing and attitudes 8,
11-13, 15,28,29-32
Augustine, St 333-4, 335,
336, 337
Austin, John 1. 513, 590
authority, first-person 290,
291, 569, 616
and Davidson 137,235-6
and Descartes, Rene 52-3,
273,291, 541-2, 551
and folk psychology 310
and incorrigibility 291, 352,
396, 514
and infallibility 56, 291,
396-7, 522, 551, 554,
618, 619
see also introspection; self
autism, as theory of mind
deficit 253,255, 256-7
automaton states, mental states
as 323-4, 326
Averroes, Ibn Rushd 334,337
Averroism 337
Avicebron (Solomon ibn
Gabirol) 337
Avicenna (Ibn Sinal 334
awareness
of agency 116-17,119,
238
and consciousness 241-2,
396, 550
pre-introspective 395
of reasons 66, 116
see also introspection;
perception
Baddeley, A. D. 438, 439
Baier, A. 270
Baillargeon, R. 372
Baker, L. R. 304, 305, 425
Baron-Cohen, S. 256
Bartlett, F. C. 440
Barwise, Jon 176
Bayesian models 271
Beakley, B. 323
behaviour
as action 416-17,472
agential 132-3, 135
and belief 261-2, 558-9
cognitive: and function
specification 180; and
proof 180-1, 184
and Dretske 263-4
logical 245
as necessary 136, 139
and observability 54-5,
104-5, 138-9
physical 132-3, 134-5,
138-9
prediction 150, 300-2, 309,
416, 499
and propositional attitudes
11, 13, 17, 20, 23,
292-3, 488, 490-1,
523-4
as sufficient 136, 138, 139
see also action; causation;
pain; Quine, Willard Van
Orman
behaviourism 132-40
analytic (logical) 134-5,
137,417
concessive 89-90, 104
contemporary 135-9
eliminative 90, 104-5,
133-4, 137,292, 542
and functionalism 99,
135-6,324
and intentionality 380
and knowledge and
experience 56-7, 505
and language 138-9, 160,
170, 522
logical 325, 579, 582
and mind and action
89-90, 231, 309,
323, 326
psychological 133-4, 170,
319, 500
and rationality 428-9
refutation 139-40
and representation 538
Rylean 132, 135, 136, 417,
542
supervenient 137-8, 139
see also Davidson, D.; Dennett,
Daniel C.; desire; Quine,
Willard Van Orman
Bekerian, D. A. 437,439
belief
attribution 36-45, 137-8,
561,
566-7
and behaviourism 89-90,
135,136-9
and consciousness 468
consistency 527
de re/de dicta 38-42, 145,
241
and Dennett 239
directionality 11-12.29-
36. 95-6. 382-3.
386-7
and emotion 270. 271. 273
and environment see
externalism; internalism
epistemology 146-51: and
folk psychology 146-9
(fallibility of 149-51);
and limits of knowledge
151-2. 162-3;
Observational Theory
149; and Simulation
Theory 147-9; and
Theory Theory 147-9
false 30-5.41. 147.226-7.
255.256-7.262.
496-7. 574. 593
in folk psychology 146-9.
301. 308. 422-8
fragmented 566
and functionalism 99.
150-1. 380. 399. 405
as indication 563
informational model
259-63
and intentionality 30.
136-7.239.375-7.
379. 382-3
and introspection 398-9
and language 36-8. 39-40.
255. 589. 591-2
as learned 227
metaphysics 140-6:
propositional theories
140-2. 144-5;
sentential theories
142-5.401-2
as natural kind 450
purpose 228. 229
as relational 30-45. 145.
328-9.404-5
and representations 401-2
styles 38-42
systematicity 329
truth conditions 226-7.
228-30. 247. 401. 426
and verification 134.227-8
see also attitudes.
propositional; causation;
content. propositional;
desire; expectation;
ontology; opinions;
perception; rationality;
reasons and causes;
self-deception
Bennett. Jonathan and
intention 194
Bentham. Jeremy and
intention 376
Berkeley. George
and concepts 188
and existence and perception
266
and imagination 364
and innateness 366
and will 133. 611. 612
blindsight 305. 328. 396. 460
and consciousness 215.218
Block. N. 189
and conceptual role
semantics 193. 196.
198
and dual aspect semantics
513
and functionalism 318. 322.
323
Bobrow. D. G. 436
Boethius. Anicius Manlius
Severimus 333-4
Bonaventure. St 335. 336
Bowers. J. M. 439
Boyd. Richard 511
Boylls. C. C. 442
brain
as algorithmically calculable
125. 126. 128-9
and anomalous monism 122
and behaviourism 134
and belief 41
and computational-
representational systems
(C-R) 153-5. 301.
595-6. 604. 606
connectionist models 154.
200. 302
event-related potentials
(ERPs) 154-5
and mind 78. 82-3. 86-7.
90-7. 105. 236-7
and network approaches
200
neural net models 154
and pain 456-8
and personal identity 555
pineal gland 87-8.268
as recursively approximable
129
split 305
as syntactic engine 237
see also computational models
Brand. M. 120
Brentano. Franz 521. 524-5
and intentionality 191. 237.
321.379.387
Broad. C. D. 278-9
broad content see content
Broadbent. Donald 170
Bruner. J. 444
Buddhism and subjectivity 572
Burge. Tyler
and concepts 190-1. 591
and content 222. 223. 286.
592-3
INDEX
and meaning and
environment 195
491
Burke. D. 119
Butler. Joseph 618
Cambell. Keith 280
Carlson. Greg 447
Carnap. Rudolf 132. 416. 417
Carroll. Lewis 11
Cartesianism
and dualism 87-9.92.
156-7. 164.265-7.
291. 317. 323. 326.
338-41. 349-50
and knowledge of experience
52. 54. 55-8. 163.
354-5
334-5. 337
and other minds 53-5. 137
see also cogito; Descartes. Rene
Cassiodorus. Flavius Magnus
Aurelius 333
Castaneda. H. N. 378
categorization and prototype
205
causation
of action 80-1.82-3.111.
113. 1 1 7 - 2 ~ 229-30.
231. 249. 277-8
agent 115-16
and algorithms 181-4
and behaviour 162. 263-4.
490-1
and belief 534
and causal semantics
226-7. 286. 299
and content of states 220.
286-7.425-30
and desire 117. 120.
229-30. 244-6. 248-9.
599-601
deviant causal chains 113.
117.120
and emotions 274
and epiphenomenalism 277.
582
as extensional relation
278-80
and Fodor 513
and folk psychology 308-9.
310.416-20.421
476. 513. 534
and Hume 231. 267-8. 397
and imagery 361
and information 262-3
136-7. 207. 292-5.
383-4.408.534-5
625
INDEX
causation (conCd)
and Kant 231. 557
and mind 84. 87-9. 95.
136
and neurophysiology 82-3.
84-6. 96
nomic subsumption
approach 279
and physicalism 281-7.
420. 472-3. 475.
481-4
and properties of mental
states 349. 351-2.
476-8
225. 226. 490-1
and psychoanalysis 494-6
reason as 376
and Searle 544-6
structuring 265
and will 614. 615-16
see also concepts;
overdetermination;
reasons and causes
central systems
as domain-general 251-2.
257.410-11
and modularity 442-6
Chalmers. D. J. 518-19
charity principle 233. 526-7.
530
Chase. W. G. 445
Cherniak. Christopher 528
Cherry. Colin 170
children. conceptions of mind
170.250-7
and beliefs 147. 254-6.
571-2. 574
development 498. 605
domain-general theories
251-2. 256
domain-specific theories
251-5. 504-5
and egocentrism 251-2.
257
and pretence 254-6
and the unconscious 602-4
Chinese Room Argument 237.
546-8
Chisholm. R. 325
Chomsky. Noam 153-67
and cognitive psychology
171. 538
competence/performance
529
and computational-
representational systems
153-5
166
and innateness 367-70.
504.512
626
and language 154-5.
157-67.171.367-70.
450. 504. 593
and methodological/
epistemological dualism
164-5
Church. A. 125. 527
Church-Turing thesis 125.
126
Churchland. P. M. 146. 150.
205. 300. 302. 304-5.
310-11
353.474.491
coextensivity and concepts
191-2
cogito 265. 266. 335. 339.
341. 396
see also Cartesianism
cognition
cognitive theory 181. 184.
210
definition 167
and emotion 168.271-3
and language use 405. 409
as pattern recognition
204-5
and perception 168
as reasoning 204. 206
as statistical/numerical
processing 177. 178.
179
as symbolic structure
processing 177. 178.
179. 204. 206. 208.
503
and systematicity 205-6
see also computational models
of mind; connectionism;
psychology. cognitive;
science. cognitive
cognitive psychology see
psychology. cognitive
Cohen. L. 529
coinstantiation
and concepts 191-2
and properties 412-13.
