The Paradox of Self-Consciousness - José Bermúdez

Preface
The phenomenon of self-consciousness provides one of the enduring

themes of philosophy. Many philosophers from the great to the insignifi-
cant have thought that a philosophical account of self-consciousness
would be an Archimedean point upon which the whole of philosophy
might stand. This book has no such pretensions. My concern is solely
with understanding the internal articulation of the phenomenon of self-
consciousness. In fact, my concern is primarily with those forms of self-
consciousness that are more primitive, both logically and ontogenetically,
than the conceptual and linguistic forms of self-consciousness to which
philosophers have traditionally devoted their attention.
My concern with these primitive forms of self-consciousness stems
from a paradox that strikes hard at traditional understandings of the rela-
tion between linguistic self-reference and self-conscious thought. I de-
velop what I term the paradox of self-consciousness in chapter 1. The
core of the paradox is the apparent strict interdependence between self-
conscious thought and linguistic self-reference. This paradox is insoluble,
I argue in chapter 2, if it is assumed that the conceptual and linguistic
forms of self-consciousness are the only forms. In chapters 3 and 4, I set
out a framework for thinking about the intentional explanation of behav-
ior that allows truly ascribing content-bearing representational states to
creatures who lack both conceptual abilities and linguistic abilities. Chap-
ters 5 through 9 are devoted to showing that among those content-
bearing representational states, there are several types that properly count
as forms of primitive self-consciousness. Finally, in chapter 10, I show
how a recognition of the existence and nature of these forms of primitive
self-consciousness can be used to solve the paradox of self-consciousness.
xii Preface
The perspective from which this book is written is philosophical natu-

ralism. I take it, for example, that philosophical accounts of concepts
must accord not just with what is involved in the fully-fledged mastery of
a concept but also with how it is possible for a thinker to acquire that
concept in the normal course of cognitive development. The distinction
between logical issues about concepts and psychological issues about con-
cepts is, of course, a real one, but to my mind it does not provide a clear
demarcation between the areas of competence of two separate disciplines,
or of two distinct levels of explanation. Throughout this book I make free
use of empirical work from various areas in scientific psychology in mak-
ing philosophical points. I hope that the book as a whole may serve as an
indirect argument for the utility of this approach.
Acknowledgments
This book was conceived and written during my tenure (1993–1996) of

a British Academy Postdoctoral Research Fellowship, hosted initially by
the University of Cambridge and subsequently by the University of Stir-
ling. I am extremely grateful to the academy for making it possible for
me to spend three years unencumbered by teaching and administrative
commitments. I also owe a very great debt to my colleagues at the Univer-
sity of Stirling for encouraging my research, most particularly by allowing
me seven months leave to complete this book before taking up my teach-
ing duties. Peter Sullivan read a complete draft of the text with great care
and is responsible for numerous improvements. While at Stirling I have
benefited enormously from daily discussions on the philosophy of mind
with Alan Millar.
I first began thinking about some of the issues in this book during the
academic year 1992/1993, when I was a member of the Spatial Represen-
tation Project at King’s College, Cambridge. Discussions during the proj-
ect with Bill Brewer, Naomi Eilan, and Anthony Marcel were very helpful.
As will be apparent from the text, I have been greatly inspired by the
writings of John Campbell and Christopher Peacocke, who were also
associated with the project. An intermittent correspondence with Susan
Hurley over the last few years has helped me to sharpen many thoughts,
and I am extremely grateful for the care and rigour with which she read
a penultimate draft of the manuscript as well as for the opportunity to
learn from her own unpublished writings.
A version of chapter 1 was delivered to the Sapientia Colloquium at
Dartmouth College, where I was a visiting scholar in the summer of 1996.
Chapters 3 and 4 draw heavily on my paper “Nonconceptual Content:
xiv Acknowledgments
From Perceptual Experience to Subpersonal Computational States”

(1995e). Chapter 7 is based on “Ecological Perception and the Notion of
a Nonconceptual Point of View,” my contribution to The Body and the
Self (MIT Press, 1995), which I edited with Anthony Marcel and Naomi
Eilan. Section 6.1 is drawn from the coauthored Introduction to that
work.
It has been a pleasure to work once again with the MIT Press. Amy
Brand has seen the project through with tact and forbearance. I am partic-
ularly grateful to her for securing the services of an excellent but sadly
anonymous reviewer who made more useful suggestions than any author
deserves. The final text has been greatly improved by the copyediting
skills of Alan Thwaits.
A Note to the Reader
This book has been written in an interdisciplinary spirit, and I very much
hope that it has things to say that will be of interest to workers in disci-
plines cognate to philosophy, particularly empirical psychology and cog-
nitive science. Some readers may well find that they can more easily grasp
the main line of my argument if they bypass some of the more narrowly
philosophical discussion involved in identifying the paradox of self-
consciousness and in setting up a possible framework for its resolution.
On an initial reading, therefore, they could omit the first four sections
of chapter 1, moving directly to sections 1.5 and 1.6. This should pro-
vide an outline of the central issues posed by the paradox of self-
consciousness, which can subsequently be filled in by referring back to
the earlier sections of the chapter. Sections 2.1 and 2.2 might also be
omitted on a first pass. Section 2.3 should be sufficient to give a sense of
the general strategy which I propose to adopt. In chapter 3, section 3.1 is
likely to be of interest primarily to a philosophical audience. After that, I
would hope to be able to hold my audience.
1
The Paradox of Self-Consciousness
It is very natural to think of self-consciousness as a cognitive state or,

more accurately, as a set of cognitive states. Self-knowledge is an example
of such a cognitive state. There are plenty of things I know about myself.
I know the sort of thing I am: a human being, a warm-blooded rational
animal with two legs. I know many of my properties and much of what
is happening to me, at both physical and mental levels. I also know things
about my past: things I have done and places I have been, as well as people
I have met. But I have many self-conscious cognitive states that are not
instances of knowledge. For example, I have the capacity to make plans
for the future—to weigh up possible courses of action in the light of
goals, desires, and ambitions. I am also capable of a certain type of moral
reflection, tied to moral self-understanding and moral self-evaluation. I
can pursue questions like, What sort of person am I? Am I the sort of
person that I want to be? Am I the sort of person that I ought to be?
When I say that I am a self-conscious creature, I am saying that I can
do all of these things. But what do they have in common? Are some more
important than others? Are they all necessary? Could I lack some and still
be self-conscious? These are central questions that take us to the heart of
many issues in metaphysics, the philosophy of mind, and the philosophy
of psychology. In reflecting on them, however, we have to start from
a paradox that besets philosophical reflection on self-consciousness. This
paradox emerges when we consider what might seem to be the obvious
thread that ties the various manifestations of self-consciousness together.
I call it the paradox of self-consciousness. It is a paradox that raises the
question of how self-consciousness is even possible. Some preparatory
work is required before it can be appreciated, however.
2 Chapter 1
1.1 ‘I’-Thoughts
Confronted with the range of putatively self-conscious cognitive states

listed above, one might naturally assume that there is a single ability that
they all presuppose. This is my ability to think about myself. I can only
have knowledge about myself if I have beliefs about myself, and I can only
have beliefs about myself if I can entertain thoughts about myself. The
same can be said for autobiographical memories and moral self-
understanding. These are all ways of thinking about myself.
Of course, much of what I think when I think about myself in these
self-conscious ways is also available to me to employ in my thoughts
about other people and other objects. My knowledge that I am a human
being deploys certain conceptual abilities that I can also deploy in think-
ing that you are a human being. The same holds when I congratulate
myself for satisfying the exacting moral standards of autonomous moral
agency. This involves concepts and descriptions that can apply equally to
myself and to others. On the other hand, when I think about myself, I am
also putting to work an ability that I cannot put to work in thinking about
other people and other objects. This is precisely the ability to apply those
concepts and descriptions to myself. It has become common to refer to
this ability as the ability to entertain ‘I’-thoughts. I shall follow this
convention.
What is an ‘I’-thought? Obviously, an ‘I’-thought is a thought that in-
volves self-reference. I can think an ‘I’-thought only by thinking about
myself. Equally obviously, though, this cannot be all that there is to say
on the subject. I can think thoughts that involve self-reference but are not
‘I’-thoughts. Suppose I think that the next person to get a parking ticket
in central Cambridge deserves everything he gets. Unbeknownst to me,
the very next recipient of a parking ticket will be me. This makes my
thought self-referring, but it does not make it an ‘I’-thought. Why not?
The answer is simply that I do not know that I will be the next person to
get a parking ticket in central Cambridge. If A is that unfortunate person,
then there is a true identity statement of the form I 5 A, but I do not
know that this identity holds. Because I do not know that this identity
holds, I cannot be ascribed the thought that I will deserve everything I
get. And so I am not thinking a genuine ‘I’-thought, because one cannot
think a genuine ‘I’-thought if one is ignorant that one is thinking about
The Paradox of Self-Consciousness 3
oneself. So it is natural to conclude that ‘I’-thoughts involve a distinctive

type of self-reference. This is the sort of self-reference whose natural lin-
guistic expression is the first-person pronoun ‘I’, because one cannot use
the first-person pronoun without knowing that one is thinking about
oneself.
We can tighten this up a bit. When we say that a given thought has a
natural linguistic expression, we are also saying something about how it
is appropriate to characterize the content of that thought. We are saying
something about what is being thought. This I term (without prejudice)
a propositional content. A propositional content is given by the sentence
that follows the ‘that’ clause in reporting a thought, a belief, or any prop-
ositional attitude. The proposal, then, is that ‘I’-thoughts are all and only
the thoughts whose propositional contents constitutively involve the first-
person pronoun.
This is still not quite right, however, because thought contents can be
specified in two ways. They can be specified directly or indirectly. An ex-
ample of a direct specification of content is (1):
(1) J. L. B. believes the proposition that he would naturally express by
saying, ‘I will be the next person to receive a parking ticket in central
Cambridge’.
A direct specification of content involves specifying what I would say in
oratio recta, if I were explicitly to express what I believe. In contrast, an
indirect specification of content proceeds in oratio obliqua. The model
here is reported speech. So the same content indirectly specified would
be (2):
(2) J. L. B. believes that he will be the next person to receive a parking
ticket in central Cambridge.
Proposition (2), however, can also be used as a report of the belief that
would be directly specified as follows:
(3) J. L. B. believes the proposition that he would naturally express by
saying, ‘J. L. B. will be the next person to receive a parking ticket in
central Cambridge’.
Nonetheless, (1) and (3) are not equivalent. This is easily seen. J. L. B.
could be suffering from an attack of amnesia and struggling to remember
his own name. So even though a kindly soul has just put it to him that
4 Chapter 1
J. L. B. will be the next person to get a parking ticket in central Cam-

bridge and he believes that kindly soul, he fails to realize that he is J. L. B.
Because of this, (1) is not a correct report of his belief, although (3) is.
Nonetheless, as it stands, (2) is a correct indirect report of both (1) and
(3). This creates a problem, because only (1) is a genuine ‘I’-thought, ac-
cording to the suggested criterion of having a content that constitutively
involves the first-person pronoun, and yet there appears to be no way of
capturing the distinction between (1) and (3) at the level of an indirect
specification of content.
˜
This problem can be solved with a device due to Hector-Neri Castaneda
˜
(1966, 1969). Castaneda distinguishes between two different roles that
the pronoun ‘he’ can play in oratio obliqua clauses. On the one hand, ‘he’
can be employed in a proposition that the antecedent of the pronoun (i.e.
the person named just before the clause in oratio obliqua) would have
˜
expressed using the first-person pronoun. In such a situation, Castaneda
holds that ‘he’ is functioning as a quasi-indicator. He suggests that when
‘he’ is functioning as a quasi-indicator, it be written as ‘he*’. Others have
described this as the indirect reflexive pronoun (Anscombe 1975). When
‘he’ is functioning as an ordinary indicator, it picks out an individual in
such a way that the person named just before the clause in oratio obliqua
need not realize the identity of himself with that person. Clearly, then, we
can disambiguate between (2) employed as an indirect version of (1) and
(2) employed as an indirect version of (3) by distinguishing between (2.1)
and (2).
(2.1) J. L. B. believes that he* will be the next person to receive a
parking ticket in central Cambridge.
Proposition (2.1) is an example of the indirect reflexive, while (2) is not.
So, we can tie up the definition of an ‘I’-thought as follows.
Definition An ‘I’-thought is a thought whose content can only be speci-
fied directly by means of the first person pronoun ‘I’ or indirectly by
means of the indirect reflexive pronoun ‘he*’.
It was suggested earlier that all the cognitive states under consideration
presuppose the ability to think about oneself. This is not only what they
all have in common. It is also what underlies them all. We can now see in
more detail what this suggestion amounts to. The claim is that what
makes all those cognitive states modes of self-consciousness is the fact

that they all have contents that can only be specified directly by means of
the first person pronoun ‘I’ or indirectly by means of the indirect reflexive
pronoun ‘he*’. I will term such contents first-person contents.
1.2 Two Types of First-Person Content
The class of first-person contents is not a homogenous class. There is an

important distinction to be drawn between two different types of first-
person contents, corresponding to two different modes in which the first
person can be employed. The existence of this distinction was first noted
by Wittgenstein in an important passage from The Blue Book.1 He gives
the following examples of the two different modes:
There are two different cases in the use of the word “I” (or “my”) which I might
call “the use as object” and “the use as subject.” Examples of the first kind of use
are these: “My arm is broken,” “I have grown six inches,” “I have a bump on my
forehead,” “The wind blows my hair about.” Examples of the second kind are: “I
see so-and-so,” “I try to lift my arm,” “I think it will rain,” “I have a toothache.”
(Wittgenstein 1958, 66–67)
The explanation he gives of the distinction hinges on whether or not they
are judgements that involve identification. The passage continues thus:
One can point to the difference between these two categories by saying: The cases
of the first category involve the recognition of a particular person, and there is in
these cases the possibility of an error, or as I should rather put it: The possibility
of an error has been provided for. . . . It is possible that, say in an accident, I
should feel a pain in my arm, see a broken arm at my side, and think it is mine
when really it is my neighbour’s. And I could, looking into a mirror, mistake a
bump on his forehead for one on mine. On the other hand, there is no question
of recognizing a person when I say I have toothache. To ask “are you sure that it’s
you who have pains?” would be nonsensical. (Wittgenstein 1958, 67)
Wittgenstein is drawing a distinction between two types of first-person
contents. The first type, which he describes as invoking the use of ‘I’ as
object, can be analyzed in terms of more basic propositions. Suppose that
the thought ‘I am f ’ involves such a use of ‘I’. Then we can understand it
as a conjunction of the following two thoughts: ‘a is f ’ and ‘I am a’. We
can term the former a predication component and the latter an identifica-
tion component (Evans 1982, 185). The reason for breaking the original
thought down into these two components is precisely the possibility of
6 Chapter 1
error that Wittgenstein stresses in the second of the quoted passages. One
can be quite correct in predicating that someone is f, even though mis-
taken in identifying oneself as that person.
One way of putting this distinction, derived ultimately from Sydney
Shoemaker, is in terms of immunity to a certain sort of error. First-person
contents in which the ‘I’ is used as subject are immune to error through
misidentification relative to the first-person pronoun. Shoemaker explains
this as follows:
To say that a statement “a is f ” is subject to error through misidentification rela-
tive to the term ‘a’ means that the following is possible: the speaker knows some
particular thing to be f, but makes the mistake of asserting “a is f ” because, and
only because, he mistakenly thinks that the thing he knows to be f is what ‘a’
refers to. (Shoemaker 1968, 7–8)
The point, then, is that one cannot be mistaken about who is being
thought about. In one sense, Shoemaker’s criterion of immunity to error
through misidentification relative to the first-person pronoun (henceforth
simply “immunity to error through misidentification”) is too restrictive.
Beliefs with first-person contents that are immune to error through mis-
identification tend to be acquired on grounds that usually do result in
knowledge, but they do not have to be. The definition of immunity to
error through misidentification needs to be adjusted to accommodate this
by formulating it in terms of justification rather than knowledge.
The connection to be captured is between the sources or grounds from
which a belief is derived and the justification there is for that belief. Beliefs
and judgements are immune to error through misidentification in virtue
of the grounds on which they are based. This is very important. The cate-
gory of first-person contents being picked out is not defined by its subject
matter or by any points of grammar. What demarcates the class of judge-
ments and beliefs that are immune to error through misidentification is
the evidence base from which they are derived, or the information on
which they are based (Evans 1982, chap. 7). So, to take one of Wittgen-
stein’s examples, my thought that I have a toothache is immune to error
through misidentification because it is based on my feeling a pain in my
teeth. Similarly, the fact that I am consciously perceiving you makes my
belief that I am seeing you immune to error through misidentification.
This suggests that we can modify Shoemaker’s definition as follows:
Definition To say that a statement “a is f” is subject to error through

misidentification relative to the term ‘a’ means that the following is pos-
sible: the speaker is warranted in believing that some particular thing is
f, because his belief is based on an appropriate evidence base, but he
makes the mistake of asserting ‘a is f ’ because, and only because, he
mistakenly thinks that the thing he justifiedly believes to be f is what ‘a’
refers to.2
First-person contents that are immune to error through misidentification
can be mistaken, but they do have a prima facie warrant in virtue of the
evidence on which they are based, because the fact that they are derived
from such an evidence base is closely linked to the fact that they are im-
mune to error through misidentification. Of course, there is room for con-
siderable debate about what types of evidence base are correlated with
this class of first-person contents, and this will be discussed further below.
It seems, then, that the distinction between different types of first-
person content originally illustrated by Wittgenstein can be characterized
in two different ways. We can distinguish between those first-person con-
tents that are immune to error through misidentification and those that
are subject to such error. Alternatively, we can discriminate between first-
person contents with an identification component and those without such
a component. For the purposes of this book I shall take it that these differ-
ent formulations each pick out the same classes of first-person contents,
although in interestingly different ways.
It will be obvious that these two classes of first-person contents are
asymmetrically related. All first-person contents subject to error through
misidentification contain an identification component of the form ‘I am
a’. Now, consider the employment of the first-person pronoun in that
identification component. Does it or does it not have an identification
component? If it does, then a further identification component will be
implicated, of which the same question can be asked. Clearly, then, on
pain of an infinite regress, at some stage we will have to arrive at an em-
ployment of the first-person pronoun that does not presuppose an identi-
fication component. The conclusion to draw, then, is that any first-person
content subject to error through misidentification will ultimately be
anchored in a first-person content that is immune to error through
misidentification.3
8 Chapter 1
This leaves us with a class of self-ascriptions that are identification-free.

How are we to explain what is going on in this class of self-ascriptions?
Shoemaker’s suggestion is that to see how such identification-free self-
reference is possible, we need to investigate the class of predicates that
can be employed in such self-ascriptions. He writes, “There is an impor-
tant and central class of psychological predicates, let us call them P*
predicates, each of which can be known to be instantiated in such a way
that knowing it to be instantiated in that way is equivalent to knowing it
to be instantiated in oneself” (Shoemaker 1968, 16). It is a matter of
some controversy just what predicates fall into this class, and indeed,
whether they are all psychological, as Shoemaker claims. This will be dis-
cussed below. For present purposes, we can just take being in pain to be
a canonical example of a P* predicate. Shoemaker’s claim is that no self-
identification is required to move from the judgement ‘There is pain’ to
the judgement ‘I am in pain’, provided that the judgement is based on
actually experiencing the pain. But how are we to flesh this out? One way
of doing so is to employ the idea of a primitively compelling inference, an
inference that a subject both can and should make without any further
reasoning or justification (Peacocke 1992). The thought is that when the
predicate ‘pain’ is applied on the basis of feeling the pain, the inference
from the existential statement ‘There is pain’ to the self-ascription ‘I am
in pain’ is primitively compelling in virtue of the status of pain as a P*
predicate. Another way (and one that might be closer to Shoemaker’s
view) would be to deny that any such inference is required at all. Knowing
that there is pain just is knowing that one is in pain, which in turn can be
understood as a disposition to think ‘I am in pain’. On this view, thinking
‘There is pain’ is just another way of thinking ‘I am in pain’ (provided, of
course, that the thought is grounded on experiencing the pain).
Neither of these suggestions, though, can explain what it is to possess
the capacity to make self-ascriptions that are immune to error through
misidentification. Both make essential reference to the capacity to think
thoughts like ‘I am in pain’, and these thoughts fall squarely into the cate-
gory of identification-free self-ascriptions, which we are trying to explain.
Clearly, we need to take an alternative tack to make progress here.
1.3 The First-Person Pronoun and a Deflationary Account of

Self-Consciousness
Since we cannot explain what immunity to error through misidentifica-

tion consists in through the class of predicates that feature in identifica-
tion-free judgements, the obvious alternative is to look at the other
component of identification-free judgements, namely the first-person ele-
ment, naturally expressed with the first-person pronoun. And in fact there
is at least one good reason to think that the semantics of the first-person
pronoun might hold the key to immunity to error through misidentifica-
tion. This emerges when one reflects that the first-person pronoun has
guaranteed reference. It is commonly and correctly held that the practices
and rules governing the use of the first-person pronoun determine what
its reference will be whenever it is employed, according to the simple rule
that whenever the first-person pronoun is used correctly, it will refer to
the person using it (Barwise and Perry 1981, Strawson 1994). It follows
from this rule that the correct use of ‘I’ is enough to guarantee both that
it has a referent and that the referent is the user. So it is impossible for the
first-person pronoun (correctly and genuinely employed) to fail to refer or
for it to refer to someone other than the person using it.
Guaranteed reference and immunity to error through misidentification
are clearly very closely related. It is natural to suggest that immunity to
error through misidentification (relative to the first-person pronoun) is
simply a function of the meaning rule for the first-person pronoun.4 The
claim here is not, of course, that the set of judgements with guaranteed
reference maps onto the set of judgements immune to error through mis-
identification. There are clear instances of judgements expressible with
the first-person pronoun that ipso facto have guaranteed reference but
that fail to be immune to error relative to the first-person pronoun. Sup-
pose, for example, that a baritone singing in a choir hears a tuneful voice
that he mistakenly judges to be his own when it is in fact the voice of his
neighbour (also a baritone). He then judges, ‘I am singing in tune’. This
is a clear instance of misidentification relative to the first-person pronoun,
because the baritone makes the mistake of assuming that the baritone
whom he justifiably believes to be singing in tune is the baritone to whom
10 Chapter 1
‘I’ refers. Nonetheless, this is not a counterexample to the suggested view,

because the evidence base upon which the baritone judges is not one of
the categories of evidence bases that generates judgements immune to
error relative to the first-person pronoun. Reformulating the issue will
make it more perspicuous. The evidence base is such that the ensuing
judgement has to be analyzed conjunctively in terms of an identification
component and a predication component (‘That baritone is singing in
tune’ and ‘I am that baritone’).
It seems natural to combine this close connection with our earlier
conclusions by proposing an account of self-consciousness in terms of the
capacity to think ‘I’-thoughts that are immune to error through misidenti-
fication, where immunity to error through misidentification is a function
of the semantics of the first-person pronoun. This would be a deflationary
account of self-consciousness. If we can straightforwardly explain what
makes those first-person contents immune to error through misidentifi-
cation with reference to the semantics of the first-person pronoun, then
it seems fair to say that the problem of self-consciousness has been dis-
solved, at least as much as solved. The proposed account would be on a
par with other noted examples of deflationism, such as the redundancy
theory of truth.
Let me break this deflationary theory down into a set of distinct claims.
The first claim is that what is distinctive about the various forms of self-
consciousness is that they involve thinking ‘I’-thoughts. As a thesis about
explanation, this comes out as the following claim:
Claim 1 Once we have an account of what it is to be capable of thinking
‘I’-thoughts, we will have explained everything distinctive about self-
consciousness.
The second claim is a claim about the form that such an account will have
to take. It stems from the thought that what is distinctive about ‘I’-
thoughts is that they are either themselves immune to error through mis-
identification or they rest on further ‘I’-thoughts that are immune in that
way. This yields the following claim:
Claim 2 Once we have an account of what it is to be capable of thinking
thoughts that are immune to error through misidentification, we will have
explained everything distinctive about the capacity to think ‘I’-thoughts.
The final claim derives from the thought that immunity to error through
misidentification is a function of the semantics of the first-person
pronoun:
Claim 3 Once we have an account of the semantics of the first-person
pronoun, we will have explained everything distinctive about the capacity
to think thoughts that are immune to error through misidentification.
The suggestion is that the semantics of the first-person pronoun will ex-
plain what is distinctive about the capacity to think thoughts immune to
error through misidentification. Of course, this needs to be fleshed out a
little. Semantics alone cannot be expected to explain the capacity for
thinking such thoughts. The point must be that all that there is to the
capacity to think thoughts that are immune to error through misidentifi-
cation is the capacity to think the sort of thoughts whose natural linguis-
tic expression involves the first-person pronoun, where this capacity is
given by mastery of the semantics of the first-person pronoun. This yields
the following reformulation:
Claim 3a Once we have explained what it is to master the semantics of
the first-person pronoun, we will have explained everything distinctive
about the capacity to think thoughts immune to error through mis-
identification.
So on this view, mastery of the semantics of the first-person pronoun is
the single most important explanandum in a theory of self-consciousness.5
One immediate question that might be put to a defender of the defla-
tionary theory is how mastery of the semantics of the first-person pro-
noun can make sense of the distinction between first-person contents that
are immune to error through misidentification and first-person contents
that lack such immunity. Both types of first-person content are naturally
expressed by means of the first-person pronoun. So how can mastery of
the semantics of the first-person pronoun capture what is distinctive
about those first-person contents that are immune to error through mis-
identification? One can see, however, that this is only an apparent diffi-
culty when one remembers that those first person contents that are
immune to error through misidentification (those employing ‘I’ as object)
have to be broken down into their two constituent elements: the identifi-
cation component and the predication component. It is the identification
12 Chapter 1
component of such judgements that mastery of the semantics of the first-

person pronoun is being called upon to explain, and the identification
component is, of course, immune to error through misidentification.
There are grave problems with the deflationary account, as will emerge
in the next section, but at this stage it is important to stress how well it
meshes with one persuasive strand of thought about self-consciousness.
Many philosophers have thought that what is distinctive about beings
who are self-conscious is that they are capable of ascribing predicates to
themselves. Those predicates are sometimes taken to be psychological (by
those of a Cartesian bent) and sometimes both psychological and physi-
cal. If it is assumed that the predicates in question have a constant sense,
whether they are applied to oneself or to others, then everything distinc-
tive about the self-conscious grasp of those predicates that apply to one
seems to fall on the act of self-ascription, and in particular, on the first-
person pronoun by which that self-ascription is effected. A move toward
the deflationary view is therefore a natural next step.
It is also important to stress how the deflationary theory of self-
consciousness, and indeed any theory of self-consciousness that accords
a serious role in self-consciousness to mastery of the semantics of the
first-person pronoun, is motivated by an important principle that has gov-
erned much of the development of analytical philosophy. This is the prin-
ciple that the philosophical analysis of thought can only proceed through
the philosophical analysis of language. The principle has been defended
most vigorously by Michael Dummett. Here is a particularly trenchant
statement:
Thoughts differ from all else that is said to be among the contents of the mind in
being wholly communicable: it is of the essence of thought that I can convey to
you the very thought that I have, as opposed to being able to tell you merely
something about what my thought is like. It is of the essence of thought not
merely to be communicable, but to be communicable, without residue, by means
of language. In order to understand thought, it is necessary, therefore, to under-
stand the means by which thought is expressed. (Dummett 1978, 442)
Dummett goes on to draw the clear methodological implications of this
view of the nature of thought:
We communicate thoughts by means of language because we have an implicit

understanding of the workings of language, that is, of the principles governing
the use of language; it is these principles, which relate to what is open to view in
the employment of language, unaided by any supposed contact between mind and
mind other than via the medium of language, which endow our sentences with
the senses that they carry. In order to analyse thought, therefore, it is necessary to
make explicit those principles, regulating our use of language, which we already
implicitly grasp. (Dummett 1978, 442)
Many philosophers would want to dissent from the strong claim that the
philosophical analysis of thought through the philosophical analysis of
language is the fundamental task of philosophy. But there is a weaker
principle that is very widely held:
The Thought-Language Principle The only way to analyze the capacity
to think a particular range of thoughts is by analyzing the capacity for
the canonical linguistic expression of those thoughts.
The Thought-Language Principle dictates an obvious methodology for
the analysis of thought. To understand what it is to be capable of thinking
a particular range of thoughts, one must first find the canonical linguistic
expression for the thoughts in question and then explain the linguistic
skills that must be mastered for the use of that linguistic expression.
When the methodology associated with the Thought-Language Prin-
ciple is combined with the point just made about predicates having a
constant sense irrespective of whether they are being used for first-,
second-, or third-person ascriptions, the deflationary theory falls out very
naturally. The weaker thesis tells us that the only way to understand self-
conscious thoughts is through understanding the linguistic expression of
those self-conscious thoughts. The fact that the paradigm cases of self-
conscious thoughts (like instances of self-knowledge) involve ascribing
certain properties to oneself means that attention must be directed to the
distinctive features of the linguistic means whereby a subject is able to
apply certain predicates to himself in the appropriate way. Since the
constancy-of-sense thesis means that the distinctive features cannot lie in
the relevant predicates, we soon arrive at the deflationary theory.
Despite these two powerful motivations for it, the deflationary theory
runs into two very serious problems that together constitute what I term
the paradox of self-consciousness. In the following section I will bring
these problems out.6
14 Chapter 1
1.4 Explanatory Circularity and Capacity Circularity
According to claim 3a, once we have explained what it is to master the

semantics of the first-person pronoun, we have explained everything dis-
tinctive about the capacity to think thoughts that are immune to error
through misidentification, and hence everything distinctive about self-
consciousness. We can start evaluating this by being a little more precise
about what the semantics are that have to be grasped. I suggested earlier
that the rule governing the use of the first-person pronoun is that it refers
to the person using it. More precisely, we can state the following token-
reflexive rule:
Token-reflexive rule, version 1 When a person employs a token of ‘I’, in
so doing he refers to himself.7
This token-reflexive rule is the key to the semantics of the first-person
pronoun. So, according to the deflationary account, mastery of this rule
is the explanatory key to the capacity to think thoughts immune to error
through misidentification.
An obvious problem with this suggestion emerges from the points
made earlier about the third-person pronoun ‘he’. It can be employed
as the ordinary reflexive ‘he’ or as the indirect reflexive, which, follow-
˜
ing Castaneda, can be written ‘he*’. Clearly in the context of the token-
reflexive rule it is the indirect reflexive that is required. Reporting
self-reference by means of the ordinary reflexive ‘he’ leaves open the possi-
bility that the person who is referring to himself might be referring to
himself without realizing that he is so doing, and this is clearly not pos-
sible with the first-person pronoun. So the token-reflexive rule needs to
be reformulated:
Token-reflexive rule, version 2 When a person employs a token of ‘I’, in
so doing, he refers to himself*.
This creates obvious problems of circularity, however, because we can
only understand how a person can refer to himself* by understanding
how a person can refer to himself by employing the first-person pronoun.
The indirect reflexive in indirect speech needs to be explained through
the first-person of direct speech, which is, of course, ‘I’.8 This circularity
appears damaging to the deflationary account of self-consciousness. Re-
call that I characterized a first-person content as one that can be specified

directly only by means of the first-person pronoun ‘I’ or indirectly only
by means of the indirect reflexive ‘he*’. If, as the deflationary account
suggests, we take the capacity to think thoughts with first-person contents
to be what we are trying to explain, then it is viciously circular to suggest
that this capacity can be explained by mastery of a rule that contains a
first-person content.
Perhaps, though, the token-reflexive rule can be formulated in a way
that avoids this difficulty. Consider the following suggestion:
Token-reflexive rule, version 3 Any token of ‘I’ refers to whoever pro-
duced it.9
This seems on the face of it to avoid the problem, in the sense that it does
not employ the third-person pronoun and therefore does not force us to
choose between direct and indirect reflexive construals of that pronoun
in oratio obliqua. But how are we to construe it? In one sense version 3
of the rule can be read as just a notational variant of version 1 and hence
subject to precisely the same ambiguity between ordinary and indirect
reflexive pronouns. But there is a way of construing it that is more
promising:
Token-reflexive rule, version 4 If a person employs a token of ‘I’, then
he refers to himself in virtue of being the producer of that token.
In version 4 it is plausible to suggest that ‘himself’ can be read as an
ordinary reflexive pronoun. The possibility that a person might refer to
himself without realizing that he is doing so is what forced us toward the
indirect reflexive in version 1. But in version 4 it seems that this possibility
is ruled out by the fact that the person is referring to himself as the pro-
ducer of the token of ‘I’.
Again, though, there is a difficulty.10 What is it for someone to refer to
himself in virtue of being the producer of a given token? It seems clear
that such a person will not succeed in referring to himself unless he knows
that he produced the token in question. But this brings us straight back
to the circularity problem. Employing a token of the first-person pronoun
in a way that reflects mastery of its semantics (as construed by the token-
reflexive rule, version 4) requires knowing that one is the producer of
the relevant token and that is a piece of knowledge with a first-person
16 Chapter 1
content.11 This is the first strand of what I term the paradox of self-
consciousness. Any theory that tries to elucidate the capacity to think
first-person thoughts through linguistic mastery of the first-person pro-
noun will be circular, because the explanandum is part of the explanans,
either directly, as in version 2, or indirectly, as in version 4. Let me call
this explanatory circularity.
It is important to keep the problem of explanatory circularity distinct
from a closely related problem. Elizabeth Anscombe has controversially
argued that ‘I’ is not a referring expression. As part of her argument for
that bizarre conclusion, she maintains that no version of the token-
reflexive rule can provide a noncircular account of the meaning of the first-
person pronoun. She supports this with considerations much like those I
have adduced. The version of the token-reflexive rule she considers is this:
Token-reflexive rule, version 5 ‘I’ is a word that each speaker [of En-
glish] uses only to refer to himself.
She points out, very much as I have done, that although version 5 truly
fixes the reference of the first-person pronoun, it does not capture what
is distinctive about ‘I’, because it fails to distinguish between genuine self-
reference and accidental self-reference. But if, she continues, version 5 is
adjusted to accommodate the distinction by replacing the direct reflexive
pronoun with the indirect reflexive pronoun, then she identifies problems
of circularity: “The explanation of the word ‘I’ as ‘the word which each
of us uses to speak of himself’ is hardly an explanation!—At least,
it is no explanation if that reflexive has in turn to be explained in terms
of ‘I’; and if it is the ordinary reflexive then we are back at square one”
(Anscombe 1975, 48). This dilemma comes about, she thinks, only be-
cause the first-person pronoun is being treated as a referring proper name.
The only solution she sees is to deny that ‘I’ is a referring expression at all.
I must stress that the notion of explanatory circularity I have outlined
in no way commits me to Anscombe’s position, although both Anscombe
and I are stressing the same features of token-reflexive accounts of the
semantics of the first-person pronoun. The conclusion I draw from those
features is that the deflationary theory of self-consciousness (and by ex-
tension, any account of self-consciousness that tries to explain what is
distinctive about self-conscious thoughts in terms of mastery of the first-
person pronoun) will end up being viciously circular because mastery of

the semantics of the first-person pronoun involves the capacity to think
first-person thoughts. This has no implications for the question Ans-
combe addresses of whether the token-reflexive rule yields an adequate
account of the semantics of the first-person pronoun. And in fact it seems
to me that, on this question, Anscombe is surely wrong. The token-
reflexive rule is an adequate account of the semantics of the first-person
pronoun precisely because it fixes the reference of any token utterance of
‘I’, and that is all that a meaning rule for ‘I’ needs to do. The token-
reflexive rule as it stands is not circular. It becomes circular only if it is
adjusted to rule out the possibility of accidental self-reference either by
replacing the direct reflexive pronoun with the indirect reflexive pronoun
or by requiring that the utterer of a token of ‘I’ should know that he
produced the relevant token. But neither of these modifications is required
for an account of the meaning of ‘I’. To hold that there is any such re-
quirement is to confuse the semantics of the first-person pronoun with
the pragmatics of the first-person pronoun. The problem of explanatory
circularity arises because it is not sufficient for the deflationary account,
or any comparable account of self-consciousness, simply to provide an
account of the semantics of the first-person pronoun. As stressed in the
previous section, the deflationary account needs to provide an account of
mastery of the first-person pronoun, and this will have to include both
the semantics and the pragmatics. Hence the modifications have to be
made, with the ensuing circularity.
The second problem with the deflationary account is closely related.
What creates the difficulty with version 4 of the token-reflexive rule is that
the reflexive self-reference achieved with the first-person pronoun requires
that the person referring to himself know that he is the producer of the
token in question (although, to repeat, I am offering this merely as a nec-
essary condition rather than as a sufficient condition). This knowledge,
of course, is knowledge with a first-person content. So the deflationary
account of self-consciousness needs to advert to precisely the first-person
thoughts it is attempting to explain if it is to provide an understanding of
the mastery of the semantics of the first-person pronoun able to anchor
the explanation of self-consciousness. This makes the deflationary theory
viciously circular, because the explanandum is required to explain the
18 Chapter 1
explanans. This is explanatory circularity. It is important to note, how-

ever, that this is not the only type of circularity created by the mutual
interdependence of first-person thought and mastery of the semantics of
the first-person pronoun. Another form of circularity arises because the
attempt is being made to explain one capacity (the capacity to think
thoughts with first-person contents) in terms of another capacity (mastery
of the semantics of the first-person pronoun). What is at stake is the rela-
tion between the relevant capacities that underlie the explanans and the
explanandum respectively. The point here is that the capacity for reflexive
self-reference by means of the first-person pronoun presupposes the ca-
pacity to think thoughts with first-person contents, and hence cannot
be deployed to explain that capacity. In other words, a degree of self-
consciousness is required to master the use of the first-person pronoun.
It is natural to describe this as an instance of capacity circularity.
But what is wrong with capacity circularity? Why should we take ca-
pacity circularity to be as vicious as explanatory circularity? It would be
in keeping with the general spirit of the deflationary account to hold that
the existence of capacity circularity just shows that we have reached the
limits of explanation. Pointing out the interconnections and interdepen-
dencies between various forms of self-consciousness is, on this view, as
far as we can go. The thought is that we are dealing with a set of abilities
that can be explained from within but not from without. No explanation
of the constituents of that set of abilities in terms of abilities that are
more fundamental can be expected, and, it might be suggested, capacity
circularity merely reflects this. On this view, capacity circularity is merely
a reflection of the fact that certain cognitive abilities form what Christo-
pher Peacocke (1979) has called a local holism.
That there are such local holisms is indisputable. And in many cases
the existence of such local holisms presents no difficulties of circularity.12
There is nothing troubling per se about an interdependence of explana-
tion. In the case of self-consciousness, however, problems of circularity
are not so easily dismissed, because they go far beyond a putative inter-
dependence of explanation. Capacity circularity has implications for the
ontogenesis of self-consciousness. In brief, if we hold that the various
abilities at the root of self-conscious thought form a local holism and
consequently can only be explained in terms of each other, then we are
ruling out the possibility of explaining how any or all of those abilities
can be acquired in the normal course of cognitive development. But this
needs to be made more explicit.
Let me begin with the following constraint, which seems to me to be
operative in all discussions of cognition.
The Acquisition Constraint If a given cognitive capacity is psychologi-
cally real, then there must be an explanation of how it is possible for an
individual in the normal course of human development to acquire that
cognitive capacity.
There is a potential ambiguity in the concept of explanation. On the one
hand, ‘explanation’ might be taken to refer to the actual account that is
offered to explain a particular phenomenon. On the other, ‘explanation’
might be thought to refer to the metaphysical basis of such an account,
that is to say, to the facts that are characterized by such an account if it is
true. I intend the Acquisition Constraint in the second of these two senses
(which might be termed the metaphysical sense as opposed to the episte-
mological sense). In this metaphysical sense the Acquisition Constraint is
truistic. It does not demand that philosophers or anybody else should be
able to provide an account of how the capacity in question is, or could
be, acquired. What it does provide, however, is a contrapositive test for
the psychological reality of any putative cognitive ability. For any pro-
posed cognitive ability, if it is impossible for an individual to acquire that
ability, then it cannot be psychologically real.
Of course, the usefulness of the Acquisition Constraint depends on
how we understand its being satisfied. Let me briefly outline one such
way, which seems to me to be paradigmatic. Every individual has an in-
nate set of cognitive capacities that it possesses at birth. Let me call that
S0. At any given time t after birth an individual will have a particular set
of cognitive capacities. Let me call that St. Now consider a given cognitive
capacity c that is putatively in St. Suppose that for any time t 2 n the
following two conditions are satisfied. First, it is conceivable how c could
have emerged from capacities present in St2n. Second, it is conceivable
how the capacities present in St2n could have emerged from the capacities
present in S0. By its being conceivable that one capacity could emerge
from a given set of capacities, I mean that it is intelligible that (in the
right environment) the individual in question could deploy the cognitive
20 Chapter 1
capacities it already has to acquire the new capacity. If those conditions

are satisfied, then we have a paradigm case of learning.
It is precisely here that capacity circularity bites, for the following rea-
son. If mastery of the first-person pronoun is to meet the Acquisition
Constraint, then clearly there must be some time t when St includes the
capacity for linguistic mastery of the first-person pronoun, and a corre-
sponding time t 2 n when St2n does not include that capacity but includes
other capacities on the basis of which it is intelligible that an individual
could acquire the capacity for linguistic mastery of the first-person pro-
noun. The implication of capacity circularity, however, is that any such
St2n will have to include the capacity to think thoughts with first-person
contents, and hence that any such St2n will have to include the capacity
for linguistic mastery of the first-person pronoun. Clearly, then, there can
be no such St2n in terms of which the Acquisition Constraint could be
satisfied.
There will be philosophers unprepared to accept that the Acquisition
Constraint, or anything like it, can have a role to play in philosophical
accounts of concepts and conceptual abilities. The most obvious basis for
such a view would be a Fregean distrust of “psychologism” that leads to a
rigid division of labor between philosophy and psychology. The operative
thought is that the task of a philosophical theory of concepts is to explain
what a given concept is or what a given conceptual ability consists in.
This, it is frequently maintained, is something that can be done in com-
plete independence of explaining how such a concept or ability might be
acquired. The underlying distinction is one between philosophical ques-
tions centering around concept possession and psychological questions
centering around concept acquisition. This distinction is, to my mind,
suspect.13 For my present purposes, though, all that I need to point out is
that, however strictly one does adhere to the distinction, it provides no
support for a rejection of the Acquisition Constraint. The neo-Fregean
distinction is directed against the view that facts about how concepts are
acquired have a role to play in explaining and individuating concepts. But
this view does not have to be disputed by a supporter of the Acquisition
Constraint. All that the supporter of the Acquisition Constraint is com-
mitted to is the principle that no satisfactory account of what a concept
is should make it impossible to provide an explanation of how that con-
cept can be acquired. The Acquisition Constraint has nothing to say

about the further question of whether the psychological explanation in
question has a role to play in a constitutive explanation of the concept in
question, and hence is not in conflict with the neo-Fregean distinction.
1.5 Capacity Circularity: An Innatist Solution?
Although the notion of capacity circularity just defined has emerged from
what might initially appear to be narrow philosophical concerns, the
problems it generates are far wider in their application.
The capacity circularity of the first-person pronoun ‘I’ arises because
mastery of the first-person pronoun requires the capacity to think certain
self-conscious thoughts that can only be expressed with the first-person
pronoun ‘I’. For genuine employment of the first-person pronoun it is not
enough that a speaker should utter ‘I’ in full knowledge of the token-
reflexive rule that a token of ‘I’ refers to the utterer of that token. The
speaker also needs to grasp that he himself is the utterer of that token,
which is a thought that can only be expressed in terms of the first-person
pronoun for reasons that I have tried to bring out. This has implications
for thinking about how the first-person pronoun can be learned. It is not
simply a matter of learning that there is an expression governed by the
token-reflexive rule that it refers to whoever utters it. Rather, it is a matter
of learning that there is an expression governed by the rule that it refers
to its utterer when that utter intends to refer to himself*. To put it in the
first person, I can only learn to employ the first-person pronoun by learn-
ing that there is an expression governed by the rule that it refers to me
when I intend to refer to myself.
The problem that this creates is a simple one. It seems to make it impos-
sible to understand how mastery of the first-person pronoun could ever
be genuinely learned. The thought that I need to grasp if I am to learn
how to employ the first-person pronoun is not a thought that I can enter-
tain before I have mastered the first-person pronoun—not, at least, ac-
cording to any theory that accepts the widely held view that a thought
cannot be entertained by a creature who lacks the ability for the canon-
ical linguistic expression of that thought. So, to master the first-person
pronoun, I must already have mastered the first-person pronoun.
22 Chapter 1
It might be thought that capacity circularity as thus described both

confirms, and can be resolved by, an innatist account of the relevant abili-
ties, because it is only threatening on a crude empiricist account of learn-
ing and concept acquisition. Much of contemporary cognitive science and
linguistics postulates the existence of innate modules as a way of ex-
plaining how cognitive abilities can be acquired.14 This is offered as an
explicit contrast to learning-based developmental theories (Gazzaniga
1992). In an innatist account the learning environment serves merely as a
source of triggers determining the selection of one of a range of circuits
or parameters. There are two areas in which such theories have become
prominent. The first and most celebrated is the nativist theory of language
acquisition pioneered by Chomsky. According to the nativist view, lan-
guages can be learned only if the language user is already innately aware
of their essential syntactic structure.15 The second is the modular version
of the nativist theory of mind in humans and (more controversially) in
some higher primates.16 Combining these two theories offers the prospect
of an innatist response to the capacity circularity of the first-person pro-
noun, by showing how the relevant cognitive abilities are present in what
I earlier termed the individual’s innate endowment.
There are several reasons, however, why this initial reaction should be
resisted. The first is quite straightforward. Despite the complexity and
sophistication of the dominant government-binding (GB) paradigm in
transformational grammar (Chomsky 1981, Haegman 1993), the theory
has nothing to say about the first-person pronoun that will help with the
problem of capacity circularity. When the first-person pronoun is em-
ployed nominatively as the subject of a sentence, it counts simply as an R
expression (referring expression) and as such is governed solely by the
third constraint of binding theory, namely, that R expressions occurring
in noun phrases are always free. When the third-person pronoun is used
anaphorically in indirect-speech reports, it is subject to the first constraint
of binding theory, namely, that anaphoric pronouns are bound in the gov-
erning category, but this is, of course, impervious to the distinction be-
tween the direct reflexive ‘he’ and indirect reflexive ‘he himself’ (or ‘he*’,
˜
as Castaneda writes it). But none of this is surprising, since the peculiari-
ties and complexities of the first-person pronoun arise at the semantic
level and GB theory is a theory about syntax.
The second point follows from this. The existence of an innate univer-
sal grammar, no matter what its degree of complexity and no matter how
its parameters are set, does not mean that language is not learned. The
postulation of an innate universal grammar is not in itself a theory of
language acquisition. Although motivated by the thought that without
such an innate grammar the acquisition of language would be impossible
(the famous “poverty of the stimulus” argument), it leaves open the cru-
cial question of how the innately given parameters are set. What governs
the transition from the initial state of the language faculty to the steady
state of adult linguistic competence? There are, broadly speaking, three
candidate theories here. According to the “no growth” theory, language
acquisition does not involve a process of learning, and the parameters are
set purely by exposure to linguistic data. The partial and extended nature
of the process of language acquisition is explained in terms of the devel-
opment of nonlinguistic cognitive capacities, such as memory or atten-
tion, that expand the range of linguistic data to which children can be
exposed. According to the “maturation” theory, on the other hand, dif-
ferent components of universal grammar mature at different times, and
this explains the appearance of linguistic learning.
Neither of these theories will help with capacity circularity, however,
for familiar reasons. Both the no-growth and maturation theories deal
solely with the acquisition of syntax, and the problems of capacity circu-
larity arise at the level of semantics. This leads into the third theory of
language acquisition. Stephen Pinker’s recognition of the need for what
he terms the “semantic bootstrapping hypothesis” (according to which
the child uses rough-and-ready semantic generalizations to map innate
syntactic structures onto natural language) is a clear recognition that the
postulation of innate syntactic structures cannot suffice to show how lan-
guage acquisition meets the Acquisition Constraint, and he attempts to
develop an alternative (Pinker 1987). The motivations that lead to his
alternative clearly indicate that advocates of innate syntactic structures
still have to answer general questions like these: How are innate syntactic
principles to be applied to the natural language that the child encounters?
How is the transition made from innate (tacit) knowledge of general syn-
tactic principles to the ability to manipulate and combine words that have
both syntactic and semantic properties? It should be clear that an answer
24 Chapter 1
to these questions will involve an answer to the question of how it is

possible for the child to learn to manipulate a pronoun with the semantic
properties of the first-person pronoun, and at this point the problems of
capacity circularity become salient once again.
1.6 The Paradox of Self-Consciousness
A paradox is an unacceptable conclusion or set of conclusions reachable

by apparently valid argument from apparently true premises. The clearest
way to appreciate a paradox is to list the incompatible propositions that
form the premises and conclusions. In the case of the paradox of self-
consciousness, they are as follows:
1. The only way to analyze what is distinctive about self-consciousness
is by analyzing the capacity to think ‘I’-thoughts.
2. The only way to analyze the capacity to think a particular range of
thoughts is by analyzing the capacity for the canonical linguistic expres-
sion of those thoughts (the Thought-Language Principle).
3. ‘I’-thoughts are canonically expressed by means of the first-person
pronoun.
4. Mastery of the first-person pronoun requires the capacity to think
‘I’-thoughts.
5. A noncircular account of self-consciousness is possible.
6. Mastery of the semantics of the first-person pronoun meets the Acqui-
sition Constraint (in the paradigmatic way defined earlier).
What I have argued in this chapter is that propositions (1) through (6)
cannot be maintained together. More specifically, neither proposition (5)
nor proposition (6) can be maintained in conjunction with propositions
(1) through (4).
Clearly, if the arguments I have offered are valid, then one or more
of these propositions must be abandoned, and possible solutions to the
paradox of self-consciousness can be classified according to which prop-
osition(s) they reject in order to restore consistency. The problem, of
course, is that each proposition is both intuitively and philosophically
appealing, which is why the term ‘paradox’ seems appropriate. Proposi-
tion (1) seems to be entailed by the thesis that self-ascribable psychologi-
cal and physical predicates have a constant sense across first- and third-
person uses. There are powerful epistemological reasons for not wanting
to reject that thesis. Proposition (2) has been described as, and widely
accepted to be, a fundamental principle of analytical philosophy. Proposi-
tions (3) and (4) are indisputable. Both propositions (5) and (6) are highly
desirable. If (5) is not true, then it follows that the capacity to think
‘I’-thoughts is unanalyzable, which is a highly undesirable result, while
the alternative to (6) is to deny that linguistic mastery of the first-person
pronoun is psychologically real in anything like the way we understand it.
Nonetheless, these six propositions cannot all be true. In the first two
sections of chapter 2, I identify the proposition that I think needs to be
rejected if the paradox is to be solved. That proposition is proposition
(2), the Thought-Language Principle. Of course, any solution to the para-
dox will be ad hoc, and hence unacceptable, unless the rejection of the
proposition in question can be independently motivated. In section 2.3 of
chapter 2 and in chapters 3 and 4, I outline in more detail what needs to
be shown if the Thought-Language Principle is to be plausibly rejected in
a way that will solve the paradox of self-consciousness. The remainder of
the book is devoted to the twofold task of, first, motivating the rejection
of the Thought-Language Principle and, second, showing how the para-
dox of self-consciousness can be solved once the Thought-Language Prin-
ciple is abandoned.
2
The Form of a Solution
One possible reaction to the paradox of self-consciousness as outlined in

the previous chapter is that it arises only because unrealistic and ulti-
mately unwarranted requirements are being placed on what are to count
as genuinely self-referring first-person thoughts. Support for such an ob-
jection will be found in those theories that attempt to explain first-person
thoughts in a way that does not presuppose any form of internal represen-
tation of the self or indeed any form of self-knowledge. The paradox of
self-consciousness arises because mastery of the semantics of the first-
person pronoun is available only to creatures capable of thinking first-
person thoughts whose contents involve reflexive self-reference and thus
seem to presuppose mastery of the first-person pronoun. If, though, it can
be established that the capacity to think genuinely first-person thoughts
does not depend on any linguistic and conceptual abilities, then arguably
the problems of circularity will no longer have purchase.
There is an account of self-reference and genuinely first-person thought
that can be read in a way that poses just such a direct challenge to the
account of self-reference underpinning the paradox of self-consciousness.
This is the functionalist account. On the functionalist view, reflexive
self-reference is a completely unmysterious phenomenon susceptible to a
functional analysis. Reflexive self-reference is not dependent upon any
antecedent conceptual or linguistic skills. Nonetheless, the functional ac-
count of reflexive self-reference is deemed to be sufficiently rich to provide
the foundation for an account of the semantics of the first-person pro-
noun. If this is right, then the circularity at the heart of the paradox of
self-consciousness can be avoided.
28 Chapter 2
In the first section of this chapter I examine the functionalist account

of self-reference as a possible strategy. I will argue that, although it is not
ultimately successful, attention to the functionalist account reveals the
correct approach for solving the paradox of self-consciousness. The sec-
ond section shows how a successful response to the paradox of self-
consciousness must reject what I call the classical view of content. In the
final section I set the agenda for the rest of the book.
2.1 The Functionalist Account of Self-Reference
The circularity problems at the root of the paradox of self-consciousness

arise because mastery of the semantics of the first-person pronoun re-
quires the capacity to think first-person thoughts whose natural expres-
sion is by means of the first-person pronoun. It seems clear that the circle
will be broken if there are forms of first-person thought that are more
primitive than those hitherto discussed, in the sense that they do not re-
quire linguistic mastery of the first-person pronoun. What creates the
problem of capacity circularity is the thought that we need to appeal to
first-person contents in explaining mastery of the first-person pronoun,
combined with the thought that any creature capable of entertaining first-
person contents will have mastered the first-person pronoun. So if we
want to retain the thought that mastery of the first-person pronoun can
only be explained in terms of first-person contents, capacity circularity
can only be avoided if there are first-person contents that do not presup-
pose mastery of the first-person pronoun.
On the other hand, however, it seems to follow from everything I said
earlier about ‘I’-thoughts that self-conscious thought in the absence of
linguistic mastery of the first-person pronoun is a contradiction in terms.
First-person thoughts have first-person contents, where first-person con-
tents can only be specified in terms of either the first-person pronoun or
the indirect reflexive pronoun. So how could such thoughts be entertained
by a thinker incapable of reflexive self-reference? How can a thinker who
is not capable of reflexively referring to himself as English speakers do
with the first-person pronoun be plausibly ascribed thoughts with first-
person contents? The thought that, despite all this, there are indeed
first-person contents that do not presuppose mastery of the first-person
The Form of a Solution 29
pronoun is at the core of the functionalist theory of self-reference and

first-person belief, which I will examine in this section.
The best developed functionalist theory of self-reference has been pro-
vided by Hugh Mellor (1988–1989). The basic phenomenon he is inter-
ested in explaining is what it is for a creature to have what he terms a
subjective belief, that is to say, a belief whose content is naturally expressed
by a sentence in the first-person singular and the present tense. The expla-
nation of subjective beliefs that he offers makes such beliefs independent
of both linguistic abilities and conscious beliefs. From this basic account
he constructs an account of conscious subjective beliefs and then of the
reference of the first-person pronoun ‘I’. These putatively more sophisti-
cated cognitive states are causally derivable from basic subjective beliefs.
Mellor starts from the functionalist premise that beliefs are causal func-
tions from desires to actions. It is, of course, the emphasis on causal links
between belief and action that make it plausible to think that belief might
be independent of language and conscious belief, since “agency entails
neither linguistic ability nor conscious belief” (Mellor 1988–1989, 21).
The idea that beliefs are causal functions from desires to actions can be
deployed to explain the content of a given belief via the equation of truth
conditions and utility conditions, where utility conditions are those in
which the actions caused by the conjunction of that belief with a single
desire result in the satisfaction of that desire. We can see how this works
by considering Mellor’s own example (1988–1989, 24–25). Consider a
creature x who is hungry and has a desire for food at time t. That creature
has a token belief b( p) that conjoins with its desire for food to cause it to
eat what is in front of it at that time. The utility condition of that belief
is that there be food in front of x at that time. Moreover, for b( p) to cause
x to eat what is in front of it at t, b( p) must be a belief that x has at t.
For Mellor, therefore, the utility/truth conditions of b( p) is that whatever
creature has this belief faces food when it is actually facing food. And a
belief with this content is, of course, the subjective belief whose natural
linguistic expression would be ‘I am facing food now’. On the other hand,
however, a belief that would naturally be expressed with these words can
be ascribed to a nonlinguistic creature, because what makes it the belief
that it is depends not on whether it can be linguistically expressed but on
how it affects behavior.
30 Chapter 2
What secures self-reference in belief b( p) is the contiguity of cause and

effect. The essence of a subjective belief is the effects that it has on actions.
It can only have those effects conjointly with a desire or set of desires,
and the relevant sort of conjunction is possible only if it is the same agent
at the same time who has the desire and the belief:
For in order to believe p, I need only be disposed (inter alia) to eat what I face if
I feel hungry: a disposition which causal contiguity ensures that only my simulta-
neous hunger can provoke, and only into making me eat, and only then. That’s
what makes my belief refer to me and to when I have it. And that’s why I need
have no idea who I am or what the time is, no concept of the self or of the present,
no implicit or explicit grasp of any ‘sense’ of ‘I’ or ‘now’, to fix the referents of
my subjective beliefs: causal contiguity fixes them for me. (Mellor 1988–1989,
24–25)
Causal contiguity, according to Mellor, explains why no internal repre-
sentation of the self is required, even at what other philosophers have
called the subpersonal level. Mellor believes that reference to distal ob-
jects can take place only when an internal state serves as a causal surro-
gate for the distal object, and hence as an internal representation of that
object. No such causal surrogate, and hence no such internal representa-
tion, is required in the case of subjective beliefs. The relevant causal com-
ponents of subjective beliefs are the believer and the time.
The necessary contiguity of cause and effect is also the key to the func-
tionalist account of self-reference in conscious subjective belief. Mellor
adopts a relational theory of consciousness, equating conscious beliefs
with second-order beliefs to the effect that one is having a particular first-
order subjective belief. It is, on Mellor’s view, simply a fact about our
cognitive constitution that these second-order beliefs are reliably, though
of course fallibly, generated so that we tend to believe that we believe
things that we do in fact believe. Certain first-order beliefs cause the rele-
vant second-order beliefs. These second-order beliefs are subjective be-
liefs, whether or not the relevant first-order beliefs are subjective, because
they are beliefs with the content that one is oneself having a particular
belief at a particular time. And they refer indirectly to the referents of the
relevant first-order beliefs. When those first-order beliefs are subjective
beliefs, then the first-order reference is to the person having them, and by
causal contiguity, so is the reference of the second-order beliefs that they
cause in that person. Neither the first-order belief nor the second-order
belief that it causes requires any internal representation of the self.
The last link in the chain is extending this account to the linguistic
capacity to master the first-person pronoun. Here, according to Mellor,
the key is the linguistic habit, which need not, of course, be conscious,
that my desire to express a belief linguistically should cause me to use the
first-person pronoun ‘I’ when the belief I want to express is first-personal.
What governs the choice of ‘I’ is not any beliefs about myself but simply
the belief that ‘I’ is the right word to express the belief I want to express.
And what makes this belief true is that everyone else shares my habit of
using ‘I’ to refer to themselves when they want to express a first-person
belief. Mellor concludes, “A knowingly shared habit is therefore all it
takes for me to use ‘I’ and ‘now’ successfully at any time t. I still don’t
need a concept of the self or of the present; nor need I believe that my
‘now’ refers to a present now, or to t, or that my ‘I’ refers to my I, or to
Hugh Mellor” (Mellor 1988–1989, 29).
What are the implications of the functionalist account of self-reference
for the paradox of self-consciousness? The paradox of self-reference
arises because of a postulated two-way interdependence between the ca-
pacity to think thoughts with the first-person contents characteristic of
self-consciousness and the capacity to understand and use the first-person
pronoun. The first dependence claim is that the first-person contents char-
acteristic of self-consciousness can be understood only in terms of their
canonical linguistic expression with the first-person pronoun and hence
that they presuppose mastery of the first-person pronoun. The second
dependence claim is that mastery of the first-person pronoun presupposes
the first-person contents characteristic of self-consciousness. The first
dependence relation opens up the possibility of what I termed the defla-
tionary theory of self-consciousness, while the second dependence rela-
tion threatens to make any such theory circular. The functionalist theory
of self-reference takes issue with the first dependence claim. It argues that
there are first-person contents (the so-called subjective beliefs) that do
not presuppose mastery of the first-person pronoun. These first-person
contents can be put to work to explain what it is to master the semantics
of the first-person pronoun. Such mastery requires only the presence of
certain subjective beliefs, some of which are conscious, a shared linguistic
habit of using the first-person pronoun in particular situations, and a be-
lief that this linguistic habit is indeed shared.
32 Chapter 2
If we adopt the functionalist theory, the paradox of self-consciousness

seems to disappear. We can accept, as the deflationary theory does, that
there are forms of first-person content (those implicated in the capacity
for moral self-evaluation, for example, or the capacity to entertain auto-
biographical memories) that presuppose mastery of the first-person pro-
noun and are to be elucidated through explaining what such mastery
consists in. When it comes to explaining what mastery of the first-person
pronoun consists in, however, the functionalist theory can avoid both ca-
pacity and explanatory circularity by appealing to subjective beliefs that
do not presuppose those higher forms of first-person content. The real
question, therefore, is whether the functionalist theory is correct and ade-
quate. This question has two sides. The first is whether the functionalist
theory provides everything that a theory of self-reference needs to pro-
vide. Second, though, we also need to ask whether the functionalist the-
ory is based on a satisfactory and plausible account of belief and, more
particularly, of what it is for a given belief to have a given content.
I begin with the first side of the question: can the functionalist theory of
self-reference accommodate the crucial distinctions that any satisfactory
theory of self-consciousness has to accommodate? One such distinction
is the distinction between what I termed genuine first-person contents
(namely, those which can only be specified directly by means of the first-
person pronoun ‘I’ or indirectly by means of the indirect reflexive pro-
noun ‘he*’) and those contents that are self-referring but are not genu-
inely first-personal. A belief content is self-referring but not genuinely
first-personal when it refers to the believer though the believer is not
aware that it does. Can the functionalist theory capture this distinction?
I introduced the distinction through the following example:
(1) J. L. B. believes that he* will be the next person to receive a
parking ticket in central Cambridge.
(2) J. L. B. believes that he will be the next person to receive a parking
ticket in central Cambridge.
To appreciate the difference between (1) and (2), consider the anomalous
situation in which J. L. B. believes that J. L. B. will be the next person to
receive a parking ticket in central Cambridge but does not realize that
he is J. L. B. This situation can be properly described with (2) but not
with (1).
How might the functionalist theory of self-reference deal with this? The
difference between (1) and (2) in the actual contents of the beliefs (when
directly specified in oratio recta) is this:
(C.1) I will be the next person to receive a parking ticket in central
Cambridge.
(C.2) J. L. B. will be the next person to receive a parking ticket in
central Cambridge.
This difference certainly seems to be one that the functionalist theory can
accommodate. According to Mellor, (C.1) is a semantic function s(i, t, x)
from whoever, x, produces a token s(i, t) to an impersonal truth condi-
tion: x will be the first person after t to receive a parking ticket in central
Cambridge. In contrast, (C.2) is a constant function s(jlb, t) to the effect
that J. L. B. will be the first person after t to receive a parking ticket in
central Cambridge. These are two distinguishable contents with clearly
distinguishable functional roles. As this makes clear, then, the functional-
ist theory of self-reference can make perfectly good sense of the thought
that (C.1) and (C.2) are different contents, and that is how it undertakes
to accommodate the distinction between (1) and (2) and, more generally,
between genuine first-person contents and contents that are self-referring
but not genuinely first-personal. According to Mellor, when J. L. B. be-
lieves that he* will be the next person to receive a parking ticket in central
Cambridge, what he believes is that whoever is thinking that thought will
be the next person to receive a parking ticket in central Cambridge. And
this is clearly a different thought from the thought that J. L. B. will be
that person, even though he is J. L. B.
The capacity to mark this sort of distinction is clearly an important
criterion of adequacy for theories of self-reference, and the fact that the
functionalist theory can do so without apparently falling afoul of the par-
adox of self-consciousness is important presumptive evidence in its favor.
But it cannot be conclusive until we have looked more closely at the bases
on which the relevant belief contents are ascribed, because the functional-
ist theory needs to provide not just an explanation of the distinction in
the abstract but also a guarantee that the different belief contents will be
ascribed in the relevant contexts. This brings us, however, to the general
issue of whether the functionalist account is based on a satisfactory ac-
count of belief ascription, and it is here that problems arise.
34 Chapter 2
Let me start by mentioning one possible line of objection to the general

concept of belief underlying Mellor’s position. For many philosophers, it
is axiomatic that beliefs are propositional attitudes to contents composed
of concepts. To such a philosopher, the functionalist account of self-
reference is in danger of committing the (putative) solecism of attributing
belief contents where it does not seem appropriate to attribute concepts.
Although in the following chapters I will defend a version of this theory,
I do not want to press it against Mellor at this point.1 One reason for
setting this line of objection aside is that it might justifiably be thought to
beg the question against Mellor. If it is indeed the case that Mellor’s ac-
count of self-reference is embedded within a theory of belief that sepa-
rates belief ascription from concept ascription, then in the absence of any
independent reasons for rejecting the functionalist account of belief, this
might well be taken by an unbiased observer as evidence against the view
that beliefs are constitutively composed of concepts. A second reason for
not pursuing this line of thought is that it is far from clear that Mellor’s
account cannot be supplemented by a theory of what would warrant at-
tributing to a creature the concepts invoked in the relevant content. There
is no reason in principle why the functionalist account of belief cannot
be matched with a functionalist account of concepts so as to allow it
to conform to the postulated requirement that beliefs be composed of
concepts. What this brings home, of course, is that this is not a line of
argument that can be pressed without a clearer understanding of what
concepts are (as well as an argument supporting the presupposed depen-
dence of beliefs on concept possession).
Let us go back to Mellor’s original example of the hungry creature fac-
ing food. The creature has a token belief that conjoins with its desire for
food to cause it to eat what is in front of it. Mellor’s analysis of truth
conditions in terms of utility conditions leads him to identify that token
belief with the subjective belief whose linguistic expression would be ‘I
am facing food now’. This belief clearly has what I have been terming a
genuine first-person content. Now, as Mellor stresses, there is a very close
connection between beliefs with genuine first-person contents and dis-
positions to actions, and this connection is one way of marking the dis-
tinction between the two classes of self-referring beliefs I have been
discussing. A genuine first-person content, when conjoined with the ap-
propriate desire(s) in the appropriate circumstances and provided there

are no countervailing factors, will lead to action. In contrast to this, belief
contents that are self-referring but not genuinely first-personal will only
issue in action when conjoined with beliefs that are genuinely first-
personal—as when my attack of amnesia lifts and it suddenly dawns
upon me that I am that J. L. B. who will shortly receive a parking ticket,
and I start running back to my car.2
It is beyond dispute, then, that all beliefs with genuinely first-person
contents have direct and immediate implications for action. To the extent
that Mellor’s theory is based on this idea, it cannot be faulted. The prob-
lem with the theory, however, is that it seems to be based on the unargued-
for assumption that something like the converse of this principle also
holds and is all that we need to guide us in the ascription of beliefs. Let
me explain. Mellor’s position seems to rest on two fundamental thoughts.
The first of these is the thought that all cases of action to be explained
in terms of belief-desire psychology have to be explained through the at-
tribution of first-person beliefs. This is what underwrites his treatment of
the example. The utility conditions, and hence truth conditions, of what-
ever belief motivates the creature facing food are, according to Mellor,
that ‘there is food in front of me’ or ‘I am facing food’. Let me call this
the first-person-motivation thesis. It will be noticed, of course, that my
statement of the first-person-motivation thesis contains an important
qualification: it is claimed to hold only for those actions for which expla-
nation in terms of beliefs and desires is required. It turns out, though, that
(at least for Mellor) there is little need for such a qualification, because he
also seems to subscribe to a thesis that I shall label the universality-of-
belief thesis, according to which all apparently purposeful behavior is to
be explained in terms of the conjunction of beliefs and desires. My reason
for holding that Mellor subscribes to the universality-of-belief thesis is
his insistence that the functionalist account of self-reference explains how
nonlinguistic animals can have first-person beliefs. His assumption seems
to be that (to stick with the example he uses) whenever a hungry animal
eats the food in front of it, it will have been caused to do so by the con-
junction of its hunger with a first-person belief of the type he discusses.
The problem with Mellor’s theory is that neither the first-person-
motivation thesis nor the universality-of-belief thesis are true. Let me
36 Chapter 2
consider them in reverse order. The universality-of-belief thesis that every

instance of apparently purposeful behavior should be explained in terms
of beliefs and desires is at odds with a very basic principle at the heart
of the psychological study of animal behavior, namely, the principle that
intentional explanations in terms of beliefs and desires are to be deployed
in explaining animal behavior only when simpler and more parsimonious
explanations are demonstrably inadequate. Apparently purposive actions
that can be explained without recourse to beliefs and desires include those
explicable through classical or Pavlovian conditioning, for example. Here
the classically conditioned associative links between stimulus and re-
sponse are what explains the fact that a creature behaves in certain ways
in certain situations. There are also various types of explanation appeal-
ing to innate, hard-wired determinants of behavior, such as the concept
of innate releasing mechanisms (preprogrammed responses to events with
a common structure). Explanation in terms of instrumental (operant)
conditioning is more complex. Instrumental conditioning is the name
given to processes of learning that make the receipt of a reward condi-
tional upon an animal’s performance of the behaviour being learned, and
some psychologists have argued that this type of learning is possible in
certain circumstances only if the animal in question has beliefs about the
contingent relation between response and reward (Russell 1980 and Dick-
inson 1988). But even if instrumental conditioning is set to one side, a
range of psychological explanations are available that are genuine alter-
natives to intentional explanations in terms of belief-desire psychology. A
methodological assumption almost universal in psychology and animal
ethology is that there is a place for intentional explanation only when
those other modes of explanation are demonstrably inapplicable.
It is, I think, a definite shortcoming in Mellor’s theory as it stands that
it is too prodigal with the ascription of beliefs. Nonetheless, it might be
thought that some modification of the theory could accommodate this
point—most obviously, specifying the circumstances in which belief-
desire explanations are to be deployed. But matters are not as simple as
this. On the functionalist theory, beliefs are causal functions from desires
to actions. This creates a problem, however, because all of the different
modes of psychological explanation mentioned also appeal to states that
fulfill a similar causal function from desire to action. Of course, it is open
to a defender of the functionalist approach to say that it is only beliefs,
and not, for example, innate releasing mechanisms, that interact causally
with desires in a way that generates actions. But this sort of response is
of limited effectiveness unless some sort of principled reason is given for
distinguishing between a state of hunger and a desire for food. It is no use,
of course, simply to describe desires as functions from belief to actions.
The moral to draw from this, I think, is that the functionalist theory
of belief needs to be expanded so as to explain how and why beliefs are
different from innate releasing mechanisms, conditioned responses, and
other such states featured in nonintentional psychological explanation
(henceforth, ‘nonintentional psychological states’). The theory as stated
does not have the resources to deal with this problem, because all these
states, including beliefs, can plausibly be described as causal functions
from desires to actions. Of course, to say that the functionalist theory
of belief needs to be expanded is not to say that it needs to be expanded
along nonfunctionalist lines. Nothing that has been said rules out the
possibility that a correct and adequate account of what distinguishes be-
liefs from nonintentional psychological states can be given purely in terms
of their respective functional roles.3 The point is just that until we have
such an account, it will be hard properly to evaluate the functionalist
theory of self-reference. The core of the functionalist theory of self-
reference is the thought that agents can have subjective beliefs that do not
involve any internal representation of the self, linguistic or nonlinguis-
tic. It is in virtue of this that the functionalist theory claims to be able to
dissolve the paradox of self-consciousness. The problem that has
emerged, however, is that it remains unclear whether those putative sub-
jective beliefs really are beliefs.
Let us assume for the moment that this problem has been solved, and
that belief states can be adequately distinguished from nonintentional
psychological states. The functionalist theory is still not out of the woods.
There remains its reliance on what I termed the first-person-motivation
thesis, according to which all cases of action to be explained in terms of
belief-desire psychology have to be explained through the attribution of
first-person beliefs. The first-person-motivation thesis is clearly at work
in Mellor’s assumption that the utility conditions, and hence truth condi-
tions, of the belief that causes the hungry creature facing food to eat what
is in front of him are first-personal—thus determining the content of
the belief to be ‘There is food in front of me’ or ‘I am facing food’. The
38 Chapter 2
problem, however, is that it is not clear that this is warranted. There are
a range of other possible contents, none of them first-personal, that such
a belief could have. One possibility would be the demonstrative content
‘That’s food’. Another would be the indexical content ‘There’s food here’.
Either of these would satisfy the requirements posed by Mellor’s equation
of truth conditions with utility conditions. Each of them would explain
why the animal would eat what is in front of it. Nonetheless, on the face
of it these are significantly different thoughts, only one of which is a genu-
ine first-person thought.
Although there is a sense in which these three beliefs are all functionally
equivalent, because they all serve a similar causal function from desires
to action, the distinction between them is not trivial. Only one of them,
for example, can plausibly be taken as expressive of any type of self-
consciousness. Of course, putting the point like this may well seem like
begging the question to a defender of the functionalist theory of self-
reference. And such a defender might press the question of whether there
really is a genuine difference between these three possible thoughts. After
all, it is arguable that no one could assent to one of them without simulta-
neously assenting to the others. It would seem to be only in very contrived
situations that someone might assert ‘I am facing food now’ while deny-
ing ‘There is food here now’ or ‘That’s food’.
But there is one very fundamental difference that the functionalist
theory seems unable to capture. To bring it out, we need to consider the
general distinction between structured and unstructured thought. A struc-
tured thought is one composed of separable and recombinable elements.
What makes a given thought a structured thought is that its distinct parts
can be separated out and put to work in other thoughts.4 So in attributing
to a creature a structured thought, one is ipso facto attributing to it the
distinct cognitive abilities associated with the elements composing that
thought. An unstructured thought, on the other hand, is not composed
of such distinguishable components. It is more like a primitive response
to the presence of a given feature, perhaps best understood on the model
of an exclamation. Now, the content of the belief that Mellor’s functional-
ist theory demands that we ascribe to an animal facing food is ‘I am fac-
ing food now’ or ‘There is food in front of me now’. These are, it seems
clear, structured thoughts. So too, for that matter, is the indexical thought
‘There is food here now’. The crucial point, however, is that the causal
function from desires to actions, which, according to Mellor, is all that
a subjective belief is, would be equally well served by the unstructured
thought ‘Food!’
This inability to mark the distinction between a structured and an un-
structured thought is a significant problem for the functionalist account
of self-reference, because the subjective beliefs that the theory is trying to
capture are, of course, structured beliefs and yet the theory is not suffi-
ciently fine-grained to pick out only structured beliefs. Again, however, I
do not put this forward as a principled argument against the possibility
of providing a satisfactory functional account of self-reference. As before,
the point is simply that the functional account under consideration will
need to be supplemented, and it remains perfectly possible that the exten-
sion will be along functional lines.
For present purposes, however, the conclusion must be that the func-
tionalist theory of self-reference is not of immediate use in dissolving the
paradox of self-consciousness. Although it is true that the capacity circu-
larity at the heart of the paradox of self-consciousness does not arise on
the functionalist theory, the price paid for this is an inadequate theory of
self-reference. What it does do, however, is to point us very clearly in the
right direction, as will emerge in the next two sections.
2.2 Rejecting the Classical View of Content
One central idea that emerges from the functionalist account of self-
reference discussed in the previous section is that content-bearing states
can be independent of language mastery. Mellor emphasizes that his the-
ory is intended to explain how nonlinguistic creatures can be in belieflike
content-bearing states, and even though he characterizes the content of a
belief in terms of the sentence that would be its natural linguistic expres-
sion, it is very clear that the creature to which the belief is ascribed need
not be capable of uttering that sentence or indeed any sentence at all.
In this respect the functionalist theory is very much at odds with what
might be described as the conventional understanding of the nature of
content and how it should be ascribed. This conventional understanding
is clearly expressed in the Thought-Language Principle, which, it will be
40 Chapter 2
remembered, was one of the inconsistent set of propositions that together

constitute the paradox of self-consciousness:
The Thought-Language Principle The only way to analyze the capacity
to think a particular range of thoughts is by analyzing the capacity for
the canonical linguistic expression of those thoughts.
The functionalist theory of self-consciousness entails the rejection of the
Thought-Language Principle for the following reason. The functionalist
account tackles the problem of first-person thoughts not by considering
their canonical linguistic expression but instead by considering their func-
tional role, particularly their role in the explanation of action. On the
functionalist view, the best way to understand first-person thoughts is
through understanding first-person beliefs, and first-person beliefs are
themselves best understood as a particular sort of function from desires
to action. An understanding of the canonical linguistic expression of first-
person beliefs comes after, not before, an understanding of those beliefs
themselves.
It seems to me that, although the functionalist account of first-person
thought, as it stands, is not adequate to solve the paradox of self-
consciousness, it provides a clear indication of both a general strategy
and a specific strategy that any adequate solution to the paradox will have
to adopt. The general strategy is a rejection of the Thought-Language
Principle. The specific strategy concerns what is to replace the Thought-
Language Principle, namely, an analysis of first-person thoughts based on
the role of first-person thoughts in the explanation of behavior. I will
develop both of these ideas in the remainder of this chapter. First, though,
we need to take a step back to consider further some of the views about
the relation between the content of thought and the content of language
that underlie the Thought-Language Principle.
There is an obvious but central question that arises in considering the
relation between the content of thought and the content of language,
namely, whether there can be thought without language as theories like
the functionalist theory suggest. The conception of thought and language
that underlies the Thought-Language Principle is clearly opposed to the
proposal that there might be thought without language, but it is impor-
tant to realize that neither the principle as outlined nor the considerations
adverted to by Dummett directly yield the conclusion that there cannot

be thought in the absence of language. According to the principle, the
capacity for thinking particular thoughts can only be analyzed through
the capacity for the linguistic expression of those thoughts. On the face
of it, however, this does not yield the claim that the capacity for thinking
particular thoughts cannot exist without the capacity for their linguistic
expression. And let me quote again the passage from Dummett cited in
the previous chapter:
Thoughts differ from all else that is said to be among the contents of the mind in
being wholly communicable: it is of the essence of thought that I can convey to
you the very thought that I have, as opposed to being able to tell you merely
something about what my thought is like. It is of the essence of thought not
merely to be communicable, but to be communicable, without residue, by means
of language. In order to understand thought, it is necessary, therefore, to under-
stand the means by which thought is expressed. (Dummett 1978, 442)
This point about thoughts being wholly communicable does not entail
that thoughts must always be communicated, which would be an absurd
conclusion. Nor does it appear to entail that there must always be a possi-
bility of communicating thoughts in any sense in which this would be
incompatible with the ascription of thoughts to a nonlinguistic creature.
There is, after all, a prima facie distinction between thoughts being
wholly communicable and it being actually possible to communicate any
given thought. But without that conclusion there seems no way of getting
from a thesis about the necessary communicability of thought to a thesis
about the impossibility of thought without language.
So at least one vital question about the relation between thought and
language is left undetermined by the Thought-Language Principle. The
missing link is provided, I think, by two fundamental principles at the
heart of what I shall term the classical view of content. The first is
the principle that the type of thoughts which can be ascribed to a creature
are directly circumscribed by the concepts that the creature possesses:
The Conceptual Requirement Principle The range of contents that one
may attribute to a creature is directly determined by the concepts that the
creature possesses.
The second principle is a requirement on concepts that a creature can
possess. It holds that the range of concepts that a creature can possess is
42 Chapter 2
a direct function of its degree of language mastery. There is, of course,

a spectrum of different ways of understanding the assumed connection
between mastery of concepts and mastery of a language. On the strongest
and simplest view of the matter, mastery of a concept is exhausted by the
capacity properly to manipulate certain terms of a language. An example
of a weaker view would be that certain linguistic abilities are essential
criteria of conceptual mastery. The following thesis does, however, seem
to be common to all the views on this spectrum:
The Priority Principle Conceptual abilities are constitutively linked with
linguistic abilities in such a way that conceptual abilities cannot be pos-
sessed by nonlinguistic creatures.
When the Priority Principle is conjoined with the Conceptual Require-
ment Principle, it follows that nonlinguistic creatures cannot be in states
with content.
With all this in mind, let me return to the functionalist theory. Mellor’s
account is clearly opposed to the classical view of content encapsulated
in these two principles. Mellor’s contents do not involve concepts, and
they are available to nonlinguistic creatures. However, his theory suffers
from the lack of discriminative power noted in the preceding section.
There seems to be a range of different and incompatible contents that can
fulfil the causal roles with which he identifies beliefs. One response to this
would be to reject the functionalist approach and to return to the classical
view of content. I do not think that this would be the correct reaction,
however, because there are good reasons for thinking that the classi-
cal view of content is wrong. As even defenders of the classical view of
content frequently accept, there are difficult borderline cases involving
animals and young children that do seem to require the ascription of rep-
resentational states with content to creatures with no linguistic abilities
and hence, given the Priority Principle, with no conceptual abilities. One
standard strategy for dealing with these has been to retain the Priority
Principle while allowing that there are protoconcepts available to nonlin-
guistic creatures (Dummett 1993, chap. 12, is a good example). But this
is an obviously unstable position. If protothoughts really are representa-
tional, then it follows on the classical view that they cannot be available
to nonlinguistic creatures.
In the following two chapters I will explain why I think that the classi-
cal view of content is wrong. My objections to it, however, are quite com-
patible with retaining the Priority Principle, because I shall be arguing
principally against the first component of the classical view of content,
that is, the thesis that content is constrained by concept possession. There
are, it seems to me, very good reasons for thinking that there are ways of
representing the world that are nonconceptual, in the sense that they are
available to creatures who do not possess the concepts required to specify
how the world is being represented. This can be accepted (and the Con-
ceptual Requirement Principle rejected), however, without any need to
revise the idea that there is a constitutive connection between language
mastery and conceptual mastery. Quite the contrary. Hanging on to the
Priority Principle allows us to make a very clear distinction between con-
ceptual and nonconceptual modes of content-bearing representation, be-
cause the connection between language and concepts gives us a clear
criterion for identifying the presence of conceptual representation, as op-
posed to nonconceptual representation. On the other hand, though, one
consequence of holding onto the Priority Principle while rejecting the
Conceptual Requirement Principle would be that we have to recognize
the restricted application of the Thought-Language Principle. The meth-
odological principle that the only way in which to analyze particular
thoughts is through analyzing the canonical linguistic expression of those
thoughts does not seem plausible when applied to thoughts whose con-
tent is not connected in any way with linguistic abilities.
2.3 The Outline of a Solution
This distinction between conceptual and nonconceptual modes of

content-bearing representation is highly relevant to the paradox of self-
consciousness, for precisely the reasons that emerged during my discus-
sion of the functionalist theory. The circularity at the heart of the paradox
arises from the assumption that the capacity to think thoughts with the
first-person contents characteristic of self-consciousness is available only
to creatures who have mastered the semantics of the first-person pronoun.
This in turn rests on the thought that first-person contents are concep-
tual in the sense under discussion. That is, the restriction of first-person
44 Chapter 2
contents to creatures who have mastered the semantics of the first-person

pronoun is an obvious consequence of the thought that content ascription
is constrained by concept possession, combined with the Priority Prin-
ciple, affirming a constitutive connection between language mastery and
conceptual mastery. The desirability of retaining the Priority Principle has
already been indicated. Suppose, however, that the conceptual constraint
on content ascription is lifted by allowing the possibility that a creature
might be in representational states despite not possessing the concepts
required to specify the relevant content. This would open up the possibil-
ity that there might be nonconceptual first-person contents that would
not require mastery of the concept ‘I’ and hence, by the Priority Principle,
of the semantics of the first-person pronoun. These would be first-person
contents that do not involve linguistic mastery of the first-person pro-
noun. If such contents do exist and if they can be put to work to explain
mastery of the semantics of the first-person pronoun, then the paradox
of self-consciousness will be solved.
Let me be more precise about how this works. Recall the following six
incompatible propositions at the root of the paradox of self-conscious-
ness:
1. The key to analyzing self-consciousness is to analyze the capacity to
think ‘I’-thoughts.
pronoun.
4. Mastery of the first-person pronoun requires the capacity to think
‘I’-thoughts.
5. A noncircular account of self-consciousness is possible.
6. The capacity to think ‘I’-thoughts meets the Acquisition Constraint
(in the paradigmatic way defined earlier).
My proposal is that proposition (2) be rejected, in other words, that it be
denied that the only way to analyze the capacity to think a particular
range of thoughts is by analyzing the capacity for their canonical linguis-
tic expression. More precisely, I shall be arguing that there is an important
class of thoughts for which the Thought-Language Principle fails to hold
because the Conceptual Requirement Principle fails to hold.
If proposition (2) is rejected, then consistency will indeed be restored.