421. 425-6
Colebatch. J. G. 119
Coleridge. S. T. 361-2
Colvius. Andreas 335
common sense see folk
psychology
communication
and Davidson 232-4
pre linguistic 2 5 3
compatibilism. and psychology
495. 496. 573
competence/performance
529-30
Complexity Theory 171. 529
compositionality. and
conceptual role semantics
198-9
computation
abstractions 176-80.
183-4
of child 254. 255
domain-specific 251
computational models of mind
101-2. 154. 176-84.
206. 225
and activation vectors 129.
311.314-15.409
and alternative abstractions
176-7
classical model 177. 206
172.285.442. 503-4
computational/empirical
adequacy 176-7.178.
179
connectionist 106. 173.
176. 180-1. 182. 200.
314-15.604: local
177.178.182.183;
parallel distributed
processing (PDP)
177-8. 179. 183. 184.
200. 491
105-6
integrated connectionist and
symbolic (rCS) 178-9.
183-4
and perception 180. 442.
462
and philosophy of mind
182-4. 296-9
physical realizability 177.
178. 183. 285-6
and qualia 515
and rationality 527
and representation 538-9.
540
and sentential theory of
belief 142-3. 410
symbolic 176. 177. 178-9.
182-4. 296
varieties 179-82
see also functionalism; Searle.
John; Turing. Alan M.;
Turing machines
computer
as algorithmically calculable
125
as symbol processor 123-4.
125.142-3.177.200.
296
see also intelligence. artificial
concepts 185-92
acquisition 261. 408. 521
and analytical philosophy
187
atomistic theory 189,
409-10,411
and behaviourism 56, 522
causal approaches 190-2
classical theory 187-8, 192
cluster 216, 424
in cognitive psychology 168
as constituents of
propositions 185
and guessing 189-90
and imagination 362
individual 186
inferential role 189, 190,
192,298, 513
informational theories 191
as innate 189, 370
and language of thought
408, 586-7
and Logical Positivism
188-9
and mental representation
186
and normativity 451
phenomenal 515
possession conditions 221
and predicates 186
primitive 188, 191
and properties 486, 521-2
and prototypes 189, 192,
477
and senses 186-7
as shareable 186, 187, 189
and Wittgenstein 188, 189,
512,617-18
see also attitudes,
propositional; content,
conceptual; disjunction
problem; Dretske, F.;
empiricism; imagery;
representation;
semantics, conceptual
role; thought
conditioning, operant 134
connectionism 200-8
and artificial intelligence
122, 123, 125, 129,
130,410-11
and associationism 296
173, 175
and cognitive
neuropsychology 172
and computation 106, 176,
177-8, 180-1, 182,
200, 313, 604:
integrated connectionist
and symbolic theory
(rCS) 178-9, 183-4;
local connectionist 177,
178, 182, 183; parallel
distributed processing
(PDP) 177-8,179,
183, 184, 200-1, 491
and eliminativism 106, 178,
183, 207, 491
and language 173, 206-7,
368-9
210, 409, 411
and models of the brain
154, 200, 302
and modularity 203, 443,
444-5,447
and networks 200-8, 422:
designs 203;
feedforward 203, 204;
pattern recognition
204-5; and reasoning
204; recurrent 203-4;
recursive auto-associative
memory (RAAM) 206;
tensor product 178,
206
philosophical implications
206-8
and representational theory
206-8,409
see also epistemology; folk
psychology; learning;
perception
consciousness 210-18, 334
access-consciousness 26,
213-16, 216-18
accessibility 518, 548
as ambiguous term 216-17
conflations 217-18
and Connection Principle
548-9
and content 219,468,
587-6
as cultural construct 217
and directionality 20-2
and eliminativism 211-12,
217
and emergentism 576-7
and functionalism 103-4,
211, 214, 322, 507
and identity theory 354-5
and imagery 363
and introspection 133-4,
396
and knowledge 48
and language 242
and Locke 344, 552
and meaning 237, 588
and memory 434-5
monitoring-consciousness
213, 216-18
Multiple Drafts model 242
neural basis 210-11
and perception 460, 462
phenomenal 26, 210-18
INDEX
and physicalism 211, 479
12-13
and psychoanalysis 493
self-consciousness 213, 216,
550, 551-2, 556
and thought 587-8
and transcendentalism 211
transitive/intransitive
213-14
as virtual machine 241-2
see also awareness; Dennett,
Daniel C.; Descartes,
Rene; experience;
intentionality; pain;
Searle, John; Strawson,
P. F.; Wittgenstein,
Ludwig
constant, logical 194
Constitution Theory 615
constructivism 574-5
content 20, 29, 33, 37-8,
102-3, 219-24, 254-5,
349,487,488-90
attribution 220-1,224-5,
238-40
and causal semantics
226-7, 286, 299
conceptual 220, 222,
226-7,469-70
and conceptual role
semantics 103
den 4 2 6 - ~ 429-30
and externalism 221-4,
228, 232, 2 8 ~ 2 8 9 - 9 ~
298-9, 321, 465-6,
511, 565, 585-6
and holism 347, 407, 526
and internalism 289-90,
329
and language 592 - 3
of mental image 361
narrow 41-2, 223-4, 241,
290, 329-30, 404,
425-7, 429-30, 491,
502, 513, 518
perceptual 51,220-1,247,
342, 365, 383-4, 462,
463-70: appearance and
content 464-7;
conceptual/non-
conceptual 469-70;
and consciousness 211,
468; and experience and
belief 468-9; and
externalism 222-5
phenomenal 214, 216
possession conditions 220
propositional 12-13, 396,
531: ascription 593;
and belief 29-38, 141-
3, 246, 380-3, 386-8,
627
INDEX
content. propositional (cont'd)
392-4. 487. 565; and
identity theories 102-3.
488-9; and
intentionality 381-2.
392-4. 487. 562; and
knowledge 47-8.49.
58-60; and teleology
221. 585; truth
conditions 254-5.
564-5. 566-7
and rationality 220. 221.
224. 232. 245. 428-9
and realism 239-40
representational 226-30.
246-8. 398. 401. 585:
and consciousness 210.
214. 216; and
intentionality 321; and
interpretational
semantics 225-6;
learning 261-4; and
success semantics
229-30; and teleological
semantics 227-9
and subjectivism 220
subpersonal 224
tensed 426
wide 41-2. 166. 228. 230.
241. 290. 321. 330.
406.423-7.429-30.
502. 581
see also attitudes.
propositional; belief;
consciousness; Davidson.
D.; Dennett. Daniel C.;
desire; DFetske. F.; folk
psychology;
functionalism; holism;
intentionality; Lewis.
David; supervenience;
Wittgenstein. Lu':wig
contention scheduling
mechanism 118-19
context and content 287.
585-6
covariation
and concepts 191-2
and supervenience 576.
577-8. 582
creationism 190. 337
Crick. Francis 210
Cummins. Robert 297.
319-20. 501. 504
dAndrade. R. 301
Davidson. D. 231-5
and action and cause 117.
231. 490. 534-5
and anomalous monism 92.
100. 122. 231. 334.
475-6. 523
628
and behaviourism 132.
137-9
and cause-meaning
compatibilism 495
and content 220. 227
and first-person authority
137.234-5
and holism of the mental
231-2.475
and intention 375.377
and interpretivism 137-9.
232. 235. 526. 527.
535
and mental properties 92.
231. 475-6
and 'other minds' 234-5
and pro-attitudes 532
488
belief 142
and supervenience 576. 578
and truth 233-4
and type epiphenomenalism
280. 351-2.475
and weakness of will 609
Davies. M. K. 214. 216. 223.
415
Davis. L. H. 115. 120
de Groot. A. D. 445
decision procedures 168
lookup 126-7. 129
semi 127-8
definition. private ostensive
619-20
Dell. Gary S. 447
Dennett. Daniel C. 236-43.
501. 504
136-8.139
and brain as syntactic
engine 237
255
212. 213. 217. 236-8.
241-2. 243
and content 220. 236.
239-41. 243
and folk psychology 40.
300. 304. 305
and homuncular
functionalism 238.
319-20
and instrumentalism 492
and intention 208. 236-8.
239-40
241. 409
and opinions 241
and qualia 242. 514. 516.
518
dependence and externalism
223
Descartes. Rene
238. 334. 339. 341.
354. 396
and dualism 87-9.265.
266-8.291. 317. 334.
338-41. 348
and emotions 270.273
52-3.273.291. 335.
541-2. 551
and imagery 356
and innateness 366. 370
on knowledge of experience
52. 163
and will 610. 611. 612.
614-15
see also Cartesianism; cogito;
scepticism
descriptionalism 356. 359-60
designators. rigid/non-rigid
326-7. 353. 418-21.
476. 519
desire 244-50
245-6
and belief 245-7
as cause of action 117. 120.
229-30. 244-6. 249.
599-601
and content 245-8. 421
directionality 11-12.20-1.
30. 382-3
and emotion 270.271. 383
308. 427-8. 601
and goodness 244-5.
247-8
245-6.248. 375-7.
379. 382-3
law of preponderant desire
249
object 245-9. 532
and perception 244. 247-8.
464
as pro-attitude 532
and psychoanalysis 497.
598
satisfaction 229-30. 245.
247-50. 532. 598-600.
604-5
strength of 249-50.271
see also attitudes.
reasons and causes
determination and physicalism
576. 578
Dev. P. 442
Devitt. Michael 193
Dilthey. Wilhelm 496
directionality
and action 21-2. 112-14
and attitudes 20-2. 29-36.
40.44-5.95-6. 102-3.
321. 387-8. 487
and experience 20-2. 25
96. 103. 381-3.
386-8
see also belief; consciousness;
desire
discourse representation theory
175
disjunction problem
and concepts 191. 226. 262
and content 226-7. 228.
423. 465. 489
and property 297-8. 420
domain generality 256.
410-11
and central systems 251-2
domain specificity 256
253-4
and input systems 251. 252.
370.410
in knowledge 171.504
and modularity 442-6
Dretske. F. 259-64
and behaviour 263-4
and beliefs 226-7. 259-63
and concepts 186. 191.
192. 260. 261. 264
and experience 259. 260.
263. 264
and externalism 259-60
and information-based
semantics 195.
259-63.489. 512-13
and knowledge 259-60.
262-3
and learning 260-2.
263-4.287
and naturalism 259-62
and perception 259-60.
469
and teleological
functionalism 261. 320
Dreyfus. H. L. 205
Dreyfus. S. E. 205
dualism 265-8
Cartesian 87-9. 92. 156-7.
164. 236. 265-8. 291.
317. 323. 326. 338-41.
348
conceptual 231
336
functional 100
medieval 333-4. 337
methodologicai/
epistemological 164.
165
property 92. 337. 349.
473-4. 521-3. 524-5.
544
and spatiality of the mental
480
substance 521
and will 611
Duhem. Pierre 188
Dummett. M. A. E.
and behaviourism 132. 137.
138
and language 163. 164
Duns Scotus 335. 337
ego 605
egocentrism of child 252.257
Ekman. P. 168.271
eliminativism 104-6. 107.
270. 300-5. 530
behaviourist 90. 104-5.
133-4. 137. 292. 542
and connectionist
computational models
106. 178. 183. 207.
491
and consciousness 211-12.
217
materialist 105-6. 133.
157. 240. 265. 270.
310-11. 315. 352-3.
471. 491-2. 522
physicalist 474. 479
Elman. J. L. 204
emergence see supervenience
emotion 270-6
and cognition 168.272-3
and content 421
as experience/attitude/action
16. 25-6
intensity 275
and intentionality 274-5.
383
neglect of 271
and pain 453
and passivity 272
and rationality 275-6.499
reason for 532
and self-knowledge 273-4
taxonomy 270-1.274.275
empathy. and Quine 524. 525
empiricism
190
369-71. 504-5
and possibility 266
and thought 295-7
INDEX
epiphenomenalism 95.
277-87
and causality 277. 582
and computational theories of
mind 181
consequences 277-8.
280-1
and content and
consciousness 286-7.
588
and information theories
263
and language rules 160-1
neurophysiological no-gap
argument 278
and physicalism 483-4
and Searle 546
type exclusion thesis 283-5
type (property) 278-1.284.
351-2.475
and universal physical causal
comprehensiveness 283.
284-6
epistemology
and connectionism 208
foundationalist 396
genetic 170
naturalized 505-6
ethics
see also will. weakness of
Evans. G. 222. 551. 554. 556
events. mental
and Davidson 231
as neurophysiological events
278-9. 281
as physical events 282
expectation
as belief 19-20
and communication 233-4.
422
experience
and accessibility 18. 24.
46-7. 51. 56. 60. 165
and action 15. 80
and belief 259. 468-9
as category of mind 28.
46-61
and consciousness 8. 9-10.
26. 550
and directionality 20-2. 25
and expressibility 19. 25.
46-52. 55
information-based theory
260
and memory 434. 436. 440
and observability 16-17.
23. 24-5. 46-7
and other minds 52-5. 61
phenomenal character 463.
466-8.469-70.
568-70
629
INDEX
experience (cont' d)
12-13
representational approach
264
as self-reflexive 553
and theoreticity 22-3, 25
veridical 267-8
see also consciousness;
Dretske, p,; Kant,
Immanuel; knowledge;
pain; qualia; sensation
experiment in cognitive
psychology 169-70,
174-5
explanans/ explanandum
288-9
explanation
and explanatory exclusion
490
explanatory gap 210-11,
518-19
intentional 185-6, 262-4,
292
philosophical and
psychological
hypothesis 164-5
and physicalism 76-9,
82-3,97,157,412-13,
481-2
psychoanalytic 27,
493-500
see also reasons and causes;
Simulation Theory;
Theory Theory
expressibility
and attitudes 19-20
and experience 19,25,
46-52, 55
and knowledge 47-8,
49-53, 55-6, 58-60
of mental states 19-20, 25
extension 339-40, 341. 374,
385, 525, 535
and concepts 186, 192
see also intension/intensional
externalism 289-90
and content 221-4,227-8,
232, 287, 289-90,
298-9, 321, 465-6,
511, 565, 585-6
and functionalism 329, 511
historical 287
and knowledge 260-1
and meaning 233-4
see also internalism
feedback mechanisms and
action 114, 116, 118,
119
Feigelbaum, E. A. 124
Feldman, Jerome 177
630
Ficinus, Marsilius 337
fiction and attitudes 31-2
Field, H. 196-7,329,417
fit, direction of 272, 381-4
Flanagan, O. and
consciousness 211. 212,
217-18
Fodor, Jerry 292-9
255, 256
and concepts 185-6, 189,
191, 192, 227, 262
and connectionism 205,
206, 296
and content 223, 228, 298,
299,489-90, 513
292-3, 295, 299, 300,
304-6, 304
and functionalism 319-20,
323, 329, 504
semantics 195, 410
and innateness 175, 367,
370-1
input systems/central
systems 251,410
and intention 186,292-9,
320, 502-3
172, 295, 329, 370-1,
391,409,410-11,
489-90,501
and learning 294
and methodological
solipsism 195-6
modularity hypothesis 171,
175,441-4
and other minds 292
and psychology 292, 501
and representational theory of
mind 206, 299, 391
and semantics 295-6, 321
and teleology 320
folk psychology 300-6, 308-
15
and behaviourism l33,
135-6
and belief attribution 146-9
and connectionism 207-8,
305-6, 312, 313
consequences 310-12
content 301-2,421-8
development 309-10
and eliminativism 104-6,
133,269, 300-5,
310-13, 353,474, 522,
530
and ethics 506
form 147-9,300, 302-3
and functionalism 102, 310,
311, 319, 325
and irrationality 493-4
421-3, 426, 474
and Lewis 416-26
origins 308-9
and rationality 427-8, 530
and reduction 311-12, 536
and self 553
as social practice 312-13
status 300, 303-6, 308-15
transformation 314-15
see also attitudes,
Chomsky, Noam;
Dennett, Daniel C.;
desire; Dretske, F.;
dualism; Fodor, Jerry;
inference; intentionality;
representation;
Simulation Theory;
Theory Theory
form and matter 336, 337,
341, 345
Forster, K. l. 445
Frege, Gottlob
and desire 250
and extension of sentence
374
and mode of presentation
222
and propositional theory of
belief 140-2,232
and public language 165-6
and sense and reference
186,219, 388, 391
and set theory 526-7
Freigius, Johannes 333
Freud, Anna 602
Freud, Sigmund 293, 493-4,
572
and the brain 604-5
and memory 600-1, 604
and the unconscious 396,
497-8, 598-604,
605-6
Frith, U. 256
functionalism 317-22,
323-31, 351
analytical 102, 135-6
and behaviourism 99,
135-6, 324, 542-3
and belief 99, 150-1. 261,
380, 399, 405
and chauvinism 330-1
and computational model of
mind 100-2, 105-6,
180, 184,298, 319,
323, 380, 596
conceptual 318, 325,
326-8, 331
211.214.322.507
321. 329. 514
and externalism 329. 511
310. 311. 319. 325
history 317-18.323
homuncular 240. 319-20
320-1. 380
machine 318.319.320-1
and meaning 329
and physicalism 323. 324.
325-8. 476
and post-functionalism
512-13
328-9
581
as utopian 510-12
see also causation;
computational models;
Dennett. Daniel C.;
Dretske. F.; Fodor. Jerry;
holism; identity theories;
Lewis. David;
psycho functionalism;
Putnam. Hilary; qualia;
teleology
future and intention 375-9
Galanter. E. 237
Gall. Franz Joseph 441
Garon. J. 207. 305
Gassendi. Pierre 348
Geach. P. T. and behaviourism
135. 136. 325
Gentzen. G. 194
Gibson. J. J. 461. 570
Ginet. C. 115. 120
Glanvill. Joseph 479
Goclenius of Marburg 333
Godden. D. R. 438
Goldman. A. 148. 302-3.
313. 327. 506
Goodman. Nelson 369. 505.
537. 539
goodness. and desire 245-6.
248-9
Gordon. Robert 272.274.