Clearly, however, to reject proposition (2), we have to find a satisfactory
way of analyzing first-person thoughts that does not proceed via their
canonical linguistic expression and yet that is compatible with the truth
of the other propositions, most particularly (5) and (6). In fact, it is hard
to see how a proposed solution to the paradox of self-consciousness could
establish its credentials without providing a clear indication both of how
a noncircular account of the capacity to think ‘I’-thoughts might be de-
rived and of how that capacity might meet the Acquisition Constraint.
Without these three things, we will in fact not know whether the Con-
ceptual Requirement Principle can be rejected in discussions of self-
consciousness.
It is in meeting these conditions that the theoretical notion of noncon-
ceptual content has a crucial role to play. I shall ultimately be arguing
that proposition (4), to the effect that mastery of the first-person pronoun
involves the thinking of thoughts with first-person contents, can be recon-
ciled both with the existence of a noncircular account of mastery of the
first-person pronoun and with the Acquisition Constraint if at least some
of the relevant thoughts can be taken to have nonconceptual contents in
the way discussed. The central idea will be that both in explaining what
mastery of the first-person pronoun actually is and in explaining how
such a capacity can be acquired in the normal course of human develop-
ment, we can appeal to nonconceptual first-person thoughts. These are
thoughts that, although first-personal in the sense that their content is to
be specified directly by means of the first-person pronoun and indirectly
by means of the indirect reflexive pronoun ‘he*’, can nonetheless be cor-
rectly ascribed to creatures who have not mastered the first-person con-
cept (as evinced in mastery of the first-person pronoun).
Of course, this blueprint for solving the paradox of self-consciousness
is highly dependent on two things. The first is the legitimacy of the general
notion of nonconceptual content, and hence of the falsity of that part of
the classical view of content captured by the Conceptual Requirement
Principle. Showing the legitimacy of the general notion of nonconceptual
content, however, is not sufficient on its own. What also needs to be
shown is that the general notion can be applied within the area we are
concerned with and that there are indeed nonconceptual first-person
thoughts.
46 Chapter 2
I undertake the first of these tasks in chapters 3 and 4. There I outline

in more detail the conception of nonconceptual content with which I wish
to work and begin the task of legitimating it by showing that we need
such a notion of nonconceptual content for a proper characterization of
the content of perceptual experience. This is only the first step, however.
Some philosophers who accept that the content of perceptual experience
has a nonconceptual component maintain this in conjunction with the
view that correct ascriptions of states with nonconceptual content can
only be made to creatures who possess at least a minimal conceptual rep-
ertoire. Against this I defend the Autonomy Principle, to the effect that
there are no dependence relations between concept possession and states
with nonconceptual content. Of course, any satisfactory theory of non-
conceptual content must be able to provide an account of the conditions
under which it is appropriate and warranted to ascribe states with non-
conceptual content, particularly since the Autonomy Principle deprives
us of the potential evidence that might otherwise have been provided by
language use. The account that I develop follows the functionalist theory
discussed earlier in stressing the importance of content-bearing states in
the explanation of behavior. Unlike the functionalist account, however, it
places constraints upon the types of behavior that require content-bearing
explanation, as well as upon the nature of those content-bearing states.
The account I offer is intended as a corrective to what I earlier termed the
universality-of-belief thesis.
In chapters 5 through 9, I move on to the second task of showing that
the notion of nonconceptual content can profitably be applied to first-
person thought. I argue that nonconceptual first-person contents are to
be found in somatic proprioception, in certain navigational abilities, and
in certain nonlinguistic social contexts. The foundations for this account
of first-person nonconceptual contents are laid in chapter 5, where I show
how a Gibsonian perspective on perception reveals the presence of non-
conceptual first-person contents in the very structure of perceptual ex-
perience. At various points in arguing for the existence of these different
types of nonconceptual first-person thought, I adopt a form of the prin-
ciple of inference to the best explanation. In such cases my modus ope-
randi is to identify certain types of behavior that, it seems to me, can best
be explained by assuming the causal efficacy of states with nonconceptual
first-person contents. I take it that this will lay to rest the worries about
indiscriminate attributions of first-person contents that arose in the con-
text of the functionalist theory of self-reference. In chapter 10, I draw the
various strands of the argument together and show how the nonconcep-
tual first-person contents identified can be put to work to defuse both the
explanatory circularity and the capacity circularity at the heart of the
paradox of self-consciousness.
Let me close this programmatic chapter with a methodological com-
ment. This book contains much closer attention to empirical work in vari-
ous areas of psychology than it is usual to find in works on the philosophy
of mind. This is in no way accidental. As has emerged in this chapter,
and as I continue to stress throughout the book, the ascription conditions
of content-bearing states, particularly first-person content-bearing states,
are closely bound up with the potential that such states have for the
explanation of different types of behavior. One central reason why philos-
ophers have generally been unprepared to countenance or even contem-
plate the existence of nonconceptual content-bearing states is that they
have had relatively little exposure to those forms of behavior that might
seem to demand explanation in terms of such states. Animal behavior and
the behavior of prelinguistic infants paradigmatically raise the problems
for which, so I believe, theoretical appeal to states with nonconceptual
content is the only solution (or at least the best so far available). But the
behavior of animals and prelinguistic infants needs to be approached ex-
perimentally if we are to have a clear sense of what the data are that need
to be explained. This is partly because it is only in certain experimental
contexts that particular problematic forms of behavior can be isolated
and identified, a good example being the infant-dishabituation experi-
ments to be discussed in chapter 3. But it is also partly because expe-
rimental work is required to establish the parameters of particular
explanations and to test their adequacy. Any explanation that one offers
of why an animal behaves thus and so in a particular situation has impli-
cations for how it might be expected to behave in certain other situations.
Experimentally producing those situations allows given explanations to
be confirmed or disconfirmed.
It should not be assumed, however, that finding a solution to the para-
dox of self-consciousness is a purely empirical matter. The empirical work
48 Chapter 2
that I adduce is more than purely illustrative, since many of the theoretical
points that it is brought in to illuminate could not be independently for-
mulated without it. Nonetheless, that empirical work cannot bring us any
closer to a solution to the paradox of self-consciousness without being
integrated into a theoretical account that is ultimately driven by concerns
that are more traditionally philosophical—concerns about the nature of
an adequate explanation, about the nature of representation, about the
structure of perception, and most important, about what it is to be self-
conscious.
3
Content, Concepts, and Language
What gives bite to the paradox of self-consciousness as developed in chap-

ter 1 is the assumption that it is incoherent to ascribe thoughts with first-
person content to an individual who is not capable of employing the
first-person pronoun to achieve reflexive self-reference. This yields the
paradoxical conclusion that any explanation of mastery of the semantics
of the first-person pronoun will presuppose mastery of the semantics of
the first-person pronoun. One of the important points that emerged from
the discussion of the functionalist theory of self-reference in the previous
chapter, however, was that this assumption can be challenged. This chap-
ter further questions the view of content on which this assumption rests,
what I termed the classical view of content. It argues that there are good
reasons for thinking that the classical view of content is mistaken in ways
that will, when properly understood, reveal how the paradox of self-
consciousness can be resolved.
I shall be outlining a theory of content on which there can exist states
with nonconceptual first-person contents. A nonconceptual content is
one that can be ascribed to a thinker even though that thinker does not
possess the concepts required to specify that content. Nonconceptual
first-person contents are those that fall into this category and that can
only be specified by means of the first-person or indirect reflexive pro-
nouns (as discussed in chapter 1). These nonconceptual first-person con-
tents offer a way of breaking both forms of circularity identified in
chapter 1. Developing and defending a theory of nonconceptual content
as a way of defusing the paradox of self-consciousness is a complicated
task, particularly in view of the power and popularity of the classical
theory of content. Any defence of nonconceptual content will, ipso facto,
50 Chapter 3
be an attack on the classical theory. In the remainder of this chapter I will

lay out some of the reasons why a notion of nonconceptual content seems
preferable to the classical theory of content. I begin in the next section
by explaining the importance of nonconceptual content in providing a
satisfactory account of the richness of perceptual experience. This will
provide an initial legitimation of the notion of nonconceptual content,
which will then be extended in the following section.
3.1 Conceptual and Nonconceptual Content: The Richness of

Perceptual Experience
A dominant view among philosophers is that the ways in which a creature

can represent the world are determined by its conceptual capacities. This
is associated with a widespread rejection of the traditional empiricist dis-
tinction between sensation and belief. Recently, though, the existence of
mental states with nonconceptual representational content has been dis-
cussed among philosophers of mind (Evans 1982, Peacocke 1992).1 The
general thought is that it is theoretically legitimate to refer to mental
states that represent the world but do not require the bearer of those men-
tal states to possess the concepts required to specify how they present the
world as being. These are states with nonconceptual content. A noncon-
ceptual content can be attributed to a creature without thereby attribut-
ing to that creature mastery of the concepts required to specify that
content.
The central impetus for legitimating a notion of nonconceptual content
has come from the study of perceptual experience. Like traditional theo-
rists who see sensations as ineliminable components of perceptual expe-
rience, theorists have been attracted to nonconceptual content by the
thought that the richness and grain of perceptual experience is not con-
strained by the concepts that a perceiver might or might not possess.
What distinguishes the theory of nonconceptual content from theories of
the sensational component of perceptual experience is that nonconcep-
tual content is representational. If a given perceptual experience has a
nonconceptual content that f, this means that the experience represents
f as holding in the world. Conceptual and nonconceptual contents are
distinguished not by whether they are representational but according to
Content, Concepts, and Language 51
how they represent. They are distinguished according to whether, in speci-

fying how they represent the environment as being, we need to restrict
ourselves to concepts possessed by the perceiver. This is in stark contrast
to the traditional distinction between sensation and belief, according to
which the sensational component has no role to play in explaining how
the experience represents the world as being.
The first stage in imposing a principled distinction between conceptual
and nonconceptual content is that perceptual experiences can represent
the world in ways fundamentally different from those in which perceptual
beliefs can represent the world. Although perceptual experiences are rep-
resentational, rather than sensational, they cannot always be assimilated
to perceptual beliefs. Perceptual experience can have a richness, texture,
and fineness of grain that beliefs do not and cannot have. A useful illustra-
tion of this has been provided by Christopher Peacocke (1986, 1992),
who stresses that perceptual experiences have contents that are analogue
in character: “To say that the type of content in question has an analogue
character is to make the following point. There are many dimensions—
hue, shape, size, direction—such that any value on that dimension may
enter the fine-grained content of an experience. In particular, an experi-
ence is not restricted in its range of possible contents to those points or
ranges picked out by concepts—red, square, straight ahead—possessed
by the perceiver” (1992, 68). As Peacocke notes, it is prima facie plausible
to suppose that the analogue nature of perceptual experience can only be
explained by assuming that the content of perceptual experience is not
constrained by the concepts that the perceiver possesses. Observational
concepts pick out salient features of experience; there are many other
discriminably different ways that experience could have represented the
world, even though those ways can only be specified through those very
same observational concepts. This fact about perceptual experience can-
not be accommodated if the content of perceptual experience is deemed
to be conceptual.
Suppose we combine the two ideas, first, that perceptual experience
has a nondoxastic component that is nonetheless representational, and
second, that the nondoxastic component of perception is analogue and
finer-grained in a way that the doxastic component is often unable to
capture. A natural next step would be to map the distinction between
52 Chapter 3
doxastic and nondoxastic onto the distinction between conceptual and

nonconceptual. On such a view, the contents of perceptual belief are nec-
essarily limited by the conceptual resources of the perceiver, while this
restriction does not apply to the content of perceptual experience.2 This
proposal seems to me to be the right way to proceed. Before it can be
defended, however, it needs to be sharpened up.
Peacocke has proposed defining a conceptual content simply as a con-
tent composed of concepts. A nonconceptual content, therefore, is a con-
tent not composed of concepts. One source of worry with this proposal
is that it unnecessarily forecloses on the possibility of contents being con-
ceptual without being composed of concepts. Contents understood in
Russellian terms are not composed of concepts, for example, but this does
not automatically make them nonconceptual. It could still be the case
that a Russellian content cannot be ascribed to an individual unless that
individual possesses the concepts associated with the individuals and
properties that make up that content. Nonetheless, Peacocke’s suggestion
that conceptual and nonconceptual contents be distinguished in terms of
their constituents is very helpful. I take it that the constituents of a given
content are identifiable in terms of the concepts used to specify that con-
tent. Those constituents do not have to be concepts themselves, but they
can be (on some theories). On this construal, then, a content is noncon-
ceptual if and only if none of its constituents are conceptual, where a
conceptual constituent is one that can feature in a content only if the
thinker possesses the concept required to specify that constituent. A con-
tent is conceptual, in contrast, if and only if any of its constituents are
conceptual.
The key question, therefore, is whether there are any good reasons to
hold that the contents of perceptual beliefs, and indeed of propositional
attitudes in general, are necessarily conceptual, that is, whether such con-
tents can properly be attributed to a thinker only if that thinker possesses
the concepts required to specify them. It seems to me that there is a per-
suasive argument here trading on certain crucial parallels between the
normative dimensions of believing a particular content and the normative
dimensions of possessing a particular concept.3 Belief contents are indi-
viduated in part according to the rational connections they have with
other propositional-attitude contents, including, but not confined to,
other belief contents. A belief has a particular content in virtue of the
inferences that it licenses to other propositional-attitude contents, where

an inference is licensed by a belief content if and only if having a belief
with that content makes it rational for the believer to draw those infer-
ences. By the same token, concepts are also individuated in normative
reason-giving terms. This emerges particularly clearly when we reflect on
the “normative liaisons” built into the conditions of possessing a particu-
lar concept (Peacocke 1992, chapter 5). Not only will any acceptable ac-
count of what it is to possess a concept have to specify the circumstances
in which it is appropriate for a thinker to apply that concept, but also
part of what it is to possess a given concept is that one should be able
to recognize circumstances that give one good reasons to take particular
attitudes to contents containing that concept. Moreover, concept mastery
is also evidenced in dispositions to make, and accept as legitimate, certain
inferential transitions between judgements.
Of course, it follows trivially from these points that the normative di-
mensions of concept possession depend on the normative dimensions of
the belief and other attitudinal contents in which they might feature. But
this does not yet warrant the conclusion that belief contents, and attitude
contents in general, must be conceptual. To derive that further conclu-
sion, we need two further lemmas. The first is the very familiar point that
the content of a belief is determined by its constituents. If a belief has a
conceptual content, then that content will be determined by the nature of
its conceptual constituents, and if it has a nonconceptual content, then
its content will be determined by the nature of its nonconceptual constit-
uents. The second lemma is that if a content constituent is individuated
in normative and reason-giving terms, then it is a concept. This follows
from the definition of a concept, because concepts just are those content
constituents that are individuated in normative and reason-giving terms.
I should stress, however, that this is not meant to imply that contents with
nonconceptual constituents cannot stand in reason-giving relations to
other contents. That would be patently false. The claim is rather that
those reason-giving relations do not serve to individuate those noncon-
ceptual content constituents.
The conclusion that belief contents must be conceptual can be derived
as follows. Since the content of a belief is determined by its constituents,
the normative dimensions of a belief content must be determined by the
normative dimensions of its constituents. Since belief contents are individ-
54 Chapter 3
uated in terms of the normative dimensions of their inferential role, it

follows that they must play their respective inferential roles essentially.
Putting these two thoughts together yields the conclusion that belief con-
tents must be determined by constituents that themselves essentially
possess a normative dimension. These constituents, by the second lemma,
must be conceptual. So, to return to the general question of how the dis-
tinctive features of perceptual experiences can best be accommodated, we
now seem to be in a position to make a clear distinction between the
(nonconceptual) contents of perceptual experiences and the (conceptual)
contents of perceptual beliefs (and propositional attitudes in general).
The possible contents of propositional attitudes are constrained by the
conceptual repertoire of the thinker in a way that the possible contents of
perceptual experiences are not.
Some philosophers would resist this suggestion that perceptual experi-
ence has nonconceptual content. There are, broadly, two strategies for
such resistance. First, it might be argued that the content of perception is
exhausted by dispositions to acquire beliefs (see Armstrong 1968 for a
classic statement of this epistemic theory of perception). Clearly, if per-
ception has no nondoxastic elements, then, given the above argument that
the contents of belief are fully conceptual, issues of nonconceptual con-
tent do not arise. I will say no more about this possible line of argument,
however, as I suspect that it ultimately collapses into a version of the non-
conceptual theory. The idea that there is no more to perceptual content
than can be given by perceptual beliefs or dispositions to acquire such
beliefs is only plausible if the notions of belief and of content possession
are weakened so far that they effectively come into the domain of noncon-
ceptual content.
The second strategy for denying the need for nonconceptual content is
more promising. Its general thrust is that the fine grain and richness of
perceptual experience can be accommodated within a fully conceptual
account of perception. John McDowell is a sophisticated proponent of
such a strategy (McDowell 1994, particularly lecture 3). Unlike propo-
nents of the first strategy, he does accept that perceptual experience has a
richness that cannot directly be captured by concepts. He agrees that a
given perceiver can, and often does, have perceptual experience that is
more fine-grained than his conceptual repertoire, but he takes this point
to hold in only a limited sense that does not warrant talk of nonconcep-
tual content. This can best be explained through the example that he him-
self discusses, color experience. It is clear that normal perceivers can
discriminate a far wider range of colors than they have concepts for. This
is the uncontroversial starting-point from which it is tempting to conclude
(as above) that perceptual experience has a nonconceptual content. Mc-
Dowell, though, thinks that the uncontroversial starting-point can be ac-
commodated without any such theoretical move:
It is possible to acquire the concept of a shade of a colour, and most of us have
done so. Why not say that one is thereby equipped to embrace shades of colour
within one’s conceptual thinking with the very same determinateness with which
they are presented in one’s visual experience, so that one’s concepts can capture
colours no less sharply than one’s experience presents them? In the throes of an
experience of the kind that putatively transcends one’s conceptual powers—
an experience that ex hypothesi affords a suitable example—one can give linguis-
tic expression to a concept that is exactly as fine-grained as the experience, by
uttering a phrase like “that shade,” in which the demonstrative exploits the pres-
ence of the sample. (McDowell 1994, 56–57)
An obvious question to ask is why this should count as a conceptual ca-
pacity, on a par with mastering the concept of red, for example. Such
demonstrative thoughts can exist only in the presence of the particular
shade of color that they pick out, and this is hardly comparable to the
capacity to identify and reidentify appropriately colored objects over
time, which is integral to the mastery of color concepts. But McDowell
has a response to this:
In the presence of the original sample, “that shade” can give expression to a con-
cept of a shade; what ensures that it is a concept—that thoughts that exploit it
have the necessary distance from what would determine them to be true—is that
the associated capacity can persist into the future, if only for a short time, and
that, having persisted, it can be used also in thoughts about what is by then the
past, if only the recent past. What is in play here is a recognitional capacity, pos-
sibly quite short lived, that sets in with the experience. (McDowell 1994, 57)
The suggestion is that a minimal apparatus, containing solely the concept
of a shade of color and ability to employ the demonstrative pronoun, can
capture all the richness and fine grain of color experience. What makes it
a proper conceptual capacity for identifying colors is that it can serve as
the foundation for recognizing particular shades of color and for thinking
about them in their absence.
56 Chapter 3
This general strategy is not, of course, confined to colors. Theorists of

nonconceptual content often appeal to spatial features of the perceived
environment that they hold to be more finely individuated than would be
possible if individuation proceeded according to the perceiver’s repertoire
of spatial concepts. It seems plausible to hold that perceivers can perceive
the differences between a circle, an ellipse, and a sphere without pos-
sessing all three relevant concepts. McDowell’s position would accommo-
date this by appeal to the conceptual capacities minimally required to
think demonstratively about shapes (that shape, for example).
McDowell’s position is unattractive, however, because it has to concede
precisely the point that it is trying to rule out: that color experience often
has a richer and finer content than can be captured in terms of concepts
possessed by the perceiver. Although he offers a sense in which one can
think demonstratively about shades of color that one can discriminate but
nonetheless cannot directly conceptualize, he nonetheless has to describe
the exercise of basic color-recognition abilities as involving “thoughts
that are not necessarily capable of receiving an overt expression that fully
determines their content” (McDowell 1994, 58). It is worth reminding
ourselves that the import of placing a conceptual constraint on perceptual
content is to capture the idea that the type of perceptual experiences a
perceiver can have is determined by the concepts he possesses.4 The
thought often canvassed in support of this is that one cannot see, for
example, a post box without possessing the concept post box, because
perceiving a post box requires the ability to classify it as a post box and
that just is a conceptual ability (or so the argument runs). It follows from
this that if we know what concepts a perceiver possesses, we will be able
to define what might be termed a perceptual space limiting the colors,
shapes, objects, etc., that he can properly be described as perceiving. It
seems clear, however, that McDowell’s minimal recognitional capacities
do not permit this. If the relevant conceptual capacities are not those asso-
ciated with the concepts of particular shapes or particular colors but
rather those associated with the demonstrative pronoun and the general
concepts shade and color, then no such definition of a perceptual space
will be possible. The range of perceivable colors and perceivable shapes
will be unlimited, in which case it is far from clear that the notion of
conceptual content is doing much work here.
The significance of this emerges when one asks what the truth condi-
tions are for such demonstrative thought in cases where the color sample
is no longer present (in cases where we really do have what McDowell is
proposing as a recognitional capacity). A particular application of the
recognitional capacity will be true if and only if the particular shade cur-
rently being perceived really is the same shade as the shade originally per-
ceived. But what about cases when the currently perceived shade is not the
same as the previously perceived shade, as when a perceiver mistakenly
confuses the lilac that he is now looking at with the maroon that he saw
previously? Suppose that a perceiver does something that we can explain
only by attributing to him just such a confusion of shade (such as, for
example, making a mistake in sorting objects). The natural way of de-
scribing what is going on would be that he is perceiving lilac and confus-
ing it with maroon, whether or not he possesses the relevant concepts.
McDowell’s demonstrative account is committed to denying this. Mc-
Dowell would presumably say that the recognitional capacity for recog-
nizing that shade that persists in thought (namely, maroon) is being
misapplied. This, however, does not allow us to describe what is going on
in sufficient detail to allow us to distinguish this case from one in which
a perceiver makes a similar mistake, but this time between maroon and
mauve. This confusion would also be described as a misapplication of the
persisting recognitional capacity for recognising that shade. Nonetheless,
McDowell is committed to a picture of perceptual experience on which
there really is a salient phenomenological difference between these two
cases. And if there is a salient phenomenological difference, then the obvi-
ous way to capture it is by identifying the colors that are actually per-
ceived, even if the perceiver is not (conceptually) capable of identifying
them as such. The reason that McDowell-type demonstrative thought is
possible about perceptually encountered shades of color is that those
shades of color are encountered in experience, and this is something that
should be reflected in an adequate account of what is actually perceived.
But this brings us back to the idea that perceptual experience has noncon-
ceptual content.
Bearing in mind, then, that the only alternative to a McDowell-type
strategy is a version of the epistemic theory, the balance of the evidence
appears to be in favor of an account of perceptual experience that allows
58 Chapter 3
it to have a component that is nondoxastic but nonetheless representa-

tional. We seem driven to the idea of nonconceptual content by the need
for an adequate characterization of color experience. But, to return to
the matter in hand, we are still a long way away from a conception of
nonconceptual content that is of obvious theoretical use in explaining
primitive forms of self-consciousness. That would require a notion of
nonconceptual content with application beyond the realm of colour expe-
rience. It is to this that I turn in the following section.
3.2 Extending the Notion of Nonconceptual Content: The Autonomy

Principle
In this section I make a start on explaining how the notion of the non-
conceptual content of perception can be put to use to solve the paradox
of self-consciousness. I distinguish two different explanatory projects
within which the notion of nonconceptual content can be deployed.
These two projects can be correlated with the different strategies required
to dissolve the two types of circularity at the heart of the paradox of
self-consciousness.
One way of putting the notion of nonconceptual content to work is in
explaining what it is to possess certain concepts. In the case of a large
and probably foundational class of concepts, an adequate explanation of
what it is for a subject to possess a given concept will have to contain a
clause stipulating how that subject responds when enjoying experiences
with an appropriate content. There is an obvious danger of circularity,
however, if the content in question is conceptual. This danger of circular-
ity makes the notion of nonconceptual content appealing, because one
promising way of overcoming it is to specify the conditions for concept
possession so as to require responding in appropriate ways (say, by
applying the concept in question) when enjoying experiences with the ap-
propriate nonconceptual content.
We can see how this might work with the example of color concepts.
Any adequate account of what it is to have mastery of a particular color
concept will obviously have to require a certain sensitivity to the presence
of samples of the color in question, and it is natural to think that this
sensitivity will involve responding appropriately when experiencing
objects of the color in question. So an account of what it is to master a

particular colour concept will mention perceptual experience of that
color. But then we seem driven to the thought that such perceptual experi-
ence has a nonconceptual content, because of the obvious circularity that
there would be in the account of conceptual mastery if the content of the
perceptual experience demands mastery of the concept we are trying to
explain. The suggestion, then, is that a specification of the conditions that
must be satisfied by any individual who can properly be ascribed mastery
of a particular color concept will involve a clause stipulating that he be
sensitive to experiences with a nonconceptual content featuring the color
in question.
This use of the notion of nonconceptual content in specifying what it
is to possess given concepts has been pressed by Christopher Peacocke
(1992, 1994). Peacocke offers a specification of the possession conditions
for certain concepts that takes two parts. The first part of the specification
for a particular concept hinges on the constraint that the subject be in a
state with an appropriate nonconceptual content and that he be willing
to apply the concept when in that state. The second part is a version of
the Generality Constraint first proposed by Gareth Evans, namely that
the subject be capable of generalized concept application (Evans 1982,
sec. 4.3). Broadly speaking, the idea is that if a subject is to be properly
credited with mastery of the concepts in the thought a is F, then that
subject must be capable of thinking a is G for any property G of which
he has a conception, and similarly of thinking b is F for any object b of
which he has a conception. When these two conditions are satisfied, then,
Peacocke holds, the subject has mastered the concept in question. That
the subject has experiences with nonconceptual content is central to this
account, and this centrality derives from the noncircularity mentioned
earlier. The specified concept will feature in the specification of the first
condition, but it will not do so in a way that presupposes that the subject
possesses the concept in question. The subject’s possession of the concept
in question is neither mentioned in, nor required by, a specification of the
content of the appropriate perceptual experience.
A possession-conditions explanation of concepts canvassed by Pea-
cocke needs to be sharply distinguished from another potential way in
which the machinery of nonconceptual content might be employed to
60 Chapter 3
elucidate conceptual content and concept mastery in general. This second

form of explanation, which one might term a developmental explanation
of conceptual content, explains the acquisition of conceptual abilities
from states with nonconceptual content in terms of a developmental pro-
gression over time. On this view, nonconceptual content comes first, as a
basis from which conceptual content can emerge. Cussins (1990) gives an
account of conceptual content along these lines, suggesting that the grad-
ual development of a conception of objectivity should be understood in
terms of a progressive construction of concepts from a basis of structured
nonconceptual content. He describes this as “the transition from a pre-
objective stage where no concepts are possessed to an objective concept-
exercising stage” (Cussins 1990, 409). The possibility of this sort of de-
velopmental explanation of conceptual content is of considerable interest
to developmental psychologists studying the ontogeny of concept devel-
opment (as will be discussed at greater length in the following section),
as well as to evolutionary biologists studying the phylogeny of concep-
tual abilities.
This broad distinction between a possession-conditions explanation
and a developmental explanation is directly relevant to the different ex-
planatory projects posed by the two strands of the paradox of self-
consciousness. The distinction there, it will be remembered, is between
the unwelcome implications that stem from explanatory circularity and
those that derive from capacity circularity. Explanatory circularity poses
problems for the project of providing a noncircular account of self-
consciousness (because of the need for first-person thoughts in an account
of mastery of the first-person pronoun). Insofar as a solution to this
strand of the paradox of self-consciousness involves deploying noncon-
ceptual first-person contents to give a noncircular account of mastery of
the first-person pronoun, it seems clearly to fall within the broad category
of a possession-conditions explanation. The solution to problems arising
from capacity circularity, on the other hand, will not be an instance of a
possession-conditions explanation. What is at stake is the possibility of
showing how mastery of the first-person pronoun can emerge in the nor-
mal course of human development, which clearly counts as an instance
of developmental explanation.
One reason why this is significant is that a possession-conditions expla-

nation and a developmental explanation involve differing degrees of com-
mitment to what we can, following Peacocke (1992, 90), term the Auton-
omy Principle:
The Autonomy Principle It is possible for a creature to be in states with
nonconceptual content, even though that creature possesses no concepts
at all.
Given what, in chapter 2, I identified as the Priority Principle (namely that
conceptual abilities are constitutively so linked with linguistic abilities
that conceptual abilities cannot be possessed by nonlinguistic creatures), it
follows that the Autonomy Principle, if true, leaves open the possibility of
a nonlinguistic creature being in states with nonconceptual content.
A philosopher who holds that the notion of nonconceptual content can
provide a developmental explanation of conceptual content is likely to be
sympathetic to the Autonomy Principle, to the extent that his project is
one of explaining how a creature possessing only states with nonconcep-
tual content can develop into a full-fledged concept user.5 A philosopher
committed only to a possession-conditions explanation, however, does
not have to accept the Autonomy Principle (although there is no require-
ment to deny it). He can, for example, maintain that the nonconceptual
content of experience will yield a noncircular account of possession con-
ditions for certain primitive concepts, even if the Autonomy Principle is
denied.6 What makes this possible is, of course, that nonconceptual con-
tent is defined as content that does not require mastery of any of the con-
cepts required to specify it; it is not defined as content that requires
mastery of no concepts whatsoever. Indeed, one can be in a state with
nonconceptual content despite possessing the relevant concepts. The
point is that being in such a state does not depend on possession of the
concepts required to specify it.
On the other hand, however, if the Autonomy Principle is rejected, then
it will ipso facto be impossible to give at least one type of developmental
explanation of conceptual content in terms of nonconceptual content,
namely the type of developmental explanation that involves explain-
ing how a creature in states with only nonconceptual content, such as,
for example, a newborn human infant, can develop into a full-fledged
62 Chapter 3
concept user. By the same token, it will also rule out any account of the
phylogenetic development of conceptual content on which it appears
at a far later stage of evolution than nonconceptual content.7 Moreover,
it is plausible that a primitive form of intentional explanation is required
to account for the behavior of creatures that one might not want to de-
scribe as concept-using, and obviously, any form of intentional explana-
tion requires attributing to the creature representations of its en-
vironment. One area in which this emerges is animal-learning theory (see,
for example, Dickinson 1988, where it is argued that certain cases of
instrumental conditioning in rats support an intentional interpretation,
as well as Premack and Woodruff 1978), but it also seems highly relevant
to the study of infant cognition. The suggestion that there are experiential
states that represent the world but do not implicate mastery of the con-
cepts required to specify them is potentially important here. Clearly,
however, it can only be of theoretical use if the Autonomy Principle is
accepted.
For present purposes, the important point to bring out is that a solution
to the strand of the paradox of self-consciousness created by capacity
circularity seems prima facie to involve a commitment to the Autonomy
Principle, because such a solution requires showing how mastery of the
first-person pronoun (and hence the first-person concept) can meet the
Acquisition Constraint and it is natural to think that this will involve
showing how mastery of the first-person pronoun can be constructed on
the basis of states with nonconceptual content that can exist in the ab-
sence of any form of concept mastery. Accordingly, in the next section I
shall offer a defence of the Autonomy Principle.
3.3 Defending the Autonomy Principle: Evidence from Developmental

Psychology
Developmental psychology has long been concerned with the question,

At what stage, in early childhood or infancy, is it appropriate to ascribe
a grasp that objects exist even when not being perceived? On the tra-
ditional view, derived ultimately from Piaget, object permanence does
not appear until relatively late in development, toward the end of the sen-
sorimotor period. Prior to the end of the sensorimotor period the infant
Habituation
(a) (b)
Test
(c) (d)
Figure 3.1
A schematic depiction of displays from an experiment on infants’ perception of
object boundaries. (From Spelke and Van de Walle 1993, 135.)
universe is almost completely undifferentiated. Much recent work in de-

velopmental psychology, however, has attacked this view (see Spelke 1990
and Spelke and Van de Walle 1993 for overviews). There is strong experi-
mental evidence that infants have the capacity from a very young age to
parse the array of visual stimulation into spatially extended and bounded
individuals that behave according to certain basic principles of physical
reasoning.
Elizabeth Spelke, whose laboratory has produced much of the interest-
ing work in this area, has identified four such principles governing the
unity of individuals in the perceived array. It will be helpful briefly to
outline the results. The first principle she terms the principle of cohesion,
according to which surfaces lie on a single individual if and only if they
are in contact. It is evidence for the principle of cohesion, for example,
that infants do not appear to perceive the boundary between two objects
that are stationary and adjacent, even when the objects differ in color,
shape, and texture (figure 3.1 gives an example). Three-month-old infants
64 Chapter 3
are habituated to two objects, one more or less naturally shaped and ho-
mogenously colored, and the other a gerrymandered object that looks
rather like a lampshade. When the experimenter picks up the objects, they
either come apart or rise up cleanly. Infants show more surprise when the
object comes apart, even if (as in the case of the lampshade) the object
does not have the Gestalt properties of homogenous color and figural
simplicity. The conclusion drawn by Spelke and other researchers is that
the infants perceive even the gerrymandered object as a single individual
because its surfaces are in contact.
The principle of cohesion clearly suggests that infants will perceive ob-
jects that have an occluded center as two distinct individuals, since they
cannot see any connection between the two parts. And this indeed is what
they do perceive, at least when dealing with objects that are stationary.
Thus it seems that infants do not perceive an occluded figure as a single
individual if the display is static. After habituation to the occluded figure,
they showed no preference for either of the test displays.
On the other hand, however, infants do seem to perceive a center-
occluded object as a single individual if the object is in motion, irrespec-
tive, by the way, of whether the motion is lateral, vertical, or in depth (see
figure 3.2). According to Spelke, this is because there is another principle
at work, which she terms the principle of contact. According to the prin-
ciple of contact, only surfaces that are in contact can move together.
When the principle of cohesion and the principle of contact are taken
together, they suggest that since the two parts of the occluded object move
together, they must be in contact and hence in fact be parts of one
individual.
Spelke also identifies two further constraints governing how infants
parse the visual array. These emerge from experiments involving hidden
objects and their motions. A distinctive and identifying feature of physical
objects is that every object moves on a single trajectory through space and
time, and it is impossible for these paths to intersect in a way that would
allow more than one object to be in one place at a time. One might test
whether infants are perceptually sensitive to these features by investi-
gating whether they are surprised by breaches of what Spelke calls the
´
continuity and solidity constraints. Renee Baillargeon’s well-known draw-
bridge experiments are often taken as evidence that infants are perceptu-
Habituation
Test
Figure 3.2
A schematic depiction of displays from an experiment on infants’ perception of
partly occluded objects. Arrows indicate the direction and relative extent of the
rod’s motion. (From Spelke and Van de Walle 1993, 140.)
ally sensitive to the solidity constraint. In one version (Baillargeon 1987),

Baillargeon habituated 41⁄2-month-old infants to a screen rotating verti-
cally 180° on a table, rather like a drawbridge. She then placed a station-
ary object behind the screen so that it was completely occluded by the
time the screen had been raised 60°. She was interested in whether infants
would distinguish between trials in which the screen stopped when it
reached the place occupied by the object and trials in which the screen
66 Chapter 3
Continuity violation
A ?
Position
B ?
Time
Figure 3.3
A schematic depiction of events violating the continuity constraint. The solid lines
indicate each object’s path of motion. (From Spelke and Van de Walle 1993, 148.)
continued rotating (apparently passing through the object). In fact, the

infants looked longer at the second type of trial, which thus indicates that
they found this novel and surprising. Spelke takes this as evidence of the
infants’ sensitivity to the impossibility of there being more than one object
in a single place at one time.
Figure 3.3 gives a schematic representation of an experiment to test
whether infants parse their visual array in accordance with the continuity
constraint. Two objects disappear and reemerge without having appeared
in the space between. This contravenes the continuity constraint because
an object can only move on a single path through space-time if that path
is continuous.
I won’t go into any more details about the experimental work on infant
perception. The point to extract is that if we accept the general methodol-
ogy of habituation followed by tests for preferential looking, we have to
conclude that the perceptual world of young infants is not at all the undif-
ferentiated chaos of noises, smells, patches of color, and flashes of light
that it was believed to be by both Piaget and the early sense-data theorists.
Suppose, then, that we accept that Spelke is right to deny Piaget’s con-
ception of the universe of early infancy as almost completely undifferenti-
ated. This raises two fundamentally important questions. The first is
whether the capacities to which she draws attention should be reflected
in accounts of the content of the perceptual experience of young infants?
An affirmative answer is compelling here. The crux of the experimental
evidence is that what young infants perceive is not a “blooming, buzzing
confusion” of sensations but rather a world composed of determinate

and bounded individuals behaving in reasonably fixed and determinate
manners. And if that is how they perceive the world, then it is clear that
such a world must feature in an account of their perceptual experiences.
This brings us to a second, considerably more controversial question:
must we conclude from an affirmative answer to the first question that
the infants in question are to be ascribed mastery of the concept of
an object, or can we say what we want to say about their perceptual ex-
perience at the level of nonconceptual content? Denying the Autonomy
Principle would certainly force us in the direction of a conceptual inter-
pretation. Moreover, anyone who does not accept the Autonomy Prin-
ciple is committed to holding that the infants in question will have to
possess the basic conceptual abilities required to support a rudimentary
conception of an objective world, because no creature that lacks these
basic conceptual abilities could count as a concept user.8 Whatever de-
tailed account is favored of what it is to conceive the world as objective,
it will clearly have to include the following: mastery of the concept of a
place, mastery of some form of the first-person concept, mastery of the
connectedness of space, and mastery of the concept of an object. These
conceptual abilities are at the core of any account of what it is to appre-
hend the objectivity of the world, in however rudimentary a way.
Any detailed theory of concept possession that is accepted will have to
reflect the close connection between possessing a concept and carrying
out certain forms of inference. Mastery of a concept is tied up with grasp
of its inferential role, where a concept’s inferential role can be understood
in terms of its contribution to the inferential powers of propositions in
which it features. It is interesting, then, that there does seem to be an
analogue for this in infant behavior. It is clear that the notion of solid-
ity or impenetrability is essential to the concept of an object, so that if
something is an object, it follows that it will be solid and hence that noth-
ing can pass through it. So we may take the following inference as a basic
example of the type of inference licensed by possession of the concept of
an object:
x is an object.
Nothing will pass through x.
68 Chapter 3
Let us return, then, to Baillargeon’s drawbridge experiments (Baillar-

geon 1987). At one level, it seems possible to describe what is going on
in the drawbridge and similar experiments in terms of a primitive applica-
tion of the inference pattern connecting objects with solidity and impene-
trability. This seems a natural way of understanding the idea that the
infants are engaged in a form of physical reasoning. On this view, they
show surprise because they have seen that something is an object, inferred
that it must therefore be solid and impenetrable, and yet seen a draw-
bridge apparently pass through it.
It seems plausible on this view to ascribe mastery of the concept of an
object to infants who demonstrated the appropriate dishabituation be-
havior in a range of contexts sufficiently wide to indicate that they had
grasped several such inference patterns. This is Spelke’s own interpreta-
tion of the dishabituation results:
I suggest that the infant’s mechanism for apprehending objects is a mechanism of
thought: an initial theory of the physical world whose four principles jointly de-
fine an initial object concept. This suggestion is motivated not only by evidence
of the centrality of the mechanism for apprehending objects, but also by a consid-
eration of the principles governing its operation. The principles of cohesion,
boundedness, substance, and spatio-temporal continuity appear to stand at the
centre of adults’ intuitive conceptions of the physical world and its behaviour: our
deepest conceptions of objects appear to be the notions that they are internally
connected and distinct from one another, that they occupy space and that they
exist and move continuously. (1988, 181)
On such a view, the basic elements of the concept are in place very early
on in life. What happens subsequently in development is their refinement
and incorporation into a body of knowledge about the physical, mechani-
cal, and dynamic properties of objects.9
There is, of course, a simple answer to this proposal. It follows from
the Priority Principle, which I adopted in chapter 2, that conceptual abili-
ties are constitutively linked with language mastery in a way that rules
out the ascription of conceptual abilities to nonlinguistic or prelinguistic
creatures. Since Baillargeon’s infants are clearly prelinguistic, this would
immediately rule out the conceptualist interpretation. This simple answer
should be treated with prima facie suspicion, however, because the Prior-
ity Principle has not yet been defended. I propose to proceed instead by
offering a general argument against the conceptualist interpretation of
the empirical evidence, showing how these general considerations can

be employed in support of the Priority Principle. What follows will, I
hope, fulfill the promissory note for a defence of the Priority Principle
issued in chapter 2.
The conceptualist interpretation blurs some very important distinc-
tions. According to the interpretation, the experimental infants are
applying a primitive inference pattern connecting the concept object with
the concept impenetrability. But the fact that young infants are surprised
when certain expectations they have are thwarted does not mean that they
have drawn inferences from the fact that something seems to be an ex-
ample of an object. This is so for (at least) two reasons.
Identifying something as an object is something that one does for rea-
sons. An individual identifies something as an object if he knows what
the criteria for objecthood are and can recognize that they are by and
large satisfied. The subject is making a judgement that can be justified or
unjustified, rational or irrational. It does not seem, however, that anything
like this is going on in the infant behavior under consideration. The fact
that infants are capable of parsing their visual array into bounded seg-
ments that (more or less) correspond to objects does not warrant the
thought that there are reasons for this that can be evaluated according to
the standards appropriate for ascriptions of justified belief. Of course,
there are reasons why such parsing abilities are in place, but they are of a
different type and are to be judged by different standards. It might, for
example, be explicable purely at the level of early visual processing. This
point can be appreciated by considering what the appropriate response
would be to an infant who shows surprise when an object passes through
something that, although bounded, is not an object (say a shadow). It
would be totally inappropriate to describe such an infant as making an
unjustified judgement. It is not the infant’s standards of rational assess-
ment that are at fault (he hasn’t got any). The problem is that his visual
processing is not sufficiently discriminating within the general category
of bounded segments.
The second problem follows from this. If a subject justifiably thinks
that something exemplifies the concept of an object, then this licenses him
in drawing certain inferences, a good example of which is the inference
from something’s being an object to its being impenetrable. But whatever
70 Chapter 3
drawing an inference means, it must mean more than having certain ex-
pectations that manifest themselves only in the presence or absence of
surprise. Drawing inferences requires grasping general rules of inference
and recognizing that their application is appropriate in a given situation.
The paradigm cases of inference are conscious reflective acts, and al-
though many of the inferences that mature concept users make are uncon-
scious and unreflective, they are exercises of capacities that could be made
conscious and reflective. There seem to be no reasons for thinking that
there is anything like this, or even approaching it, going on in dishabitua-
tion behavior.10
The contrast between genuine inference and what might best be termed
sensitivity to the truth of inferential transitions can easily be illustrated.11
Piaget’s well-known discussions of searching behavior (1954) provide
good examples of a case where inference is obviously absent. Consider
the A, not-B error (the stage 4 error). An object that an 8-month-old
infant has successfully found at place A is moved in full view of the infant
and hidden at place B. Instead of searching at B, the infant searches again
at A. It would clearly be inappropriate to describe this behavior in terms
of the infant making any inferences about the object still being where it
was first found. Consider, on the other hand, the classic experiments by
Susan Carey (1982) on 4-year-olds. Carey told the children that people
had a greenish internal organ called a ‘spleen’. She then showed them a
toy mechanical monkey and a live earthworm and asked them which was
more likely to have a spleen. Although a toy mechanical monkey obvi-
ously looks much more like a human being than an earthworm does, the
children decided that the worm was more likely to have a spleen. How
did they come to this conclusion? It seems natural to say that they made
use of inference patterns linking the concepts human being, living animal,
and internal organs. To explain what is going on in these experiments, in
stark contrast to the drawbridge experiments, we really do need to talk
about concepts and inferences.
It is not coincidental that the example of genuine inference comes from
the domain of linguistic competence. A clear understanding of the
grounds for drawing a sharp contrast between genuine inference and sen-
sitivity to the truth of inferential transitions supports the idea that there
is a constitutive link between language mastery and concept mastery. The
sort of inferences that can plausibly be taken as good evidence of concept

mastery are carried out for reasons, as is the classification of objects un-
der the relevant concepts. Indeed, there are obvious rational connections
between classifying objects in certain ways and being prepared to accept
certain inferences on the basis of those classifications. The plausibility of
the Priority Principle emerges when one starts to reflect on the constraints
on being able to appreciate rational grounds for certain inferences. Cer-
tainly, it is possible to be justified (or warranted) in making a certain infer-
ential transition without being able to provide a justification (or warrant)
for that inferential transition. It is a familiar epistemological point, after
all, that there is a difference between being justified in holding a belief
and justifying that belief. What does not seem to be true is that one can
be justified in making an inferential transition even if one is not capable
of providing any justifications at all for any inferential transitions. But
providing justifications is a paradigmatically linguistic activity. Providing
justifications is a matter of identifying and articulating the reasons for
a given classification, inference, or judgement. It is because prelinguistic
creatures are in principle incapable of providing such justifications that
the priority thesis is true. Mere sensitivity to the truth of inferential transi-
tions involving a given concept is not enough for possession of that con-
cept. Rational sensitivity is required, and rational sensitivity comes only
with language mastery.
For all these reasons, then, the phenomena that developmental psychol-
ogists interpret in terms of object permanence are not a sign of mastery
of the concept of an object. Maintaining the Autonomy Principle is the
only way in which we can reconcile this with the earlier thought that
young infants’ capacities to register object permanence should feature in
an account of the content of their perceptual experience. Denying the
Autonomy Principle leaves one with the choice between either ascribing
to the infants mastery of certain basic objective concepts or denying that
the infants have experience with representational content and that this
content has a role to play in explaining their behavior. Neither of these
possibilities is plausible. The first has already been discussed, while the
second seems to be vitiated by the truth of counterfactuals like the follow-
ing: ‘If the infants had perceived an array of unstructured visual sense
data rather than bounded segments, they would not have shown surprise
72 Chapter 3
when an apparently continuous surface comes apart’. The infants in dis-

habituation experiments show surprise because they see things a certain
way, and it is natural to describe how they see things by saying that they
parse the visual array into bounded segments, even though they have no
conceptual grasp of what those bounded segments are.
Nonetheless, some caution is needed here. Even if it is granted that the
infants’ perceptions need to be explained at the nonconceptual level, there
is still the important question as to how the nonconceptual content of
those perceptions is to be elucidated. The question can be focused by
asking whether it follows from the preceding discussion that the infants
perceive the world as divided up into objects. The reason I have formu-
lated the discussion so far in terms of bounded segments and individuals
has been precisely to leave this question open. It is a substantive question
whether the bounded segments into which the infant universe is divided
up map at all cleanly onto the objects into which the actual universe is
divided up. If those bounded segments do map onto objects, then it will
be right to describe the experimental infants as nonconceptually perceiv-
ing objects. That is to say, specifications of the content of the infants’
visual perceptions will have to be made in terms that include the concepts
pertaining to objects. If, on the other hand, the bounded segments into
which infants parse the visual array turn out not to be equivalent to ob-
jects, then some other concepts will have to be employed in specifying
how the infants perceive the world.
How might we settle this question? What conditions would have to be
satisfied for the contents of infant perceptions to be properly described in
terms of objects? The necessary conditions, it seems to me, are ultimately
derivable from a specification of the properties that can plausibly be seen
as essential to objects. Let me call these object properties. Object proper-
ties are properties constitutive of objects, so that any individual of which
a suitable set of object properties can be truly ascribed will ipso facto
count as an object. Examples of object properties are these:
• The property of following a single continuous trajectory through
space-time
• The property of continuing to exist when unperceived
• The property of having a determinate shape
• The property of being impenetrable
• The property of being subject to gravity