302.313
Graber. M. 372
Graham. G. L. 304. 305
grammar
generative 160
harmonic 179
transformational 367
universal 157.159-60.
179. 367-9
Grice. H. P.
and experience 267
and intention 194. 377.
590
Grossberg. S. 200
Gunderson. K. 322
Habermas. Jiirgen 496
Hammersley. R. H. 437
Hanko-Summes. S. 372
Hanson. A. R. 442
H a r ~ R . M . 576.578-9
Harman. Gilbert 189. 592
and conceptual role
semantics 193. 196.
198. 323
and innateness 367
and perception 468
Haugeland. John 501
Hebb. Donald 237. 296
Heidegger. M. 313
Hellman. G. 576. 578
Hempel. C. G. 132. 134-5
hermeneutics and psychology
496
heuristics and rationality 529.
530
Hinton. G. 203
history. medieval and
Renaissance 333-7
humanism 333. 337
scholasticism 334-7. 341.
345. 379
history. seventeenth and
eighteenth centuries
338-46
mind-body problem 338-43
and personal identity 343-6
Hobbes. Thomas 334. 343.
348
holism 347. 523
confirmation 188. 190
and content 347. 407. 526
and Davidson 231-2.475
and functionalism 329-30.
516
405.406-7
meaning 188-9. 190. 198.
328. 347
and physicalism 475. 477
see also semantics. conceptual
role
Hopfield. John 201
Horgan. T. 304. 305. 518-19
Hornsby. J. 120
humanism. Renaissance 333.
337
Humberstone. I. L. 415
Hume. David
and causation 231.267-8.
397
and concepts 188
and desire 247. 248
INDEX
and imagination 266.
362-3
and innateness 366
and knowledge 155
and personal identity 345.
551. 570
and will 611. 612. 614
Humphrey. N. 518
Humphreys. G. 214. 216
Husser!. Edmund 363.471
Huxley. T. H. 210
hypnosis and the unconscious
598
icon and representation
536-7. 538. 539. 540
identification of words 172.
173
identity. personal 526. 551-2.
555-6
in Hume 345. 551. 570
in Kant 345-6. 551.
552-3. 554. 556-7
in Locke 343-5. 552. 555.
574
and memory 344. 552
identity theories 348- 5 5. 597
arguments against 97. 240
and computational theory
184
and content of attitudes
102-3. 488-9
352-3. 418-20
507-8
and necessary identity
353-4
and physicalism 323. 348.
471. 481. 482
and properties of mental
states 348-52. 432
subjective features 354-5
token 96. 350-1. 355: and
functionalism 100. 318;
and Lewis 420; and
physicalism 283. 318.
326. 475. 481. 483; and
488-9; and Quine 523;
and supervenience
93-5
type 90-2.93.96.350-1.
355. 432: and
functionalism 97-101.
317-18.321; and
Lewis 420; and
physicalism 21-2.
326-7.472-3.475.
477.481-3. 536. 578.
631
INDEX
identity theories (cont'd)
582; and propositional
attitudes 489-90; and
Quine 522-3; subtype
420
see also Leibniz's Law; Lewis.
David; Quine. Willard
Van Orman
illusion
argument from 616
and perception 460-1.
465-6.467.468-9
imagery 355-61
in cognitive psychology 168
and concepts 188
content 361
and Dennett 242. 244
descriptionalism 357. 360-1
and imagination 364.
365-6
picture theory 355-60.
364-5. 540
and representation 539-41
see also concepts
imagination 361-6
and mental images 363.
365-6
and perception 362-3. 366.
463
and picture theory 364-5
and picturing 363-4
and possibility 266-8
and pretence 364
primary/secondary 361-2
and 'seeing-as' 363
incorrigibility see authority,
first-person
Indices, and representation
536, 538, 539
individualism, and language of
thought 406
infallibility see authority. first-
person
inference
505
and concepts 189, 190,
192, 298, 513
and conceptual role
semantics 198-9. 347
and developmental
psychology 256
existential 385
and folk psychology 147-8,
301-2
ideal 528-9, 530
and inferential role 51, 196,
298
rules 182
and Searle 545
632
belief 143-4, 407
information
encapsulated 443-5, 446
and knowledge 259-64.
489
perception as information-
processing 461-2
see also semantics,
information-based
information theory
l70-1
and concepts 191
and content 260-4,
512-13, 565-8
and meaning 194-5
and memory 436
innateness 175,366-72,512
and children's theory of
mind 251, 253-5
and language acquisition
367-70, 504-5
370-1,408
input systems. as domain-
specific 251.252,257,
410. 442-3
instrumentalism 208, 404,
492, 528
intelligence, artificial 101,
122-31. 225
algorithmicity assumption
124, 125-31
171-2,319
frame problem 410
and General Problem Solver
171
and intentional stance 239
Logical Theorist 124, 125,
171
and procedural semantics
193. 298
'strong' 380, 546-7
symbol system hypothesis
123-4, 125. 182. 503.
539
see also computational
models; connectionism;
modularity
intelligibility and content of
states 219-20, 221-3
intension/intensional 374-5,
380, 385-6, 387
and concepts 186,192
see also extension
intention 375-9.488
and all-out judgment 377
and belief 375-7, 379
and commitment to action
376
desire-belief model 375-6
as directed 12, 30
and future 375-9
in-action 384
prior 384
shared 379
intentionality 379-94
and action 112-14, 116,
117, 120, 166. 235.
375-6. 383-4, 408
and agent 64-5, 68-9.
113. 277
and akratic action 120-1
attribution 533
background 384-5. 549-50
co-reference/no-reference
problems 388-91, 488.
492-3
239, 321, 501-3
and consciousness 120,
236,237-8. 380,
383-4. 587-8
and content 286-7, 376.
381,488, 514, 535.
564
and deduction 566-7
and directionality 30, 34,
96, 103, 381-3, 386-8
and epiphenomenalism 277,
286-7
304, 308, 311. 375
and functionalism 103. 239,
320-1. 380
and information 263
and intensionality 380,
385-6, 387
and intentional stance 240
irreducibility 191,237, 321.
380
and meaning 194, 386
in medieval philosophy 335
and mental properties
349-50
network 384
and objects of emotion
274-5
and rationality 65-6. 239,
295. 379, 532-3,
535
and representation 207,
321, 386. 391-4,
537-8,539,561-3
and satisfaction conditions
382, 3 8 3 - ~ 548, 549
and self 557
semantic coherence 295-6,
298-9, 584
and sensation 398
structure of intentional
states 380-3
and teleology 320
and thought 194.298.
585-8
two-tiered accounts 392-4
unconscious 238. 380. 384
and will 377. 611: and
weakness of 609
see also attitudes.
propositional: belief:
causation: content.
propositional: Dennett.
Daniel C.: desire:
emotion; Fodor. Jerry:
memory: perception:
Searle. John: Stalnaker.
Robert
interaction. mind and body
267-8
internalism 289-90
and language 158--60. 163
see also externalism
interpretation
radical 526. 527
and reasons 533. 535
interpretivism 137-8.232.
235. 526. 527. 535
introspection 395-9
and distinct existences
argument 396-7
234. 237. 310. 396.
542
and imagery 356
and knowledge of experience
52. 396. 397. 571-2
and mental images 360
and modularity hypothesis
445-6
object-perception model
397-9
and pain 18. 396-7
as perception 21. 397-9.
464-6. 470
and psychology 133-4
and theoreticity 22 - 3
see also accessibility:
authority. first-person:
consciousness
irrationality
and Davidson 232. 235
and self-deception 559-60
Isidore of Seville 333
Jackson. Frank 420. 518-19
and knowledge argument
322.413.479.514-15
Jacobs. R. A. 203
James. William
and agent causation 115.
116
and emotion 273
and memory 433
and psychological
behaviourism 133
Jaynes. J. 217
Jespersen. Otto 160
John of St Thomas 334
Jordan. M. I. 204
Jung. C. G. 496
Kahneman.D. 446.529
Kamp. Hans 1 75
Kant. Immanuel
and anomalous monism 122
and causality 231. 557
and desire 244-5
and experience 372. 552 - 3.
556
and imagination 362-3
and knowledge 310. 372.
556
and personal identity
345-6. 551. 552-3.
554. 556-7
and representation 552-3.
554
and semantic coherence 296
and subjectivity 573
Kaplan. David 195. 389-90
Karmiloff-Smith. Annette
408-9.443
Kenny. A. 248
Kim. J. 327.352.473.490-1
kinds. natural 449-50.
476-7. 479. 535. 585
and functionalism 327
and identity theories 184
person as 556
and Putnam 120. 191
Klein. Melanie 498. 602
knowledge
and accessibility 59. 60
of agency 116-17
attribution 47. 58
of beliefs 147-9. 151.
396-7
de se 426
and experience 48-61.
371-2. 504: knowledge
'how-to' 58-61. 208.
371.413: 'no-
knowledge'view 55-8
and expressibility 47-8.
49-53, 55-6. 58-60
first-person 6. 52. 55-6. 59.
273-4.291. 310.
396-7
informational content
259-60
and innateness 366-7,
371-2. 504-5
linguistic 171.175
INDEX
and observability 59. 119
of possibility 266
propositional content 47-8.
49. 58-60
and qualia 322, 413.
514-15
and sense-perception 459
tacit 27.48. 60-1. 224.
360. 416. 530
third-person 52-4. 59.
291
see also Dretske. F.:
information: innateness:
other minds: subjectivity
Koch, Christoff 210
Kohonen. T. 200
Kosslyn. Stephen and imagery
356-60. 365
Kripke. Saul
and concepts 190-1
and content of belief 222.
429
and meaning and
environment 195
and necessary identity
353-4. 355
and rigid designators
419-20, 519
and theories 369
Kuhn. T. 189.217. 312.
313
language
acquisition 159-60.
367-71. 504-5
and behaviourism 138-9.
160. 170. 522
169. 171
and cognitive sCience 172
cognitive system 158-9
computational-
representational studies
154-5, 161. 164-5
and connectionism 173,
206-7. 368-9
Derivational Theory of
Complexity 171
and experience 50-1
holism and atomism 406-7
504-5
intensional theory 158
internalist approach
158-60. 163
and memory 433-4
neuropsychology 172
performance systems 158-9.
162. 166
private 136. 569-70, 619
productivity 198-9. 205
633
INDEX
language (cont'd)
36-8. 39-40. 44-5.
589-93
psychology of 171.172
shared (public) 138. 158-9.
163. 165-6. 592-3
systematicity 198 - 9. 205
and tacit knowledge 27
usability 161-2
see also belief; Chomsky.
Noam; content;
representation; rules;
semantics. conceptual
role
language and thought 36.
165. 172. 234. 564.
586-8. 589-93
and concepts 408. 586-7
and connectionism 210.
409.411
dependence of language on
thought 589-91
dependence of thought on
language 564. 591-3
and Freud 605
semantics 194-5.
259-63. 590-1
and Wittgenstein 620-1
language of thought 401-7.
408-11. 586-7
and adverbialism 403-4
broad/narrow content 166.
406
and cognitive science 172.
295.409
and conceptual role
semantics 194. 195-6
409.411
and dependence of thought
onlanguage 591-2
426. 474
and holism 405.406-7
hypothesis 102
408
and instrumentalism 404
and Mentalese 142-3. 194.
196. 294. 401. 402-3.
405-6
objections 403-5
and productivity 403
and representationalism
401-5
and sententialism 401-7.
540-1. 583-4
in symbolic theory 182
and systematicity 402 - 3
see also Dennett. Daniel C.;
634
law
Fodor. Jerry; Stalnaker.
Robert
bridge 91. 92. 231. 472-4.
579-80
of cause and effect 231.
351-2
in folk psychology 309-10.
312. 313
hedged 490
implementation 294-5.
296-9
intentional 293-9
psychophysical 231. 232.
475-6
learning
and amnesia 434. 439
and belief 227
178.180-1.184.
201-3.207. 314-15.
368-9
as evolution in brain 238
and innateness 369-72
and meaning 233-4. 235.
371
see also Dretske. F.; Fodor.
Jerry
Legendre. G. 178
Leibniz. G. W. von
and identity theory see
Leibniz's Law
and materialism 343
and pre-established harmony
341-2
Leibniz's Law 97. 265. 322.
349.385-6.431-2.582
Lenat. D. B. 124. 130
LePore. E. 189. 329
Leslie. A. M. 253-4. 255. 256
Levine. J. 210. 322. 518-19
Lewis. David 412-30
and analytical functionalism
135-6.239
137.417-18
and conceptual
functionalism 319. 323.
326-7. 331
and content 421-30
and folk psychology 416-
19. 512
and functional reductionism
100-1
421-2. 512
and identity theory 349.
351.354.418-20.473.
481-2
and physicalism 326-7
and qualia 514-15
and realism 412
and reductive materialism
412-15.419
and supervenience and
analysis 412-21
Libet. B. 114-15. 120
Lindskold. S. 306
linguistics. computational 182
Loar. Brian 328. 515
and conceptual role
semantics 193. 195.
198
Locke. John
and concepts 188
and imagery 356
and introspection 397
and inverted spectrum 103
and materialism 343
and personal identity 343-
5. 552. 555. 574
and thought 340
and will 611. 612
Loewer. B. 513
Loftus. E. F. 438-9. 440
logic
extensional 186
first-order quantifier 12 7.
134.525
modal 186
second-order quantifier 128.
129
and thought 587
truth-functional 127
Logical Positivism
and concepts 188-9
Ludlow. P. 323
Lycan. William
and conceptual role
semantics 193
and connectionism 207-8
515
and intention 240
McCarthy. John 122
McClelland. J. L. 200-1. 204.
206
McCloskey. T. 1. 119
McCulloch. W. S. 177.200.
237
McDowell. J. 220. 222. 533
McGinn. C. 211. 518
machines see intelligence.
artificial; Turing machines
MacKay. Donald 237
Mackie. J. L. 344. 484
McLaughlin. B. P. 280
Maclean. Paul 275
Malebranche. N. 582
Mally. Ernst 388. 391
Manktelow. K. l. 168
Marchman. V. 201-3
Margolis. H. 205
Marr. David 172. 179. 239
and perception 462
and theory of vision 359
Marslen-Wilson. W. D. 443.
444. 445
materialism 471-8
eliminative 133. 157. 265.
270. 471. 522: and
computational models
105-6; and folk
psychology 310-11.
315. 353; and identity
theory 240. 352-3.
418-20; and
491-2
335
non-reductive 311. 471.
490
and psychology 292
and qualia 517
reductive 412-15.419.
472-3. 535. 544
17th-century 342-3
see also Davidson. D.;
naturalism; physicalism
matter
and form 336-7. 341. 345
and mind see mind and body
Matthews. Gareth 250
Maxwell. James Clerk 130
meaning
and behaviour 138-9
and consciousness 237. 588
and environment 191. 195
and externalism 233
holism 188-9. 190. 328.
347
and intention 194. 386
and learning 233-4. 235.
371
and reference 194-5
speaker 590-1
andtruth 195.198
as use 323
verificationist theory 134-5.
188-9
and word-world relation
193-4.198
see also semantics. conceptual
role
Mechanical Philosophy 341.
343
medieval philosophy 333-7.
379
Medin. D. 189
Meltzoff. A. N. 256
memory 433-40
and amnesia 434. 439-40
associative network
approach 435
buffer 434
causal theory 277.278
168. 169-71
false 440. 600-1
Headed Records model
436-40
and imagination 364. 366
implicit/explicit 434
383-4.433
long-term 172. 173. 433-4.
441
and multiple personality
disorder 439-40
552
and phenomena 437-9
recognition 437-8
schema approaches 173.
435-6.438
sensory 171
short-term 170. 433-4. 439
see also consciousness; Freud.
Sigmund
mentalism 520. 521-5.
620-2
and action 112-13
Mettrie. De La. J. O. 157
Mill. J. S . and desire 249
Miller. George 170.171.237
Miller. J. L. 444
Millikan. R.
192
semantics 195
and teleology 319.489
mind. mapping 3-5.24-7.
489
mind and body
and Cartesianism 87-9.
156-7
in Chomsky 156-7
dual aspect theory 92.
342
and extension 268
and identity theories 90- 7.
481
in Leibniz 341-2
in Searle 544-6
and supervenience 581-3
see also dualism; materialism
Minsky. Marvin 122. 130.
200
INDEX
misrepresentation see belief.
false
Miyata. Y. 178
modality
and dualism 266
and possible worlds 484-5
see also worlds. possible
modularity 441-7
and breakdown 446
in cognitive development
253. 255. 256-7. 370.
442
and Fodor 171. 175. 441-4
in generation of desires 247
and localization 446. 447
in network approaches 203.
443.444-5.447
and speed 444-5. 446
monads
in Leibniz 342
monadic see property
monism
anomalous 92-5. 100. 123.
231. 311. 334
and dualism 122. 525
ontological 231. 326
as physicalism 122. 475-6
see also Davidson. D.; Quine.
Willard Van Orman;
supervenience
monopsychism 337
Moore. G. E. 576. 578-9
Morton. J. 437
mysterianism. new. and
consciousness 211
Nagel. T. 210.211. 216. 322
and explanatory gap 210.
519
and subjectivity 569. 570.
572-3. 574-5
names. hidden descriptions
theory 389-90. 391
narrow content see content
nativism
256
and language 371.408-9.
411
and rationality 530
natural kind see kinds. natural
naturalism 449
biological 321. 544-6
and concepts 191
and content 240. 488-90
ethical 578-9. 582
502-3. 563
335. 337
635
INDEX
naturalism (cont'd)
metaphysical 153. 157
methodological 153-9.