• The property of being internally causally connected
• The property of having a certain mass
• The property of posing resistance to touch
• The property of having its state of motion or rest explicable in terms of
the mechanical forces acting upon it
• The property of having its state of motion or rest exert forces on other
objects
There are no doubt other object properties, and some of the properties
on this list may well be reducible to others. Correctly determining the set
of object properties is obviously a significant philosophical task. For pres-
ent purposes, however, all that is required is that there actually be such a
set and that we can uncontroversially identify at least some of its core
members.
Given that object properties define objecthood, it is natural to hold that
a perceiver perceives objects to the extent that he is perceptually sensitive
to object properties.12 This is, of course, an application of a general prin-
ciple governing all attributions of perceptual content, namely, that a per-
ceiver cannot properly be described as perceiving a given thing unless he
can perceptually discriminate that thing. To return to the original ques-
tion, the following seems a plausible answer. Infants, and indeed any
other perceivers at the nonconceptual level, can properly be described as
perceiving objects to the extent that they are perceptually sensitive to ob-
ject properties. Of course, there is an important distinction between those
object properties that are, and those that are not, perceptually manifest
at a time. The object property of having a determinate shape is an obvious
example of one that is perceptually manifest at a time. By definition,
though, the object property of existing unperceived is not perceptually
manifest at a time. The evaluation of perceptual sensitivity to object prop-
erties that are not perceptually manifest at a time will need to take into
account behavioral dispositions and expectations. A good example (and
one rather familiar to developmental psychologists) of the sort of behav-
ior that would be prima facie evidence of perceptual sensitivity to unper-
ceived existence would be searching for an object that is hidden. Here,
as always when dealing with nonconceptual content, considerations of
parsimony must be paramount.
74 Chapter 3
There seems to me to be no doubt that an important element in infant

cognitive development is increasing sensitivity to an increasing range of
object properties. Although this is a purely empirical matter (and a
profitable avenue for research in developmental psychology 13), it seems
natural to conjecture that the most basic and primitive ability is the ability
to parse the visual array into bounded segments that are, as cognitive
development proceeds, perceptually endowed with progressively richer
properties, among which will feature an increasing range of object prop-
erties. When sensitivity to object properties is sufficiently developed the
bounded segments which are perceived will map closely onto the objects
which really exist in the environment. However, at some stages in infant
cognitive development there will not be perceptual sensitivity to a suffi-
cient number of object properties for it to be appropriate to describe
infants at that stage of development as perceiving objects, even at the
nonconceptual level. Clearly, therefore, a different conceptual framework
is required to specify the nonconceptual content of perception for infants
at that stage, and this is an issue to which I shall return at the end of the
next chapter.
Enough has been said, however, for it to be clear not only why the
notion of autonomous nonconceptual content is legitimate but also how
it can have a crucial role to play in a developmental explanation of con-
cept acquisition. As the passage quoted earlier from Spelke and Van de
Walle stresses, the operation of the cohesion, contact, and other similar
principles will eventually help infants towards a position in which they
can pick out objects as objects. Young infants who have acquired suffi-
cient perceptual sensitivity to object properties will be able to represent
the world in a way that divides it up into objects. Accounting for this in
terms of autonomous nonconceptual content gives us the basic raw mate-
rial from which the process of concept acquisition develops. It seems
plausible that a (theoretical) understanding of the reasons for thinking
that something exemplifies the concept of an object, together with an
understanding of what inferences it is legitimate to draw from that fact,
emerges from certain basic representational abilities that permit the sub-
ject to pick out the extension of the concept and that support certain
expectations about how the things that fall under that extension will
behave.
An adequate account of how the concept of an object meets the Acqui-

sition Constraint will involve tracing this process of making representa-
tional abilities explicit. How does the infant move from the primitive
recognitional abilities tapped in dishabituation experiments to explicit
and articulated mastery of the concept of an object? As suggested above,
this process is partly a process of increasing sensitivity to a wider range
of object properties. But that can only be half of the story. The infant
must also move from an implicit understanding to an explicit understand-
ing, from a cognitive level at which understanding is manifested in
thwarted expectations and reactive behavior to a cognitive level at which
the reasons and grounds of those expectations and reactive behavior are
themselves appreciated.
Let me end this section by offering a hypothesis about a vital stage in
this process of moving from an implicit to an explicit understanding. As
already noted, an important reason for the shift in developmental psy-
chology away from the Piagetian perspective has been the increasing use
of dishabituation experiments. This is because dishabituation experi-
ments do not tap the motor abilities required for success on Piaget’s
favored tests for understanding of object permanence. The picture that
emerges from comparing performance on Piaget’s search tasks with per-
formance on dishabituation paradigms strongly suggests that the first
stage in the move from implicit to explicit understanding involves devel-
oping the ability for behavioral, rather than purely reactive, manifestation
of the relevant representational abilities (Russell 1996). There seem to be
two significant developments in motor competence that ground successful
performance on Piagetian tasks like the A, not-B search task (Diamond
1991). Infants need to be able to inhibit their prepotent action responses
if they are to engage in appropriately sophisticated manipulatory behav-
ior, and they also need to be able to develop sequences of actions and to
link information over a separation in space and/or time. The acquisition
of these abilities is known from lesion studies to be correlated with matu-
rational changes in the frontal cortex, particularly in the supplementary
motor areas (including callosal connections between the supplementary
motor areas in the left and right hemispheres) and in the dorsolateral
prefrontal cortex (Diamond 1991).
76 Chapter 3
Accordingly, it seems plausible to suggest that a crucial bridge between,

on the one hand, the implicit nonconceptual understanding of object
properties implicated in infant performance in dishabituation experi-
ments and, on the other, the rational explicit understanding implicated
in full mastery of the object concept is an increasing ability to put that
nonconceptual understanding to work in the manipulation and explora-
tion of the environment.14 This initial stage in the process is relatively well
understood at both developmental and neurophysiological levels. Subse-
quent stages in the process are less well understood, although I hope to
shed some light on them in later chapters of this book. In the final section
of this chapter, I turn to reviewing further evidence for the explanatory
utility of autonomous nonconceptual content.
3.4 The Autonomy Principle: Further Evidence and Applications
Developmental psychology, particularly infancy research, provides the

clearest and most unambiguous examples of the explanatory power of
the notion of nonconceptual content. But there are other areas in which
the notion of nonconceptual content also seems applicable, including cog-
nitive ethology (which I will discuss in greater detail in the next chapter).
I will conclude this chapter more speculatively by pointing to the role that
the notion of autonomous nonconceptual content can play in understand-
ing, first, the phylogeny of language and, second, the visual agnosias (dis-
orders of visual perception following brain damage). I will conclude by
noting how a version of the distinction between conceptual and noncon-
ceptual content seems to be built into the computational theory of vision
developed by David Marr.
The phylogeny of language

It is often suggested that ontogeny recapitulates phylogeny (Brown 1988).
In this spirit it makes sense to move from developmental psychology to
evolutionary theory. The solution that I am proposing to the paradox of
self-consciousness is that we need to recognize the existence of primitive,
prelinguistic forms of self-consciousness if we are to explain the possi-
bility of mastery of the first-person pronoun emerging naturally in the
normal course of development. This in turn might be viewed as an appli-
cation of the general thought that an organism already needs to be think-

ing before it can acquire the capacity to employ a language. That general
thought, in turn, might be defended in one of (at least) two ways. It might
be defended on conceptual grounds, by maintaining that learning a lan-
guage is itself a process of thought, and hence that it would be impossible
for a language to be learned by a creature not capable of some sort of
language-independent thought. Some such argument is often held to be a
powerful motivation for a language of thought (Fodor 1975). Alterna-
tively, it might be defended empirically. The two obvious avenues for such
an empirical defence would be ontogenetic and phylogenetic. A possible
ontogenetic defence for the claim would be the familiar developmental
view of the language-learning child as a little linguist, for example. Com-
parable phylogenetic evidence would emerge if it could be shown that the
evolutionary emergence of language was preceded by a stage at which
early humans were clearly capable of thinking.
Of course, because of the paucity of evidence, it is hard to imagine
anything being shown about the phylogeny of language. But the evidence
that exists from paleoneurology and comparative primate neuroanatomy
is certainly consistent with the hypothesis that significant cognitive abili-
ties emerged before the evolution of language. Researchers are largely in
agreement that the two brain regions most closely involved in language
mastery are Broca’s area and the parieto-occipital-temporal cortex (in-
cluding the region known as Wernicke’s area). The parieto-occipital-
temporal cortex serves to integrate visual, auditory, and somatosensory
information and then feeds structured and integrated information for-
ward to Broca’s area. Neither of these areas is present in primate brains.
Nor, more interestingly, do they appear to be present in the early homi-
nids, that is, the early members of the primate lineage which separated off
between 10 and 4 million years ago from the African apes and eventually
evolved into Homo sapiens sapiens. The first genus of hominid, Austra-
lopithecus, is, from endocast studies, thought likely to have had an ape-
like, rather than a human-like, parietal cortex (Donald 1991, Corballis
1991).15 There is some dispute about dating the emergence of a human-
like neural organization, but recent authors are largely in agreement that
the crucial event in the emergence of language was the speciation of ar-
chaic Homo sapiens approximately 300,000 years ago, following a mas-
78 Chapter 3
sive expansion in brain size and, perhaps most significant, the descent of
the larynx (Lieberman 1991, Corballis 1991, Donald 1991).16 Only once
the larynx has descended is anything like the full range of human speech
sounds available.
Now this view of the evolution of language poses a prima facie problem
for the opponent of the Autonomy Principle for the following reason.
There is considerable evidence that early hominids, considerably prior to
the speciation of archaic Homo sapiens, were cognitively highly evolved,
certainly far more so than any nonhuman primate. Homo habilis, the first
species in the genus Homo (approximately 2 million years ago), is known
to have created and used stone tools (scrapers and choppers) and so is
believed to have had precursors of Broca’s area and a developed parietal
cortex (Tobias 1987). However, a major cognitive breakthrough is be-
lieved to have been made in the transition period leading to the speciation
of Homo erectus approximately 1.5 million years ago. The relative brain
size of Homo erectus was much larger than previous hominids (approxi-
mately 70 percent of the modern human brain), and it continued to in-
crease until the speciation of Homo sapiens. The brain of Homo erectus
is believed to have expanded particularly in the association cortex, the
hippocampus, and the cerebellum. This is associated with well-confirmed
evidence of significant cognitive evolution. Homo erectus is known to
have developed sophisticated stone tools, to have employed complex long-
distance hunting strategies (some of them tool-based), and to have moved
out of Africa to cover much of Eurasia.
Tool-making obviously requires sophisticated instrumental reasoning,
as does the social coordination involved in implementing large-scale mi-
grations. If the modern consensus of opinion is right that the emergence
of language (in the sense of a communication system marked by gen-
erativity and a lexicon containing thousands of items) did not emerge
until nearly 2 million years after the start of tool-making and social coor-
dination, then a nonlinguistic framework is required to make sense of
these more primitive, and yet undeniably cognitive, abilities.17 This is a
task that the theory of nonconceptual content that I will be developing in
this book is suited to perform. Such a framework would still be required
even if it were held that language neither did evolve, nor could have
evolved, by natural selection (Piatelli-Palmarini 1989). It would still be
necessary to explain the cognitive abilities that preceded language even if

language was held to have emerged independently of them. But on the
plausible assumption that some sort of natural-selection account is indeed
correct (Pinker and Bloom 1990), the theory of nonconceptual content
also offers scope for a phylogenetically developmental account of the ac-
quisition of conceptual mastery to match the ontogenetically develop-
mental account sketched in the previous section.
Visual agnosias
Neuropsychologists have long been familiar with a range of deficits in
visual object recognition that are collectively termed the visual agnosias.
These deficits (as opposed, for example, to disorders like blindsight) are
characterized by the relative preservation of elementary visual functions,
such as acuity, brightness discrimination, and color vision. As is often the
case with neuropsychological disorders, the first problem that confronts
workers in the area is providing a workable taxonomy of the range of
deficits. The taxonomy is particularly important because how visual
deficits are classified has obvious implications for (and is equally in-
formed by) the analysis of normal visual processing. What I suggest is
that the distinction between conceptual and nonconceptual content offers
scope for developing and refining what is by now a standard classification
of visual agnosias.
Students of visual agnosias have generally drawn a broad distinction
between the so-called apperceptive agnosias and the so-called associative
agnosias (Brown 1988, Farah 1990). In very broad terms, the distinction
is between deficits in object perception (the apperceptive agnosias) and
object recognition (the associative agnosias). Apperceptive agnosias are
due to impairments in visual perception that, although at a higher level
than, for example, visual-field deficits, nonetheless appear to be percep-
tual rather than recognitional. Objects are not properly perceived, and
hence are not recognized. Classic examples of apperceptive agnosias in-
clude simultagnosia, in which patients seem to be incapable of recogniz-
ing more than one stimulus at a time and unable to make even the
simplest forms of shape discrimination.18 In the associative agnosias, on
the other hand, objects do seem to be properly perceived but are nonethe-
less not recognized. Such deficits appear to be more cognitive than
80 Chapter 3
perceptual. Standard examples of associative object agnosias involve im-

pairment in the recognition of (certain categories of) visually presented
objects, coupled with normal recognition of those objects in nonvisual
modalities and with intact visual perception (as measured, for example,
by the ability to draw a fair copy of the stimulus).19 Impaired visual recog-
nition is manifested both verbally and nonverbally (e.g., in abilities to
mime the use of the relevant objects or to sort pictures of visually dissimi-
lar but functionally similar objects). Prosopagnosia, the inability to recog-
nize faces, is usually classified among the associative agnosias, as is alexia,
an inability to read, coupled with a preserved ability to write.
It is standard to take the two contrasts between normal vision and ag-
nosic vision and between the apperceptive and associative agnosias to
show that there are several different levels of visual processing (Marr
1982, Farah 1990). The first level is often described as the level of purely
sensory processing, where dedicated sensory analyzers and filters, gener-
ally thought to be located in anatomically distinct but nonetheless func-
tionally integrated areas of the visual cortex, yield information about
features of the environment, such as color, motion, luminance, etc. (Ts’o
and Roe 1995). This level is preserved in both the apperceptive and asso-
ciative agnosias. A second level of visual processing might be identified
programmatically as what is lacking in the apperceptive agnosias al-
though nonetheless present in the associative agnosias (and, of course, in
normal vision). The third processing level can be equally programmati-
cally specified as what is lacking in the associative agnosias but not in the
apperceptive agnosias. The associative and apperceptive agnosias yield a
dissociation between two distinct components of the visual-processing
system. One standard way of describing the functional architecture re-
vealed by this dissociation is in terms of a distinction between object per-
ception and object recognition or between form perception and meaning
perception (Brown 1988). Other more complex classifications have been
proposed. Martha Farah has proposed a tripartite analysis of higher vi-
sion. Grouping operations on the outputs of purely sensory processing
yield a grouped array operated on in parallel by the spatial-attention and
object-recognition systems (Farah 1990). The apperceptive agnosias re-
veal impairments in the initial grouping processes, while the associative
agnosias implicate one or another of the spatial-attention and object-
recognition systems.
The existing classifications can usefully be supplemented by thinking

about the functional differences revealed by the visual agnosias in terms
of the distinction between nonconceptual and conceptual content. The
grouping operations impaired in the apperceptive agnosias and preserved
in the associative agnosias produce representations of the world at the
level of nonconceptual content. The grouping operations parse the visual
array into spatially extended and bounded individuals that stand in spa-
tial relations to each other, and that, of course, is how the world seems
to be represented in the instances of infant perception that were taken as
paradigm examples of nonconceptual content. The visual world of the
infant and the visual world of the associative agnosic can be understood
in terms of each other. Of course, the comparison is of necessity a re-
stricted one, since associative visual agnosias are always messy and often
highly localized ( prosopagnosia is an obvious example). But there seems
to be a clear analogy to be drawn between how the infant sees the world
and how the world would be seen by an idealized global associative ag-
nosic—a patient whose representations of the world are devoid of seman-
tic features but for whom the elementary grouping and parsing elements
of higher vision are fully operational. Conversely, those functionally dis-
tinct elements of higher vision that are absent or severely impaired in the
associative agnosias are best viewed as generating representations of the
world at the conceptual level. Although there is considerable evidence
that the associative agnosias are not simply impairments in naming, and
hence purely linguistic, it is highly relevant that they are associated with
damage to the left hemisphere (the hemisphere specialized for language),
particularly to the occipital-temporal junction.
The computational theory of vision

Since one of Marr’s motivations came from reflecting on what disorders
in visual processing revealed about the functional architecture of vision,
it is perhaps not surprising that a subpersonal analogue of the personal-
level distinction between conceptual and nonconceptual content seems to
be built into Marr’s pioneering computational theory of vision (Marr and
Nishihara 1978, Marr 1982).20
Marr’s account of vision describes the stages by which variations in
illumination are parsed to yield an image of objects in space, in terms of
82 Chapter 3
a series of representations, termed ‘sketches’, each of which is the product

of a limited series of specialized processes dedicated to discriminating dis-
tinct features of the available information. In the first stage, the primal
sketch, a scene is represented in terms of groupable edges, bars, and blobs
with a range of basic properties (orientation, contrast, length, width, and
position). In the second stage, cues for depth, such as stereopsis and con-
tour occlusion, are added to the primal sketch. In the third stage, a 3D
structural description of the object in object-centered space is computed
using a catalog of stored structural descriptions. The fourth stage involves
the imposition of a semantic interpretation upon the structured object.
For the theorist of nonconceptual content, the crucial point here is the
sharp separation between the third and fourth stages. The third stage
yields a parsing of the perceived array into spatially extended and spa-
tially related bounded individuals of the sort characteristic of representa-
tions with nonconceptual content. This is a stage carried out in complete
independence of the semantic and categorizing steps in the fourth stage.
Of course, the 3D sketch is part of the mechanism that subserves vision.
It is neither the object nor the content of vision.21 Nonetheless, Marr’s
key idea that the structural description of objects is computed indepen-
dently of the semantic classification and conceptualization of those ob-
jects lends support both to the distinction between conceptual and
nonconceptual content and to the Autonomy Principle.
4
The Theory of Nonconceptual Content
In chapter 3, I provided an existence proof of autonomous nonconceptual

content. If we are to do justice both to the differences and to the similari-
ties between infant and adult cognition then we will have to recognize the
existence of states that represent the world in a way that is independent
of concept mastery and, moreover, that can be ascribed to creatures who
possess no concepts whatsoever. What we now need is a closer under-
standing of the circumstances in which such states can be attributed, as
well as a more detailed specification of the form that such attributions
will take.
4.1 Attributing States with Autonomous Nonconceptual Content
On many philosophical construals, content is whatever is specified in a

‘that’ clause when attributing a propositional attitude. In attributing be-
lief, desire, fear, or any other propositional attitude to an individual, we
do three things: identify a particular person; attribute a particular atti-
tude to that person; and specify the proposition to which that attitude is
held. Specifying the proposition to which an attitude is held is equivalent
to giving the content of that attitude, i.e., what is believed, desired, or
feared.
If this familiar conception of propositional content is deemed the only
appropriate one, then it seems natural to resist the suggestion that there
might be forms of content that are nonconceptual in the sense discussed
in section 1.2, because propositional attitudes are paradigmatically
conceptual (for reasons that I brought out in chapter 3). To make any
84 Chapter 4
progress, then, we need an account of what it is for a state to have content

that will not rule out the notion of nonconceptual content as a matter of
definition. A promising start is provided by Peacocke’s (1992) suggestion
that a state has content if and only if it has a correctness condition. On
Peacocke’s view (the minimal account of content possession), a state pre-
sents the world as being in a certain way if and only if there is a condition
or set of conditions under which it does so correctly, and the content of
the state is given in terms of what it would be for it to present the world
correctly. This certainly leaves open the possibility of nonconceptual con-
tent, but it leaves several important questions unanswered.1
Suppose that we distinguish two different tasks that a general account
of content possession might be called upon to perform. The first (consti-
tutive) task would be to give an account of what the content of a given
state is, on the assumption that such a state has content. The second (cri-
terial ) task would be to explain how one might tell whether a state has
content at all. The minimal account certainly fulfils the constitutive task,
but it is not clear that it satisfies the criterial task. Correctness conditions
are conditions of some state of affairs being represented correctly. So be-
fore we can determine whether a state has a correctness condition, we
must determine whether it is a representational state.
Consider, for example, how the minimal account would deal with the
question of whether carrying information betokens the presence of con-
tent. As standardly construed (Dretske 1981), a state of affairs carries
information about another state of affairs if and only if there is nomologi-
cal covariance between the respective types of which they are tokens. Fa-
miliar examples are the rings of a tree, which carry information about the
age of the tree, and fossils, which carry information about the bone struc-
ture of extinct animals.
In both cases it would seem that a correctness condition can be pro-
vided. In the first example the correctness conditions would be ‘The rings
on the tree correctly indicate the age of the tree if and only if the number
of rings 5 the number of years the tree has been in existence’. But would
these be genuine correctness conditions? Many would think not. They
might argue as follows. No state could count as a representational state
unless it were possible for it to misrepresent the environment. But it is the
lawlike connection between, for example, the number of rings and the
The Theory of Nonconceptual Content 85
number of years that makes it plausible to speak of the former as carrying

information about the latter, and what makes it a lawlike connection is
the fact that the number of rings and the number of years invariably coin-
cide.2 Such invariable coincidence, however, clearly rules out the possibil-
ity of misrepresentation.3
In cases like these, then, the minimal account is not sufficient to settle
the question of whether we are dealing with genuine content or not. We
need to supplement it with an account of representation—of what it is
for a creature to be representing the environment. As is required by the
argument rejecting the Autonomy Principle (section 3.3), this account
will have to identify the conditions that must be satisfied for nonconcep-
tual content to be ascribed to non-concept-using creatures. On the other
hand, such an account should also be able to explain what is going on in
ascriptions of conceptual content. Conceptual and nonconceptual con-
tent are both forms of content because there is a single notion of repre-
sentation applicable to both of them. What is needed is an account of the
general conditions under which it makes theoretical sense to hold that
creatures are representing the environment.
What is at stake here is the appropriateness of distinct types of expla-
nation. Theorists have traditionally distinguished between two different
categories of explanation of behavior: mechanistic explanations and in-
tentional explanations (Taylor 1964). Intentional explanations have sev-
eral distinguishing characteristics. First, they are teleological. That is, they
explain an organism’s behavior in terms of the purposes and desires that
the behavior is intended to satisfy. Second, intentional explanations can-
not be eliminated in favour of nonintentional explanations. Third, inten-
tional explanations appeal to desires and purposes in conjunction with
nonmotivational representational states. Simply specifying a desire that
an organism could satisfy by performing a particular action cannot pro-
vide a satisfactory explanation. Two different types of representation
have to be involved: perceptual representations of the environment and
representations of how performing that action can satisfy the desire in
question. It is not the case, however, that there have to be two different
representations. Quite often a single perceptual representation of the en-
vironment will also illustrate how performing a given action will satisfy
the desire in question.4
86 Chapter 4
Explanations of behavior, particularly when dealing with the cognitive

abilities of nonlinguistic creatures, quite rightly operate with a principle
of parsimony. Appeals to representational states should be made only
where it is theoretically unavoidable, where there is no simpler mechanis-
tic explanation of the behavior ( pace Fodor [1987, Preface]). This raises
the question, of course, of how ‘behavior’ is being understood here. It is
natural to ask whether what is at stake is behavior narrowly construed
in terms of independently characterizable bodily movements or behavior
broadly construed in terms of bodily movements related in certain ways to
the environment. This is a complicated question, and I state my position
without argument. It seems to me that the distinction between narrow
and broad construals of behavior maps cleanly onto the distinction be-
tween mechanistic and intentional explanations of behavior. Mechanistic
explanations of behavior are appropriate when, and only when, the be-
havior to be explained can be specified in terms of nonrelational bodily
movements. Intentional explanations, in contrast, go hand in hand with
relationally characterized bodily movements. So, to return to the issue of
parsimony, this provides a further way of conceptualizing the difference
between behavior that needs to be explained through representational
states and behavior that does not. The latter, but not the former, can be
satisfactorily characterized nonrelationally.
As a first step towards building on this, consider the thesis that repre-
sentational states are intermediaries between sensory input and behav-
ioral output (theoretically) required to explain how behavioral output
emerges on the basis of sensory input. We can identify situations, such as
the operation of primitive evolutionarily hard-wired connections, where
the connections between sensory input and behavioral output do not de-
mand the attribution of representational states. Consider a creature ge-
netically programmed to respond in a certain way to a certain stimulus,
say to move away from the perceived direction of a particular stimulus.
One can explain why it made such a movement in any particular situation
without saying that it was representing the distal cause of the stimulus.
Registering the relevant stimulus causes the appropriate response, and
this can be fully understood, explained, and predicted without any appeal
to an intermediary between stimulus and response. By the same token,
the movement can be satisfactorily described in bodily terms. It does not
have to be described relationally as a movement away from whatever the

distal cause happens to be.
So for situations in which the connection between sensory input and
behavioural output are invariant in this way, we do not need to bring in
representational states. This does not just rule out primitive hard-wired
forms of reflex behavior. Consider the classic situation of hungry rats for
whom the sound of a tone regularly signals the delivery of food to a maga-
zine. Suppose (as is in fact the case) that after exposure to this set-up the
rats acquire an approach response—they approach the magazine during
the sound of the tone. There seem to be two ways of describing this be-
havior. The first and perhaps the most intuitively appealing would be to
describe the situation in representational terms. This would involve two
different representations: a representation of the goal (the food) and a
representation of the instrumental contingency between approaching the
magazine and attaining the goal. The goal is taken to be intrinsically mo-
tivating, and what drives the approach response is a comprehension that
the response will lead to the goal. Alternatively, a completely nonrepre-
sentational account might be given in which there is appeal simply to
lawlike connections, built up through associative conditioning, between
the rat hearing the sound and his approaching the magazine.
How can we decide between these two options? One way of doing so
discussed in the psychological literature (Holland 1979, Dickinson 1980)
would be to employ an omission schedule. An omission schedule inverts
the connection between the approach response and the reward. If the rat
approaches the food source during the sound of the tone, no food is deliv-
ered. There is no reason to think that this would interfere with the rat’s
initially learning the association between tone and food, because in the
initial trials it would not approach the magazine (there being no reason
for it to approach the magazine before it has learned the association).
Once it has learned the association, however, the rat will approach the
magazine in the familiar way. But by the omission schedule, the approach
will lead to the food being withheld. What will the rat do now? The
operative idea here is that if the rat’s response (approaching the maga-
zine) is really determined by a representation of the instrumental con-
tingency between approaching the magazine and gaining food, then the
response will not occur when the instrumental contingency ceases to
88 Chapter 4
hold, especially when it actually results in nonachievement of the goal.

Experiments on rats show that rats do indeed maintain the approach re-
sponse even in omission schedules (Holland 1979).5
There are two comments to be made here. The first is that the persis-
tence of the original behavior militates strongly against the idea that the
rat’s behavior is driven by a representation of the instrumental con-
tingency between performing the action and attaining the goal, since it
continues to perform the action even when the goal is not attained. The
second is that the invariance of the behavior is an integral part of its being
explicable solely in terms of lawlike connections between stimulus and
response built up by associative conditioning.
This is closely connected with the earlier point about the importance of
the connection between representation and misrepresentation. The reason
that tropistic and classically conditioned behavior can be explained with-
out reference to representational perceptual states is, as already stressed,
that the response is invariant if the creature in question registers the rele-
vant stimulus. What this means, of course, is that the behavior can be
explained by how things are in the immediate environment. There is,
therefore, no need to appeal to how things are taken to be (how they are
represented as being). The need to appeal to how things are taken to be
comes in only when the law-governed correlation between stimulus and
response breaks down. This can come about either when the response
occurs in the absence of the stimulus or when the stimulus occurs and is
registered without the response following. In the first of these cases one
might say that there is a representation of the stimulus and that this is
what generates the response, and in the second case one might say that
even though the stimulus is there, it is not represented so as to bring about
the appropriate response. Both of these are different types of misrepresen-
tation, and the existence or possibility of misrepresentation is a good cri-
terion for the necessity of bringing in representational perceptual states.
It will be helpful, I think, if I situate my position relative to the account
of “registration” developed by Jonathan Bennett in the detailed account
of teleological explanation of animal behavior that he develops in the
justly influential Linguistic Behaviour (Bennett 1976). The technical no-
tion of registration that Bennett develops has many features that make it
appear that when a creature registers something, it is ipso facto in a per-
ceptual state with representational content. According to Bennett, a crea-

ture c registers that p if c is in a sensory state that is sufficiently like some
p-operative state, where a p-operative state is one that implies that c’s
acting on p is noncoincidental (Bennett 1976, 56). This is considerably
weaker than the criteria that I favour for the attribution of representa-
tional states. The notion of registration is, for example, weak enough to
be involved in conditioned behavior.6 If, as a result of Pavlovian condi-
tioning, an animal salivates whenever it hears a tone, then its actions will
certainly be noncoincidental. But, although there is a case to be made for
viewing instrumental conditioning as involving representational states,
Pavlovian conditioning seems fundamentally different. What is significant
is that, although Bennett stresses that there is no entailment between p
and ‘c registers that p’, his notion of registration is perfectly compatible
with the existence of laws of the form ‘Whenever a c-like creature registers
that p, it will f ’.7 In such a case, then, Bennett’s account fails to distin-
guish between cases in which creatures react invariantly to a stimulus and
those in which they do not. I take it to be an advantage of my own ac-
count that it respects this important distinction.
Jerry Fodor has made some interesting suggestions about how the pos-
sibility of misrepresentation might be explained by the lack of a lawlike
correlation between stimulus and response (Fodor 1986, 14–16). Ex-
plaining why there should be an invariant connection between stimulus
and response requires postulating the existence of transducers (devices
that transfer sensory stimuli into physical output in a suitable form to
interact with the motor system). We need transducers to explain how an
organism can pick up stimuli at all, and when there is an invariant con-
nection between stimulus and response, we need to postulate the exis-
tence of transducers sensitive to specific properties (what are often known
as dedicated transducers). The existence and operation of these dedicated
transducers, hooked up in the appropriate ways to the motor system, is
all that we need to explain the invariant behavior in question. Now in the
case of intentional behaviour, which does not involve a lawlike connec-
tion between stimulus and response, it is clear that we are not dealing
with dedicated transducers. Creatures who are behaving intentionally are,
of course, sensitive to particular properties of the stimulus, but these are
not properties that can be detected by dedicated transducers. So how are
90 Chapter 4
they detected? One answer is that their presence is inferred from the pres-
ence of properties that can be picked up by transducers. The need for
such a process of inference, however, brings with it the possibility of error,
because it is inductive, not deductive, inference. Hence the possibility of
misrepresentation.
This connects with another strand in the notion of representation.
When we appeal to representational states to explain behavior, we rarely,
if ever, appeal to single states operating in isolation. The behavior of or-
ganisms suitably flexible and plastic in their responses to the environment
tends to be the result of complex interactions between internal states. As
Robert Van Gulick notes, “Any adequate specification of the role which
a given state or structure plays with respect to the system’s behaviour will
have to be in terms of the partial contributions which that state makes
to the determination of behaviour in conjunction with a wide variety of
internal state combinations” (1990, 113–114). There are several reasons
for this.
First, interaction between internal states is one possible way of ex-
plaining why there do not exist lawlike correlations between input and
output. It is a familiar point from the philosophy of perception that the
same state of affairs can be perceptually represented in different ways de-
termined by the influence of varying beliefs governing what one expects
to see or to what one has just seen.
Second, it seems right to draw a broad distinction within the general
class of representational states between “pure” representational states on
the one hand and motivational states on the other. In many situations in
which it is appropriate to appeal to representational states in explaining
behavior, a theorist will be compelled to appeal to instances of both types
of representational states. And, of course, it is not just the conjunction of
representational states and motivational states that is relevant, but their
interaction.
Third, representational states have to interact with other representa-
tional states. Organisms respond flexibly and plastically to their environ-
ments partly in virtue of the fact that their representational states respond
flexibly and plastically to each other, most obviously through the influ-
ence of stored representations on present representations. The possibility
of learning and adapting depends on past representations’ contributing to
the determination of present responses, and hence interacting with them.

Moreover, representing a particular feature of the environment does not
have effects on behavior just when a creature is confronted by that very
feature. It can also be effective when the creature is confronted by some-
thing that is relevantly similar to the represented feature, or when the
feature is absent and there is nothing relevantly similar in the environment
(as when a creature determines to leave its shelter because it cannot detect
either a predator or anything predatorlike in the vicinity).
Two separate constraints follow from this. First, there must be path-
ways enabling a given representational state to connect up with other
states, both representational and motivational. There must be cognitive
integration of the relevant states. This point is most familiar in the con-
text of propositional attitudes. Part of what it is, for example, to have a
belief is that it should be open to modification from newer, incompatible
beliefs, or that it should hook up, in appropriate cases, with desires in
order to bring about actions, or with fears in order to prevent action.
Second, part of what is involved in the integration of representational
states is that a creature representing the environment should be capable
of registering when the environment is relevantly similar over time. The
simplest example would be when the represented current environment is
identical in all respects to the represented previous environment. But the
appropriate sort of flexibility requires being able to register in what re-
spects there is a match in the environment as represented and in what
respects there is no such match, so that a suitable response can be deter-
mined by integrating the relevant match with previous experience and
current motivational states.
This means that representational states must be structured so that they
can be decomposed into their constituent elements, which can then be
recombined with the constituent elements of other representational
states. Suppose, for example, that a creature represents its environment
as containing food in an exposed place with a predator within striking
distance. If its ensuing behavior is not to be driven either by the associa-
tion food-eat or by the association predator-flee, the creature will need to
evaluate whether it can reach the food without the predator noticing,
whether the food is worth the risk, whether having gained the food it can
then escape to safety, etc. These primitive forms of inference require being
92 Chapter 4
able to distinguish the various elements in the original representation and

to integrate them with comparable elements in other representational
states.
This requirement of compositional structure is at the core of one im-
portant recent discussion of the conditions for ascribing conceptual
content. Evans (1982) understands thoughts as essentially structured,
composed of distinct conceptual abilities that are constitutively recombin-
able in indefinitely many distinct thoughts. This is the Generality Con-
straint, which he imposes on genuine concepts. The idea of concepts as
the essentially recombinable constituents of thought is familiar enough
not to require further discussion, falling naturally, as it does, out of the
parallelism widely assumed to hold between conceptual thought and lan-
guage. What is perhaps less clear is how this requirement of structure is
met by the nonconceptual perceptual states discussed in the previous
chapter. What can count as the recombinable elements of a perceptual
state? It seems to me that the key here is the thought that nonconceptual
perceptual experience nonetheless involves parsing the perceived array
into distinct elements. In the case of vision, this involves the ability to
parse the array of visual stimulation into spatially extended bodies. Plac-
ing this parsing ability in the context of perceptual recognitional abilities
that allow such spatially extended bodies to be reidentified in subsequent
experiences makes the structure of nonconceptual perceptual experiences
much clearer.
The requirement of compositional structure is closely related to the
generativity usually taken to be the defining feature of language. It is hard
to see how generativity could be explained without compositionality. This
is not to say, however, that wherever there is compositional structure,
there must be generativity. That would be to obscure the distinction be-
tween conceptual and nonconceptual content. Linguistic generativity
comes with the conjunction of compositional structure and global recom-
binability. It is not just that grammatical sentences can be decomposed
into their constituent elements. The crucial point is that those grammati-
cal elements can be recombined to yield infinitely many grammatically
well-formed new sentences. It is this feature of language that Evans’s Gen-
erality Constraint aims to capture at the level of thought. Theorists have
frequently held that structured content must be recombinable content.
What leaves space open for the possibility of nonconceptual content,
however, is that compositional structure can exist in the absence of global

recombinability. And this is exactly how we should understand noncon-
ceptual content.
For the reasons given above, nonconceptual thought must be structured
to allow recognition of partial similarities and to allow primitive forms
of inference. But the structural components of nonconceptual content
cannot be recombined at will. At the nonconceptual level there are only
limited behavior-driven and environment-constrained possibilities for
generating further nonconceptual contents out of a given set of noncon-
ceptual contents. It is plausible to think that the transition from noncon-
ceptual content to conceptual content is at least in part a matter of
moving from partially recombinable content constituents to globally re-
combinable content constituents. Part of the motivation for the Thought-
Language Principle is the thought that only through some form of syntax
can the possibility of global recombinability be explained. This may well
be true, but it clearly leaves conceptual space open for structured content
that has only limited combinatorial properties.
It should not be thought that a requirement of compositionality at the
level of content imposes a strong requirement of compositionality in the
vehicle of that content. Fodor and Pylyshyn (1988) have famously sug-
gested that when this compositionality requirement is properly under-
stood, it entails that the computational theory of cognition (CTC) is the
only viable cognitive architecture. They argue that the main alternative to
CTC, the parallel-distributed-processing (PDP) paradigm, is faced with
either being a mere implementation of CTC (if it does satisfy the composi-
tionality requirement) or failing to provide an adequate account of cogni-
tion (if it does not). The debate that this challenge has raised is enormous,
and this is not the place to discuss it in any detail. I have two brief points,
however. First, it does not seem to me that Fodor and Pylyshyn’s challenge
can be defused by the sort of instrumentalist moves made in Clark (1989).
At a minimal level, the compositionality requirement entails that repre-
sentational states have constituents that are causally efficacious, because
the only alternative is to make constituent structure epiphenomenal. The
vehicle of content does have to be structured. On the other hand, however,
and this is the second point, it is far from obvious that this minimal level
of structure ipso facto counts as an implementation of CTC. Smolensky
(1991) has argued that there is an important class of PDP representations
94 Chapter 4
(which he terms tensor product representations) that are compositionally

structured in a way that approximates rather than implements a CTC
architecture (see also Horgan and Tienson 1992). If he is right about this,
then this would provide good reasons for thinking that the composition-
ality requirement that flows from the nature of intentional explanation is
neutral with regard to cognitive architecture.8
The discussion in this section suggests that the following four criteria
need to be satisfied before given states can properly be described as repre-
sentational states.
• They should serve to explain behavior in situations where the connec-
tions between sensory input and behavioral output cannot be plotted in
a lawlike manner.
• They should admit of cognitive integration.
• They should be compositionally structured in such a way that their ele-
ments can be constituents of other representational states.
• They should permit the possibility of misrepresentation.
As has emerged at various points during the discussion of these four

marks of content, they are all satisfied paradigmatically by the conceptual
contents that are the objects of folk-psychological propositional attitudes.
On the other hand, it does not follow from them that conceptual proposi-
tional content is the only genuine form of content. This gives us a way of
determining the issue of nonconceptual content. When confronted with
states that are candidates for the ascription of nonconceptual content, we
can ask whether these states satisfy the four conditions of content and
thus qualify as representational states. If the four conditions are satisfied,
then this will show that the states in question are contentful in virtue of
features shared with the conceptual contents of folk psychology.
4.2 The Form of Autonomous Nonconceptual Content
As I have stressed, the type of autonomous nonconceptual content we are

interested in is perceptual. We need an understanding of perceptual con-
tent that will capture how the world is perceived by creatures who need
not have the conceptual apparatus needed to conceptualize what they per-
ceive. Given the irrelevance of conceptual mastery for the matter in hand,
standard accounts of content in terms of senses and other propositional

objects are clearly irrelevant. A radically different understanding of per-
ceptual content is needed.
The framework within which such an understanding can be developed
has been provided by Christopher Peacocke (1989, 1992). Peacocke de-
velops an account of the content of perceptual experience based on
the idea that a given perceptual content should be specified in terms of
all the ways in which space could be filled out around the perceiver that
are consistent with that content being a correct representation of the
environment:
The idea is that specifying the content of a perceptual experience involves saying
what ways of filling out a space around the origin with surfaces, solids, textures,
light and so forth, are consistent with the correctness or veridicality of the experi-
ence. Such contents are not built from propositions, concepts, senses, or continu-
ant material objects. (Peacocke 1989, 8–9)
In specifying such a content, we need to start by organizing the space
around the perceiver. This is done by specifying origins and correspond-
ing axes based on the perceiver’s body. There will be a range of such ori-
gins, and they will vary according to the modality in question. In the case
of touch, for example, the possibilities include origins at the center of the
palm of each hand, as well as certain other origins that will capture the
sensitivity to touch of the remainder of the body surface. Once we have
selected a particular origin and defined a set of axes in terms of them, we
can identify points (or rather point types) in terms of those axes. This
provides a framework within which the account of how space is “filled
out” can be completed as follows:
For each point (strictly, I should say point-type), identified by its distance and
direction from the origin, we need to specify whether there is a surface there and,
if so, what texture, hue, saturation and brightness it has at that point, together
with its degree of solidity. The orientation of the surface must be included. So
must much more in the visual case: the direction, intensity, and character of light
sources; the rate of change of perceptible properties, including location; indeed,
it should include second differentials with respect to time where these prove to be
perceptible. (Peacocke 1992, 63)
The conceptual resources with which this specification is carried out are
not, of course, attributed to the perceiver. This is what makes it an ac-
count of nonconceptual content.
96 Chapter 4
The account is completed as follows. Selecting an origin and axis and

specifying the relevant values for each of the points in the perceivable
space around that origin determines a spatial type, which Peacocke terms
a scenario. This spatial type determines a class of ways of filling out the
space, any of which could be instantiated in the real world. The next step
is to assign to the origin and axes real places and directions in the world,
and to assign a time. This yields a content that Peacocke terms a posi-
tioned scenario. This positioned scenario is correct if and only if the vol-
ume of perceptually discriminable space around the perceiver at a
specified time instantiates the spatial type specified by the scenario.
It is straightforward how this notion of scenario content helps with the
problem with which I began this chapter. Some aspects of what I de-
scribed as the fine-grainedness of perceptual experience can be accommo-
dated without difficulty. This is particularly clear with regard to color
perception. Specifying the precise color and shade of particular points on
a surface is an integral element in specifying the scenario content for any
creature sensitive to color. Once this is done, justice will have been done
to the idea that a creature can perceive shades of color that it cannot
conceptualize. What is not yet clear, however, is how this notion of sce-
nario content can help us with the far more complicated idea that an
account of the contents of infants’ perception should reflect their percep-
tion of object permanence, which they cannot conceptualize.
To make progress here, we need to consider a further level of noncon-
ceptual representational content that Peacocke introduces. Peacocke
notes that it is possible for two experiences to represent the world in dif-
ferent ways at the nonconceptual level, despite having identical scenario
contents. He gives the familiar example of perceiving an object that is
in fact square as a square, as opposed to perceiving it as a diamond.
Whether the object is perceived as a square or as a diamond, the scenario
content is the same (because the same values hold at each of the points
identified by their distance and direction from the axes). But there re-
mains a significant phenomenological difference. One way of capturing
this difference is in terms of what symmetries are perceived. Perceiving
something as a square involves perceiving symmetries about the bisectors
of its four sides. If the same figure is perceived, but the perceived symme-
try is about the bisectors of its four angles, then it will be perceived as a
diamond. One way of dealing with this difference would be at the level
of conceptual content, by invoking the perceiver’s possession of some or

other of the relevant concepts, such as the concept of a square, the con-
cept of a diamond, or the concept of something being symmetrical. On
the other hand, however, it does seem that the difference between perceiv-
ing an object as a square and perceiving it as a diamond is available in
the absence of such concept mastery. Moreover, it is desirable to be able
to appeal to it in giving a possession-conditions explanation of what such
conceptual mastery consists in. Accordingly, Peacocke offers a further
level of nonconceptual representational content that he terms protoprop-
ositional content:
The contents at this second layer cannot be identified with positioned scenarios,
but they are also distinct from conceptual contents. These additional contents I
call protopropositions. These protopropositions are assessable as true or false. A
protoproposition contains an individual or individuals, together with a property
or relation. When a protoproposition is part of the representational content of
an experience, the experience represents the property or relation in the protoprop-
osition as holding of the individual or individuals it also contains. (Peacocke
1992, 77)
Just as scenario content is built up out of real space-time points, proto-
propositional content is built up out of real individuals and real proper-
ties. Again, this qualifies as nonconceptual representational content
because the perceiver need not have any conceptual grasp of those individ-
uals and properties.
It is, I shall suggest, at the level of protopropositional content that we
can capture the phenomenology of object permanence. Before going into
this, however, it is important to stress that this is an employment of Pea-
cocke’s theoretical machinery that he himself is not prepared to counte-
nance. Peacocke opposes the Autonomy Principle and does not believe
that either scenario or protopropositional content can be ascribed to crea-
tures who lack the conceptual skills necessary to generate a rudimentary
conception of an objective world. I shall not take issue with his argument
here. He claims that a primitive form of the first-person concept must be
possessed by any creature capable of being in states with scenario content.
My own position on this is that the work that Peacocke thinks can only
be performed by the first-person concept can in fact be done by the first-
person nonconceptual contents that I shall be discussing in the following
chapters. At the moment I can only issue a promissory note (see section
8.2 below).
98 Chapter 4
One point that does need to be made with regard to Peacocke’s own
development of the notion of protopropositional content is that it con-
tains a crucial ambiguity that might well play into the hands of the
defender of autonomous nonconceptual content. Protopropositional
content, as we have seen, contains an individual or set of individuals,
together with one or more property or relation. It is natural to ask what
these individuals are, since they are obviously not the space-time points
picked out in the positioned scenario. According to Peacocke, “These
properties and relations can be represented as holding of places, lines, or
regions in the positioned scenario, or of objects perceived as located in
such places” (Peacocke 1992, 77). This raises a very important issue. If a
property is being represented as holding of an object in the positioned
scenario, then it is clear that the object itself is being represented. But the
account of scenario content so far given does not provide any clues as to
what this might amount to. Even though a positioned scenario might con-
tain objects, that fact is not manifest in the scenario content, because
scenario content is specified in terms of points. So, whatever basis there
might be for thinking that objects are represented as such in the proto-
propositional content of a given experience (for thinking, for example,
that that object is being represented as a square rather than as a dia-
mond), it cannot come from the scenario content.
One possible position here (which may be Peacocke’s own) would hold
that an object is perceived as such in virtue of certain forms of conceptual
mastery. It is the conceptual content of the experience in question that
permits objects to be represented as objects, thereby making possible the
relevant protopropositional content. Thus, to continue with Peacocke’s
own example, when a subject perceives a particular object as a diamond
rather than a square, that is a fact about that subject’s experience that
needs to be explained at the level of protopropositional content. The ex-
planation takes the following form: the perceiver’s experience represents
the object in question as instantiating the property of being symmetrical
about the bisectors of its four angles (rather than the property of being
symmetrical about the bisectors of its four sides). This is an explanation
at the level of nonconceptual content because a perceiver’s experience can
take this form without the perceiver having to possess all or any of the
concepts symmetrical, bisector, or angle (or for that matter, diamond ).
Nonetheless, the perceiver’s experience represents the object in question

as instantiating those properties, and on the current view, that is only
possible if the perceiver possesses at least the concept of an object in gen-
eral. So, although to perceive an object as instantiating various properties
(like being diamond-shaped) there is no need for the perceiver to have a
conceptual grasp of those properties, the perceiver must at least have the
concept of an object.
The position just sketched out is perfectly consistent, but it is not com-
pulsory. Indeed, it seems prima facie incompatible with the possibility of
autonomous nonconceptual content at the protopropositional level, since
it makes the possibility of an important class of protopropositional con-
tents depend on the perceiver’s possession of at least one concept. To that
extent, therefore, the arguments in the previous chapter defending the
notion of autonomous nonconceptual content clearly count against it.
The question arises, therefore, of how the defender of autonomous non-
conceptual content can accommodate the possibility of protopropo-
sitional content. The obvious alternative is simply to deny the general
principle that objects can only feature in the content of perception if the
perceiver possesses the concept of an object in general. However, consid-
erable caution is required here. The denial of the general principle can be
developed in both a narrow and a broad manner. Let me go back briefly
to Peacocke’s characterization of protopropositional content:
A protoproposition contains an individual or individuals, together with a prop-
erty or relation. When a protoproposition is part of the representational content
of an experience, the experience represents the property or relation in the proto-
proposition as holding of the individual or individuals it also contains. . . . These
properties or relations can be represented as holding of places, lines or regions in
the positioned scenario, or of objects perceived as located in such places. (Pea-
cocke 1992, 77)
Suppose we ask, what different types of thing can occupy the subject posi-
tion in a protoproposition? In the final sentence quoted, Peacocke offers
four candidates, namely, places, lines, spatial regions, and perceived ob-
jects. Now, after rejecting the general principle that the perception of ob-
jects depends on possession of the concept of an object, one way to
proceed would be to take Peacocke’s four candidates as exhausting the
field. On this (narrow) construal, there can be no protopropositional indi-
viduals that are neither places, lines, spatial regions, or perceived objects.
100 Chapter 4
The discussion of infant cognition in the previous chapter provides

clear reasons for thinking that the narrow construal cannot be correct.
As I noted there, infants from an early age seem to represent the world as
divided into bounded segments. However, the basic perceptual capacity
to parse the perceived array into bounded segments is not in itself equiva-
lent to the perception of objects, even at the nonconceptual level. The
extent to which infants (and, of course, prelinguistic and nonlinguistic
creatures in general) can appropriately be described as perceiving objects
is determined by the extent of their perceptual sensitivity to what I earlier
termed object properties, that is, those properties that constitutively mark
out individuals as objects. There are levels of cognitive development at
which there is insufficient perceptual sensitivity to object properties for it
to be appropriate to describe perceivers as perceiving objects. This can be
marked in two ways. First and most straightforward, there are perceptu-
ally discriminable features that are clearly object properties but to which
infants at relatively early stages of development fail to be perceptually
sensitive. For example, it seems to be the case that neonates do not per-
ceive continuously moving center-occluded objects as a single individual
with a hidden part (Slater et al. 1990). Second, infants can be perceptually
sensitive to properties that are not object properties and that in fact divide
up the perceived array in a way that does not map at all cleanly onto how
the perceived real-world situation is composed of objects. For example,
it is well documented that three-month-old infants parse the perceived
array in accordance with the principle that single individuals are spatially
connected bodies that retain their connectedness when they move. But
this often leads them to take to be a single object what is really a
composite.
Since, however, even in such cases the requirements of inference to the
best explanation will often require the ascription of states with noncon-
ceptual content to such perceivers and such states will often have proto-
propositional content, it follows that there must be protopropositional
individuals to which properties and relations can be ascribed other than
objects (and lines, places, and regions of space). Some technical vocabu-
lary is required to specify these protopropositional individuals. Let me
introduce the notion of an object*. An object* is defined as a bounded
segment of the perceived array. The notion of an object* is intended to
capture minimal differentiation of the perceived array of the sort that is
to be found in early infancy. By saying that objects* feature as individuals

in the protopropositional content of perceived experience, I mean to dis-
tinguish them from protopropositional properties and relations. An ob-
ject* is a unified and coherent segment of the perceived array that can be
perceived as having certain properties and as standing in certain relations
to other objects*. Some of those properties will be what I have termed
object properties (such as the property of having a determinate shape),
and others not (such as the property of having a particular color). Indeed,
it is plausible to hold that for a given sense modality, certain object prop-
erties have to be perceived if a perceiver is to perceive an object* in that
modality. In the visual modality, for example, it seems clear that objects*
will have the object property of possessing a determinate shape. In the
tactile modality, objects* will have to possess the dynamic object property
of resistance to touch.
With this notion of an object*, Peacocke’s account of scenario and pro-
topropositional content can be developed so as to accommodate the pos-
sibility of autonomous nonconceptual content, and in particular so as to
allow us to specify how the infants discussed in the previous chapter are
perceiving the world. The first component will be a specification, formu-
lated in terms of scenario content, of how the very general contours of
the environment are perceived. This will specify what colors are seen
where, for example. The second component is the novel one. It uses the
notion of objects* to specify how the perceived array is parsed into spa-
tially extended and bounded segments, and then specifies the properties
that those spatially extended and bounded segments are perceived as hav-
ing, as well as the relations that are perceived as holding between them.
By now the form that a specification of autonomous nonconceptual
content should take should be clear. The modification that I have pro-
posed to Peacocke’s theory of protopropositional content allows us to do
justice to the powerful evidence that the best explanation of the behavior
of infants and other nonlinguistic and prelinguistic creatures requires at-
tributing to them states with nonconceptual content, even though they
are not properly described as perceiving objects. We now have a theoreti-
cal framework in play that should overcome resistance to the idea of
autonomous nonconceptual contents. The next stage is to employ this
theoretical framework to avoid the paradox of self-consciousness. This
project begins in the following chapter.
5
The Self of Ecological Optics
The ground is now clear for a start to be made on defusing the paradox
of self-consciousness. I suggested in the last two chapters that the key to
defusing the paradox lies in primitive forms of self-consciousness with
nonconceptual first-person contents, and most of those chapters was
taken up with establishing the legitimacy of the general notion of noncon-
ceptual content. The next stage is to illuminate what these primitive forms
of self-consciousness are and to show how the theory of nonconceptual
content can be brought to bear on them. In this chapter I discuss some of
the distinctive claims of the theory of ecological optics developed by J. J.
Gibson, to see how they might serve as the basis for a form of primitive
self-consciousness.
5.1 Self-Specifying Information in the Field of Vision
It is open to a sceptic about the possibility of nonconceptual first-person

contents to concede everything said in the previous two chapters about
the legitimacy of a theory of autonomous nonconceptual content while
objecting in principle to the idea that there are forms of primitive self-
consciousness that such a theory can help us to understand. The whole
ˆ
raison d’etre of using the theory of nonconceptual content to explain in-
fant understanding of object permanence is provided by the considerable
experimental evidence that shows that infants are capable of parsing the
visual array into bounded segments. But, a sceptic might argue, even if
there is such a thing as primitive self-consciousness, it cannot be under-
stood on these lines for the simple reason that the self is not phenomeno-
logically salient in a way that demands to be incorporated into an account
104 Chapter 5
of the content of perceptual experience. The theory of nonconceptual

content is a theory of what is perceived, whereas the self is not something
that can be perceived—so how can there be nonconceptual first-person
contents? If the self is not perceived, then how can it feature in the content
of perception?
It is important to distinguish the claim at the heart of this objection
from Hume’s familiar claim that the self is not directly encountered in
introspection. As he famously puts it in the Treatise,
For my part, when I enter most intimately into what I call myself, I always stumble
on some particular perception or other, of heat or cold, light or shade, love or
hatred, pain or pleasure. I never catch myself at any time without a perception,
and can never observe anything but the perception. (Hume 1739–1740/1978,
252)
What Hume is worried about is his introspective failure to encounter the
putative owner of his sensations. The sceptical worry expressed in the
previous paragraph is related but importantly different. It is a worry
about the contents of perception. Whereas Hume is noting that he cannot
find the self in which his perceptions are supposed to inhere, the worry at
stake here is about the contents of those perceptions. The self, it is being
suggested, cannot appear in the content of any ordinary, outwardly di-
rected perception (I leave aside the question of whether it can feature as
the owner of such outwardly directed perceptions in the content of any
inwardly directed introspective perception).
It is not to Hume that we must turn to find the view under discussion,
but to Schopenhauer, and after him to Wittgenstein. Here’s a passage
from Schopenhauer’s well-known essay “On Death and Its Relation to
the Indestructibility of Our Inner Nature”:
But the I or the ego is the dark point in consciousness, just as on the retina the
precise point of entry of the optic nerve is blind, and the brain itself is wholly
insensible, the body of the sun is dark, and the eye sees everything except itself.
Our faculty of knowledge is directed entirely outwards in accordance with the
fact that it is the product of a brain-function that has arisen for the purpose of
mere self-maintenance, and hence for the search for nourishment and the seizing
of prey. (Schopenhauer 1844/1966, 2: 491)
A similar theme occurs during the discussion of solipsism in Wittgen-
stein’s Tractatus Logico-philosophicus, which many commentators have
held to be directly inspired by Schopenhauer:
The Self of Ecological Optics 105
5.633 Where in the world is a metaphysical subject to be found?

You will say that this is exactly like the case of the eye and the visual field. But
really you do not see the eye.
And nothing in the visual field allows you to infer that it is seen by an eye.
(Wittgenstein 1921/1961, 57)
According to this view, the self or subject cannot appear in the content of
perceptual experience, simply because the self or subject is never per-
ceived. In Wittgenstein’s useful phrase, the self does not feature in the
visual field, because it is not in the world but limits it. Both Schopenhauer
and Wittgenstein deployed this thought about the perceptual invisibility
of the subject to rather different and idiosyncratic ends, into which there
is no need to digress. For present purposes, we need only note that they
provide a very clear statement of the view that the self cannot be per-
ceived. This view is strong prima facie support for the objection that there
is no place for the self in the content of perceptual experience. In the
following, I will term the conjunction of the view that the self cannot be
perceived with the view that there is no place for the self in the content
of perceptual experience as the Schopenhauer/Wittgenstein view. Obvi-
ously, I appeal to these two writers purely for illustrative purposes. The
Schopenhauer/Wittgenstein view is suggested by the passages I have
quoted, but it matters little whether either of them actually held it.
This is where the relevance of Gibson’s theory of ecological optics ini-
tially emerges. It is relevant in several ways, the first of which is the crudest
but also the easiest to grasp. Gibson’s concept of a field of view is, I take
it, more or less the same as Wittgenstein’s notion of the visual field. It is
the solid angle of light that the eyes can register. This is not the same as
the images that appear on the two retinas, and nor is it to be understood
as a melange of visual sensations. It is the panorama that reveals itself to
the two eyes working together. Gibson stresses certain peculiarities of the
phenomenology of the field of vision. Notable among these is the fact
that the field of vision is bounded. Vision reveals only a portion of the
panorama to the perceiver at any given time (roughly half in the human
case, because of the frontal position of the eyes). The fact that the field of
view is bounded is itself phenomenologically salient. The boundedness of
the field of vision is part of what is seen, and the field of vision is bounded
in a way quite unlike how spaces are bounded within the field of vision:
106 Chapter 5
The edges of the field of view hide the environment behind them, as those of a
window do, and when the field moves there is an accretion of optical structure at
the leading edge with deletion of structure at the trailing edge, as in the cabin of
a steam shovel with a wide front window and controls that enable the operator
to turn the cabin to the right or the left. But the edges of the field of view are
unlike the edges of a window inasmuch as, for the window, a foreground hides
the background whereas, for the field of view, the head of the observer hides the
background. (J. J. Gibson 1979, 112)
He continues, evocatively,
Ask yourself what it is that you see hiding the surroundings as you look out upon
the world—not darkness surely, not air, not nothing, but the ego! (J. J. Gibson
1979, 112)
From a Gibsonian perspective, then, the Schopenhauer/Wittgenstein view
misreads the data. Gibson agrees with Wittgenstein that the self is the
limit of the visual field, or field of view, but he thinks that it is precisely
in virtue of this that the self features in the content of perception. The
self appears in perception as the boundary of the visual field, a moveable
boundary that is responsive to the will.
The boundedness of the visual field is not the only way in which the
self becomes manifest in visual perception, according to Gibson. The field
of vision contains other objects that hide, or occlude, the environment.
These objects are, of course, various parts of the body. The nose is a par-
ticularly obvious example, so distinctively present in just about every vis-
ual experience:
The nose is here. It projects the largest possible visual solid angle in the optic
array. Not only that, it provides the maximum of crossed double imagery or
crossed disparity in the dual array, for it is the furthest possible edge to the right
in the left eye’s field of view and the furthest possible edge to the left in the right
eye’s field of view. This also says that to look at the nose one must converge the
two eyes maximally. Finally, the so-called motion parallax of the nose is an abso-
lute maximum, which is to say that, of all the occluding edges in the world, the
edge of the nose sweeps across the surfaces behind it at the greatest rate whenever
the observer moves or turns his head. (J. J. Gibson 1979, 117)
The cheekbones and perhaps the eyebrows occupy a slightly less domi-
nant position in the field of vision. And so too, to a still lesser extent, do
the bodily extremities: hands, arms, feet, and legs. They protrude into the
field of vision from below in a way that occludes the environment and yet
differs from the way in which one nonbodily physical object in the field
of vision might occlude another. They are, as Gibson points out, quite
peculiar objects. All objects, bodily and nonbodily, can present a range of
solid angles in the field of vision (where by a solid angle is meant an angle
with its apex at the eye and its base at some perceived object), and the
size of those angles will, of course, vary according to the distance of
the object from the point of observation. The further away the object is,
the smaller the angle will be. This gives rise to a clear and phenomeno-
logically very salient difference between bodily and nonbodily physical
objects:
The visual solid angle of the hand cannot be reduced below a certain minimum;
the visual solid angle of a detached object like a ball can be made very small by
throwing it. These ranges of magnification and minification between limits link
up the extremes of here and out there, the body and the world, and constitute
another bridge between the subjective and the objective. (J. J. Gibson 1979, 121)
Of course, the further away a particular body part is from the point of
observation and the more moveable it is, the greater the range of solid
angles will be. This leads Gibson to another provocative conclusion:
Information exists in the normal ambient array, therefore, to specify the nearness
of the parts of the self to the point of observation—first the head, then the body,
the limbs and the extremities. The experience of a central self in the head and a
peripheral self in the body is not therefore a mysterious intuition or a philosophi-
cal abstraction but has a basis in optical information. (J. J. Gibson 1979, 114)
Perceived body parts are, according to Gibson, “subjective objects” in the
content of visual perception.
At a crude level, then, Gibson’s theory of ecological optics offers an
account of the phenomenology of visual experience that denies the Scho-
penhauer/Wittgenstein claim that the self is not, and cannot be, per-
ceived. Of course, no defender of the Schopenhauer/Wittgenstein view
will accept the lessons that Gibson draws from the phenomenology, for
obvious reasons. All Gibson’s points about the boundedness of the field
of vision and the peculiarities of bodily surfaces can be accepted without
accepting his gloss in terms of the ego being what blocks out the unper-
ceived hemisphere. What is clear, however, is that Gibson’s account of the
phenomenology of visual experience cannot be ignored. The way in which
he describes the situation is accurate and important (and it is remarkable
that the points he makes should have had to wait so long to come into
108 Chapter 5
the open). But once this is recognized the way is open for a less crude
deployment of ecological optics against the Schopenhauer/Wittgenstein
position.
The crucial claim in the Schopenhauer/Wittgenstein position is that be-
cause the self is not directly perceived, there is no place for the self in
accounts of the content of perceptual experience. It is this inference,
rather than the initial premise, that the theory of ecological optics pro-
vides us with the material to challenge. The crucial thought is that even
if it is conceded, pace Gibson, that the self is not directly perceived, it is
still the case that the self has a place in the content of perceptual experi-
ence in virtue of the self-specifying information that is an integral part of
that perceptual experience. Before going into more detail here, though, a
little more background about ecological optics is required.
Perhaps the most basic notion in ecological optics is the notion of a
perceptual invariant. One of Gibson’s central complaints against tradi-
tional theories of perception is that they fail to accommodate the fact that
perception is an active process that involves movement and takes place
over time. Whereas traditional theories have tried to understand vision by
simplifying it, Gibson’s starting point is not the laboratory but the real
environment. What happens when an animal moves around the world?
What does it see? The starting point is obvious. The animal sees a huge
number of surfaces illuminated from a range of different directions. Even
if there is just a single source of light, the light from that source will illu-
minate all the surfaces from different directions, depending on the path
that the animal takes. Complicated enough at any given moment, the
array of these illuminated surfaces obviously changes as the animal moves
around. How can the animal make sense of this confusion of surfaces
under illumination? How can a constantly changing pattern of stimula-
tion yield constant perceptions?
Traditional empiricist theories of perception resort to constructive and
serial accounts in terms of information processing and stored memories,
starting from information on the retina and building up to a three-
dimensional representation of the world (Marr 1982). Gibson, in con-
trast, thinks that the relevant information is already there in the field of
vision. Gibson’s approach to perception shifts the emphasis from the reti-
nal image to the changing patterns in the optic array that we experience
as we move around the world.1 He stresses that the illuminated surfaces

in the field of vision do not change completely at random. Consider the
relatively straightforward example of moving toward a wall and looking
ahead of one to the point of impact. As one approaches the wall, the array
of illuminated surfaces will obviously change. But there is a certain order
to the change. The part of the wall at which one is looking remains sta-
tionary, although, of course, the magnification of the solid visual angle
will accelerate drastically as the wall is approached (the phenomenon of
looming). But around the stationary part of the wall there are textured
surfaces radiating outward in what Gibson terms patterns of optic flow.
These textured surfaces expand in a lawlike manner as one approaches
them (and, obviously, contract if and when they pass beyond the head).
Both the immobility of the part of the wall where one will impact and the
patterns of optic flow are perceptual invariants. They are higher-order
patterns that remain constant during change in the field of vision brought
about by the perceiver’s movement, the movement of the environment, or
a combination of the two. The immobility of the wall is an example of
a structural invariant, while the changing patterns in the optic flow are
transformational invariants. According to Gibson, it is the underlying in-
variant structure of the field of vision that allows perceptual order to
emerge from what should, by all rights, be complete chaos.2
This sketchy background should be enough to allow the points relevant
to the current discussion to emerge. The first point to make is that all
the phenomenological points mentioned earlier can be glossed in a less
provocative manner. What was formerly described as the self being di-
rectly perceived can be reinterpreted in terms of the existence of self-
specifying structural invariants in the field of vision.3 Whether or not it is
correct straightforwardly to deny the Schopenhauer/Wittgenstein view
on the basis of the ecological analysis of what is perceived, this analysis
does show beyond reasonable doubt that there is self-specifying infor-
mation available in the field of vision. The outline and contours of the
body impose a high-order invariant structure on the field of vision, which
will vary, of course, across individuals as well as across species. But this
is only a fraction of the self-specifying information available in visual
perception. There are two more important types of self-specifying
information.
110 Chapter 5
The ecological emphasis on explaining the perceptions of creatures as

they move about in the world brings out the first type of self-specifying
information. The very fact that motion through the environment is per-
ceptually controlled poses a very significant problem for psychologists, as
this passage brings out:
Drivers of cars see where they are going, if they pay attention. Viewers of a Ciner-
ama screen see where they are going in the represented environment. A bee that
lands on a flower must see where it is going. And all of them at the same time see
the layout of the environment through which they are going. This is a fact with
extremely radical implications for psychology, for it is difficult to understand how
a train of signals coming in over the optic nerve could explain it. How could
signals have two meanings at once, a subjective meaning and an objective one?
How could signals yield an experience of self-movement and an experience of the
external world at the same time? How could visual motion sensations get con-
verted into a stationary environment and a moving self? (J. J. Gibson 1979, 183)
The mass of constantly changing visual information generated by the sub-
ject’s motion poses an immense challenge to the perceptual systems. How
can the visual experiences generated by motion be decoded so that sub-
jects perceive that they are moving through the world? Gibson’s notion of
visual kinesthesis is his answer to this traditional problem. Whereas many
theorists have assumed that motion perception can only be explained by
the hypothesis of mechanisms that parse cues in the neutral sensations
into information about movement and information about static objects,
the crucial idea behind visual kinesthesis is that the patterns of flow in
the optic array and the relations between the variant and invariant fea-
tures make available information about the movement of the perceiver, as
well as about the environment:
Vision is kinesthetic in that it registers movements of the body just as much as
does the muscle-joint-skin system and the inner ear system. Vision picks up both
movements of the whole body relative to the ground and movement of a member
of the body relative to the whole. Visual kinesthesis goes along with muscular
kinesthesis. The doctrine that vision is exteroceptive, that it obtains “external”
information only, is simply false. Vision obtains information about both the envi-
ronment and the self. (J. J. Gibson 1979, 183)
As an example of such a visually kinesthetic invariant, consider the opti-
cal flow in any field of vision when the perceiver is moving. The optical
flow starts from a center, which is itself stationary. This stationary center
specifies the point that is being approached. That is, the aiming point of
locomotion is at the vanishing point of optical flow. In combination with

the phenomenon of looming mentioned earlier, this is a powerful source
of information about the direction of movement.
A good way of appreciating the significance of the claim that the move-
ment of the perceiver is directly perceived is by distinguishing it from the
common explanation of the phenomena in question in terms of feedback
mechanisms. In cases of voluntary movement, the operation of efference
copy ensures that information about the motor command to move the
relevant limbs, sent to the motor center from the higher centers of the
brain, is processed by the perceptual centers in the brain, and then well-
known proprioceptive mechanisms provide feedback on the motor action
actually being performed.4 When further information reaches those per-
ceptual centers indicating movement in the field of vision, the calibration
of that information with the efference copy and proprioception allows
those changes in (reafferent) input that are due to the perceiver’s own
movement to be separated out from those changes in input due to the
movement of the environment. On this view, the information that the per-
ceiver is moving is not actually present in the field of vision. Visual input
just specifies movement, and the movement due to the perceiver’s own
movement is separated out and compensated for on the basis of nonvis-
ual information.
The most obvious difficulty with this as a complete account of the dis-
crimination of self-motion is that it cannot explain the perception of pas-
sive movement (when the perceiver moves but is not responsible for this
movement). Striking experiments have brought out the significance of
passive movement. In the so-called “moving room” experiments, subjects
are placed on the solid floors of rooms whose walls and ceilings can be
made to glide over a solid and immoveable floor (Lishman and Lee 1973).
If experimental subjects are prevented from seeing their feet and the floor
is hidden, then moving the walls backwards and forwards on the sagittal
plane creates in the subjects the illusion that they are moving back and
forth. This provides strong support for the thesis that the movement of
the perceiver can be detected purely visually, since visual specification of
movement seems to be all that is available. An even more striking illustra-
tion emerges when young children are placed in the moving room, be-
cause they actually sway and lose their balance (Lee and Aronson 1974).
112 Chapter 5
Adults will not normally lose their balance in the moving-room experi-
ments. But they do lose their balance if they are in unfamiliar postures,
like standing on a beam for example. The optical flow yields the informa-
tion that they (the subjects) are moving, and when they compensate for
this apparent movement they fall over.
The significance of visual kinesthesis is that information specifying the
movement of the perceiver is present in visual perception. This is, of
course, self-specifying information, and as such it means that the self has
a place in the content of visual experience. Of course, much remains to
be said about how the self can feature in visual contents. Before going
into this, however, I should mention that there is a further important form
of self-specifying information available in the field of vision, according to
the theory of ecological optics. This is due to the direct perception of a
class of higher-order invariants that Gibson terms affordances. It is in the
theory of affordances that we find the most sustained development of
the ecological view that the fundamentals of perceptual experience are
dictated by the organism’s need to navigate and act in its environment,
that the organism and the environment are complementary. The uncon-
troversial premise from which the theory of affordances starts is that ob-
jects and surfaces in the environment have properties that are relevant to
the abilities of particular animals, that allow different animals to act and
react in different ways:
If a terrestrial surface is nearly horizontal (instead of slanted), nearly flat (instead
of convex or concave), and sufficiently extended (relative to the size of the animal)
and if its substance is rigid (relative to the weight of the animal), then the surface
affords support. It is a surface of support, and we call it a substratum, ground or
floor. It is stand-on-able, permitting an upright position for quadrupeds and bi-
peds. It is therefore walk-on-able and run-over-able. It is not sink-into-able like a
surface of water or a swamp, that is, not for heavy terrestrial animals. Support
for water bugs is different. (J. J. Gibson 1979, 127)
This much is indisputable. The bold and interesting claim that Gibson
then goes on to develop is that affordances like these can be directly per-
ceived. Information specifying affordances is available in the structure of
light to be picked up by the creature as it moves around in the world.
The possibilities that the environment affords are not learned through
experience, nor are they inferred. They are directly perceived as higher-
order invariants. And, of course, the perception of affordances is a form
of self-perception, or at least a way in which self-specifying information

is perceived. The whole notion of an affordance is that of environmental
information about one’s own possibilities for action and reaction.
An important point to bear in mind is that affordances are properties
that objects and surfaces have relative to the organisms that perceive
them, or at least could perceive them; they are not properties of the per-
ceiving organisms.5 They are to be sharply distinguished from what psy-
chologists often refer to as valences, the subjective properties of pleasure
and pain that can come to be associated with supposedly neutral percep-
tions of objects and surfaces. According to the ecological theory, there are
no such neutral perceptions. The meanings and values that objects and
surfaces have for perceivers are directly perceived, not learned by associa-
tion. This emerges particularly clearly in the striking visual-cliff experi-
ments carried out by Eleanor Gibson with a range of collaborators (E. J.
Gibson 1969).
The visual-cliff experiments were done with a glass floor strong enough
to support the weight of the animals and young infants placed on it. One
half of the glass floor was backed with an opaque surface, so that it could
not be seen through. A similar opaque surface was placed some distance
below the other half of the glass floor. Thus a clifflike effect was created,
with a sharp divide between a “shallow” half and a “deep” half. A range
of animals were placed on the glass floor (including infant rats and snow-
leopard cubs). They all showed a preference for the shallow side over the
deep side. Infants and young children were also placed in the experimen-
tal situation. When young children capable of walking or crawling were
placed on the glass floor, they crawled and walked normally when on the
shallow side, but froze and showed distress on the deep side. When very
young infants of 6 weeks or so were placed on the floor, their cardiac
responses showed that they too were distressed.
The crucial point about the visual-cliff experiments is that visual infor-
mation conflicts with haptic information (from the sense of touch). The
animals and infants are in contact with a surface of support, whether they
are on the shallow side or the deep side of the visual cliff. Wherever they
are on the glass floor, the haptic information tells them that they are on a
surface that will bear their weight. As far as vision is concerned, however,
this information exists only when they are on the shallow side. On the
114 Chapter 5
deep side the visual information is that the supporting surface is far below
them (just as if one were on a glass-bottomed balcony on the side of a
precipice). The distress behavior manifested in the experimental situation
shows that the visual information dominates over the haptic information.
The conclusion that Eleanor Gibson and her coworkers drew from this is
that the capacity to perceive depth is innate. They reasoned that what
causes the distress is the visual perception of depth together with a fear
of heights.6 James Gibson, in contrast, has an explanation in terms of
affordances:
But the sight of a cliff is not a case of perceiving the third dimension. One per-
ceives the affordance of its edge. A cliff is a feature of the terrain, a highly signifi-
cant, special kind of dihedral angle in ecological geometry, a falling-off place. The
edge at the top of the cliff is dangerous. It is an occluding edge. But it has the
special character of being an edge of the surface of support, unlike the edge of a
wall. One can safely walk around the edge of a wall but not off the edge of a cliff.
To perceive a cliff is to detect a layout but, more than that, it is to detect an
affordance, a negative affordance for locomotion, a place where the surface of
support ends. (J. J. Gibson 1979, 157)
It is not that the infants and animals perceive depth and then are driven
to distress behavior by their fear of heights. Rather, they show distress
because they directly perceive that they are in an environment that af-
fords falling.
On the basis of Gibson’s theory of ecological optics, then, we can iden-
tify the following three different types of self-specifying information in
visual experience:
• Information about bodily invariants that bound the field of vision
• Information from visual kinesthesis about the movement of the
perceiver
• Information about the possibilities for action and reaction that the envi-
ronment affords the perceiver
Putting these three types of self-specifying information together provides
a powerful counterbalance to the view that perceptual experience pro-
vides information only about the external world. Instead, we find that
information about the ambient environment is inextricably combined
with self-specifying information, without which the former would be of
little use. This duality of exteroceptive and proprioceptive information in
perceptual experience is at the core of the theory of nonconceptual first-
person contents.
5.2 The Content of Ecological Perception
It would be natural at this stage in the argument to suggest that the in-
sights of ecological optics into the phenomenology of perception show
that perceptual experience is itself a source of nonconceptual first-person
contents. Subject to certain qualifications, I think that this is correct. This
section explains why and explores some of the qualifications.
The self-specifying information that ecological optics discerns in per-
ceptual experience is clearly available to a wide range of creatures, as well
as to humans from the earliest days of infancy.7 There is no question that
concepts are required to pick up this self-specifying information. Here we
are clearly in the realm of autonomous nonconceptual content, if we are
in a realm of content at all. In chapters 3 and 4, I noted that there are
considerable difficulties in attributing states with autonomous noncon-
ceptual content to non-language-using creatures, and I pointed out cer-
tain general constraints on such attributions. It is from these constraints
that we must start.
States with representational content are intermediaries between sen-
sory input and behavioral output that are (theoretically) required to ex-
plain how behavioral output emerges on the basis of sensory input. The
theoretical requirement to deploy states with content emerges only when
the connections between sensory input and behavioural output are not
invariant. As was discussed above, states with autonomous nonconcep-
tual content play a role in intentional explanation, and intentional expla-
nations are teleological. This means that they explain behavior in terms
of a conjunction of representational states and motivational states. Now
there is no reason to think that all creatures whose perceptual systems can
be correctly described as picking up self-specifying information behave in
ways that can only be explained intentionally in this way. Quite the con-
trary. The relevant self-specifying information can be deployed in the sort
of conditioned behavior that is entirely explicable by stimulus-response
psychology. By the same token, there is no absurdity in describing tropis-
tic behavior, like that of the Sphex wasp, as guided by self-specifying in-
variants and visual kinesthesis in the wasp’s limited field of vision. Only
a small subset of creatures sensitive to self-specifying information in the
field of vision behave intentionally. Picking out the members of that small
116 Chapter 5
subset is an experimental task, to be achieved by determining whether

their behavior can be explained nonintentionally. If it can be explained
nonintentionally, then obviously the parsimonious nonintentional expla-
nation should be adopted. But if nonintentional explanation is impos-
sible, then inference to the best explanation drives us to an intentional
explanation. It is only here that the issue of nonconceptual first-person
contents arises. This is the first constraint. Perceptual experience cannot
be a source of nonconceptual first-person contents if the behavior of the
creature in question can be explained in a nonintentional manner. But
how does the coperception of self and environment in perceptual experi-
ence contribute to intentional explanations of behavior in instances
where such explanation is required?
The need for intentional explanation arises in the absence of lawlike
correlations between environmental stimulus and behavioural response—
when, for example, the response occurs in the absence of the stimulus.
This, in turn, is connected with the possibility of misperception: of per-
ceiving things as being where they are not, or of perceiving something
that is there as having properties that it does not have. Instances of mis-
perception like these are what make it impossible to formulate lawlike
correlations between environmental stimulus and behavioral response,
because misperception generates behavioral responses that cannot be pre-
dicted from the layout of the environment. Part of the significance of the
ecological analysis of perception is that it makes comprehensible both
how such misperception can occur and how it can directly feed into ac-
tion. A possible example is provided by the moving-room experiments
discussed in the previous section. The movement of the room creates de-
ceptive visual-kinesthetic information. Subjects misperceive the move-
ment of the room relative to them as their own movement relative to the
room, and they behave accordingly (most spectacularly in the case of the
young children who fall over). The visual cliff is another potential case in
point. The misperception of an affordance generates distress behavior
that is not a function of the presence of any real danger. As Gibson puts
the point, “If information is picked up perception results; if misinforma-
tion is picked up misperception results” (1979, 142).8
There is a broader point to be extracted here. The direct perception of
visual kinesthesis and affordances has immediate salience for the launch-
ing and control of action. The optical information for visual kinesthesis
specifies the perceiver’s movement relative to the environment. It specifies
the imminence of collisions and the consequences of maintaining a partic-
ular trajectory. This obviously has immediate implications for how the
perceiver behaves—whether he modifies or maintains his trajectory, for
example. The situation is even clearer with the perception of affordances.
Perceiving an affordance just is perceiving the possible actions and reac-
tions that the environment affords. This immediate salience of perception
for action is at the core of ecological optics. And this leads us to the idea
that when we are dealing with behavior that supports an intentional ex-
planation, perception is a source of nonconceptual first-person contents.
An analogy with indexical beliefs will illustrate the point. It is widely held
that indexical beliefs are required to explain why an individual behaves
as he does.9 Among those indexical beliefs, beliefs with first-person con-
tents are particularly important. The thought is that we will not be able
to explain why an individual behaves as he does unless we understand the
beliefs that support and drive his behavior, and that we can explain why
those beliefs support and drive his behavior only if they are beliefs about
himself and his possible courses of action. The first person is, in Perry’s
famous phrase, an essential indexical. It is essential, he argues, in order
to capture the immediate salience of an agent’s beliefs to what he actu-
ally does.
Of course, indexical beliefs in Perry’s sense occur at the level of concep-
tual content. They are available only to thinkers who have attained a so-
phisticated mastery of concepts, including mastery of the concept of the
first-person. But Perry’s central point can be carried over to the level of
nonconceptual content. The salience of an agent’s beliefs to what he actu-
ally does, which, according to Perry, requires the thinking of first-person
thoughts, is directly parallel to the salience of perception to action that is
highlighted in the ecological theory of perception. A vital part of what it
is to describe a belief as a first-person belief is that the belief makes a given
action immediately comprehensible (as my fleeing becomes immediately
comprehensible in light of my belief that a bear is about to attack me, to
continue with Perry’s familiar example). This is directly applicable to the
perceptual case. Perception is directly salient to action in virtue of the
self-specifying information that is coperceived with information about
118 Chapter 5
the environment, and this is what makes it appropriate to describe the

content of perception as a first-person content. Perceptual contents are
first-person contents in virtue of their immediate connections with behav-
ior. The conclusion to draw, then, is that in cases where the behavior of
non-language-using creatures demands an intentional explanation, such
explanations can draw on first-person perceptual contents just as expla-
nations of the behavior of more conceptually sophisticated creatures can
draw on first-person beliefs.
It might well be asked how perceptual contents can play a part in inten-
tional explanation at the nonconceptual level, since intentional explana-
tion is generally understood to proceed in terms of beliefs and desires,
which are, of course, propositional attitudes with paradigmatically con-
ceptual content. Let me first introduce terms for the analogues of beliefs
and desires at the level of nonconceptual content.10 I will call these proto-
beliefs and protodesires, to avoid the conceptual implications of the no-
tions of belief and desire. I take it that the possibility of nonconceptual
protodesires is relatively uncontroversial. I have not encountered any de-
fenders of the view that all motivational states must be conceptual. The
possibility of protobeliefs is more controversial, however.
Let me start by recalling that protobeliefs fall into two broad classes:
perceptual protobeliefs and instrumental protobeliefs. A perceptual pro-
tobelief is a representation of the layout of the immediate environment.
An instrumental protobelief has to do with how a particular protodesire
might be satisfied in a given situation. I briefly suggested in chapter 3 that
these two types of belief can be combined in a single perception. Gibson’s
notion of an affordance provides an excellent illustration of this. To say
that affordances are directly perceived is precisely to say that instrumental
relations can feature in the content of perception, or alternatively, that a
single protobelief can be both perceptual and instrumental. Nonetheless,
this unity is not necessary. Instrumental and perceptual protobeliefs can
be separated at the nonconceptual level.
Perceptual protobeliefs are canonically present-tense (which is, of
course, equivalent to saying that they are not tensed at all). They have to
do with here-and-now features of the occurrent environment. The most
basic form of instrumental protobeliefs are either to be understood on the
model of Gibsonian affordances or to be modeled as expectations about
possible outcomes.11 These may be hard-wired, or they may be acquired

through some form of exposure to the phenomena in question (and thus
may be the result of conditioning). One important feature distinguishing
both perceptual and instrumental protobeliefs from their conceptual ana-
logues is that there is no distinction at the nonconceptual level between
entertaining a protobelief content and holding it to be true. The ability to
entertain a representational state without assenting to it is a high-level
cognitive ability. It is associated with, inter alia, the capacity for delibera-
tion and what is often termed practical reasoning. These are not, I take
it, operative at the nonconceptual level. Of course, a creature may have
more than one instrumental protobelief in a situation in which he can
only act on one of them. But to say that he can only act on one of them
is not to say that he can decide which of them he is going to act on. It
is natural to assume that the weighting of instrumental protobeliefs is
determined by, among other things, the comparative strengths of the rele-
vant motivational states.
There is a delicate balance to be struck in positing protobeliefs as ex-
planatory tools at the nonconceptual level. If the differences between
protobeliefs and conceptual beliefs are stressed, then one can raise the
question of how two such dissimilar states can play comparable roles in
the explanation of behavior. On the other hand, assimilating protobeliefs
and conceptual beliefs leaves one open to the objection that crucial differ-
ences are being overlooked. On the plus side, protobeliefs and conceptual
beliefs are thus far comparable in that they both satisfy the four criteria
for genuinely representational states offered in chapter 3. On the debit
side, some crucial differences have just been mentioned. There is one im-
portant issue that has not yet been touched upon, however.
It is as unanimously agreed among philosophers as anything is that
intentional mental states are nonextensional, where this means that in
specifying the contents of an intentional mental state, coreferential terms
cannot always be substituted salva veritate (without changing the truth
value of the content ascription). The nonextensionality of mental states
is connected to the fact that they involve thinking about objects under
particular modes of presentation. Thought about an object is not about
that object tout court, but rather about that object as grasped or con-
ceived in a certain way. The question, then, is whether the states that I am
120 Chapter 5
describing as protobeliefs can exhibit anything analogous to this property

of nonextensionality.
It has seemed obvious to some philosophers, even those most sympa-
thetic to the idea that a form of cognition can exist in the absence of
language, that thinking about objects under particular modes of presenta-
tion is only available to language-using creatures. Consider the following
passage from Charles Taylor:
Now it is a feature of our language that, in applying a description to something,
in picking something out under a concept, we are picking out something to which
other descriptions also apply, which falls under other concepts as well. . . . Thus
for language users the expression ‘thinking of something as an X’ has a specific
force, for of these it is true that they could also think of (what they recognize as)
the same thing under some other description. . . . But for an animal, thinking of
or fearing something “as a so-and-so” cannot be given this sense. For of an animal
we could never say that he was conscious of this same thing (that is, what to him is
the same thing) under two different descriptions, either at one time or on different
occasions. . . . This means that the only type of consciousness of the objects
around them that we can attribute to animals is a consciousness of their immedi-
ate relevance to their behaviour. The red object is something to fly from, the meat
something to be eaten, and so on, but the animal is not aware of the object as
something he must fly from, or of the meat as something to be eaten. (Taylor
1964, 67–68)
Taylor’s reasoning here seems clear. Thinking of an object under a nonex-
tensional mode of presentation is, he claims, equivalent to thinking of it
under a particular description. And thinking of an object under a particu-
lar description is only available to language users, because descriptions
are linguistic phenomena. This leaves Taylor with only a restricted sense
in which the cognition of animals and infants is cognition at all. Ac-
cording to Taylor, an animal “is not aware of the object as something he
must fly from.” What happens, presumably, is that the animal perceptu-
ally registers the object in a way that triggers a certain reaction. On Tay-
lor’s view, it is the type of action or reaction triggered that determines
whether we are dealing with behavior that requires intentional explana-
tion or not. It is the directedness of actions that renders them appropriate
for intentional rather than mechanical explanation. He explains his
position:
In short, we have at least prima facie grounds for classing them as agents. . . .
That is to say, we want to class them as beings who direct their behaviour, such
that attributing a given direction to a given segment of an animal’s behaviour is
not simply saying that this segment has a certain end-product, or that it is to be
explained by certain laws, but rather that, whatever the explanation, and whether
it achieves its end or not, it is given this direction by the animal. (Taylor 1964, 67)
So if an animal perceptually registers an object and acts in a manner that
it is appropriate to describe as fleeing from the object, then it is engaging
in behaviour that needs to be explained intentionally. Thus does Taylor
explain the possibility of intentional behavior even in creatures incapable
of thinking about objects under particular modes of presentation.
This compromise position has its difficulties, however. If behavior of
this type really is to be significantly analogous to the intentional behavior
of language-using human beings, then it must render the creature’s behav-
ior comprehensible as a response to its perceptually registering the object
that triggered the appropriate response. The explanation will be of the
form ‘The animal fled because it perceptually registered that object’. But
for this explanation to be informative, it must also explain why perceptu-
ally registering that object should make fleeing an appropriate response.
It must explain why the perception generates the response of fleeing. This
cannot be done, however, if what the animal perceives (the content of the
animal’s perception of the object) does not include the object as some-
thing potentially dangerous. Without this, there will be no (intentional)
explanation of why the animal flees from the object, rather than ap-
proaching or ignoring it. But the problem, of course, is that this is impos-
sible if, as Taylor maintains, the animal “is not aware of the object as
something he must fly from,” simply because to be aware of an object as
something dangerous is to be aware of it under a nonextensional mode
of presentation.
So if Taylor is right and the cognitive repertoire of non-language-users
does not include something analogous to the capacity to think of objects
under nonextensional modes of presentation, then the possibility of giv-
ing satisfactory intentional explanations of their actions seems to be
severely compromised. This would be a very unwelcome conclusion. For-
tunately, though, taking seriously the ecological account of perception
provides a way of making sense of the idea that non-language-using crea-
tures are capable of thinking about objects under nonextensional modes
of presentation. The concept of an affordance is, of course, the key. The
central idea at work in analyzing perception in terms of affordances is
that the environment is not perceived in neutral terms. What are perceived
122 Chapter 5
are the possibilities that the environment affords for action and reaction.
These include, of course, the potential of various locations for providing
shelter, concealment, or nourishment. And, of course, the fact that a par-
ticular place is perceived as affording shelter explains why an animal will
seek shelter there. The animal does not perceive a place neutrally and then
seek shelter there. What the animal perceives is a possibility of shelter,
and it acts accordingly. The visual cliff offers a dramatic example, as do
the more day-to-day examples of affordances that emerge from visual kin-
esthetics, such as looming.
The perception of affordances, therefore, provides a way in which we
can understand how perceptual protobeliefs can have an analogue of the
property of nonextensionality. Perceptual protobeliefs are not neutral.
They involve perceiving objects under modes of presentation, where these
modes of presentation are given by the possibilities for action and reac-
tion that those objects afford. This analogue of nonextensionality can be
pressed even further. We can, for example, make perfectly good sense of
two or more perceptual protobeliefs being about the same object under
different modes of presentation. An object can present different affordan-
ces, depending on the angle from which it is approached and perceived.
As an animal moves around the object, its perceptual protobeliefs will
change accordingly. The same object can also be perceived on different
occasions under different modes of presentation. Suppose that an animal
catches sight of a predator. The predator is perceived, of course, as af-
fording danger, and the animal responds appropriately by fleeing. Sup-
pose that some time later the animal encounters the same predator, only
this time the predator is lying dead. Now the predator is perceived as
affording nourishment, and the animal feeds accordingly. Here we have
two different affordances, which yield two different modes of presenta-
tion, and the difference between the modes of presentation is what ex-
plains the difference in behavior.
To draw the strands of this section together, what we learn from Gib-
son’s ecological approach to perception is that perceptual experience is
itself a source of nonconceptual first-person contents. These nonconcep-
tual first-person contents are crucial to the intentional explanation of the
behavior of non-language-using creatures. Such intentional explanation
proceeds in terms of protodesires and protobeliefs, and the protobeliefs
have nonconceptual first-person contents derived from perception that

render them immediately salient to the perceiver’s behavior. These proto-
beliefs with nonconceptual first-person contents have properties impor-
tantly analogous to the nonextensionality of the mental states of
language-using creatures, particularly the property of involving objects
under particular modes of presentation.
5.3 The Ecological Self in Infancy
It will be remembered from chapter 1 that part of the paradox of self-

consciousness is how to explain how young children can bootstrap them-
selves into self-conscious thought. If self-consciousness is understood in
terms of the capacity for first-person thoughts involving reflexive self-
reference and yet the capacity for reflexive self-reference presupposes
the capacity to think first-person thoughts, then we are confronted with
a circle of interlocking concepts. This seems to breach certain plausible
constraints on the acquisition of cognitive abilities. If the interdependence
of reflexive self-reference and first-person thought is complete, it fol-
lows that there are no more-basic cognitive abilities in terms of which
self-consciousness can be explained, and hence that there are no more-
basic cognitive abilities from which self-consciousness can emerge in
the normal course of cognitive development. The acquisition of self-
consciousness then becomes a complete mystery. Self-consciousness
depends upon the capacity to take a first-person perspective (to think
first-person thoughts), but we have no understanding of how entry into
that first-person perspective is possible.
The discussion of ecological optics in this chapter presents us with the
possibility of a radically new perspective on the problem. When percep-
tion is understood in ecological terms, and subject to the constraints al-
ready discussed, perception itself is a source of first-person contents.
These first-person contents are autonomous nonconceptual contents, in
the sense discussed in chapter 3, and as such are available to young infants
as well as to animals. This opens up the following possibility. If it can be
shown that these nonconceptual first-person contents are available more
or less from the start of life, then the whole problem of how entry into
the first-person perspective is possible fails to arise. If very young infants
124 Chapter 5
are capable of entertaining perceptual states with nonconceptual first-

person contents, then they are already in the first-person perspective, and
the apparent paradoxicality of trying to explain how a creature can ac-
quire the capacity for first-person thought without already possessing
that capacity disappears (although, of course, there will remain questions
regarding the transition to the conceptual level).
Of course, this is a stage in the argument at which empirical work must
come to the fore, because the question now arises of whether there is
evidence to support the proposal that nonconceptual first-person con-
tents are available from more or less the beginning of life. Fortunately, a
considerable amount of work has been done by developmental psycholo-
gists inspired by Gibson’s theories of perception. Some workers in the field
have found the evidence so convincing that they have introduced the term
‘ecological self’ to capture how self-specifying information is picked up
from the beginning of life. Consider the following passage from an impor-
tant paper by Ulric Neisser:
The information that specifies the ecological self is omnipresent, and babies are
not slow to pick it up. They respond to looming and optical flow from a very
early age, discriminate among objects, and easily distinguish the immediate conse-
quences of their own actions from events of other kinds. The old hypothesis that
a young infant cannot tell the difference between itself and its environment, or
between itself and its mother, can be decisively rejected. The ecological self is
present from the first. (1988, 40)12
What Neisser and others mean by‘ecological self’ is not quite the same as
what I have been discussing as nonconceptual first-person contents, but
it is sufficiently close that the evidence adduced to support theories of
the ecological self also supports the thesis that nonconceptual first-
person contents can be present from the beginning of life. In this section
I briefly review some of the important developmental evidence for this
conclusion.13
Neonatal distress crying