162-7
479. 502-3
and psychology 495. 498
and Quine 520. 521-2
and semantics 405-7
see also Dretske. F.;
materialism; physicalism;
Quine. Willard Van
Orman
necessity. a posteriori 414-15
Neisser. Ulric 171
Nelson. K. 367
Nemirow. 1. 413
neo-Platonism 337
neurophilosophy 175
neurophysiology
and belief 96
172
545. 548-9. 550
and identity theory 91-2.
133. 317-18
and intention 321
and 'no-gap' arguments
120.278.281-2
and pain 99-100.317-18.
454-9. 522-3
and physicalism 78. 82-3.
84-5.277-9. 281-4
and reductionism 106
see also causation; pain
neuropsychology
and action control 119-20
cognitive 172
310
Newell. Allen 124. 125.
171-2.237.239
Newton. Isaac 155. 156-7.
536
Nichols. S. 302
nociceptors. and pain 454-6.
458
nominalism 452
Norman. D.A. 118-19.436
normativity 450-1
see also rules
Nowlan. S. J. 203
object constancy. and
innateness 367.371-2
objectivity
251
636
and Davidson. D. 232-4
and subjectivity 234.
552-4. 556-7. 570-3
observability
and attitudes 16-17.23-4
and behaviour 54-5.
104-5. 138-9
and existence 22
and experience 16-17.23.
24-5.46-7
and knowledge 59. 119
observation. and cognitive
psychology 174
ontology 452
and belief 150.452
and folk psychology 309-10
and properties 92. 486
and reductionism 122. 231
and the soul 335
see also monism
opinions. in Dennett 241
Optimality Theory 179
OShaughnessy. Brian 13. 342
'other minds'. problem of
and analytical functionalism
135. 136. 139
children's awareness of 253.
256
313
and language use 167
and scepticism 52 - 5. 61
see also Davidson. D.; Fodor.
Jerry; Searle. John;
Sellars. Wilfrid;
Wittgenstein. Ludwig
Over. D. E. 168
overdetermination. and
causation 85-7. 89.
94-5. 100. 104. 281.
481. 482
pain 452-60
and accessibility 4. 18. 24.
46-7. 396-7
and behaviour 452. 453
325. 326
and consciousness 9. 10.
454. 588.618-19
and directionality 20-1. 25
as experience 24-5.46-7.
53.419.452.569
and expressibility 19. 25.
46-7. 53-4
291. 552. 569
326-8. 330. 508-9
and identity theories 99-
100. 349. 353-4
meaning 453. 459
nature 452
and neurophysiology
99-100. 317-18.
454-9
and observability 24-5.
46-7
and rigid/non-rigid
designators 325. 352.
418-21
as sensation 452-3
as subjective 452-4.457
and theoreticity 25
Papert. S. 200
Parfit. D. 555. 574
Pargetter. R. 420
passivity. and emotion 272
pattern recognition. in
connectionist networks
204-5
Peacocke. C. A. B. 185-6. 468
Peirce. Charles Sanders 156.
536-7
Penfield. W. 217
Penrose. R. 128
perception 459-63
as activity of mind 7. 9-10
and belief 220. 225. 247.
398. 427. 430. 464-5.
468-9
and cognition 168
in computational theories
180.442.462
and concepts and activity
461. 462
and connectionism 444- 5
and consciousness 460. 462
and desire 244. 247-8. 464
and directionality 21
and emotion 272
epistemic/non-epistemic 259
and existence 266
and imagination 362-3.
366. 463
and information processing
461-2
as input 266
383-4. 394. 464-70.
561
as phenomenal 464. 466.
468.471
and representation 462.
501. 570
requisites 462-3
and self 551
sense-perception 459-60
veridical 465-7
see also content. perceptual;
Dretske. F.; introspection;
perception (cont'd)
Quine. Willard Van
Orman; sensation; sense
datum
Pemer. Josef 255.256. 574
Perry. John 175
person see identity. personal; self
phenomenology 471
phenomenon/phenomenal/
phenomenological 471.
568-70
philosophy and psychology see
psychology. and philosophy
physicalism 47l-8
317
and belief 151
and Completeness 481-2.
483-4
and consciousness 211.479
critique of 479-84
definition 76. 479-80
eliminative 474. 479
and explanation 76-9.
82-3.97.157.412-13.
481-2
325-8.476
591
and monism 122. 334
and naturalization 476-7.
479. 502-3
non-reductive 474-8.
482-4. 576. 578-80
post-analytic 477-8
and qualia 514. 515-17
reductive 472-4. 476
and will 611
see also causation;
epiphenomenalism;
identity theories;
materialism; monism;
naturalism; property.
mental; Quine. Willard
Van Orman
Piaget. Jean
and developmental
psychology 170.
251-2. 254. 256-7.
408
and object constancy 371-2
and subjectivity 568. 570-2
Pinker. Steven 201.256. 359
Pitts. W. H. 177.200
Place. U. T.
and identity theory 317
and materialism 412
and physicalism 135
Plato
and dualism 265
and emotion 270. 272
and knowledge 530
333. 334. 337
and modularity 441
and morality and action 608
and thought and symbol
299
pleasure and emotion 25-6
Plotinus 333. 337
Plunkett. K. D. 201-3
pluralism and monism 122
Pollack. J. 206
Pomponazzi. Pietro 337
possible worlds see worlds.
possible
Potter. E. K. 119
practition. and intention 378
predicate
extension 374
intension 186. 372
and property 486
predication in folk psychology
421-2
Premack. D. 253
presentation. modes of 37-8.
43. 221
pretence
in child's development
254-5
and imagination 364
Pribam. K. 237
Price. H. J. 468
Prichard. H. A. 612
Priestley. Joseph 157
Prince. A. 179. 201
Prior. E. W. 420
privacy of the mental 52-3.
569. 574
and Searle 545
and Wittgenstein 618-20.
622
see also authority. first-person;
subjectivity
private language argument see
probability. subjective 196-7
problem. and mystery 156.157
problem solving
in cognitive psychology
171-2
in connectionism 207
propagation
back 173.202-3
property
and natural kind 449-50
and ontology 452
synthetic identity 508
property. mental 486
and Davidson 92. 231.
475-6
INDEX
and epiphenomenalism
279-80
476. 508
and identity theories 91-2.
348-52.432
in intentional psychology
296-9
and materialism 472-4
multiple realizability 324-5
as non-linguistic 486
and physicalism 326. 472-
8. 479. 481-4
qualitative see qualia
and Quine 522
and relations 486
as second-order property
331
see also causation; concepts;
supervenience
proposition 486-7
abstract 141-2. 143. 145
declarative sentence as 12.
232.486-7
and emotion 271-2
and intentionality 487. 562
necessary/contingent 141-2
singular/general 389-90.
394.422.425-7.429
structured 31 n.1
see also attitudes.
propositional; content.
propositional;
intentionality; Stalnaker.
Robert
proto-declaratives in infant
communication 253. 255
proto-imperatives in infant
communication 253
prototypes
and categorization 205. 315
477
and Freud 604- 5
psychoanalysis
and cause and meaning
494-6
and explanation 27.
493-500
explananda 493-4. 497
legitimation 499-500
and the unconscious 27.
496-9. 598-604
and wish and phantasy
496-9. 598-605. 606
see also Freud. Sigmund
psychofunctionalism 102.
3 1 7 - 2 ~ 325. 327-8
see also functionalism
psycholinguistics. and cognitive
psychology 171
637
INDEX
psychology
133-4
human experimental
170-1
ideal theory 510-12
introspective 133-4.317
materialist 292. 296
and philosophy 291.
500-6.521
practical 524
psychology. cognitive 105.
143. 167-75
171-2
and cognitive science 1 72.
175. 520
and computational models
176.285.319.503-4
175. 200. 210
303-6
and generative grammar
171
and human experimental
psychology 170-1. 500
501-2
methodology 174-5
and perception 168. 501
philosophy 175
and philosophy of science
500-6
and rationality 529
scope 168
social 170
and social/developmental
psychology 169-70
psychology. developmental
147. 250-7
169-70
and concepts 187
domain-general theories
251-2. 256
domain-specific theories
251-5
and representational theory
408-9
and subjectivity 568. 573.
574-5
and unconscious 498
psychology. ecological. and
cognitive psychology
169-70
psychology. intentional 292-6
property theories 296-8
psychology. social
638
169-70
302. 306
psychosemantics 321
Putnam. Hilary 507-13
and belief attribution 40-1.
139
and concepts 190-1
and content 220. 222. 286.
321
and identity theory 317-18.
351. 507-8
and innateness 367. 368.
369
and machine functionalism
318. 323. 329
and natural kinds 120. 191
and post-functionalism
512-13
Twin-Earth theory 40-1.
191. 195. 216. 222.
321. 423-4. 501. 585.
596
Pylyshyn. Z. W. 205. 206.
359-60. 365. 411
pyschology. computational.
subpersonal 223-4
qualia 50-1.52-4.211.
467-8. 471. 514-19
absent 242. 515
and explanatory gap
518-19
321-2.330-1.515-19
and imagery 361
and Inverted Earth 517-18
and inverted spectrum
103-4.317.515-17.