There is strong experimental evidence suggesting that infants have an in-
nate tendency to cry in response to the crying of other infants at more or
less the same age (Simner 1971). When Martin and Clark (1982) repli-
cated some of these experiments using infants with a mean age of 18.3
hours, they noticed by chance that five crying male infants stopped crying
when they heard a recording of themselves crying. To investigate this

more systematically, they tested 47 infants within a day after birth. Both
calm and already crying babies were played tape recordings either of
themselves crying or of another newborn baby crying, and then the total
amount of crying over a 4-minute period was measured. The experiment-
ers found that infants who were calm when the test began cried signifi-
cantly more when they heard another infant crying than when they heard
themselves crying, while infants who were already crying cried less after
hearing themselves cry than after hearing another infant crying.
Here we seem to have a clear case of self-perception in newborn infants.
The infants can discriminate their own crying from the crying of other
infants of the same age. What is particularly interesting is that the sense
of hearing is the relevant modality. Although most of the discussion of
self-specifying invariants in perception has focused on vision, these exper-
iments suggest that auditory information is equally self-specifying.
Neonatal imitation
On the traditional view of cognitive development, associated with Jean
Piaget, the capacity for facial imitation is a high-level cognitive ability,
occurring only towards the end of the second year. What makes facial
imitation so important is the fact that awareness of the behavior to be
imitated and the awareness of the imitating behavior occur in different
sensory modalities. Facial imitation involves matching a seen gesture with
an unseen gesture, since in normal circumstances one is aware of one’s
own face only haptically or proprioceptively. If successful facial imitation
is to take place, a visual awareness of someone else’s face must be appre-
hended so that it can be reproduced on one’s own face.14
Meltzoff and Moore (1977) found that infants between 12 and 21 days
old could successfully imitate three distinct facial acts (lip protrusion,
mouth opening, and tongue protrusion) and one manual act (sequential
finger movement). The infants were prevented by a dummy placed in their
mouths from responding before the model gesture was complete. The
dummy was then removed and the infants were videoed close up. Inde-
pendent judges then decided what gestures the infants were trying to
make. The judges did this without knowing what model gestures the in-
fants were responding to. The results of the experiment seemed clearly to
126 Chapter 5
show that imitation behavior was going on here. Meltzoff and Moore
(1983) then looked for evidence of imitation capabilities in infants just
after birth. Here too they were successful. The average age of the infants
was 32 hours, with the youngest infant only 42 minutes old. Two gestures
were employed (mouth opening and tongue protrusion) and once again
the scorer was looking at videotapes of the infants without any informa-
tion about the gesture the experimenter had made. Here too the evidence
was clearly in support of imitation. These results have been interestingly
replicated and extended by T. M. Field and his coworkers (1982), who
found that newborn infants, 2 days old, could consistently imitate the
facial expressions of an adult model when that model either smiled,
frowned, or displayed an expression of surprise.
There are very good reasons to think neonatal imitation behavior needs
to be explained in intentional terms. In the first place, the behavior in-
volves memory and representations. As mentioned earlier, the infants had
a dummy placed in their mouths to prevent them from responding before
the experimenter had completed the model gesture. By the time the in-
fants are in a position to imitate the gesture, the experimenter has reverted
to his normal facial expression. Clearly, then, there must be some sort of
stored representation of the model gesture, and this stored representation
triggers the imitation response in a way that seems incompatible with
many accounts of the functioning of innate reflexes and innate releasing
mechanisms. Furthermore, the infants correct their response over time
until their own gesture has homed in on the model gesture. Such a trial-
and-error approach is not characteristic of reflex behavior.
What is most striking about neonatal imitation behavior, though, is the
richness of self-specifying information that it implicates in the perceptual
experience of infants just after birth. The infants are perceiving the exper-
imenter as a being just like themselves, not only as structurally similar but
also as capable of behaving in similar ways. It is not simply a case of
grasping the intermodal equivalence of, say, a perceived tongue protrusion
and their own act of protruding their tongue, because they also have to
grasp that the perceived tongue protrusion is the sort of thing that they
can imitate, as opposed, for example, to the movement of a nipple, to
which they respond by sucking rather than by trying to imitate it. The
point can be put in more straightforwardly Gibsonian terms by saying
that when these infants perceive the face of the experimenter, they per-
ceive their own possibilities for acting as the experimenter is acting. They
perceive a set of affordances, but these are no ordinary affordances, be-
cause they reveal to the infants information not only about how they can
behave but also about their physical makeup.
Infants’ reaching behavior

It is not only the complicated affordances revealed in the perceptions driv-
ing neonatal imitation behavior that involve self-specifying information.
Perceptions of even the most basic affordances involve self-specifying
information. Reaching behavior is a case in point. Reaching behavior is
driven by the perception that an object is within reach (by the perception
that the object affords reaching), and this is, of course, a form of self-
specifying information. It is interesting, therefore, that the capacity for
fairly accurate perception of whether or not objects are within reach ap-
pears early on in infancy, particularly since infants’ reaching distance is
in a constant state of flux, because of not only their rapid growth but also
rapid changes in their motor capacities. An interesting illustration comes
from experiments performed by Jeffrey Field (1976). He found that the
reaching behavior of 15-week-old infants is adjusted to the physical dis-
tance separating them from the object, as measured by the reduction in
the probability of their making relevant movements as their distance from
the object increased. Earlier work by T. G. R. Bower (1972) suggested
that this adjustment is in place as early as two weeks. He found that when
an object is placed out of all possible reach, there is a marked reduction
in the frequency with which infants extend their arms toward it.
Reaching behavior is a good index of how infants perceive affordances.
Also relevant in this context are the catching skills of very young infants.
Even newborn infants try to intercept objects that are moving past them
and within reach, and the experimental evidence shows that babies are
capable of adapting the speed of their reach to the speed of the moving
object (Von Hofsten 1982). Here we are dealing with behavior that in-
volves reaching toward objects in motion, rather than at rest, but it is
appropriate to describe what is going on in similar terms. Infants are per-
ceiving the affordance of reachability in a way that has immediate impli-
cations for their behavior.
128 Chapter 5
Visual kinesthesis in prelocomotive infants

Visual kinesthesis is an important instance of the pick-up of self-
specifying information, and there is a range of experimental evidence sug-
gesting that the capacity to pick up this type of information is present
very early on in infancy. Looming is a case in point, and much work has
been done on how young infants react in moving-room situations, where
information from the optic flow specifies the imminence of a nonexistent
collision. The original work was done in babies who were already able to
walk, and as I mentioned in the previous section, the babies fell over when
they tried to compensate for what they registered as a loss of balance (Lee
and Aronson 1974). When these experiments were extended to infants
who could crawl but were unable to walk or stand, compensatory move-
ments were also observed (Butterworth and Hicks 1977). The same
was found in completely prelocomotive infants from the age of 8 weeks
(Pope 1984). In prelocomotive infants the compensatory movements
were head movements—the infants were seated in a baby seat and their
head movements were measured in relation to the movements of their
surroundings.15
5.4 Moving beyond Perception
The conclusion to be drawn from the experimental evidence briefly dis-

cussed in the previous section is that nonconceptual first-person contents
are available from the beginning of life. This has extremely interesting
implications for one of the strands in the paradox of self-consciousness.
If perception itself can be a source of nonconceptual first-person contents
by supplying self-specifying information, and if the pick-up of self-
specifying information starts at the very beginning of life, then there
ceases to be so much of a problem about how entry into the first-person
perspective is achieved. In a very important sense, infants are born into
the first-person perspective. It is not something that they have to acquire
ab initio. Of course, the first-person perspective that infants have at birth
is extremely limited. The range of self-specifying information that they
are capable of picking up is tiny in comparison with that of older children
and adults. But it is there all the same and provides a foundation on which
a greater sensitivity to different types of self-specifying information can
be built as the perception of simple affordances grows into the perception

of more complex ones and awareness of visual kinesthesis becomes more
subtle and differentiated.
Of course, this is still only the very beginnings of a solution. I have
identified a class of primitive first-person contents. These contents are
primitive both conceptually and ontogenetically. There are no more-
primitive types of first-person content in terms of which the first-person
contents of perceptual experience can be understood.16 Nor are there any
contents that are ontogenetically more primitive. But this still leaves us
with the theoretical task of explaining how this fits into a fully articulated
theory of self-consciousness. What is required is an account of how these
primitive forms of first-person content are built up into full-fledged self-
consciousness. In giving such an account, I will need to be sensitive to
the factual dimension—to the actual developmental progression by which
children acquire the capacity for different forms of first-person thought.
Clearly, though, it is no use thinking that a properly articulated theory of
self-consciousness can be gained by simply describing the developmental
progression from first-person perceptual contents to full-fledged first-
person thought. A properly articulated theory of self-consciousness is re-
quired to give the conceptual framework for any such description. None-
theless, developmental considerations do circumscribe the form that such
an account will take, in virtue of the Acquisition Constraint, as formu-
lated in chapter 1. A theory of self-consciousness that starts with the
primitive first-person contents that can exist in perceptual states must
conform to the Acquisition Constraint. It must show how full-fledged
self-consciousness can be constructed from such primitive first-person
contents in a way that makes it comprehensible how each increasingly
developed form of self-consciousness can emerge ontogenetically from its
predecessors. Before embarking on that project, however, I need to draw
attention to a further class of primitive first-person contents, and I will
discuss such contents in the next chapter.
6
Somatic Proprioception and the Bodily Self
The picture that emerged from the previous chapter is of a form of primi-
tive self-consciousness operative in the very structure of perception. Gib-
son’s analysis of the pick-up of information about the self through the
interplay of self-specifying and other-specifying invariants provides evi-
dence for a primitive form of self-consciousness implicated in the basic
mechanisms of perception and action from the earliest days in infancy
onward. The self that is thus perceived (what some psychologists have
called the ecological self ) is an embodied self. Most of the self-specifying
information implicated in ecological perception is information about the
body: the disposition of bodily parts and their movement. But it is not
only through ecological perception that information is acquired about the
embodied self. Gibson’s great insight was that vision (and the other ex-
teroceptive sense modalities) are only “windows onto the world” to the
extent that they provide relational information about the perceiver’s posi-
tion, movement, and other bodily properties. But there are other sources
of such information. The highly complex somatic proprioceptive sys-
tem provides even more fine-grained and detailed information about
the perceiver’s position, movement, limb disposition, and other bodily
properties.
No less than visual perception, somatic proprioception is a form of
experience with nonconceptual first-person contents. Like the self-
specifying information in ecological perception, somatic proprioception
is a form of primitive self-consciousness. Also like the self-specifying in-
formation in ecological perception, somatic proprioception of the embod-
ied self is available from birth (and indeed before). These two forms of
primitive self-consciousness are the foundations for the higher levels of
132 Chapter 6
self-consciousness that will be discussed later on in this book. They are

the building-blocks for the bootstrapping process that will eventually re-
sult in mastery of the first-person concept and the capacity for full-fledged
self-consciousness.
The significance of somatic proprioception for the analysis of self-
consciousness has not been properly recognized by philosophers. There
are several reasons for this. The first is that philosophers have tended to
operate with a narrow and restrictive conception of somatic propriocep-
´
tion (but see the essays in Bermudez, Marcel, and Eilan 1995). This re-
strictive conception of somatic proprioception is partly responsible for an
unwillingness among philosophers to consider somatic proprioception as
a perceptual modality. But an important role has also been played by phil-
osophical arguments purporting to show that somatic proprioception
cannot be a form of perception. In this chapter I will attempt to lay these
conceptions to rest. I will develop an account of somatic proprioception
on which it qualifies as a form of perception whose object is the embodied
self. Section 6.3 explains why somatic proprioception is a form of percep-
tion with the embodied self as its object, and section 6.4 outlines the
criteria on which this qualifies as a form of self-consciousness. In section
6.5, I offer an account of the content of somatic proprioception that ex-
plains in more detail the form of consciousness of the self that is impli-
cated in somatic proprioception. The first step, though, is to get clear on
the phenomena under discussion.
6.1 The Modes of Somatic Proprioception
Somatic proprioception is not a simple or unitary phenomenon.1 The best

place to begin is by considering some of the internal information systems
´
involved. The general introduction to Bermudez, Marcel, and Eilan 1995
offers the following as some of the principal types of information de-
ployed in somatic proprioception and an indication of their physiologi-
cal sources:
• Information about pressure, temperature, and friction from receptors
on the skin and beneath its surface
• Information about the relation of body segments from receptors in the
joints, some sensitive to static position, some to movement
Somatic Proprioception and the Bodily Self 133
• Information about balance and posture from the vestibular system in

the inner ear and the head/trunk dispositional system and information
from pressure on any parts of the body that might be in contact with a
gravity-resisting surface
• Information about bodily disposition and volume obtained from skin
stretch
• Information about nutrition and other homeostatic states from recep-
tors in the internal organs
• Information about muscular fatigue from receptors in the muscles
• Information about general fatigue from cerebral systems sensitive to
blood composition
• Information about bodily disturbances derived from nociceptors
These somatic information systems vary along several dimensions.

First, they vary according to whether the information they provide is in-
formation solely about the body (e.g., the systems providing information
about general fatigue and nutrition), as opposed to information about
the relation between the body and the environment (e.g., the vestibular
system, which is concerned with bodily balance). Other systems can be
deployed to yield information either about the body or about the environ-
ment. Receptors in the hand sensitive to skin stretch, for example, can
provide information about the hand’s shape and disposition at a given
time or about the shape of small objects. Similarly, receptors in joints and
muscles can yield information about how the relevant limbs are distrib-
uted in space or, through haptic exploration, about the contours and
shape of large objects.
Second, not every system yields information that is consciously regis-
tered. Most information about balance and limb position, for example,
feeds directly into the control of posture. There are also significant differ-
ences in how different types of information come to consciousness. It is
important to distinguish two different types of conscious somatic propri-
oception (mediate somatic proprioception and immediate somatic propri-
oception), distinguished according to the role played by bodily sensation.
The experience of pain is a paradigm example of mediate somatic propri-
oception, because the sensation of pain is a constitutive part of the experi-
ence. Pain is something we feel, as are itches, tickles, and so forth. But
this cannot be the model for conscious somatic proprioception in general.
There are ways of finding out about our bodily posture and movements
134 Chapter 6
that count as instances of somatic proprioception (because they are de-

rived from somatic information systems) and yet that are not based on
sensations. The systems involved tend to be discussed by psychologists as
joint-position sense and kinesthetic awareness respectively. Joint-position
sense is the awareness that we have of how our body parts are distributed
in space and relative to each other. Kinesthetic awareness is the awareness
of limb movement.
There are at least two compelling reasons for denying that kinesthetic
awareness and joint-position sense provide a mediate sensation-based
awareness of limb position and movement.2 That sensations are not neces-
sary is shown by the many occasions when we are capable of reporting on
limb position and movement without being able to identify any sensations
yielding the awareness on which such reports are based. That sensations
are not sufficient emerges when one reflects that bodily sensations are
rarely of sufficiently fine a grain to yield the relatively precise awareness
we have of posture and movement, let alone the extremely precise aware-
ness required to maintain balance and control action. Bodily sensations—
pain is again a paradigm example—are rough indicators of large-scale
events in the body, too rough to provide a detailed awareness of posture
and movement.
The classification that emerges from these distinctions is as follows.
Somatic proprioception must be understood in terms of the operations of
internal somatic-information systems. Some of these information systems
operate completely below the threshold of conscious awareness. Other
information systems do yield information that is consciously registered.
Of this second category, some information systems operate through the
medium of bodily sensations, such as pain, hunger, and fatigue. These are
instances of mediate somatic proprioception. They contrast with immedi-
ate somatic proprioception, where bodily sensations are at best concomi-
tant (if present at all). The paradigm examples of immediate somatic
proprioception are kinaesthetic awareness and joint-position sense.
6.2 Somatic Proprioception and the Simple Argument
Despite the range of different phenomena that come under the heading
of somatic proprioception, there is a relatively simple argument showing
the direct relevance of an important class of the different modes of so-

matic proprioception to philosophical accounts of self-consciousness.
The class in question is, of course, the class of conscious modes of somatic
proprioception and the argument is as follows:
1. The self is embodied. [Premise]
2. Somatic proprioception provides perceptions of bodily properties.3
[Premise]
3. Somatic proprioception is a form of self-perception.
4. Therefore, somatic proprioception is a form of self-consciousness.
I shall call this the simple argument. I am simply assuming premise (1) of
the simple argument. It is something that only a dualist would deny. The
parts of the simple argument that require more discussion are premise (2)
and the conclusion (4). Clearly, if somatic proprioception is a form of
perception, then it will provide perceptions of bodily properties. What
might not appear so obvious, however, is that somatic proprioception
counts as a form of perception at all. In the next section I explain how
somatic proprioception satisfies a plausible and stringent set of con-
straints for anything to count as an instance of perception, thereby estab-
lishing the truth of (2). It will be apparent that (3) follows from (2) in at
least the following sense: proprioception counts as self-perception be-
cause it involves the perception of properties that are in fact properties of
one’s self. The real question, of course, is whether proprioception quali-
fies as an instance of self-perception in any sense sufficiently rich to count
as a form of self-consciousness and hence to underwrite the conclusion
(4). In section 6.4, I offer an affirmative answer to this question.
6.3 Somatic Proprioception as a Form of Perception
First I need to substantiate the claim that somatic proprioception is a

form of perception. This will obviously depend upon one’s chosen ac-
count of perception. The most developed arguments in this area have been
formulated by Sydney Shoemaker (particularly 1994, but see also 1986),
and I will take his discussion as the framework for this section.
Shoemaker is explicitly concerned with the question of whether intro-
spective self-knowledge is a form of inner perception. He proceeds by
identifying the defining features of two distinct models or stereotypes of
136 Chapter 6
perception and then asking whether introspective self-knowledge pos-

sesses those features. These two models of perception are the object-
perception model and the broad perceptual model, with the defining fea-
tures of the second being a subset of the defining features of the first. The
first applies paradigmatically to vision and rests on the idea that percep-
tion is a relation in which one stands to objects. On the second model,
perception is a relation not to objects but to facts. Clearly, to motivate
lemma (3) in the simple argument, it needs to be shown that somatic
proprioception falls under the object-perception model.4
Here are the features that Shoemaker finds distinctive of the object
model of perception (1994, 252–254). (I have taken the liberty of naming
the features.)
The object constraint While sense perception provides one with awareness of
facts, this awareness of facts is a function of awareness of the objects involved in
these facts in a sense-experience distinct from the object of perception.
Shoemaker characterizes the second feature in these terms:
The identification constraint Sense perception involves “identification informa-
tion” which allows one to pick out the object of perception through its relational
and nonrelational properties. Such information enables the “tracking” of the ob-
ject over time, and its reidentification from one time to another.
Within this constraint I think it is helpful to distinguish the identification
constraint proper (that sense perception involves identification informa-
tion) from the following precondition of its being satisfied:
The multiple-objects constraint Ordinary modes of perception admit of our per-
ceiving, successively or simultaneously, a multiplicity of different objects, all of
which are on a par as nonfactual objects of perception. (Shoemaker 1986, 107)
Shoemaker’s argument that the self cannot be the object of any form of
inner perception corresponding to the object-perception model concen-
trates on these three constraints.
Shoemaker takes it as uncontroversial that there is no form of self-
perception satisfying the object constraint:
No one thinks that in being aware of a sensation or sensory experience one has
yet another sensation or experience that is “of” the first one, and constitutes its
appearing to one in a certain way. . . . And no one thinks that there is such a thing
as an introspective sense-experience of oneself, an introspective appearance of
oneself that relates to one’s beliefs about oneself as the visual experience of things
one sees relate to one’s beliefs about those things. (Shoemaker 1994, 254–255)
The first denial is indeed uncontroversially true, but the second is more
problematic. Shoemaker clearly thinks that if the object constraint were
satisfied, there would be two separate things going on: ordinary sensory
experiences and a putative introspective sense-experience of oneself. But
this is precisely what the defender of the simple argument denies, because
the claim of that argument is just that instances of somatic proprio-
ception are both ordinary sensory experiences and introspective sense-
experiences. Somatic proprioception can provide sensory experience of
the body. There is a clear impasse here. Both sides have a way of interpret-
ing the object constraint that supports their own position, and it seems
best to make the interpretation of the object constraint depend on the
outcome of the other constraints.5
The identification and multiple-object constraints are the most prob-
lematic for somatic proprioception, because it is tempting to argue that
somatic proprioception necessarily yields information solely about the
body. This would entail that the multiple-objects constraint cannot be
met by somatic proprioception. But if there is necessarily only one object
of somatic proprioception, then there are good reasons for thinking that
the identification constraint cannot be satisfied either, since it is not clear
what it would mean to have identifying information about an object that
is necessarily the only object of the type of awareness in question.
An initial counter to this would be that the sense of touch is a source
of somatic proprioception (as stressed in the first section). It is obviously
not the case that the body is the sole object of the sense of touch, because
touch can be put to work to yield proprioceptive or exteroceptive in-
formation (information about the body or information about nonbodily
objects in space). And when the sense of touch is being used to yield ex-
teroceptive information, it obviously involves the possibility of perceiving
a range of different objects both simultaneously and successively. None-
theless, it might still be possible for us to draw a clear distinction between
exteroceptive and proprioceptive instances of tactile perception, parti-
tioning instances of touch into two mutually exclusive groups, and then
to argue that the multiple-objects constraint is only satisfied for extero-
ceptive touch and that this has no implications for those occasions when
the information systems that subserve the sense of touch are being put to
a proprioceptive use.
138 Chapter 6
However, the phenomenology of tactile awareness does not generally

support this sort of partition (Gordon 1978, O’Shaughnessy 1989, Mar-
tin 1992). Consider an instance of exploratory haptic perception, as when
one discerns the shape of an object in the dark by running one’s fingers
over it.6 The following two things are both true of such a perception.
First, through one’s awareness of the changing spatial properties of one’s
fingers, one gains an awareness of the spatial properties of the object.
That is to say, there is an important sense in which exploratory touch is
representational. This is a very basic point, following from the fact that
the vehicle of exploratory touch is experienced stimulation at or near the
surface of the skin.7 Second, tactile awareness of the spatial properties of
the explored object is not gained by conscious inference from a prior and
independent awareness of the contours and movement of our fingers. In
paradigm cases of haptic exploration, the perceiver attends directly to the
shape and other spatial properties of the explored object. Understanding
the phenomenology of touch is understanding how these can both be
true.
Although in haptic exploration the perceiver’s attention is paradigmati-
cally directed toward properties of the nonbodily objects being explored,
this need not be the case. An attention shift is all that is required for the
properties of the limb doing the exploring to become the focus of atten-
tion, and this does not involve a complete change in the structure and
content of awareness. Relevant here is some form of the distinction be-
tween focal awareness and peripheral awareness, in particular, the possi-
bility of awareness of something even though one is not attending to it.
This idea is most familiar from vision, where it seems clear that items at
the boundaries of the visual field are consciously registered, even though
they are not being attended to. Peripheral awareness seems to require
at least the following. First, items in peripheral awareness can generally
be brought into focal awareness, either by moving the body into an ap-
propriate perceptual relation to an object (by turning one’s head, for
example) or by focusing on one element in a complicated perceptual ex-
perience (as when one tries just to listen to the violins in a symphony).
Second, instances of peripheral awareness can have implications for ac-
tion and reaction without bringing focal awareness to bear, as when one
flinches from something seen out of the corner of the eye.
The best description of the phenomenology of touch is that tactile ex-

perience is always both exteroceptive and proprioceptive. Attention can
be directed either proprioceptively or exteroceptively, and it can be shifted
from one to the other, but this should be viewed as an alteration of the
balance between focal and peripheral awareness. When attention is di-
rected exteroceptively toward the spatial properties of an object, the per-
ceiver remains peripherally aware of the spatial properties of the relevant
limb, and vice versa. And this shows how the tactile elements of somatic
proprioception can meet the multiple-objects constraint. Even when the
attention is fixed firmly on the proprioceptive dimension of tactile aware-
ness, the exteroceptive dimension remains phenomenologically salient in
background awareness (Martin 1995). And since it is uncontroversial that
deploying the sense of touch exteroceptively permits both the simultane-
ous and successive perception of a range of distinct objects, it follows that
the multiple-objects constraint is satisfied in all instances of tactile so-
matic proprioception, although, of course, in different ways.
Nonetheless, tactile somatic proprioception is only a part of somatic
proprioception as a whole, and even if the multiple-objects constraint is
satisfied in this limited sphere, it is illegitimate to extrapolate from this to
somatic proprioception as a whole. Fortunately, though, there is another
independent line of argument that can be adduced to show that the
multiple-objects constraint is satisfied by somatic proprioception. This is
also an argument from the phenomenology of somatic proprioception,
but it emerges from reflection on the phenomenology of sense experience
in general.
It is wrong to think that the senses form five distinct modes of experi-
encing the world (or six, if proprioception is included), each with its own
proper objects. One reason for discomfort with this suggestion is that it
creates epistemological problems with the cross-modal integration of
what, ever since Aristotle, have been known as the common sensibles.
Objects have spatial properties that can be detected by more than one
sense, corresponding more or less to what philosophers have classified as
primary qualities. These common sensibles are obviously the key to any
experience of an objective spatial world, since they define the spatial
properties of such a world, but the possibility of such experience seems
to be cast in doubt if, for example, the senses of touch and vision have
modality-specific objects (Evans 1985).
140 Chapter 6
These problems, moreover, are not simply epistemological. The very

possibility of acting on the world demands an integration of visual and
tactile information. To reach out for an object successfully, for example,
one has to integrate visual perceptions of the object (and of one’s own
hand) with the proprioceptive feedback, tactile and otherwise, about limb
position and movement and with the experienced shape of the object.
Even movements as simple as this would pose enormous computational
demands if the thesis of modality specificity were correct.
There is, however, powerful empirical evidence that the perceptual in-
formation systems subserving perceptual experience are not, in fact,
modality-specific. One important set of evidence comes from the experi-
mental work on neonatal imitation discussed in earlier chapters. The abil-
ity of neonates to imitate facial movements like mouth opening and
tongue protrusion can be interpreted in a way that clearly implies that
tactile proprioception and visual perception do not have different proper
objects.8 Facial imitation requires a complex cross-modal integration of
tactile proprioception and vision. And the particular significance of neo-
natal imitation is that because the imitation behavior can take place
almost immediately after birth, this integration cannot have been the
product of association or empirical correlation.
Another important set of empirical evidence comes from choice
reaction-time tasks, tasks involving two or more stimuli to which the sub-
ject has to make a selective response that is timed. Some such tasks are
matching tasks, where the spatial properties of the stimulus are isomor-
phic (to at least a limited degree) with the spatial properties of the re-
sponse. In mapping tasks, however, there is no such isomorphism. The
relation between spatial properties is arbitrary. The experimental evi-
dence shows fairly conclusively that matching tasks are performed with
much greater fluency and speed than mapping tasks. This has been taken
to show that there is a common coding structure between visual input and
motor output (Prinz 1990, Brewer 1993). Since motor output depends
on kinesthetic feedback from tactile proprioception, this in turn clearly
implicates a common coding structure between vision and tactile
proprioception.
A third set of empirical evidence comes from recent work in the neuro-
physiology of space perception. It is well known that lesions to the poste-
rior parietal cortex produce spatial deficits in humans and nonhuman

primates, and the inference frequently drawn is that the posterior parietal
cortex is the brain area where the representation of space is computed.
Recent neurophysiological work based on recordings from single neurons
has suggested that the distinctive contribution of the posterior parietal
cortex is combining information from various modalities to generate
head-centered, body-centered, and world-centered coordinate systems
(Andersen 1995). Information about visual stimuli is initially transmitted
in retinal coordinates. Calibrating this with information about eye posi-
tion yields head-centered coordinates, and further calibration with infor-
mation about body position (perhaps derived from the vestibular system)
yields a body-centered frame of reference.9 The distal targets of reaching
movements are encoded on this modality-free frame of reference, as are
motor commands and proprioceptive information about limb position.10
We need to move a step further, though, from considering the structure
and coding of information systems to the phenomenology of the percep-
tual awareness that those systems subserve. The crucial point here is that
our perceptual awareness of the world is cross-modal. Our perceptions of
the world form what Michael Ayers has described as an integrated sen-
sory field. He captures the phenomenology of the situation as follows:
A judgement which links the objects of different senses may itself be, and very
often is, an immediate perceptual judgement, directly grounded on the deliverance
of sense. Thus it is not normally as a result of inference, habitual association or
the like (although in a few peculiar cases of disorientation some form of inference
may be to the point) that I judge the object I feel with my hand to be the object I
see. Quite simply, I perceive it as the same: the identity enters into the intentional
content of sensation, and of my total sensory ‘field’. (Ayers 1991, 2: 187)
It seems to me that Ayers’s description is absolutely right. What we expe-
rience in sense perception is a presentation of the world that integrates
information from all modalities. It is in fact very rare that we have
modality-specific perceptions. This is clearly the case when we are dealing
with the perception of common sensibles. The shape that I see is seen as
the very same shape that I could close my hand around. But it also applies
to the special sensibles. As Ayers points out, when I taste something in my
mouth, my gustatory awareness of its taste is integrated with my tactile
awareness of its taste and my proprioceptive awareness of its heat. These
are different, although often attentively distinguishable, components
142 Chapter 6
of a single perceptual experience. The same applies to auditory experi-

ence of sounds as coming from a particular direction. Such experiences
are inherently relational, and only one of the relata is itself heard. Somatic
proprioception needs to be brought to bear to provide the directionality.
A striking result of this cross-modal integration is that all experience
emerges as inherently spatial. One way of capturing this idea of an inte-
grated sensory field would be through the concept of an egocentric
space.11 Sense perception generally involves localizing what is perceived
within an egocentric frame of reference centered on the perceiver’s body.
Within this space the body, of course, occupies a privileged position, be-
cause objects are located relative to axes whose origin lies in the body, and
this is part of the overall significance of somatic proprioception. Somatic
proprioception provides the sort of information that makes possible ori-
entation and action within egocentric space. An example of the signifi-
cance of somatic proprioception is provided by familiar experiments with
lenses that invert the visual field. As is well known, the brain adjusts to
the inverting lenses after a relatively short period of time, and the per-
ceiver starts to see the world the right way up again. The information
about correct orientation that results in the visual field being successfully
reinverted clearly does not come from vision, because all visual informa-
tion is inverted. It must be information derived from the known orienta-
tion of the body, and hence with its source in somatic proprioception.
The same holds for the integration of vision and action. The locus of
action is egocentric, or body-relative, space, and the possibility of action
depends both on correct information about the location of objects relative
to body-centered axes and on correct information about the location and
movement of body-parts. Somatic proprioception is crucial for both of
these tasks.
The picture that is emerging is of perceptual experience involving a
sensory field that is integrated across modalities and in which somatic
proprioception plays a crucial role. Now, according to the multiple-
objects constraint, genuine instances of sense perception permit the per-
ception of a range of objects, either simultaneously or successively. It was
initially suggested that somatic proprioception could not possibly satisfy
this constraint, because somatic proprioception has the body as its sole
object of awareness. We have already explored how this seems false in the
case of tactile somatic proprioception. But it now seems that we can

extend this point to somatic proprioception as a whole. What has been
confusing the issue here is the thought that somatic proprioception
should be considered in isolation, as a distinctive set of information sys-
tems subserving several different forms of awareness. When propriocep-
tion is so considered, it does seem natural to think that its sole object of
awareness is the body. But, although somatic proprioception does give us
information about our bodies, it does so as one of a range of ways we
have of finding out about the properties of our bodies in the context of
an overall perceptual field, which most emphatically does satisfy the
multiple-objects constraint. Quite possibly, if we were so constituted as
to experience the world in the fragmented and disjointed way assumed
by many of the early empiricists, then it would be proper to deny that
somatic proprioception can meet the multiple-objects constraint. But be-
cause of how we are constituted and how somatic proprioception is inte-
grated into the general sensory field, the multiple-objects constraint is
satisfied.
I turn now to the identification constraint. One reason for denying that
it can be satisfied by somatic proprioception would be that genuine in-
stances of identification involve the possibility of misidentification and
hence cannot be immune to error through misidentification relative to the
first-person pronoun (in the sense identified in section 1.2). Since somatic
proprioception is immune to this sort of error, it might be thought that
somatic proprioception cannot satisfy the identification constraint. A fun-
damental problem, however, with interpreting the identification con-
straint in terms of being subject to error through misidentification is that
it seems to exclude paradigm cases of perceptual awareness. Perceptual
demonstrative judgements are themselves immune to error through mis-
identification (relative to the demonstrative pronoun). When I perceive an
object and judge ‘That is green’, I cannot misidentify the reference of the
pronoun ‘that’, although I may, of course, fail to refer (Shoemaker 1986,
Cassam 1995). It follows from this, of course, that being subject to error
through misidentification cannot be a necessary feature of perceptual
awareness, and hence that a broader reading must be given of what is to
count as perceptual identification for the purpose of the identification
constraint.
144 Chapter 6
A plausible proposal here would be to view perceptual identification

in terms of the capacity to keep track of an object over time. The rele-
vant type of error then becomes losing track of an object, as opposed to
misidentification. But it is far from clear that somatic proprioception does
exclude losing track of an object, pace Cassam 1995. It does seem that
one can lose track of one’s body over time, or at least of various parts of
one’s body. This happens, for example, when one absent-mindedly walks
home on automatic pilot instead of to the shops or taps one’s foot in time
to a piece of music without noticing. The best reason for describing these
as cases of losing track of what one’s body is doing is the feeling of sur-
prise that comes when one notices what has been going on. What can
confuse matters here is a failure to make the distinction stressed earlier
between proprioceptive awareness and the operation of the propriocep-
tive information systems. In the two examples just given, the proprio-
ceptive information systems are functioning as they must if actions like
walking are even to be possible. But just because they are functioning,
this doesn’t mean that the subject is keeping track of his body, any more
than does the fact that the proprioceptive information systems continue
to function while one is asleep. What counts is the lack of the appropriate
sort of proprioceptive awareness.
This is not to imply that somatic proprioception is subject to error
through misidentification relative to the first-person pronoun.12 Quite the
contrary. Although it is true that if, per impossibile, one were propriocep-
tively to misidentify another’s body for one’s own one would have lost
track of one’s own body, it does not follow (nor is it true) that losing track
of one’s body means misidentifying another’s body for one’s own, or that
when I am in touch with my body I can be mistaken about whether it is
mine or not.13 This is very important because of the close connection be-
tween judgements that are immune to error through misidentification rel-
ative to the first-person pronoun and genuinely first-person judgements—
a connection stressed by Shoemaker himself (1968), as well as by other
philosophers (Evans 1982, chap. 7).14 As I shall argue in the next section,
the fact that somatic proprioception is itself immune to error through
misidentification relative to the first-person pronoun provides vital
support for the claim that somatic proprioception counts as a genuine
form of self-consciousness. For present purposes, the important point to
note is that, although somatic proprioception can count as a form of

perception only in virtue of meeting the multiple-object and identifi-
cation constraints, this does not mean that it is subject to error through
misidentification.
Moreover, this conception of how the identification constraint is satis-
fied through keeping track of one’s body fits well with the general picture
of the phenomenology of sense perception sketched earlier in terms of an
integrated sensory field. If the phenomenology of sense perception is in-
deed integrated rather than fragmented, then the body will appear as one
object among the range of objects that appear in the integrated sensory
field. Of course, even though this might be enough to ensure that the body
appears perceptually in a way that satisfies the multiple-objects con-
straint, it is undeniable that (at least in normal subjects) the body appears
as a highly distinctive object. This is due partly to certain aspects of the
phenomenology of visual proprioception but principally to the normal
functioning of the proprioceptive awareness that keeps track of the body
as the center and focal point of body-relative egocentric space. Proprio-
ceptive awareness is critically responsible for the body appearing as a dis-
tinctive object in the integrated sensory field.
That leaves just the object constraint, and since neither of the other
constraints are effective against the simple argument, it seems justifiable
to interpret the object constraint in a manner favorable to the thesis that
somatic proprioception is a form of self-perception. As far as the simple
argument is concerned, lines (2) and (3) have been vindicated. In the fol-
lowing section I turn to line (4) and the claim that the self-perception
yielded by somatic proprioception is a form of self-consciousness.
6.4 Somatic Proprioception as a Form of Self-Consciousness
The move from (3) to (4) in the simple argument might be attacked as
resting on the fallacy of equivocation. What makes somatic propriocep-
tion a form of self-perception, it might be argued, is that it involves per-
ceiving certain bodily properties that are properties of one’s self. But this
is not enough to make it a form of self-consciousness. Genuine self-
consciousness requires the further feature that those bodily properties
should be perceived as properties of one’s self. And then one might
146 Chapter 6
suggest both that there are no reasons to think that somatic propriocep-
tion involves anything along those lines and that genuine self-
consciousness of this sort is only available to creatures who have mastered
the first-person pronoun.
There are two considerations, however, that militate against this sort
of objection. The first is that if the distinction between self-perception
and genuine self-consciousness is to apply to somatic proprioception of
one’s own bodily properties, then it should also apply to introspection of
one’s own psychological properties. This would mean that introspection
of one’s own psychological properties could also fail to count as a form
of self-consciousness, and that seems to be a rather bizarre conclusion.
Of course, it is open to the objector to bite the bullet here, but there is a
prima facie implausibility in any view that denies that my awareness of
my own psychological properties is a form of self-consciousness. It is
equally implausible, I feel, to deny that there is a parallel between somatic
proprioception and introspection. What the objector is claiming is that
an individual can perceive certain bodily properties through somatic pro-
prioception without perceiving that those properties are his own proper-
ties. Such a perception might be specified as follows:
(1) x thinks ‘That leg is flexing’.
The objector maintains that (1), as contrasted, for example, with (2), is
not a genuine form of self-consciousness.
(2) x thinks ‘My leg is flexing’.
Note, however that an exactly parallel distinction can be made for the
introspection of psychological properties. Compare (3) and (4):
(3) x thinks ‘That’s a feeling of jealousy’.
(4) x thinks ‘I’m jealous’.
Both (3) and (4) are introspectively based judgements. If the distinction
between (1) and (2) is telling against the view that somatic proprioception
is a form of self-consciousness, the parallel distinction between (3) and
(4) should be telling against the view that introspection is a form of self-
consciousness.15
But, of course, neither the distinction between (1) and (2) nor that be-
tween (3) and (4) is relevant here. Neither somatic proprioception nor
introspection can support as sharp a distinction between “mere” self-

perception and “genuine” self-consciousness as the objection requires. To
see why, consider the following example, where there really is precisely
such a sharp distinction. Suppose, for example, that I catch sight of some-
body in a mirror without realizing that the person is myself and I think
‘That person has got clenched fists’. This would be a situation that might
perhaps be described as an instance of self-perception, but it is clearly not
an instance of self-consciousness, and this, of course, is precisely
the distinction that the objection is trying to capture. What is not clear,
however, is that this distinction carries over in any straightforward way
to the case of somatic proprioception. What is distinctive about the
mirror case (and what makes it so obviously not an example of gen-
uine self-consciousness) is the fact that I misidentify myself. I correctly
identify certain bodily properties but fail to recognize that they are
properties of my own body and misattribute them to somebody else. But
nothing like this goes on in somatic proprioception. In fact, nothing like
this can go on in somatic proprioception. One of the distinctive fea-
tures of somatic proprioception is that it is subserved by information
channels that do not yield information about anybody’s bodily proper-
ties except my own (just as introspection does not yield information
about anybody’s psychological properties except my own). It follows
from the simple fact that I somatically proprioceive particular bodily
properties and introspect particular psychological properties that those
bodily and psychological properties are my own. This feature of somatic
proprioception (that, as mentioned earlier, it is derived from information
that is immune to error through misidentification relative to the first-
person pronoun) militates strongly against the idea that the proposed
contrast between (putatively accidental) self-perception and genuine self-
consciousness is sharply defined in either somatic proprioception or
introspection.16
Still, granting that somatic proprioception can be a source of thoughts
that are about oneself in a way that provides no purchase for the distinc-
tion between accidental and genuine self-reference does not explain
why somatic proprioception is properly described as a form of self-
consciousness. There is a substantive issue here, which is whether so-
matic proprioception falls within the same cognitive kind as those forms
148 Chapter 6
of “full-fledged self-consciousness” that center around the competent

employment of the first-person pronoun. But how can we reach a nonsti-
pulative solution to this question? The strategy I propose is to identify
what might be termed the core requirements of genuinely self-conscious
thought. If those core features can be specified without presupposing
mastery of the first-person pronoun and if those features are also shared
by somatic proprioception, then we will have a principled reason for
holding that somatic proprioception is properly described as a form of
self-consciousness.
So what are the core requirements for genuinely self-conscious thought?
It seems to me that there are two such requirements. The first has already
been discussed in this section. This is the requirement that genuinely self-
conscious thought should be about oneself in a way that is nonaccidental.
It cannot be the case that genuinely self-conscious thought ascribes prop-
erties to an individual who is in fact oneself without one’s being aware
that the individual in question actually is oneself. The second requirement
is closely related to this requirement of nonaccidental self-reference. Sup-
pose that we ask why accidental self-reference fails to count as genuine
self-consciousness. The obvious answer is that thoughts that accidentally
self-refer fail to have the immediate implications for action found in genu-
˜
inely self-conscious thoughts, as has been made familiar by Castaneda
(1966) and Perry (1979). This second identifying feature of genuinely
self-conscious thoughts—that they feed directly and immediately into ac-
tion—is an essential part of the functional characterization of first-person
thoughts (as emerged in chapter 5).
I have already suggested that somatic proprioception cannot give rise
to thoughts that are accidentally about oneself in this manner. This flows
out of the discussion of immunity to error through misidentification rela-
tive to the first-person pronoun. And it is equally clear that somatic pro-
prioception does have immediate and direct implications for action. This
holds true both of self-directed action (like scratching itches) and of ac-
tion whose target is not the body. And the role of the various somatic
information systems in providing feedback about limb movement and
balance provides sufficient testimony to these implications for action. I
will return to the connections between somatic proprioception and action
in the final section.
That somatic proprioception satisfies these two requirements does not,

however, exhaust the sense in which somatic proprioception counts as a
form of self-consciousness. To make further progress here, we need to
distinguish between two distinct modes of self-consciousness yielded by
somatic proprioception: the narrow and the broad. What I have been dis-
cussing so far is properly described as narrow self-consciousness. It might
be described as a form of self-knowledge, knowledge of one’s bodily prop-
erties. In addition to this, however, somatic proprioception yields a form
of broad self-consciousness. At the core of the notion of broad self-
consciousness is the recognition of what developmental psychologists call
self-world dualism. Any subject properly described as self-conscious must
be able to register the distinction between himself and the world. Of
course, this is a distinction that can be registered in a variety of different
ways. The capacity for self-ascription of thoughts and experiences, in
combination with the capacity to conceptualize the world as a spatial and
causally structured system of mind-independent objects, is a high-level
way of registering this distinction. But there are more primitive ways of
registering the distinction, provided by somatic proprioception. Somatic
proprioception provides a way, perhaps the most primitive way, of reg-
istering the distinction between self and nonself. This is a weaker dis-
tinction than the distinction between self and world, of course, but it is
certainly a necessary component of it. There are two key elements to the
self/nonself distinction yielded by somatic proprioception. The self here
is, of course, the material self, and it is useful to characterize these two
elements in terms of the different ways of grasping the bodily self that
these elements involve and provide.
The first element of broad self-consciousness that somatic propriocep-
tion provides is an awareness of the limits of the body. The sense of touch
is primarily responsible for this. The felt boundaries of the body define
the limits between self and nonself. This is directly a function of the points
about the joint proprioceptive and exteroceptive structure of tactile
awareness mentioned earlier in the context of the multiple-objects con-
straint. Both the proprioceptive and the exteroceptive dimension of touch
are phenomenologically salient (and possible objects of attention) in tac-
tile experience. Because of this, there is an important sense in which the
distinction between self and nonself is itself phenomenologically salient.
150 Chapter 6
This offers a way of grasping the body as a spatially extended and

bounded object.
A second element of broad self-consciousness emerges when one re-
flects that somatic proprioception is one of the key ways by which one
becomes aware that the body is responsive to one’s will. The feedback
gained through kinesthesia, joint-position sense, and the vestibular sys-
tem makes one aware that the body is responding to motor commands.
In combination with the previous mode of broad self-consciousness, this
yields a way of grasping the body as an object that is responsive to the
will. True, much of the body is not at all responsive to the will—the inter-
nal organs are obvious examples. But strikingly, the scope of the will does
seem clearly to encompass all bodily surfaces and extremities (whether
directly or indirectly). Although not every portion of the bodily surface
can be moved at will, every portion of the bodily surface can nonetheless
be experienced as moving in response to an act of the will. Take an arbi-
trary area on the top of the head, for example. Although I cannot move
that area at will (in isolation), I can nonetheless experience it as moving
when I move my head as a whole. It seems fair to say that the limits of
the will mark the distinction between the self and the nonself just as much
as does the skin, although in a different way.
Philosophers dealing with bodily awareness have to balance two facets
that pull in opposite directions. There is, first, the obvious fact that the
body is a physical object in the world, which leads to the plausible
thought that awareness of the body must be somehow awareness of the
body as an object in the world. Second and equally obvious is the fact
that the body is (at least from a first-person perspective) quite unlike any
other physical object, which leads to the equally plausible thought that
awareness of the body is somehow fundamentally different from aware-
ness of the body as an object in the world. The two different facets just
isolated offer a way of reconciling the apparent conflict. What somatic
proprioception offers is an awareness of the body as a spatially extended
and bounded physical object that is distinctive in being responsive to
the will.
It is important to stress, however, that narrow self-consciousness and
broad self-consciousness are not distinct existences. On the contrary,
broad self-consciousness features ineliminably in the content of narrow
somatic self-consciousness. This is the final step in the argument that so-
matic proprioception counts as a genuine form of self-consciousness. To
see why, reflect that the critic of the simple argument who finds a fallacy
of equivocation is effectively drawing precisely the distinction between
broad and narrow self-consciousness. He is pointing out that narrow self-
consciousness is all that follows from the conjunction of (1) and (2),
whereas something more like broad self-consciousness is required for (4).
To appreciate why this suggestion is fundamentally mistaken, however, a
closer look at the content of somatic proprioception is required.
6.5 The Content of Somatic Proprioception
It has been noted by several authors that the content of proprioception is

a form of spatial content (O’Shaughnessy 1980, Brewer 1993). The
bodily states that form the objects of proprioception are proprioperceived
in a way that locates them within a space that is usually (although not
always) bounded by the limits of the body. The qualification is needed
because of two well-documented phenomena from clinical medicine. Pa-
tients with amputated limbs often report feeling “phantom limbs,” as do
some patients with congenital absence of limbs. In cases of phantom
limbs the limits of the proprioperceived body extend beyond the limits of
the physical body, with sensations located elsewhere in the body being
referred to and experienced in the phantom (Melzack 1992, Ramachan-
dran 1994).17 Phantoms are often highly incorporated into the body
scheme. There are widespread reports of patients being able to move their
phantoms voluntarily, and the patient F. A. studied by Ramachandran
regularly attempted to employ his phantom right arm to ward off blows,
grasp objects, or break a fall. The example of phantom limbs clearly
shows that the space of the proprioceived body sometimes extends be-
yond the bounds of the physical body. A converse set of disorders, the
so-called somatoparaphrenic delusions, shows that the space of the pro-
prioceived body can be narrower than the bounds of the physical body.
Patients who have suffered from right-hemisphere stroke damage (usually
in conjunction with damage to the right parietal lobe) are often anosag-
nosic about their deficit (that is, they deny its existence). A small subset
of such patients are led by their anosagnosia to deny that the paralyzed
152 Chapter 6
limb belongs to them.18 Nonetheless, in both these pathological cases,

proprioceptive states are still experienced within the frame of a body-
relative space. The first question that arises in thinking about propriocep-
tive content, therefore, is how we should think about the frame of refer-
ence that determines how proprioceptive states are located within body-
relative space.
Proprioception is not, of course, the only source of body-relative spatial
contents. It is highly plausible to view perception and basic bodily actions
as respectively providing or based on body-relative spatial contents, and
it is natural to suggest that perception, proprioception, and the intentions
controlling basic bodily actions must have spatial contents coded on com-
parable frames of reference. This would make it relatively easy to answer
the question about the frame of reference of proprioceptive content, since
the spatial content of perception and basic bodily intentions is relatively
well understood. The spatial locations of perceived objects and objects
featuring in the contents of intentions are given relative to axes whose
origin lies in the body—what is often termed an egocentric frame of refer-
ence.19 An initial proposal, therefore, might be that the axes that de-
termine particular proprioceptive frames of reference are centered on
particular body parts, just as are the axes determining the frames of refer-
ence for perceptual content and basic intentions.20
Despite its appealing economy, however, this account is ultimately un-
acceptable, because of a fundamental disanalogy between the bodily
space of proprioception and the egocentric space of perception and ac-
tion. To appreciate the problem, consider what reason there could be for
locating the origin of proprioception in one body part rather than an-
other. In the case of vision or exteroceptive touch it is natural to think
that there is a perceptual field bounded so as to determine a particular
point as its origin. If, for example, the visual field is described as the solid
angle of light picked up by the visual system, then the origin of the visual
field can be taken to be the apex of that solid angle. Similarly, one might
suggest that the origin of the frame of reference for exploratory touch
is a point in the center of the palm of the relevant hand. But somatic
proprioception is not like this at all. It is not clear what possible reason
there could be for offering one part of the body as the origin of the pro-
prioceptive frame of reference.
In contrast with vision, audition, and the other canonically exterocep-

tive modalities, there are certain spatial notions that do not seem to be
applicable to somatic proprioception. For any two objects that are visu-
ally perceived, it makes obvious sense to ask both of the following ques-
tions: (a) Which of these two objects is farther away? (b) Do these objects
lie in the same direction? Moreover, the possibility of asking and answer-
ing these questions is very closely bound up with the fact that visual per-
ception takes place relative to a frame of reference that does indeed have
an origin. Posing question (a) is more or less equivalent to asking whether
a line between the origin and one object would be longer or shorter than
a corresponding line between the origin and the other object. Question
(b) is just the question of whether, if a line were drawn from the origin to
the object farther away, it would pass through the nearer object. This, of
course, is an important part of the explanation of why we can normally
answer questions like (a) and (b) at a glance.
Neither of these questions makes sense, however, with respect to pro-
prioception. One does not find oneself asking whether this propriocep-
tively detected hand movement is farther away than this itch, or whether
this pain is in the same direction as that pain. Nor, of course, is this merely
a peculiarity of our language. What I am really asking when I ask which
of two objects is farther away is which of the two objects is farther away
from me, and a similar tacit self-reference is included when I ask whether
two objects are in the same direction. The reason one cannot ask these
questions with regard to somatic proprioception is that the events about
which one learns through somatic proprioception take place within the
confines of the body, and there is no privileged part of the body that
counts as me for the purpose of discussing the spatial relations they bear
to each other (in the way that, for example, the origin of the visual frame
of reference can count as me when those questions arise for visually pre-
sented objects).
The conclusion to draw from this is that the spatial content of somatic
proprioception cannot be specified within a frame of reference that takes
the form of axes centered on an origin. So if somatic proprioception is to
have a content at all, it must be understood in a fundamentally different
way. In the remainder of this chapter I will develop an account that cap-
tures these distinctive features of proprioceptive content.
154 Chapter 6
Let me begin by distinguishing two distinct aspects of proprioceptive

content: the descriptive and the spatial. Roughly speaking (and to antici-
pate somewhat) the content of somatic proprioception is that certain
events are taking place at particular bodily locations. The nature of such
an event is specified by the descriptive aspect of proprioceptive content,
while the spatial aspect specifies the bodily location at which such an
event takes place. I will begin by discussing the spatial aspect of proprio-
ceptive content.
A satisfactory account of any form of spatial content needs to explain
how places are individuated by providing criteria for sameness of place.
In the case of somatic proprioception, of course, we need criteria for
sameness of bodily location. What complicates matters here is that on the
face of it there seem to be several different forms of criteria for sameness
of bodily location that individuate places in different ways. Consider the
following two situations:
1. I have a pain at a point on the sole of my right foot when I am standing
up, and my right foot is resting on the ground in front of me.
2. I have a pain at the same point on the sole of my right foot when I am
sitting down, and my right ankle is resting on my left knee.
According to one set of criteria, the pain is in the same bodily location in
(1) and (2), that is to say, it is at a given point on the sole of my foot.21
According to another set of criteria, however, the pain is in different
bodily locations in (1) and (2) because the sole of my foot has moved
relative to my other body parts. Let me term these the A location and B
location respectively. Note that the B location is independent of the actual
location of the pain in objective space. The B location of the pain in (2)
would be the same if I happened to be sitting in the same posture five feet
to the left.
With bodily sensations still as paradigms of somatic proprioception,
both A location and B location are ineliminable components of the phe-
nomenology of bodily sensations. I take it that the phenomenology of (1)
is fundamentally different from the phenomenology of (2), which shows
that specifications of proprioceptive content must be sensitive to B loca-
tions. But it is also true that it would be hopelessly incomplete to report
the phenomenology of (2) by saying merely that I feel a pain just above
and to the left of my left knee down and slightly left of my left shoulder,
although (some version of) that would be sufficient to pick out the B
location. This shows that specifications of proprioceptive content must
also be sensitive to A locations.
Now both the A location and B location need to be specified relative
to a frame of reference, which requires taking a fixed point or set of fixed
points relative to which spatial location is to be fixed. Since it emerged
earlier that it would not be right to take a single fixed point as an origin,
it follows that we must look for a set of fixed, or at least relatively fixed,
points in terms of which we can fix the A location and B location of a
given bodily event.
In thinking about the possibility of such a frame of reference we need
to bear in mind that the human body has both moveable and (relatively)
immoveable body parts. On a large scale, the human body can be viewed
as an immoveable torso to which are appended moveable limbs: the head,
arms, and legs. Within the moveable limbs there are small-scale body
parts that can be moved directly in response to the will (like the fingers,
the toes, and the lower jaw) and others that cannot (like the base of the
skull), but all the small-scale body parts within the moveable limbs are
moveable in the limited sense that their position can be altered relative to
any given point in the (relatively) immoveable torso.22 Let me now intro-
duce the technical concept of a hinge. The intuitive idea that I want to
capture with this term is the idea of a body part that allows one to move
a further body part. Examples of hinges are the neck, the jaw socket, the
shoulders, the elbows, the wrists, the knuckles, the finger joints, the leg
sockets, the knees, and the ankles. The distinction between moveable and
immoveable body parts, together with the concept of a hinge, creates the
following picture of how the human body is segmented. A relatively im-
moveable torso is linked by hinges to five moveable limbs (the head, two
legs, and two arms), each of which is further segmented by means of fur-
ther hinges.
It is, I think, to the hinges that we need to turn to find the fixed points
in terms of which we can give the particular A location and B location
of individual body parts at a time. One good reason for this is that an
awareness of the location of the hinges, as well as of the possibilities for
movement that they afford, can plausibly be viewed as an inevitable con-
comitant of learning to act with one’s body in early childhood. Another
156 Chapter 6
good reason is that individual body parts are paradigmatically identified

in terms of hinges. The forearm, for example, is the volume between the
elbow and the wrist. The palm of the hand is the volume bounded by the
wrist and the knuckles. The calf is the volume of leg that falls between
the knee and the ankle. Using hinges provides a nonarbitrary way of seg-
menting the body that accords pretty closely with how we classify body
parts in everyday thought and speech.
Let me start by explaining how the hinges can be deployed to determine
any given A location. A particular bodily A location is specified relative
to the hinges that bound the body part within which it is located. A par-
ticular point in the forearm is specified relative to the elbow and the wrist.
It is the point that lies on the surface of the skin at such and such a dis-
tance and direction from the wrist and such and such a distance and direc-
tion from the elbow.23 This mode of determining A locations secures the
defining feature of an A location, which is that a given point in a given
body part will have the same A location irrespective of how the body as
a whole moves or of how the relevant body part moves relative to other
body parts. The A location of a given point in a given body part will
remain constant in both those movements, because neither of those move-
ments brings about any changes in its distance and direction from the
relevant hinges. Note, however, that this holds true only if the fixed points
are restricted to those hinges that bound the relevant body part. If the num-
ber of fixed points were expanded to include nonbounding hinges, this
would have the undesirable consequence that the A location would vary.
I turn now to the more complicated issue of B location. The first point
to observe is that it is individual A locations that have particular B loca-
tions. This is clear from situations (1) and (2) described above, in which
a pain retains a constant A location in two different B locations. In each
situation what has the relevant B location is the pain that is A-located at
a particular point in the sole of my foot. It is not the pain simpliciter that
in (2) is B-located just above and to the left of my left knee and down and
slightly left of my left shoulder, but the pain in the sole of my foot. This
provides an important clue as to how the account of B location ought to
proceed. If what has a given B location is an A location and A locations
are identified within given body parts (relative to the hinges that bound
those body parts), then it seems we can map an A location onto a B loca-
tion if we can plot the location of the given body part. How might this
be done?
Body parts have been defined as body segments bounded by hinges.
These hinges afford a range of movements in three dimensions (with some
hinges, like the shoulder, affording a greater range of movements than
others, like the elbow). A hinge such as the wrist allows the hand to be
positioned in one of a range of orientations relative to the lower arm. The
basic idea in mapping A locations onto B locations is to specify what
orientation the body part is in relative to the hinge. Take a point in the
center of the palm of the hand. That point has its A location relative to
the hinges that bound the palm of the hand, while its B location is given
by the orientation of the palm of the hand relative to the wrist. Of course,
this is not enough to fix a unique B location, for it fails to register the
changing B location of a point in the center of the palm of my hand as I
keep my hand in the same orientation relative to my wrist but raise it
above my head. Accommodating such changes in B locations is possible
only if the orientation of the relevant body part is relativized more widely.
In the example of the point in the palm of the hand, the orientation of
the wrist needs to be fixed relative to the elbow, and the orientation of
the elbow relative to the shoulder. That would provide a unique B loca-
tion within the arm and would also specify the location of the arm relative
to the torso.
The general model, then, for the identification of B locations is as fol-
lows. A particular constant A location is determined relative to the hinges
that bound the body part in which it falls. That A location will either fall
within the (relatively) immoveable torso, or it will fall within a moveable
limb. If it falls within the (relatively) immoveable torso, then its B location
will also be fixed relative to the hinges that bound the torso (neck, shoul-
ders, and leg sockets). If, however, that A location falls within a moveable
limb, then its B location will be fixed recursively relative to the hinges
that lie between it and the immoveable torso. This enables any given B
location to be calibrated with any other. Suppose that I want to scratch
an itch in my left arm with the tip of the middle finger of my right hand.
Both A locations have B locations recursively specified as above. Each of
those recursive specifications will relativize the B location to the respec-
tive shoulder. Thus, all that is needed for those B locations to be fixed
158 Chapter 6
relative to each other (and hence for me to locate the itch in my arm with
the middle finger of my right hand) is a specification of the position of
each shoulder relative to the other.
This account remains faithful to the phenomenology of bodily expe-
rience. An important part of the phenomenology of bodily experience is,
as several writers have stressed, that bodily sensations such as pain are
presented not only as being at a particular point relative to other parts of
the body but also as located within a particular body part (Martin 1993,
Brewer 1993). This is captured very clearly by the distinction between
and interdependence of A locations and B locations.24 A bodily sensation
has a particular B location (relative to the body as a whole) only insofar
as it has a particular A location (within a particular body part), because B
locations are fixed on the basis of A locations. Moreover, the interrelation
between A locations and B locations captures the further phenomenologi-
cal point that body parts are perceived as belonging to the body as a
whole (Martin 1995). But explaining properly why this is so brings us to
what I have termed the descriptive dimension of proprioceptive content.
Earlier I contrasted characterizing the nature of a proprioperceived
event within the body (the descriptive dimension) with specifying where
that event is in the body (the spatial dimension). The first question to be
asked of the descriptive dimension of proprioceptive content concerns the
nature of the proprioperceived event. What is the direct object of proprio-
ception? The answer to this question can be read off from the understand-
ing that we now have of the spatial dimension of proprioceptive content.
The direct object of proprioception is the state of the body at a particular
location. In terms of A location and B location, this means that the direct
object of proprioception will be the state of the body at a particular loca-
tion in a given body part, which is itself located relative to the rest of the
body. Note one consequence of this, to which I will return at the end of
this section. The body as a whole features in the descriptive dimension of
proprioceptive content, just as it does in the spatial dimension.
The descriptive aspect of proprioceptive content is that the body at a
particular location is in a particular state. But what sort of state? It may
be helpful to consider the range of bodily states that feature in proprio-
ceptive content under the familiar headings of ‘quality’ and ‘quantity’.
The qualitative states are those familiar from the phenomenology of
bodily sensation. Obvious examples are the states of being bruised, being
damaged, being tickled, being itchy, being tender, being hot. Most of the
qualitative states featuring in proprioceptive content share the feature of
departing from what one might term bodily equilibrium. One prominent
reason for the body to obtrude on consciousness is that it is in an abnor-
mal condition of one form or another. This is a respect in which the quali-
tative states generally differ from the quantitative ones. The sensation that
a limb is moving in a particular direction or the feeling that one’s legs are
crossed are paradigm quantitative states. Generally, these are not, of
course, departures from bodily equilibrium but rather ways in which one
keeps track of one’s body in its normal operations and activities. Of
course, the states that are the objects of particular proprioceptive contents
can have both qualitative and quantitative features, as when one feels that
one’s bruised ankle is swollen. They can also have several qualitative and/
or quantitative features at once, as when one clasps one’s hands together
and moves them (perhaps to hit a volleyball).
Now any acceptable account of content must explain the correctness
conditions of the relevant content-bearing states. Of course, any given
proprioceptive content will be correct if and only if there is an event tak-
ing place at the appropriate bodily location with the relevant qualitative
and/or quantitative features. This is true but uninformative, however.
What we really need is some indication of the criteria by which one might
recognize whether these correctness conditions are satisfied or not. This
is, of course, an epistemic rather than a constitutive issue. Let me make
some brief comments in this direction.
The key to understanding how the correctness conditions of proprio-
ceptive content can be applied is the functional role of content-bearing
proprioceptive states with regard to the events that cause them and the
actions to which they give rise. Some of these actions are explicitly di-
rected towards the body (e.g., scratching an itch). Others are implicitly
directed towards the body (e.g., snatching one’s hand away from a flame).
Others are not body-directed at all (e.g., the role of what I termed quanti-
tative features in controlling action). In each of these cases, however, it is
possible to employ the concept of an appropriate action to illuminate the
correctness conditions. The correctness conditions for explicitly body-
directed actions, like scratching itches, are that the action should be
160 Chapter 6
appropriate to the disturbance that causes the proprioceptive state. This

requires, of course, that there actually be a disturbance at the bodily loca-
tion at which the action is directed (which would satisfy the correctness
condition for the spatial aspect) and that the action be appropriate to the
disturbance there (which would satisfy the correctness condition for the
descriptive aspect). For proprioceptive states that cause implicitly body-
directed actions, the correctness conditions are similar. One might view
an appropriate action in both of these cases as an action that would tend
to restore bodily equilibrium, for example, by relieving the pain.
The correctness conditions of proprioceptive states that cause non-
body-directed actions are slightly more complicated in that (unlike bodily
sensations) they are not necessarily linked to isolated and easily identifi-
able events taking place in the body. Typically, these are proprioceptive
states reporting what I termed quantitative features, like general limb dis-
position and movement. Here, though, we can see how the correctness
conditions might be brought to bear by considering the model sketched
earlier, in which two different types of proprioceptive information (about
initial limb disposition and then feedback about limb movement) are im-
plicated in paradigm cases of intentional action where a motivational
state and a perceptual state jointly produce an intentional command. The
obvious fact that correctness of the proprioceptive information is a neces-
sary condition of the success of the intentional action illustrates how one
might recognize whether the correctness conditions are satisfied: the cor-
rectness conditions are satisfied if the relevant perception is true and the
relevant motivational state is satisfied.25
This brief sketch leaves many questions unanswered, but I take it that
the account so far is secure enough for me to bring this discussion of
proprioceptive content to bear on the argument that somatic propriocep-
tion is a genuine form of self-consciousness. In the previous section I dis-
tinguished between broad self-consciousness (awareness of the material
self as a spatially extended and bounded physical object distinctive in its
responsiveness to the will) and narrow self-consciousness (the material
self’s knowledge of its bodily properties as given in proprioceptive
contents of the type I have been discussing). Instances of narrow self-
consciousness will have contents with both spatial and descriptive dimen-
sions. The body as a whole features in both of these dimensions of
content. This is particularly clear in the case of the spatial dimension,

which involves an awareness of the body as an articulated structure of
body parts separated by what I have termed hinges. This implicates an
awareness of the self as spatially extended and bounded. But it also impli-
cates the second component of broad self-consciousness (awareness of the
bodily self as responsive to the will). This is because awareness of the
hinges is closely bound up with awareness of the body’s possibilities for
action. The body presents itself phenomenologically as segmented into
body parts separated by hinges because those are the natural units for
movement. In the descriptive aspect of proprioceptive content, moreover,
it is once again the body as a whole that features. The content of somatic
proprioception is that the body, at a particular A location and a particular
B location, is in a particular state with certain qualitative and/or quantita-
tive features. Each such content exemplifies both broad and narrow self-
consciousness.
6.6 Summary
In section 6.2, I offered the simple argument to show that somatic propri-
oception is a form of primitive self-consciousness. The crucial claims in
the simple argument were, first, that somatic proprioception is a form of
perception and, second, that as self-perception it is a form of self-
consciousness. These claims were defended in sections 6.3 and 6.4 respec-
tively. The soundness of the simple argument shows that somatic proprio-
ception is a source of what in earlier chapters I described as first-person
contents. Like the first-person perceptual contents discussed in the previ-
ous chapter, these contents are primitive and available more or less from
birth.26 In at least one respect, the contents of somatic proprioception are
richer than the first-person contents of perceptual protobeliefs. As has
emerged from the discussion of the content of somatic proprioception,
they encompass a genuine awareness of the limits and responsiveness to
the will of the embodied self. Somatic proprioception is a source of infor-
mation not just about particular properties of the embodied self but also
about the nature of the embodied self itself.
The pick-up of self-specifying information in ecological perception and
the nonconceptual first-person contents of somatic proprioception are the
162 Chapter 6
most primitive forms of self-consciousness. One implication of this is that

it widens the scope of what might be termed the first-person perspective
far beyond the domain of humans, and even the higher mammals. This is
particularly significant for any philosopher who shares the plausible view
that self-consciousness, even in its primitive forms, carries with it a degree
of moral significance. A further implication more relevant to present con-
cerns is that the first-person nonconceptual contents of ecological percep-
tion and somatic proprioception are the basic building blocks from which
the full-fledged self-consciousness associated with mastery of the first-
person pronoun will eventually emerge, and hence are the raw material
from which to construct both an account of how the acquisition con-
straint can be met for full-fledged self-conscious thought and a noncircu-
lar explanation of the conditions of mastery of the first-person pronoun.
7
Points of View
The previous two chapters have shown how somatic proprioception and
the structure of exteroceptive perceptual experience can be a source of
nonconceptual first-person contents from the very beginning of life. This
is an important start in dissolving the paradox of self-consciousness, be-
cause it disproves the central assumption generating the paradox, namely,
that it is incoherent to ascribe thoughts with first-person content to an
individual who is not capable of employing the first-person pronoun to
achieve reflexive self-reference. But it is just a start, and what is required
now is an understanding of how we get from the primitive first-person
contents implicated in perceptual experience at the ecological level and
in somatic proprioception to the various high-level forms of first-person
thought that involve reflexive self-reference. How do we build up from
the pick-up of self-specifying information to full-fledged self-conscious-
ness? One promising strategy here is to identify certain fundamental di-
mensions of self-consciousness that neither the perceptual pick-up of self-
specifying information nor the degree of self-awareness yielded by so-
matic proprioception can do justice to, and then to consider how such
basic-level information pick-up needs to be augmented if those dimen-
sions of self-consciousness are to come into play.
7.1 Conceptual Points of View and Nonconceptual Points of View
The previous chapter showed how somatic proprioception offers an

awareness of the body as a spatially extended and bounded object that is
responsive to the will. A major part of the significance of this is that it
offers the subject a way of registering the distinction between self and
164 Chapter 7
nonself. It does this by marking the boundaries of the self. The bound-
aries of the self emerge in somatic proprioception both as the limits of the
will and as the limits of felt feedback about the disposition and movement
of body parts. Crucial to this emergence is the sense of touch, which,
because it is simultaneously proprioceptive and exteroceptive, provides
an interface between the self and the nonself. As was also pointed out in
the previous chapter, registering the distinction between self and nonself
is a very primitive form of self-awareness. This is because self-awareness
has a very significant contrastive dimension. A central way of considering
a creature’s degree of self-awareness is in terms of its capacity to distin-
guish itself from its environment, and no creature would be able to dis-
tinguish itself from its environment if it did not have a minimal degree
of self-awareness to begin with. But once that minimal degree of self-
awareness is in place, the richness of the self-awareness that accompanies
the capacity to distinguish the self from the environment is directly pro-
portionate to the richness of the awareness of the environment from
which the self is being distinguished. There is, of course, very little such
richness in somatic proprioception, which, although it has an exterocep-
tive dimension, provides relatively little information about the organiza-
tion and structure of the world. The world that manifests itself in somatic
proprioception is a world of surfaces, textures, and resistances—primi-
tive indeed in comparison with the causally structured world of physical
objects. And the form of self-awareness that emerges from somatic pro-
prioception is correspondingly primitive.
Thus, to make progress beyond the minimal registering of the distinc-
tion between self and nonself that comes with somatic proprioception, we
need to consider the subject’s awareness of its environment. Philosophers
discussing self-awareness often place a lot of weight upon metaphors of
‘perspective’ and ‘point of view’. The point of these metaphors is to cap-
ture certain characteristic aspects of how the environment is disclosed to
self-conscious subjects, as opposed to how it is disclosed to impartial sci-
entific investigation. There is more at stake here than a simple metaphori-
cal restatement of traditional and rather vague distinctions between
subjectivity and objectivity. The metaphors themselves put a distinctive
spin on the notion of subjectivity involved. In particular, they bring out
the centrality of spatiality in understanding self-conscious thought, be-
Points of View 165
cause the most obvious way of understanding how the environment can
be apprehended from a particular perspective or point of view is in spatial
terms. Having a distinctive perspective or point of view on the envir-
onment is something that a self-conscious subject does at least partly in
virtue of occupying a particular spatial position. Of course, though, as I
stressed in my discussion of Gibson’s ecological optics, it is a mistake to
view this in an atomistic way. It is only in the experimental laboratory
that cognition takes place from a fixed spatial point. Perspectives and
points of view move through the world, retaining continuity through
change. The real challenge is understanding what this sort of perspectival
movement involves.
Kant is the philosopher who has had the clearest grip on the relation
between self-awareness and awareness of the environment, and the whole
of the Transcendental Analytic in the Critique of Pure Reason can be
viewed as a sustained working out of the interdependence between them.
To go into the details of Kant’s treatment, although fascinating, would
take us into murky waters far from the matter at hand. I propose to use
P. F. Strawson’s (1966) discussion of Kant’s transcendental deduction of
the categories as the philosophical point of departure. Strawson felici-
tously uses the notion of a point of view and places it at the very heart
of Kant’s treatment of the interdependence between self-awareness and
awareness of the environment.
To appreciate what is going on here, we need to step back a little.
Strawson’s Kant is concerned with the structure of self-conscious experi-
ence. What, he asks, are the necessary conditions that must be fulfilled
for self-conscious experience to be a genuine possibility? What must hold
for a series of thoughts and experiences to belong to a single, unitary self-
conscious subject? He begins thus:
It is a shining fact about such a series of experiences, whether self-ascribed or
not, that its members collectively build up or yield, though not all of them contrib-
ute to, a picture of an 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. (Strawson 1966, 104)
According to Strawson this “shining fact” is the key to understanding the
conditions that Kant places upon the possibility of genuinely self-
conscious experience. The central condition is that a subject’s experiences
166 Chapter 7
must “determine a distinction between the subjective route of his expe-

riences and the objective world through which it is a route” (Strawson
1966, 104). And it is this, of course, that Strawson terms the subject’s
point of view on the world. He explains:
A series of experiences satisfying the Kantian provision has a certain double as-
pect. On the one hand it cumulatively builds up a picture of the world in which
objects and happenings (with their particular characteristics) are presented as pos-
sessing an objective order, an order which is logically independent of any particu-
lar experiential route through the world. On the other hand it possesses its own
order as a series of experiences of objects. If we thought of such a series of experi-
ences as continuously articulated in a series of detailed judgements, then, taking
their order and content together, those judgements would be such as to yield, on
the one hand, a (potential) description of an objective world and on the other
the chart of the course of a single subjective experience of that world (Strawson
1966, 105).
Strawson’s fleshes out the notion of a point of view in a way that draws
together two distinct (sets of) conceptual capacities: the capacity for self-
ascription of experiences and the capacity to grasp the objectivity of the
world.
Let us take a further step back and ask why Strawson’s Kant thinks that
these two sets of conceptual abilities need to be brought together. The first
point is that the question of the possibility of self-conscious experience is
inextricably bound up with the question of the possibility of any form of
experience. Strawson brings this out by considering the hypothesis that
there might be an experience whose objects were sense data: “red, round
patches, brown oblongs, flashes, whistles, tickling sensations, smells”
(Strawson 1966, 99). This would not count as anything recognizable as
experience, he maintains, because it would not permit any distinction to
be drawn between a subject’s experiences and the objects of which they
are experiences. In such a case, the esse (being) of the putative “objects
of experience” would be their percipi (being perceived). There would be
no distinction between the order of experiences and the order of the “ob-
jects of experience,” and this, according to Strawson, effectively means
that we cannot talk either about objects of experience or about a subject
of experience, and hence we cannot talk about experience at all.
Strawson’s central claim is that no creature can count as a subject of
experience unless it is capable of drawing certain very basic distinctions
between its experiences and the objects of which they are experiences.
Points of View 167
Experience that reflects a temporally extended point of view on the world

will ipso facto permit these basic distinctions to be drawn in virtue of
possessing the “double aspect” outlined in the passage quoted earlier.
The matter can be further clarified through the concept of a nonsolipsis-
tic consciousness, which Strawson introduces in Individuals. There he
refers to “the consciousness of a being who has a use for the distinction
between himself and his states on the one hand, and something not him-
self or a state of himself, of which he has experience, on the other”
(Strawson 1959, 69). The notion of a point of view elucidates what any
creature’s experience must be like for it to count as nonsolipsistic in this
sense.
Nonetheless, Strawson’s conception of a point of view needs some mod-
ification before it can be fitted into the current framework. As Strawson
understands the matter (unsurprisingly, given his Kantian inspiration),
experience reflecting the double aspect required for the development of
a point of view is available only to creatures who have mastered both
the sophisticated conceptual skills involved in being able to ascribe expe-
riences to themselves and the basic concepts required for understanding
the objectivity of the world. Although he does not offer an explicit argu-
ment for why the notion of a point of view must be defined in that way, I
think the following captures his position:
1. Genuine nonsolipsistic consciousness requires the capacity to distin-
guish one’s experiences from the objects experienced.
2. That requires the capacity to identify one’s experiences as one’s
experiences.
3. Identifying one’s experiences as one’s experiences is a form of self-
ascription.
4. The self-ascription of experiences requires reflexive self-reference.
The discussion of nonconceptual first-person contents in the preceding
two chapters, however, should have generated a certain suspicion of this
line of argument. The previous chapter in particular showed that nonso-
lipsistic experience both can and does exist in the absence of the sort of
conceptual skills that Strawson takes (without argument) to be necessary
for it. A creature who, in virtue of somatic proprioception, can register
the distinction between self and nonself has clearly arrived at a form of
nonsolipsistic consciousness.
168 Chapter 7
We can keep the essential thrust of Strawson’s position by reformulating

it as follows. Taking the central distinction to be between subjective expe-
rience and what that subjective experience is experience of allows us to
capture the crucial feature missing in the hypothesized purely sense-
datum experience without immediately demanding any relevant concep-
tual capacities on the part of the subject. Using this as the central notion,
we can reformulate the original characterization: having a temporally ex-
tended point of view on the world involves taking a particular route
through space-time in such a way that one’s perception of the world is
informed by an awareness that one is taking such a route, where such an
awareness requires being able to distinguish over time between subjective
experience and what it is experience of. For obvious reasons I term this a
nonconceptual point of view.
7.2 Self-Specifying Information and the Notion of a Nonconceptual

Point of View
Once the notion of a point of view has been formulated so that no rele-
vant conceptual requirements are built into it, one might naturally think
that the analysis of perceptual experience and somatic proprioception in
chapters 5 and 6 already shows us how to make sense of the notion of a
nonconceptual point of view. No special argument is needed to show that
it is possible to have a nonconceptual point of view, it might be suggested,
because such a nonconceptual point of view is built into our experience
of the world from the very beginning. Much of the self-specifying infor-
mation picked up in perceptual experience is information about the spa-
tiotemporal route that one is taking through the world. This is, of course,
particularly apparent in kinesthetic experience, both visual and proprio-
ceptive. It might thus seem that one does in fact have a continuous aware-
ness of oneself taking a particular route through the world that does not
require the exercise of any conceptual abilities, in virtue of having a con-
stant flow of information about oneself as a physical object moving
through the world. The suggestion is that the ecological coperception of
self and environment, together with the proprioceptively derived distinc-
tion between self and nonself, is all that is needed for experience from a
nonconceptual point of view. This suggestion is a very natural one. But it
Points of View 169
is misplaced, and seeing why it is misplaced is an important first step in

the project of moving from basic-level information pick-up to full self-
consciousness.
We need to begin by exploring in more detail why a point of view has
been described as temporally extended. Is it the nonconceptual point of
view itself that is temporally extended, or simply the lifespan of the crea-
ture that has the point of view? The issue can be focused with a thought
experiment. Suppose that we imagine a creature whose experience takes
place completely within a continuous present and lacks any sense of past
or future. The hypothesis is that every experience that this creature has
is completely novel. Could such a creature have experience reflecting a
nonconceptual point of view? If the answer is affirmative, then it follows
that a nonconceptual point of view has to be temporally extended only
in the limited sense that it is enjoyed by creatures who move through the
world over time. If negative, we will have to look for a new sense of tem-
poral extension.
Reflection on this thought experiment leads, I think, to the conclusion
that the capacity to make the basic distinctions at the heart of the notion
of a nonconceptual point of view would be absent in the case under dis-
cussion. A creature whose experience takes place completely within a
continuous present cannot draw the fundamental nonsolipsistic distinc-
tion between its experience and what it is experience of. A minimal re-
quirement on being able to make such a distinction is that what is being
experienced should be grasped as existing independently of any particular
experience of it. What we are trying to avoid, it will be remembered, is a
situation in which the esse of the putative “objects of experience” is their
percipi, and this requires “a component of recognition or judgement
which is not simply identical with, or wholly absorbed by, the particular
item which is recognised, which forms the topic of the judgement”
(Strawson 1966, 100). There must be recognition that the object of expe-
rience has an existence transcending the particular occasion on which it
is apprehended. At the most basic level, such a grasp of independent exis-
tence itself involves an understanding that what is being experienced at
the moment either has existed in the past or will exist in the future, that
it has an existence transcending the present moment. By definition, how-
ever, a creature that experiences only a continuous present cannot have
any such understanding.
170 Chapter 7
There is a possible confusion here that it is important to clear up. The

distinction between self and nonself that has been seen to be present in
somatic proprioception is emphatically not equivalent to the distinction
between experience and what it is experience of that is currently at issue.
What could potentially confuse matters is that both distinctions appear
to qualify as instances of what Strawson terms nonsolipsistic conscious-
ness. A nonsolipsistic consciousness is “the consciousness of a being who
has a use for the distinction between himself and his states on the one
hand, and something not himself or a state of himself, of which he has
experience, on the other” (Strawson 1959, 69). A creature who, in virtue
of somatic proprioception, has a grasp of the boundaries and limits of the
self, and hence of the distinction between self and other, can be described
as “having a use for the distinction between himself and his states, on the
one hand, and something not himself or a state of himself on the other.”
The same holds for a creature who has grasped the distinction between
experience and what it is experience of. But it would be a mistake to think
that a grasp of the first distinction would automatically bring with it a
grasp of the second. The distinction between self and nonself has nothing
to do with the concept of experience. Moreover, there is an important
difference in the respective temporal dimensions of the two distinctions.
The distinction between self and nonself is available purely synchroni-
cally. It does not require taking into account times other than the present,
unlike the distinction between experience and what it is experience of.
That second distinction is diachronic, as will emerge shortly.
The important question as far as this second distinction is concerned,
therefore, seems to be this: What form must experience take if it is to
incorporate an awareness that what is being experienced does not exist
only when it is being experienced? And in particular, what must the tem-
poral form of any such experience be? Clearly, such an awareness would
be incorporated in the experience of any creature that had a grasp of the
basic temporal concepts of past, present, and future, but we are looking
for something at a more primitive nonconceptual level. What I would
suggest is that certain basic recognitional capacities offer the right sort of
escape from the continuous present without demanding conceptual mas-
tery. Consider the act of recognizing a particular object. Because such an
act involves drawing a connection between current experience of an ob-
Points of View 171
ject and a previous experience of it, it brings with it an awareness that

what is being experienced has an existence transcending the present mo-
ment. Of course, not all such acts of recognition will involve an awareness
that what is being experienced exists independently of being experienced.
Reference to a previous experience can occur in the absence of any aware-
ness that the object of experience exists unperceived, as is the case, for
example, when a particular sensation or subjective feeling is recognized.
It is certainly possible to recognize an experience as one previously experi-
enced without any grasp of the distinction between experience and what
it is experience of. Clearly, the recognitional capacities must be exercised
on something extraneous to the experiences themselves. But what?
Physical objects are obvious candidates, and any creature that could
recognize physical objects would have experience that involves drawing
the right sort of distinctions. However, one might wonder whether the
distinctions could be drawn at a level of experience that does not involve
objects, and in fact it is hard to see how there could be such a thing as a
nonconceptual point of view if the distinctions could not be drawn at a
more primitive level. Strawson (following Kant) is surely right to assume
that a full understanding of the nature of objects comes only with a core
set of conceptual abilities. It is important, then, that there should be
“things” that can be recognized but that are not physical objects, but
what are they? One natural suggestion to make here would be that such
“things” are places. The thought here is that the fundamentally spatial
nature of the experienced world provides the most basic material for the
exercise of basic recognitional capacities. The experience of creatures
who cannot manipulate concepts in the manner required to think about
an objective spatial world of mind-independent objects can nonetheless
reflect the spatiality of their lived environment in terms of the primitive
capacity to think about places.
There is an obvious problem, though, that this proposal has to deal
with. The identification of places and the identification of objects seem
so inextricably bound together that it is not clear how there can be reiden-
tification of places without thought about objects. The connection be-
tween objects and places has been thought to hold in both directions. It
is certainly clear that the criteria of identity for any given material object
involve spatiotemporal continuity, and that if I want to establish whether
172 Chapter 7
an object at one time is identical with another object at another time,

then one of the criteria that has to be satisfied is that they should either
be at the same place or that there should be a plausible spatial route be-
tween the two places that they occupy. But many have also argued that
the interdependence holds also in the opposite direction, that the criteria
of identity for places necessarily depend on the criteria of identity for
objects (Strawson 1959, Quinton 1973). It is important, however, to be
clear on what the terms of the debate are here. One question that might
be asked here is whether we, as concept-exercising and language-using
creatures, could have anything like the understanding of space and spatial
relations that we actually do have without having the capacity to think
about objects and to employ those thoughts about objects in identifying
places and their relations. But what is relevant here is the more general
question of whether any capacity for place reidentification could be avail-
able to creatures who lack the capacity to think in terms of enduring ma-
terial objects occupying spatial positions (Campbell 1993).
The problem arises, of course, because places themselves cannot be rec-
ognised simpliciter. They must be recognised in terms of what occupies
them. So, to rebut the objection that there cannot be place reidentification
without object reidentification, it will be necessary to offer an alternative
account of the possible occupants that could make place reidentification
possible. It seems to me that, far from material objects being the only
such occupants, there are (at least) two distinct alternatives.
One alternative has already been discussed at some length in chapter
3, where I noted that recent work in developmental psychology strongly
supports the view that young infants parse their visual perceptions into
bounded segments, about whose behavior they have certain expectations,
which expectations nonetheless do not qualify as the perception of ob-
jects (still less as involving mastery of the concept of an object). I sug-
gested in the final section of chapter 4 that by defining the concept of an
object* we could capture at a nonconceptual level how the perceived array
can be parsed into spatially extended segments that are more primitive
than objects. Drawing on this, I find it natural to propose that objects*
could have an important role to play in place reidentification.
We can make a start on the second alternative by noting the view that
there are levels of language use that do not involve identifying reference to
Points of View 173
particular objects (Strawson 1959, Campbell 1993). What is often termed