581
as non-relational 514
and physicalism 514.
515-17
as relational 514. 515
and sensations 355. 453.
561
516-18. 576. 581
see also Dennett. Daniel C.:
sensation
qualities. secondary. and
content 220
Quine. Willard Van Orman
520-5
and analytic/synthetic
distinction 199. 407
and anomalous monism
523. 525
137. 139. 522
and causation 268
and concepts 186. 188-9.
190.192.410.521
and Dennett 237.239.
242-3
indeterminacy of translation
134. 137. 139. 158.
164.237. 239. 524-5.
526-7
and linguistic theory 165.
520
and naturalism 520. 521-2
and perception 523
491. 523-5
and underdetermination of
theory by data 369
Ramsey. F. P. 264. 416. 422
Ramsey. W. 207. 305
rationalism. and innateness
366. 370
rationality 526-31
agent-constitutive 526-31
and attitude attribution
450-1
and beliefs 220. 399.
401-2. 526-7. 532-4
bounded 528
cognition as 204. 206. 299.
334
50S
as coherence 232-3. 526.
528
competence and
performance 529-30
constitutive 428-30. 530
and content 20. 221. 223.
232.245.428-9
and emotions 275-6.499
530
and intention 65-6. 295.
379. 532-3. 535
and knowledge 504
normative 526. 528
pluralist models 530
realistic models 528-9
and sententialism 401-2
shared 232-3
standard idealization models
526-8. 529
and thought 587
Read. Allen Alker 233
realism
critique 239-41. 242
and emotions 275
and folk psychology 239-
40.305.428.429
466
'mild' 242
naive 466-7
and properties 522
realizability
142-3. 177. 178. 181.
183. 285-6
multiple: and functionalism
324-5.476-7. 515.
581: and identity
theory 100-1. 350-1.
355: and properties
295-6. 581
reasoning
analogical 251. 410
theoretical 248. 485
reasoning. practical 168. 169.
235. 245. 248. 485
in folk psychology 308
and intention 376. 377-8.
532
practical reasoning system
(PRS) 302-3
and psychology 496-7.499
reasons and causes 531- 5
and action 65-7. 79-81.
84-6. 113. 117-20.
229-30. 231. 531-4
awareness of 66. 116
and behaviour 263-4
and beliefs 245. 246.
401-2. 531-3
and content of states
220-1. 387. 531
and desire 246-50. 531-2
explanatory role 533-5
and intention 65-6. 295.
376. 531
reason-giving relation
531-2. 533
see also causation
reduction 535-6
definitional 579
of mental to physical 92-5.
100. 350. 474-5. 536
nomological 579
ontological 122. 231. 536
scientific 76-7. 154.480.
518-19.535-6
reductionism
behavioural 231. 579
175
and consciousness 211. 21 7
and content 223
functional 100-1. 380
materialist 412-15. 419
and naturalism 449
phenomenalist 342
physicalism 76-7.472-4.
576. 578-80
see also identity theories
reference
direct 186
failure 552. 553
and language 166.194-5.
198. 522
Reid. Thomas 460-1
reinforcement. positive 250
Renaissance philosophy 333.
334. 337
representation 536-40
analogue/digital
(propositional) theories
365-6. 538-40
and beliefs 401-2
causal theories 226-7. 228.
408.421-3.441
257
in cognitive psychology 168.
170. 501
in computational theories
1 7 ~ 178-9. 1 8 ~ 299.
462. 540
and concepts 186-7.407.
408. 486
decoupled 254
474
and imagery 539-40
382. 386. 391-4. 537-
8. 539. 561-3
as interpretational semantics
225
and language 159. 241.
305.359.401-3.422.
429. 538-9. 541
as map 422-3
mental 401. 441. 620-2
and motor schema 118
and naturalism 538
and phenomenal states
215-16
propositional content 143
and resemblance 537. 540
teleological theory 227-9
theory of signs 536-7
and the unconscious 497
and vector codings 314
see also concepts:
connectionism: content.
representational: Fodor.
Jerry: icon: imagery:
Kant. Immanuel:
language of thought:
perception: sententialism:
symbol: thought:
Rey. G. 190. 213. 516
INDEX
Rorty. Richard 352-3
Rosch. E. 189. 205
Rose. G. F. 129
Rosenblat. F. 200
Rosenfeld. P. 306
Rosenthal. D. 323
rules
and algorithms 181-4
and Davidson 233-4
language 160. 165. 171.
368-9. 450. 504
learning 202
as norms 450
and Wittgenstein 233
Rumelhart. David 118.200-1
Rumelhart-Norman model
118
Russell. Bertrand 123. 171.
342. 386. 387
and content 219
389-91. 393
Paradox 527. 529
theory of descriptions
389-91
Ryle. Gilbert 307. 541-3
136. 417. 542
and category mistakes 541
and Dennett 237.242-3
541-2
and imagination 363-4
influence 412. 541
and knowledge 208. 541-2
371
belief 142
and topiC neutrality 349
and will 612
St John. M. F. 204. 206
Samuel. Arthur 123. 124
Sartre. Jean-Paul 363
Saussure. F. de 160
scanning. internal. and
monitoring-consciousness
213
scepticism
Cartesian 339. 459
and other minds 52-5
Scheler. M. 272
schema. motor 118
activation trigger schema
118-19
and memory 173.435-6.
438
selection 119
639
INDEX
Schiffer. Stephen 195. 328.
482
scholasticism. medieval 334-7.
341. 345. 379
Schopenhauer. Arthur
and imagination 364
and will 610. 612. 614
science
in Chomsky 156
and mentalism 524-5
philosophy of. and cognitive
psychology 500-6
physical 479-80
see also physicalism
science. cognitive
172.173.175
176. 210. 297-9. 410
and connectionism 200
162. 172. 295. 409.
423
and Quine 520
as reverse engineering 239.
242
see also modularity; Searle.
John
scientism 495. 509. 510
Scruton. Roger 363
Searle. John 544-50
and algorithmicity
assumption 125.
503-4
and cognitive science
546-9. 550
and computational model
544. 546-8
217-18. 544-7. 548-9.
550
and direction of fit 272.
381-2
and epiphenomenalism 546
547. 548. 549-50
and linguistic theory 165
and mind-body problem
544-6
and other minds 545
and speech act theory
380-2
and unconsciousness 544.
548-9. 550
self 550-7
and action 15.26. 550-1.
556-7
and autism 257
251-2. 571-2
and control 571-3
640
and experience 10. 550-1.
556-7
and imagination 364
and mapping of mind 27
550-1. 552. 555-6.
570. 573
and proto-self 5 71
see also agent; authority. first-
person; identity;
introspection; Kant.
Immanuel; subjectivity;
will
self-consciousness 213. 216.
550. 551-4. 556
self-deception 396. 494. 532-
3. 558-60
in Davidson 235
and emotions 273-4
and the unconscious 559
self-intimation see authority.
first-person
self-knowledge 52. 55-6. 59.
291. 554
and emotion 273-4
and introspection 310.
396-7
see also authority. first-person
Selfridge. O. G. 200
Sellars. Wilfrid
and conceptual role
semantics 193.196.
309. 323
and content 220
310-11
371
and other minds 309
semantics
dual-aspect 198. 513
formal 186
inferential role 299
information-based 194-5.
259-63.489. 512-13.
590-1
internalist 159
interpretational 226-7
401-3. 405-6. 409.
421.426-7
procedural 193.298. 329
situation 175
success 229-30
teleological 227-9
see also Fodor. Jerry; syntax
and semantics
semantics. conceptual role
(CRS) 187. 193-9
and compositionality 198-9
and conceptual role 196-7
and content 103
and dual aspect semantics
198
epistemological! ontological
problems 197
and functionalism 323. 329
194
and logical connectives 194
and methodological
solipsism 195-6
and semantics/metaphysics
194-5
and truth and reference 197
and word-world relations
193-4
semiotics. structuralist 193
sensation 560-1
accessibility 46
act-object approach 398
adverbial theory 398
in folk psychology 308
and introspection 365.
397-9
and language 619-20
and mental properties
349-50. 354-5
observability 46
and perception 459.460-1.
463. 471. 560
projected 458
see also pain; qualia
sense datum
464-8. 471. 560-1
and qualia 51. 561
senses
and concepts 186-7
and content 219.391
and Dretske 264
sententialism
and belief 142-5. 401-2
401-7. 540-1. 583-4
and naturalistic semantics
405-6
and representationalism
401-3
sets
. fuzzy set' theory 189
non-decidable 127-9
non-semidecidable 129
theory 526-7. 529
Shaffer. J. 473
ShaIlice. T. 118-19
Shannon. Claude 170
Shaw. Cliff 123.171
Shepard. Roger 174
Sherrington. C. S. 242
Shoemaker. S. 468. 518. 555
Siger of Brabant 337
Simon. Herbert 124. 171.
237. 505-6. 528
Simulation Theory 147-9.
302-3. 313. 561
Skinner. B. F.
133-4.170.171.522.
542
and primitive concepts 191
Smart. J. J. C.
327-8
and identity theory 317.
349
and materialism 412. 418
Smith. E. 189
Smolensky. P. 178. 179. 206
Sober. E. 261. 319
solipsism 61. 195-6. 197.
569
Somin. H. A. 445
soul
in Descartes 341
in Kant 345
in Locke 344-5
in medieval philosophy
333-7. 341
spatiality of the mental 479-80
spectrum. inverted see qualia
speech acts. and intentional
states 380-2. 544. 591
Spelke. E. S. 372
Sperling. George 215
Spinoza. Baruch 231. 243-4
and dual aspect theory 342
and will 612
Stalnaker. Robert 561-8
and content 241.511-12.
564-8
and intentionality 561-8
564
and necessity 415
and propositions 562-8
Stich. S. P.
207. 300. 302. 304-5.
304. 491
391
and rationality 530
Stillings. N. 442
Strawson. P. F. 166. 362
and consciousness 553-4
'Stroop effect' 120-1
structuralism. and language
160. 193. 298
Suarez. Francisco 334
subjectivism. and content 220
subjectivity 568-75
251
and compatibilism 573
and concepts 486
354. 551
and experience 287.
464-70. 499-500
and knowledge 52.479
and mental images 360-1
and objective world 234.
552-4. 556-7. 570-3
and pain 452-4.457
and phenomenality 568-70
and physicalism 479
and point of view 568.
574-5
and reason-giving relation
532
and the will 570-74
see also self
substance
and Aristotle 340. 343
and Descartes 340-1. 348
and Locke 343-5
and Spinoza 342
super-ego 605
supervenience 536. 575-83
and analysis 412-21
139
of content 222. 227-8.
286. 321. 490
and dependence 576. 582
and emergentism 576-7
global 576. 578. 580.
581-2
and identity theories 93 - 5.
352
576
local 581
and moral theory 576.
578-9. 582
and non-reductive
physicalism 576.
578-80
481.483. 576. 578
and property covariation
576. 577-8. 582
and qualia 514. 516-18.
576. 581
of soul 335
strong 578. 579-80.
581-2
as theory of mind 581-2
weak 577-8. 579. 581
see also monism. anomalous;
reduction
INDEX
symbol
in Fodor 295-6. 299
and representation 536-40
symbol system hypothesis see
intelligence. artificial
synapses. and pain 455-6.
457
syntax
development 256
and semantics 546-7. 548.
583-4
systems. supervisory
attentional 119. 121
Tannenhaus. M. A. 447
Taylor. C. 237
Tedeschi. J. T. 306
teleology 585
biological 227-8. 500
and content 220.227-9.
581
319-21. 322. 331.489.
515
representational 227-9
temporality. and action
68-74. 112. 114-16. 614
text comprehension. mental
models 175
theoreticity. and mental states
22-4
Theory Theory 147-9. 302-3.
308-12. 561
Thompson. F. 576. 578
thought 585-9
and behaviour 292-3
in Descartes 340
and experience 54
higher-order 213.215
and imagination 362
and language see language
and thought
in Locke 340
and logic 587
and meaning 194-5. 309
opacity 403
physical base 210. 324
primary process 497
productivity 402-3.408
as representational 143.
241. 255. 261. 264.
401-7.408.411.441
semantic properties 401-3.
405
as symbols 295-6. 299
systematicity 205-6.
402-3. 408. 411
and the world 585-6
see also concepts; content;
intentionality; language
of thought
641
INDEX
Tichener. E. B. l33
Tichy. P. 415
tractability 477-8
traits. attribution. in folk
psychology 301-2. 306
transcendentalism. and
consciousness 211
transitivity
and concepts 191
and preference 527
translation. indeterminacy
l34. l37. l39. 158. 164.
165.237.239. 524-5.
526-7
Treisman. Anne 1 74
truth
and meaning 198
and shared expectation
233-4
trying
and action 115-16. 120.
245
and will 6l3-14
Turing. Alan M. 101. 122-3.
125. 128. 164.296-7.
299. 594-6
Turing machines
125-6. 128-9
and computational theory
101-2. 296. 508-10.
594-6
Tversky. A. 446. 529
Twin-Earth theory 596
and concepts 191. 198
216.222.406.423-4.
426. 581. 585-6
and externalism/internalism
290. 321. 329. 476-7
501-2
and supervenience 222
Tyler. 1. K. 443. 444. 445
type/token 93. 596-7
and action 111
and concepts 186
and event 279
and thoughts 295
Ullian. J. S. 129
unconscious. the 598-606
642
and the ego 605
and emotions 273
384
and metapsychology
604-6
468
and psychoanalysis 27.
496-9. 598-604
and repression 601-2
in Searle 544. 548-9. 550
and self-deception 559
and super-ego 605
and wish fulfilment
598-601. 603-5
unification problem. and mind
154.156-7.164
Updike. John 9. 10. 25
Van Gelder. T. 205
Van Gulick. R. 320
Velleman. J. D. 377. 379
verilicationism
and behaviourism l34-5
and belief l34. 226-8
and concepts 188-9. 190
volition
action theory 120. l33
and agent causation 115-
16.117
see also will
Von Humboldt. Wilhelm 160
Wason selection task 167. 168
Wasserman. S. 372
Watson. J. B. 356
and behaviourism l33-4.
542
Wellman. H. 301
White. S. L. 241. 327-8. 516
Whitehead. A. N. 123. 171
Wiggins. D. 555-6
will 610-17
and action 73-4. 115-16
and desire 245
and emotion 270
identifying conditions
611-13
and intention 377. 611
and physicalism 611
reality 6l3-17
and schemata selection
control 119
and subjectivity 570-4
weakness of 120-1.235.
270. 377.494. 532-3.
608-10: and morality
608. 610; and
rationality 608-9
see also Descartes. Rent);
intentionality;
William of Moerbeke 334
William of Ockham 335. 337
Williams. B. 555
Wimmer. Heinz 255. 574
Wittgenstein. Ludwig 617-22
and behaviourism l32. l36
and content 522.620-21
and essential privacy
618-20. 622
and family resemblance
theory 189
and imagination 363. 365
and intention 549. 588.
620-1
and language games 621-2
and meaning as use 323
and mental representation
620-2
and other minds 149.
234-5. 619-20
and private language
argument l36.
569-70. 619
and rule-following 233.
451
and self 552
and will 612-13. 618
see also concepts
Woodward. J. 304. 305
word identification 172.173
worlds. possible 31 n.1. 141.
159. 484-5
and intension 186. 372
and Lewis 412. 413
and propositions 487.
564-5
Wundt. Wilhelm 133
Young. J. Z. 238
Compiled by Meg Davies
Blackwell Companions to Philosophy
This student reference series is centred on analytic philosophy but also covers
important aspects of the continental tradition and of non-Western
philosophies. It offers a comprehensive survey of philosophy as a whole. The
entries in each volume combine summarized information on names, terms, and
movements, with critical personal essays contributed by leading figures. Each
essay is cross-referenced and supported by a selected bibliography.
A Companion to the Philosophy of Mind
The philosophy of mind is arguably the leading area of philosophical research,
not least because of its connections with cognitive science and related areas of
psychology, linguistics, and computation. This Companion provides an
alphabetically-arranged guide to the subject , firmly rooted in the philosophy of
mind, but with entries that survey adjacent fields of interest.
Written by an international assembly of the leading philosophers, the volume
includes extensive cross-referencing, glossary entries, detailed bibliographies,
and a comprehensive index. Among the entries themselves are series of 'self-
profiles' by key figures of today, including Noam Chomsky, Donald
Davidson, Daniel Dennett , Fred Dretske, Jerry Fodor, David Lewis,
Hilary Putnam, John Searle, and Robert Stalnaker.
'Guttenplan's extended and well-written introductory essay
stands on its own as a fine introduction to current philosophy of
mind.' The Times Literary Supplement
Already published in this series
A Companion to Ethics
Edited by Peter Singer
A Companion to Aesthetics
Edited by David Cooper
A Companion to Contemporary Political Philosophy
Edited by Robert E. Goodin and Philip Pettit
A Companion to Epistemology
Edited by Jonathan Dancy and Ernest Sosa
A Companion to Metaphysics
Edited by Jaegwon Kim and Ernest Sosa
Samuel Guttenplan is Senior Lecturer in Philosophy at Birkbeck College,
London.
Jacket illustration: P aul Klee, Angelus dubiosus, watercolour on paper, 29.5 X 21 em, private
collection.
Jacket design by Workhaus Graphics
ISBN 0-631-19996-9
90000>
113 BLAC I<WE LL
o 9 7
Reference
1-98
$ 29.95

A Companion To The Philosophy of Mind

Uploaded by

Copyright:

Available Formats

A Companion To The Philosophy of Mind

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

A Companion To The Philosophy of Mind

Uploaded by

Copyright:

Available Formats

What is the structure and purpose of the book?

What rights does the publisher reserve regarding the content?

RIa ck lPcl1

You might also like