“feature-placing” discourse characterizes the world in terms of features
in a manner notably different from the more familiar subject-predicate
discourse. Whereas subject-predicate discourse depends upon identifying
reference to material particulars, and these are then ascribed qualities or
properties (which are sortal universals, appropriate for identifying and
classifying material particulars), feature-placing discourse deploys uni-
versal terms that are not sortal universals and that are not predicated of
material particulars. ‘It is raining’ is a case in point, as is ‘There is food’.
In neither of these cases is there a particular thing of which certain prop-
erties are predicated, nor are the universal terms involved the sort of terms
that can be used to classify particulars. Quine’s distinction between mass
terms and count terms is very much to the point here (1960, sections
19–20). Mass terms are terms that refer cumulatively to general types of
stuff, like water, wood, or rain. Any bit of wood is wood, as is any collec-
tion of bits of woods, and the question of how many is not an appropriate
one to ask once the presence of wood has been identified. In contrast,
count terms must be provided in order for things to be counted and num-
bered. They pick out particular things.1 We can understand feature-
placing discourse as a form of language use that operates at the level of
mass terms and has no count terms, with the important proviso that the
range of available mass terms in such feature-placing discourse is not re-
stricted to those terms that function as mass terms in our own far more
complex language.2
This notion of primitive feature-placing discourse can be modified to
help us understand how place reidentification might take place without
thought about objects. Let us assume a fairly primitive creature that none-
theless possesses the capacity to recognize features, such as warmth or
food. Let us assume, moreover, that such a creature can navigate its way
around the environment in a manner that is not entirely stimulus-driven
(according to the criteria discussed in chapter 4). In navigating its way
around the environment, it returns to particular places, and this can be
made comprehensible in terms of certain features (warmth, food, water,
etc.) that exist at those places and that it is appropriate to describe the
creature as detecting. Such a situation it is compelling to describe as
one in which spatial behaviour is driven by information about features
174 Chapter 7
holding at places. This in turn opens the way to the thought that we can
understand some forms of place reidentification in terms of feature
reidentification.
So we can make sense of place reidentification in terms either of the
notion of an object* or that of feature reidentification, or of course in
terms of both. This answers the worry that place reidentification depends
on object reidentification by showing that at least a primitive form of
place recognition can be prior to object identification.3 But there is an
important dimension not yet discussed. We can draw a broad distinction
between two types of memory. There are, on the one hand, instances
of memory in which past experiences influence present experience but
without any sense on the part of the subject of having had the relevant
past experiences and, on the other, instances in which past experiences
not only influence present experience but also the subject is in some
sense aware of having had those past experiences. In the former case
what licenses talk of memory is the fact that a subject (or an animal or
even a plant) can respond differentially to a stimulus as a function of
prior exposure to that stimulus or to similar stimuli. This is not to deny
that such a memory can be extremely complex. Quite the contrary, it
seems to be central to the acquisition of any skill, even the most devel-
oped. The “differential response” does not have to be simple or repetitive.
Nonetheless, we can draw a contrast between memory at this level and
the various forms of memory that do involve an awareness of having had
the relevant past experiences, as, for example, when a memory image
comes into one’s mind or one successfully recalls what one did the previ-
ous day. Clearly, there are many levels of such memory (which I shall term
‘conscious memory’), of which autobiographical memory is probably the
most sophisticated. One thing that these forms of conscious or explicit
memory all have in common, however, is that previous experience is con-
sciously registered, rather than unconsciously influencing present
experience.4
This distinction of forms of memories is a crude one, not least because
such terms as ‘conscious’ and ‘explicit’ are so hazy, but it can be illus-
trated through some striking experimental work carried out with am-
nesic patients (Schacter, McAndrews, and Moscovitch 1988). As is well
known, amnesia is an inability to remember recent experiences (even from
Points of View 175
the very recent past) and to learn various types of information, and it
results from selective brain damage that leaves perceptual, linguistic, and
intellectual skills untouched. Memory deficits in amnesic patients have
traditionally been studied using techniques designed to elicit explicit
memories. So, for example, amnesic patients might be instructed to think
back to a learning episode and either recall information from that episode
or say whether a presented item had previously been encountered in the
learning episode. On tests like these, amnesic patients perform very badly.
What is interesting, though, is that the very same patients who perform
badly on these tests of recall and recognition can perform successfully
(sometimes as successfully as normal subjects) on tests that do not require
thinking back to specific episodes or consciously bringing to mind earlier
experiences. The acquisition of skills is a case in point, and there is con-
siderable experimental evidence showing that amnesic patients can ac-
quire both motor and intellectual skills over a series of learning episodes
even though they cannot explicitly remember any of the learning epi-
sodes. A striking example is the densely amnesic patient who learned how
to use a personal computer over numerous sessions, despite declaring at
the beginning of each session that he had never used a computer before
(Glisky, Schacter, and Tulving 1986). In addition to this sort of capacity
to learn over a succession of episodes, amnesics have performed well on
single-episode learning tasks (such as completing previously shown words
when given 3-letter cues). Experiments like these in amnesic patients
clearly reveal the difference between conscious and nonconscious mem-
ory, but similar dissociations can be observed in normal subjects, as when
performance on indirect tasks reveals the effects of prior events that are
not remembered.
Nonetheless, the distinction of forms of memories is not yet as clear as
it needs to be, since the distinguishing feature of conscious memories, as
so far defined, is that they issue in verbal reports, which is of limited use
at the nonconceptual level. To sharpen the distinction further it is helpful
to recall that memory is a causal notion (Martin and Deutscher 1966).
This holds both of conscious and unconscious memory, although, of
course, in different ways. At a very general level, the notion of memory
has the following core role to play in the explanation of behavior: we need
to appeal to the operations of memory to explain a creature’s current
176 Chapter 7
behavior where we cannot fully explain that behavior without assuming

that it is being being causally affected by previous experiences and/or
thoughts (henceforth, ‘experiences/thoughts’). But the satisfaction of this
core condition counts as an instance of conscious memory if and only
if the previous experiences/thoughts are causally efficacious through the
medium of occurrent experiences/thoughts that refer back to those previ-
ous experiences/thoughts. In the case of unconscious memory, in con-
trast, this further requirement is not satisfied.
Note that there are two ways in which the further requirement for con-
scious memory could fail to be satisfied. First, it could be the case that
there are no distinctive occurrent experiences/thoughts. Of course, this
would not mean that there are no occurrent experiences/thoughts at all.
Rather, this case arises when a creature’s occurrent experiences/thoughts
are precisely those that it would have had if the previous experiences/
thoughts had not been causally efficacious. As an example, consider
Glisky’s amnesic patient encountering his computer at the beginning of
his tenth lesson (which, of course, he thinks is his first). Despite the effects
of the previous nine lessons, the patient has more or less the same experi-
ences/thoughts that he had when he first encountered the computer. No
distinctive experiences/thoughts are caused by his previous encounters
with the computer. Nonetheless, those previous encounters could be caus-
ally efficacious if, for example, the patient’s first action was to switch on
the computer and take hold of the mouse. A second way of failing the
further requirement would occur if, although the subject in question did
have distinctive occurrent experiences/thoughts caused in the appropriate
way, those experiences/thoughts do not refer back to the previous experi-
ences and/or thoughts that caused them. One common way in which this
occurs is when perceptual discriminative capacities have been sharpened
through long experience. Here a creature’s occurrent perceptions could
be significantly altered by previous experiences/thoughts without there
being any reference back to the experiences/thoughts that are causally
responsible for the increased discrimination manifest in its occurrent
perceptions.
It is important to recognize that neither form of potential failure can be
present if we are plausibly to maintain the presence of conscious memory.
Conscious memory requires more than that past experience/thoughts
Points of View 177
causally affect present experiences/thoughts. It requires that they do so in

a way that makes them recognized as past experience/thoughts. The grasp
of a primitive notion of the past is built into the ascription conditions of
conscious memory: to attribute conscious memory is ipso facto to attri-
bute a grasp of times other than and prior to the present.
So, the distinction is between, on the one hand, the processing of re-
tained information derived from past experience and, on the other, the
registration of past experience as past experience in conscious awareness.
When we apply this distinction to the matter in hand, our working notion
of a point of view demands that conscious memory be involved in the
recognitional abilities at the heart of having a point of view upon the
world. Reflect on the case of a creature, perhaps a swallow, that is per-
fectly capable of performing complicated feats of navigation (like finding
its way back to its nest or to the warmer climes where it spends the winter)
but that nonetheless has no conscious memories of repeatedly encoun-
tered features or objects* at particular places. Such a creature is clearly
sensitive to certain facts about the places, features, or objects* in ques-
tion, because those facts determine behavior. Moreover, it is quite pos-
sible that its occurrent experiences should be causally affected by its
previous encounters with those places, features, or objects*. Nonetheless,
it is important to recognize that we are applying the concept of reidentifi-
cation here in a restricted sense. In the case of a creature capable
only of nonconscious memory, that creature’s capacity to reidentify a
place is exhausted by its capacity to find its way back to that place.
But when we consider place reidentification that involves the opera-
tions of conscious memory, the situation becomes fundamentally differ-
ent. When a creature finds its way back to a particular place that it then
consciously remembers (or when it consciously remembers an object* or
feature at a place to which it has found its way back), it is emerging from
a continuous present and moving toward possession of a temporally ex-
tended point of view. This is so for two reasons. First, consciously to
remember a particular place, object*, or feature is to be aware of having
been at that place or encountered that object* or feature before. A crea-
ture who can remember having been somewhere before has the begin-
ning of an awareness of movement through space over time. And that in
turn is the beginning (and certainly a necessary, although not sufficient,
178 Chapter 7
condition) of being able to draw what Strawson terms the distinction be-
tween the subjective route of one’s experiences and the world through
which it is a route. Second, recall that part of the significance of the no-
tion of a point of view is that it provides a sense of self-world dualism
richer than the minimal registering of the distinction between self and
nonself that emerges from somatic proprioception. As I mentioned ear-
lier, the richness of the self-awareness that accompanies the capacity to
distinguish the self from the environment is directly proportionate to the
richness of the awareness of the environment from which the self is being
distinguished. And what is significant about place reidentification that
involves conscious memory is that it makes possible a richer form of
awareness of the environment. Without conscious memory, no creature
would be able to conceive of their environment as composed of items
that have an existence transcending the present moment. What conscious
memory offers is the possibility of conceiving of objects as having existed
in the past.
These two reasons for why place reidentification involving conscious
memory is so central here depend on how conscious memory implicates
a form of temporal orientation toward the past. There are, of course, dif-
ferent types and degrees of temporal orientation toward the past, and
one question that arises is, what type of understanding of the past does
conscious memory contribute to the acquisition of a nonconceptual point
of view upon the world? An important distinction in this area has been
made by John Campbell (1994, section 2.1). Campbell distinguishes be-
tween what he terms temporal orientation with respect to phase (hence-
forth phasal orientation) and temporal orientation with respect to
particular time (henceforth particular orientation). He illustrates the dis-
tinction as follows:
Consider an animal that hibernates. Through the part of the year for which it is
awake, it regulates its activity depending on the season. Such an animal certainly
has a use for temporal orientation. It can recognize that it is late spring, perhaps
by keeping track of how long it has been since winter, and realize that soon it will
be summer. But it may not have the conception of the seasons as particular times;
it may be incapable of differentiating between the autumn of one year and the
autumn of another. It simply has no use for the conception of a particular au-
tumn, as opposed to the general idea of the season. So while the animal is capable
of orientation with respect to phases, it is not capable of orientation with respect
to particular times. (Campbell 1994, 38)
Points of View 179
At issue here is the capacity to think about event tokens, as opposed to the
capacity to think about event types. Phasal orientation does not involve
distinguishing between event types and event tokens. Phasal orientation
is concerned simply with where one is on a given temporal cycle and does
not provide the resources for thinking about the relations, temporal or
otherwise, between that temporal cycle and another. In a particular orien-
tation, in contrast, event tokens are discriminated by means of a temporal
frame of reference that places all event tokens within a single linear series
ordered by the relations of temporal priority and temporal simultaneity.
Does the work to which I am putting the concept of conscious memory
require a phasal orientation or a particular orientation toward the past?
Conscious memory offers the beginning of both an awareness of move-
ment through space over time and an awareness that the environment is
composed of items that have an existence transcending the present mo-
ment. Temporal orientation toward the past is at the heart of both of
these forms of awareness. There seems no good reason to think that pha-
sal orientation will not be sufficient here. The creature described by
Campbell in the passage just quoted has both an awareness of movement
through space over time and an awareness that the environment is com-
posed of items whose existence transcends the present moment, despite
its inability to orient itself toward particular times. Although place reiden-
tification involves reidentifying a particular place, it does not follow that
if reidentification is to involve conscious memory, then the memory in
question must be a memory of having previously encountered that place
at a particular time. A memory of having previously encountered the
place at a particular stage in a temporal cycle would be quite sufficient.
Certainly, a conscious memory of having previously encountered a par-
ticular place is a memory of a particular happening or event that occurred
at a particular time. But it does not follow that the memory must refer to
a particular time within a temporal frame of reference, as associated with
the particular orientation toward the past. This is tied up with the fact
that memory is a causal notion. The causal dimension of memory is cap-
tured particularly clearly in the influential and widely accepted account
of memory offered by Martin and Deutscher (1966), which I will use
to illustrate the point. The core of the account is given by the following
three conditions:
180 Chapter 7
A remembers a particular event or thing f if and only if

i. within certain limits of accuracy he represents f as having been en-
countered in the past;
ii. if f was public, he observed f; if f was private, then it was “his”;
iii. his past experience of f was (causally) operative in producing a state
or succession of states that ultimately produced his representation of f.
(Martin and Deutscher 1966, 166)
I take this to be a relatively uncontroversial theory of memory. Conditions
(i) to (iii) can be satisfied whether or not f is located by A within a tempo-
ral frame of reference associated with the particular orientation toward
the past. If A is capable only of phasal orientation toward the past, then
it follows, of course, that he is not capable of representing the temporal
difference between f in temporal phase P and f9 in temporal phase P9,
and hence the particular times at which they occur. But if A currently
represents f and his current representation is causally derived from his
experience of f, then, given that f is a particular event occurring at a
particular time, it follows from conditions (i) to (iii) that A is enjoying a
memory of a particular event that occurred at a particular time.
It might be objected, though, that A is not properly described as repre-
senting f in the situation described. Let us assume that two particular
events f and f9 within temporal cycles P and P9 are phenomenologically
indistinguishable for A. If this were the case, then a subject would only
be able to distinguish them by referring them to different points in a single
temporal series that includes both P and P9, and this, ex hypothesi, is
something that A cannot do. In such a case, the objection runs, A cannot
be properly described as representing f rather than f9, and hence would
not satisfy condition (i). The point of the causal theory, however, is that
whether A is representing f or f9 is determined by the causal history of
A’s current representation. It is, of course, perfectly possible that there is
no single determinate causal history, and that A is really representing a
generic f-type event or object in a manner causally traceable to the con-
junction of its experiences of f, f9, f0, . . . . But it would be a mistake to
think that this is all that A can properly be described as doing. Memory
is best analyzed as a form of perception, and it cannot generally be a
requirement upon perceiving a particular f that one be able to distinguish
that f from any other f-type event or object that might have been caus-
ally responsible for one’s current perception.
Points of View 181
This is a good moment to comment on some potential ambiguities in

the concept of episodic memory. Campbell defines ‘episodic memory’ so
that it counts as an instance of the particular orientation: “I will be inter-
preting ‘episodic memory’ to mean memory of a past happening con-
ceived as having a particular past time at which it took place” (Campbell
1994, 40). I have no objection to this stipulation. But it is important to
bear in mind that many psychologists operate with a simple binary
distinction between memory of things happening and memory for facts
and skills (semantic and procedural memory). If memory is classified
in this binary way, then accepting Campbell’s definition will have the
consequence that all memories of things happening will count as in-
stances of particular orientation toward the past. This is an unwelcome
conclusion, because it makes analytically true a claim that is controversial
and in any case (as I have just argued) false. It would be helpful to enrich
the vocabulary here by distinguishing the class of autobiographical mem-
ories within the more inclusive class of episodic memories.5 Autobio-
graphical memories are those memories of particular events that are
temporally indexed so that they fit within a linear, autobiographical frame
of reference.
Let me take stock. I have used the notion of a temporally extended
point of view to characterize the structure of experience in a creature
capable of making a rudimentary distinction between its experiences and
what those experiences are experiences of. When we looked more deeply
at this idea, it emerged that the capacity to make such a distinction is
available only to creatures who are capable of exercising certain basic
recognitional abilities: the interlocking abilities to recognize places and
features. Moreover, these recognitional abilities must involve the exercise
of explicit or conscious memory. Putting the matter like this, however,
brings out into the open two important problems that need to be con-
fronted. The first is a practical problem. It is relatively straightforward to
establish the presence of explicit or conscious memory in language-using
subjects, since their subjective reports are usually a reliable source. But
when we are dealing with nonlinguistic animals, all we have to go on are
observable facts about differential responses to stimuli as a function of
previous exposure. How can we decide whether to explain these re-
sponses in terms of implicit or explicit memory? The second problem is
182 Chapter 7
theoretical, and more radical. It emerges because we are looking for

cognitive abilities that it makes sense to ascribe to non-concept-using
creatures, like young infants and many animals. But many philosophers
are skeptical about the very possibility of nonconceptual recognitional
abilities. At the very least, then, this skepticism must be set to rest.
To begin with the skeptical worry, one reason for thinking that the no-
tion of nonconceptual recognition does not make sense is the thought that
recognition is essentially an activity of classification that involves relating
objects of experience under concepts. There is, I think, a simple reply to
this. If there were no nonconceptual recognition, then it would be impos-
sible to see how any concepts could ever be acquired. Whatever detailed
theory of concept acquisition we adopt, it is obviously true that the acqui-
sition of observational and perceptual concepts takes place on the basis
of perceived resemblances between items.6 But there can be no perceived
resemblances without basic recognitional abilities. Recognition is the
most fundamental step in categorization and classification, and it is corre-
spondingly absurd to make it depend on more sophisticated cognitive
abilities.7 But in contrast to the very clear sense we have of what it is to
classify things under concepts, we have very little understanding of what
nonconceptual recognition might consist in at the level of conscious or
explicit phenomenal experience (i.e., when we are dealing with recogni-
tional abilities that cannot be understood solely in terms of nonconscious
information processing). There are three points that it is worth making in
this context.
First, it is a mistake to hold that there can be no organized experience
without the experience of conceptually classifiable particulars. On the
contrary, we have already looked at two ways in which experience can be
organized without the application of concepts. The first emerged in my
first discussion of nonconceptual content in chapter 3, where I noted that
infants seem to be capable of parsing their experience into bounded seg-
ments that support certain basic expectations. This is a clear example of
experience being organized without the application of concepts. Compara-
ble types of organization emerged earlier in this section with feature-
placing modes of thought. Of course, neither of these types of experience
can count as experience of particular items, if ‘particular items’ is read to
mean ‘enduring material objects’, but that would be a question-begging
reading.
Points of View 183
A second point is that there seems room for an unanalyzable feeling of

familiarity as a basic building block for cognition. The notion of familiar-
ity is at the heart of the theory of memory that Russell offered in The
Analysis of Mind, and it has by and large not received good press among
philosophers (Russell 1921, Pears 1975). According to Russell, the feeling
of familiarity integral to the content of memory images gives rise to the
belief that the image refers to the past. One significant, perhaps conclu-
sive, difficulty for this theory is describing the feeling of familiarity in a
way that clearly distinguishes it from the belief to which it is supposed to
give rise, and the result has been a general distrust of appealing to feelings
of familiarity in the context of memory. But this general distrust must be
mistaken. An enormous part of learning, and hence of cognition as a
whole, depends upon grasping that one event is similar to another, that
something is the same as something else, and at the most basic level, it is
hard to see how this could ever be apprehended without some sort of
feeling of familiarity. Learning depends on the appreciation of similarity,
as Quine has noted on many occasions, and familiarity is just an appreci-
ated similarity. This is certainly not compromised by granting the anti-
Russellian point that, in the case of belief-forming and language-using
creatures, this feeling of familiarity is not clearly distinguishable from be-
liefs about the pastness of the remembered event to which it gives rise,
since we are interested in familiarity as experienced by creatures incapable
of entertaining beliefs about the pastness of past events.
Third, there is a close relation among recognitional abilities, features,
and basic inductive generalizations. When an animal recognizes a particu-
lar place through recognizing a particular feature, that feature will, of
course, be something in which it has a direct interest and that is directly
relevant to its behavior. Recognizing a given feature is closely bound up
with recognizing possibilities for action relative to that feature, and when
this is performed over a period of time, it is appropriate to speak of the
formation of a basic inductive generalization about what to expect at a
particular place. The induction is generated by the experience of repeated
conjunctions (which is itself presumably underwritten by feelings of fa-
miliarity). This gives us a further important element in nonconceptual
recognition, because the expectancies to which such inductive generaliza-
tions give rise are presumably part of what it is like to recognize a place
in terms of a feature.
184 Chapter 7
These basic inductive generalizations are to be distinguished from

the instrumental protobeliefs discussed in chapter 5. Those instrumental
protobeliefs, involved in explaining the simplest and most basic forms of
intentional activity, were elucidated in terms of present-tense expectancies
about the behavior of objects in the environment. They do not count as
instances of conscious memory. In fact, they fail to satisfy the criteria for
conscious memory in the second of the two ways identified earlier. Al-
though instrumental protobeliefs have, of course, been produced by past
experiences, they contain no reference back to the past experiences caus-
ally responsible for those effects.
There is room for a strong claim about the connection between recogni-
tional abilities and basic inductive generalizations at the nonconceptual
level. There is a very important question about how memories can have
have implications for current behavior. Jonathan Bennett has posed the
question and solved it (note, though, that his terminology is rather differ-
ent from mine):
Non-linguistic behaviour is essentially a manipulation of a present environment,
and the immediately relevant beliefs must be ones about the present. If an item of
behaviour of this kind is also to show that the agent has a certain belief about the
past, this must presumably be because the attribution of the past belief helps to
explain the present one. But that requires that the two be somehow linked by the
agent, and I do not see how they can be linked except through a general belief.
(Bennett 1976, 104)
It is through the medium of basic inductive generalizations that a creature
can bring its conscious memories to bear upon current behavior. Here is
Bennett’s illustration:
Suppose that a chases a victim v up a tree, guards the foot of the tree for a while,
and then climbs up after v. We could discover that a climbs the tree because he
believes that v is in the tree. Can that belief be explained as arising from the belief
that v was earlier in the tree? Only, I suggest, if a is also credited with the linking
belief that v-like animals leave trees only by climbing down them (or that v-like
animals never fly, or some such). (Bennett 1976, 105)
In his earlier book Rationality Bennett argued that the interdependence
of basic inductive generalizations and conscious memories (in my terms)
meant that neither could be ascribed to nonlinguistic creatures. Only lin-
guistic creatures, he claimed, could manifest these two cognitive abilities
separately. Only in language can one manifest a conscious memory with-
Points of View 185
out a related inductive generalization, and vice versa. This yields his
restriction of these abilities to language users when coupled with the as-
sumption that a creature that cannot manifest two different types of cogni-
tive ability separately cannot properly be credited with either of them.8
As Bennett subsequently realized, however, the fact that basic inductive
generalizations and conscious memories always coexist in the explanation
of behavior if they feature in it at all does not mean that they cannot be
separated out on behavioral grounds. Suppose that we are faced with a
particular piece of behavior that appears to demand explanation in terms
of a conscious memory of a particular place and a concomitant inductive
generalization. Different inductive generalizations dictate different forms
of behavior in different circumstances. By examining what a creature does
in different circumstances, we can work out which inductive generaliza-
tions are governing its behavior and apply this back to the original behav-
ior. Similarly, a given conscious memory of a place may, in conjunction
with different inductive generalizations, dictate identifiable forms of be-
havior relative to that place. By comparing different candidate conscious
memories with different candidate inductive generalizations, it should
be possible in principle to narrow the field down to a single memory-
generalization pair.
This interdependence of conscious memories and basic inductive gener-
alizations is very significant in dealing with the practical problem identi-
fied earlier: how are we to establish the presence of explicit or conscious
memory in nonlinguistic creatures, given the importance of verbal reports
in deciding what memories to ascribe to linguistic creatures? Let me set
the scene by describing two sets of experiments in the animal learning
literature. The first is a set of experiments on monkeys (rhesus and pig-
tailed macaques) carried out by Gaffan (1977). Monkeys were trained to
watch a screen on which 25 colored slides were presented twice at ran-
dom during a testing session. The monkeys were tested for their capacity
to detect whether a given slide was being presented for the first or second
time during a training session. Pressing a lever during the second presen-
tation of a slide was rewarded with a sugar pellet, but pressing the lever
during the first presentation was not. The two rhesus monkeys tested were
able (after training) to discriminate with 90 percent accuracy between
first and second presentation, although an average of 9 slides intervened
between the two presentations.
186 Chapter 7
Contrast Gaffan’s experiments with the superficially more complicated

memory experiments performed by Davis and Fitts (1976) on rhesus
monkeys and pig-tailed macaques. In these experiments, pictures pasted
onto boards were used to cover wells in which food rewards could be
placed. The monkeys were presented with a sample picture, which they
pushed off the well to discover whether the sample was rewarded or not.
After a short delay, two pictures were presented simultaneously: one the
sample and the other completely novel. If the sample had originally been
rewarded it was rewarded again, while if the well under the sample had
originally been empty, then the novel picture was rewarded. The monkeys
mastered this principle.
In some ways this second set of experiments seems to involve more
complicated memory operations than the previously described experi-
ments. The animals are being asked to remember not just whether a pic-
ture has been presented before but also whether it was rewarded or not.
This is a two-way classification. On the other hand, however, there is cor-
respondingly less reason to hypothesize that conscious recognition and
familiarity are driving the behavior in the experimental situation. Mas-
tering the principle can be explained in terms of the direct reinforcement
of associations. The first such association is the straightforward associa-
tion between the sample and the reward. The second is the association
between an unrewarded sample and the novel picture. To revert back to
the original characterization of conscious memory, it does not seem neces-
sary in explaining what is going on to appeal to more than the causal
efficacy of previous experience on present behavior. But there does not
seem to be a similar explanation of what is going on in Gaffan’s recogni-
tion experiments. That is because the relevant association there involves
familiarity and recognition. It is an association between rewards and the
familiarity of a presented image. Successful performance in the Gaffan
experiments rests upon successful mastery of a basic inductive generaliza-
tion along the lines of ‘Pressing a lever when something seen before is seen
again leads to a reward’ in conjunction with specific individual conscious
memories of the form ‘This has been seen before’. Specific individual con-
scious memories are required to learn this basic inductive generalization.
I offer Gaffan’s experiments as a clear example of behavior that demon-
strably requires explanation in terms of conscious memories.
Points of View 187
Let me relate this back to the concept of a nonconceptual point of view.

The conclusion so far is that any creature who has a temporally extended
point of view on the world must possess conscious recognitional abilities.
This enables us to return to the issue with which this chapter began. We
set out to identify certain fundamental dimensions of self-consciousness
to which the perceptual pick-up of self-specifying information cannot do
justice, and then to consider how basic-level information pick-up needs
to be augmented if those dimensions of self-consciousness are to come
into play. I put forward the notion of a nonconceptual point of view as
just such a fundamental dimension of self-consciousness, and I suggested
at the beginning of this section that the perceptual pick-up of self-
specifying information might be sufficient for experience to reflect a point
of view on the world. We are now in a position, though, to see why it
cannot be sufficient.
On the ecological view, perception is fundamentally a process of ex-
tracting and abstracting invariants from the flowing optical array. Or-
ganisms perceive an environment that has both persisting surfaces and
changing surfaces, and the interplay between them allows the organism
to pick up the sort of information that specifies, for example, visual kines-
thesis. The key to how that information is picked up is the idea of direct
perception. The mistake made by existing theories of perception, ac-
cording to Gibson, is construing the process of perception in terms of a
hierarchical processing of sensory inputs, with various cognitive processes
employed to organize and categorize sensations. A crucial element of this
serial processing is bringing memories to bear upon present experience.
As emphasised in chapter 5, Gibson rejects this whole picture of percep-
tion. Accepting that present experience is partly a function of past experi-
ence, he firmly denies that this sensitivity to past experience is generated
by processing memories and sensations together. His alternative account
rests on the idea that the senses as perceptual systems become more sensi-
tive over time to particular forms of information as a function of prior
exposure.9 Although Gibson was rather polemical about what he termed
“the muddle of memory,” it would seem that his account seems to involve
no more than a differential response to stimuli as a function of past
experience. Gibson’s position seems to be that conscious recognition is
not implicated in ecological perception, although it might or might not
develop out of such ecological perception. It is perfectly possible for a
188 Chapter 7
creature to have experience at the ecological level without any conscious

recognitional capacities at all. If, then, a capacity for conscious place rec-
ognition is a necessary condition of having experience that involves a tem-
porally extended point of view, then it seems that the dual structure of
experience involved in the ecological coperception of self and environ-
ment must be significantly enriched before yielding a point of view.
The point, of course, is not that this creates any problem for the basic
idea of ecological perception; rather, it is that the materials offered by
Gibson’s own account need to be supplemented if they are to be employed
in the theoretical project under discussion, and it is perfectly possible that
Gibson’s concepts of information pick-up and direct “resonance” to in-
formation in the ambient environment could have a crucial role to play in
such an extension of the basic way in which ecological perception is sensi-
tive to past experience (as they are, for example, in the account of percep-
tion and memory developed in Neisser 1976). Gibson’s account alone
cannot do all the work it was earlier suggested it might be able to do,
because the capacity for conscious place recognition needs to be added
onto the ecological coperception of self and environment.
7.3 Three Intersecting Distinctions and the Acquisition Constraint
In chapter 5, I argued that the pick-up of self-specifying information in

perceptual experience was a source of nonconceptual first-person con-
tents. In chapter 6, I argued the same for somatic proprioception. I also
suggested that these are the most primitive types of first-person con-
tents—the building blocks from which we can construct a principled ac-
count of both the nature and the acquisition of self-consciousness. One
obvious question that this raises is, are we committed to the view that
any creature whose perceptual experience is a source of nonconceptual
first-person contents ipso facto counts as self-conscious? This would, in
the eyes of most theorists, be a reductio ad absurdum of the position un-
der discussion. My discussion of the notion of a nonconceptual point of
view in this chapter has offered a way to avoid this conclusion. We now
have the conceptual machinery to impose a division within the class of
creatures whose perceptual experience supports nonconceptual first-
person contents. This division divides those creatures that have a noncon-
ceptual point of view on the world from those creatures that do not. This
Points of View 189
distinction can be employed to mark a significant substantive sense of

self-consciousness.
Still, several important issues remain to be discussed. First, more needs
to be said about the notion of a nonconceptual point of view. The discus-
sion in this chapter has been devoted principally to establishing conscious
place recognition as a necessary condition for possession of a nonconcep-
tual point of view, and it should be obvious that it is far from a sufficient
condition. In the next chapter I will discuss in more detail how the capac-
ity for conscious place recognition is put to work to provide a nonconcep-
tual point of view on the world. A second set of issues emerges when we
ask how the distinction stressed in this chapter maps onto certain other
distinctions, one of which has already been discussed. In particular, some-
thing needs to be said about how the distinction between conscious place
recognition and nonconscious place recognition relates, first, to the dis-
tinction between conscious and nonconscious creatures and, second, to
the distinction discussed in chapter 4 between behavior that supports an
intentional interpretation and behavior that does not. A third point that
needs to be discussed is how the conclusions of this chapter mesh with
the Acquisition Constraint, which governs the overall project of showing
how full-fledged self-consciousness can be built up from the pick-up of
self-specifying information in perceptual experience.
Describing the type of memory implicated in experience reflecting a
nonconceptual point of view as conscious memory can make it tempting
to think that all and only creatures capable of conscious place recognition
are conscious. This is certainly suggested by the contrast drawn between
conscious place recognition and unconsciously generated differential re-
sponses due to prior exposure to stimuli. Nonetheless, there is nothing
incoherent about the idea that a creature inhabiting what I termed the
continuous present could have conscious experiences. Of course, it could
not have conscious experiences if we stipulate, as Strawson does, that
conscious experiences are available only to creatures with a point of view
on the world. But there are no good grounds for making this stipulation.
Perceptual experiences with first-person nonconceptual contents derived
from the pick-up of self-specifying information can be entertained by
creatures who lack a point of view on the world in virtue of lacking con-
scious recognitional abilities. I can think of no non-question-begging
reason to deny that these are experiences. And in any case, even someone
190 Chapter 7
determined to deny that such states could count as experiences in some

rich and as yet unspecified sense of the word would nonetheless have to
accept that such a creature could still be conscious in virtue of having
sensations.10
We should not assume that only creatures possessed of the appropriate
conscious recognitional capacities can behave in ways that demand inten-
tional explanation. The operational criteria for the appropriateness of in-
tentional explanation discussed in chapter 4 neither make any reference
to, nor imply the existence of, conscious recognitional abilities. These cri-
teria derive from the basic thought that representational perceptual states
are intermediaries between sensory input and behavioral output to which
appeal needs to be made when there is no invariant connection between
input and output. There is no reason to think that a creature moving
around the world and acting on the basis of self-specifying information
in the various ways described by Gibson could not be capable of behavior
that is plastic and flexible enough to require intentional explanation. And
if the behavior is suitably plastic and flexible, then, as discussed in chapter
5, the first-person nonconceptual contents derivable from the pick-up of
self-specifying information are particularly well suited to capture the con-
tents of perceptual protobeliefs in way that explains their direct relevance
to action. Intentional explanation can enter into the picture prior to the
emergence of the conscious recognitional abilities necessary for posses-
sion of a point of view on the world.
This point can be brought home by reflecting on the strong pressures
to concede that at least some instrumentally conditioned behavior is in-
tentional (Russell 1980, Dickinson 1988). Instrumental conditioning cer-
tainly does not require conscious recognitional capacities, being perfectly
explicable on Gibsonian lines in terms of increasing perceptual sensitivity
to a particular affordance. All that instrumental conditioning requires is
consciousness in the very weak sense of sentience, since reinforcement
depends on the capacity to feel sensations. It is reasonable, I think, to
hold that the capacity for intentional behavior is restricted to creatures
that are conscious in the weak sense of being able to feel sensations. But
the converse does not hold. Creatures whose behavior is not intentional
can be perfectly capable of feeling sensations. This is presumably what
makes possible the various types of instrumental conditioning that do not
result in intentional behavior.
Points of View 191
The conclusion to draw, then, is that that we are dealing here with three
intersecting distinctions, rather than three different ways of describing the
same distinction. None of the three distinctions map cleanly on to any of
the others. Nonetheless, it does seem that these three distinctions give us
a hierarchy of cognitive abilities, ordered according to the relations of
dependence among these abilities. Basic consciousness is clearly the most
primitive of the three abilities. It is presupposed by both the others, while
itself presupposing neither of them. In fact, it is very plausible to think
that it is innate (Gazzaniga 1995). Next comes the capacity for inten-
tional behavior, which requires consciousness in the weak sense of sen-
tience but is independent of the conscious recognitional abilities that we
have seen to be necessary conditions for possession of a nonconceptual
point of view. It is perfectly possible, as I have stressed, for the first-person
nonconceptual contents yielded by the pick-up of self-specifying informa-
tion to occur at this level. Subject to the conditions discussed in this chap-
ter, such contents can interact with the highest of the three cognitive
abilities, which is the capacity to entertain conscious recognitional abili-
ties of the sort discussed in this chapter.
It seems reasonable to think that the categorization of these three fac-
tors in terms of their degrees of primitiveness provides useful clues to
answering the more general developmental question of how the capacity
for conscious recognition can arise, both ontogenetically and phylogenet-
ically. It will be remembered that the following constraint was suggested
as a means of ensuring that an articulated theory of self-consciousness
is suitably sensitive to the developmental progression from first-person
perceptual contents to first-person thought:
The Acquisition Constraint If a given psychological capacity is psycho-
logically real, then there must be an explanation of how it is possible for
an individual in the normal course of development to acquire that
capacity.
The application of the acquisition constraint to the present case is
straightforward. We have a modest hierarchy of three stages, intended to
capture relations of logical dependence. But how plausible is this hierar-
chy in developmental terms? Do the abilities associated with each level
provide sufficient resources for the acquisition of the abilities associated
with higher levels? The first level is basic sentience. This really presents
192 Chapter 7
no ontogenetic difficulties, since the capacity to feel sensations is innate

and certainly present long before birth. There is no need to explain how
the capacity to feel sensations arises in the developing individual.11 I have
also discussed how positing a primitive feeling of familiarity can help with
understanding the transition from intentional behavior without conscious
recognitional abilities to intentional behavior that does incorporate con-
scious recognitional abilities. The real problems come with the transition
from basic sentience to intentional behaviour. How can this transition
take place?
The first point to make is that the key to the whole procedure will be
provided by associations between stimuli and sensations. A vital part of
the evolutionary purpose of the capacity to feel sensations must presum-
ably be to make it possible for organisms to learn from their environment.
In its most basic form, this will take the form of correlations between
painful stimuli and harmful sensations and corresponding correlations
between pleasurable stimuli and beneficial sensations. The mechanisms
that allow such correlations to emerge support the basic form of learning:
learning that enables the organism to avoid things that are harmful to it
while actively pursuing things that are beneficial to it. At this level we are,
of course, dealing with stimulus-response behavior and the question is,
how and why does stimulus-response behavior develop into behavior
showing the flexibility and plasticity characteristic of intentional behav-
ior? The answer I tentatively offer is that intentional behavior emerges
when the number of stimulus-response correlations potentially relevant
to any given action becomes so great that the need arises to choose be-
tween a range of possible responses to a given situation. The real point
about intentional behavior is that a creature behaving intentionally be-
haves in a way that it need not have (which is why one important criterion
for intentional behavior is whether or not there is a lawlike correlation
between environmental stimulus and behavioral response). It is plausible
to see this sort of situation arising because there are so many ways in
which the creature could have behaved. The most primitive form of inten-
tional action surely arises when a decision procedure emerges for deciding
between competing possible courses of action, each of which may itself
have a nonintentional ancestry. Intentional behavior should be under-
stood at least in part as a response to a computational problem.
8
Navigation and Spatial Reasoning
Previous chapters have uncovered some of the basic elements of primitive

self-consciousness. In chapter 5, I used ecological theories of percep-
tion to show how nonconceptual first-person contents can feature in the
structure of perception, particularly visual perception. Chapter 6 ex-
plored conscious somatic proprioception as a comparable source of
self-specifying contents. Both of these are ground-level forms of self-
consciousness in the sense that, although they are not on their own suffi-
cient to warrant the ascription of self-consciousness in anything but a
derivative sense, they will be crucial building blocks in states that are
properly described as self-conscious. In chapter 7, I began to show how
one works up to self-conscious states by developing the notion of experi-
ence that reflects a nonconceptual point of view on the world, under-
standing this to involve taking a particular route through space-time in
such a way that one’s perception of the world is informed by an awareness
that one is taking such a route. I have so far examined only one aspect of
what it is to have a nonconceptual point of view on the world, namely,
that it must involve conscious or explicit memory as a condition of con-
scious place recognition. This chapter will complete the account.
8.1 From Place Recognition to a Nonconceptual Point of View:

Navigation and Spatial Awareness
At the heart of the notion of a nonconceptual point of view is registration

of the nonsolipsistic distinction between experience and what is experi-
enced. A minimal requirement of being able to make such a distinction is
that what is being experienced should be grasped as having an existence in-
194 Chapter 8
dependent of any particular apprehension of it. This in turn requires recog-

nition that what is experienced exists at times other than the particular
occasion on which it is apprehended. This is where conscious memory
deployed in recognition enters the picture, as a basic way of recognizing
this independent existence, because recognizing something involves corre-
lating past and present experience of that object. I then suggested that
recognizing places in terms of features present at those places is the most
basic way in which this can be carried out. This already takes us beyond
the account given of somatic proprioception and beyond Gibson’s account
of first-person contents gained through the pick-up of self-specifying in-
formation, but it still leaves us short of the notion of a nonconceptual
point of view on the world for reasons that will emerge in this section.
Although conscious place recognition has been offered as a way of reg-
istering the distinction between subjective experience and what it is expe-
rience of, it is not yet clear that we have an account of place recognition
rich enough for the job. In the previous chapter, place recognition was
explained through the recognition of features or objects* at particular
places. This indirect way of thinking about places needs to be distin-
guished from the more direct way of thinking about places that registers
that they are places. Only the second of these ways manifests what can
properly be described as an understanding of space, because understand-
ing the nature of space involves understanding that a network of spatial
relations holds between places independently of the spatial relations that
hold at any given time between the various objects* or features found at
those places. It is potentially misleading to say, as many philosophers have
said, that space is a single system of spatial relations. Certainly, there is
only a single system of spatial relations in the sense that all places are
interconnected, that is, in the sense that there cannot be two or more
spaces that are spatially insulated from each other. Nonetheless, for prac-
tical purposes, there are two importantly different types of spatial rela-
tions. There is the static and relatively unchanging system of spatial
relations holding between places, and there is the highly changeable sys-
tem of spatial relations holding between things always located at a partic-
ular place but not always at the same place.
The essence of navigation, and spatial thought in general, is bringing
these two types of spatial relations into harmony—being able to calcu-
Navigation and Spatial Reasoning 195
late, for example, that if a creature C has moved a certain distance at a

certain speed relative to creature D and if creature D is now at place P,
then creature C will be found at place P*. Spatial thought depends on a
grasp of the fact that spatial relations between places will stay constant
even though the spatial relations between objects vary, because a signifi-
cant component of it uses those constant spatial relations as a means of
mapping and keeping track of the changing spatial relations. Conse-
quently, any understanding of the nature of space must respect the distinc-
tion between these two different systems of spatial relations. Moreover,
this distinction also seems to be integral to the notion of a nonconceptual
point of view. Possessing a nonconceptual point of view on the world
involves being aware that one is taking a particular route through space-
time as one navigates that route, and an adequate awareness that one is
taking a particular route through space-time clearly presupposes a degree
of understanding of the nature of space.
Obviously, though, the account of place recognition in the previous
chapter does not respect this distinction, or more cautiously, the demands
placed upon place recognition could be satisfied by a creature with no
inkling of the distinction. It is perfectly possible for a creature to be cap-
able of recognizing features as features that it has seen before and to put
these recognitional abilities to work in navigating from one feature to
another, without apprehending that there are spatial relations holding be-
tween the places at which those features are instantiated and that those
spatial relations hold independently of the spatial relations between the
features. Place recognition in the minimal sense discussed in the previous
chapter can exist without an understanding of space. Such minimal place
recognition involves the recognition of things that are places but does not
demand that they be recognised as places. This indicates one dimension
along which my minimal account of place recognition needs to be de-
veloped if it is to be built up into the notion of a nonconceptual point
of view.
There is a second, related limitation in the minimal account of place
recognition. The exercise of conscious memory in the recognition of
places reflects a sense of the nonsolipsistic distinction between experience
and what is experienced, but I have said nothing about the functions that
this serves within the life of any creature capable of it. We are trying to
196 Chapter 8
capture the basic degree of self-consciousness implicated in a creature’s

having a point of view or perspective on the world, and it is reasonable
to expect that this degree of self-consciousness will be implicated across
a broad spectrum of cognitive and practical abilities. There seem to be
three important questions at issue here, none of which has yet been dis-
cussed. First, what part does this nonsolipsistic distinction play within
cognitive and practical life more widely construed? Second, what further
cognitive and practical abilities are made available by the possibility of
conscious place recognition? Third, what is the relevance of any of these
cognitive and practical abilities to the possession of a nonconceptual
point of view on the world and to primitive self-consciousness in general?
There are, then, two dimensions along which I am obligated to develop
the minimal account of place recognition, and there are good reasons to
think that the concept of navigation will be centrally implicated in both
of them. Different types of navigation involve different levels of spatial
awareness, with correspondingly different modes of understanding
places. It is highly plausible that the move from the minimal grasp of place
already in play to the fuller spatial understanding implicated in the notion
of a nonconceptual point of view can be profitably analysed as a move
from one form of navigational competence to another. Thus the concept
of navigation seems very helpful in fulfilling the first obligation. More-
over, the ability to navigate through a spatial environment is itself at the
intersection of a range of other cognitive and practical abilities, and hence
seems equally promising with respect to my second obligation.
In the following sections of this chapter I will be primarily concerned
with substantiating this general idea that careful attention to different
types of navigational ability will help to bridge the theoretical gap be-
tween minimal place recognition and full-fledged possession of a noncon-
ceptual point of view. Before moving on to this, however, it is important
clearly to articulate the different elements that have emerged in my discus-
sion of the notion of a nonconceptual point of view so that we can be
clear about the end point to which we are moving.
The most general characterization of the notion of a nonconceptual
point of view is this: having a nonconceptual point of view on the world
involves being aware that one is taking a particular route through space-
time as one navigates that route. But what exactly does this entail? A
minimal requirement upon any creature to whom a nonconceptual point

of view can correctly be ascribed is that it be capable of registering the
distinction between experience and the experienced. This is the nonsolip-
sistic requirement that, as I argued in the previous chapter, can be satisfied
only by forms of experience that do not take place wholly within a contin-
uous present. The temporal implications of the nonsolipsistic requirement
arise because grasping that an object of experience has an existence tran-
scending the particular occasion on which it is experienced implies grasp-
ing that what is being experienced at a particular time has either existed
in the past or will exist in the future. The possession of recognitional
abilities involving conscious memory provides the most basic way in
which the nonsolipsistic requirement can be satisfied, and I suggested that
the recognition of places in terms of features is the most basic such recog-
nitional ability.
In addition to this nonsolipsistic dimension of the notion of a noncon-
ceptual point of view, there is a second dimension. A nonconceptual point
of view requires awareness that one is taking a particular route through
space-time, and this brings with it the need for a degree of spatial aware-
ness. Recognizing places places cannot take place in isolation or purely
passively; it goes hand in hand with navigating from one place to another.
The navigational capacities that this involves will be a crucial element in
possessing a nonconceptual point of view on the world. When the partic-
ular route through space-time that a creature takes is determined by its
navigational abilities, its awareness of that route (demanded by its posses-
sion of a nonconceptual point of view) will have to reflect the fact that it
is navigating through the world. This in turn demands a degree of under-
standing of the nature of space. In particular, it demands an understand-
ing of the distinction between the spatial relations that hold between
things and the spatial relations that hold between places.
Let me then provisionally break down the notion of a nonconceptual
point of view into the following constituent elements:
The nonsolipsistic component (the distinction between experience and
what that experience is experience of) requires
a. grasping that an object of experience exists independently of a particu-
lar apprehension of it, which requires
198 Chapter 8
b. grasping that what is experienced exists at times other than those at

which it is experienced, which requires
c. the exercise of recognitional abilities involving conscious memory,
which is most primitively manifested in
d. feature-based recognition of places.
The spatial awareness component (putting place recognition to work for
navigational purposes) requires
a. awareness that one is navigating through the environment, which
requires
b. an understanding of the nature of space, which requires
c. a grasp of the distinction between the spatial relations that hold be-
tween places and the spatial relations that hold between things.
8.2 Spatial Awareness and Self-Consciousness
The issues raised by the close relationship between spatial awareness, self-
consciousness, and an understanding of the mind-independence of the
objects of experience has been discussed in the philosophical literature
relatively frequently since Kant opened the debate in the Critique of Pure
Reason. Two recent and influential treatments of the issues will be found
in Christopher Peacocke’s A Study of Concepts (1992) and John Camp-
bell’s Past, Space, and Self (1994). The juxtaposition of their respective
positions poses an important challenge, which it will be the aim of the
rest of this chapter to answer.
Christopher Peacocke has argued that any well-founded attribution to
a creature of a capacity for the reidentification of places presupposes that
the creature in question has mastery of the first-person concept.1 If sound,
this argument will have disastrous implications for the argument of this
book, for it directly entails the incoherence of the notion of a nonconcep-
tual point of view. Let me schematize his argument:
1. The attribution of genuine spatial representational content to a crea-
ture is justified only if that creature is capable of reidentifying places
over time.
2. Reidentifying places requires the capacity to identify one’s current lo-
cation with a location previously encountered.
3. Reidentifying places in this way involves building up an integrated rep-
resentation of the environment over time.
4. Neither (3) nor (4) would be possible unless the subject possessed at
least a primitive form of the first-person concept.
To appreciate the significance of (4), it is important to remember the Pri-
ority Thesis, discussed and endorsed in chapter 2:
The Priority Principle Conceptual abilities are constitutively linked with
linguistic abilities in such a way that conceptual abilities cannot be pos-
sessed by nonlinguistic creatures.
According to the Priority Principle, concept possession goes hand in hand
with linguistic mastery. Thus if Peacocke’s argument is sound, it entails
that what he is calling place reidentification cannot be available either to
nonlinguistic creatures or to linguistic creatures that lack mastery of the
first-person pronoun. Hence, on the assumption that Peacocke’s place re-
identification is an element in what I term possession of a nonconceptual
point of view on the world, his argument is a direct threat to the position
that I am developing.
Let us look more carefully at how Peacocke understands place reidenti-
fication. Place reidentification is the ability to identify a current place with
a previously encountered place. This seems comparable to what I have
been describing as place recognition involving conscious memory. In (3),
however, the notion of place reidentification is made rather richer with the
suggestion that place reidentification is available only to those creatures
capable of building up an integrated representation of the environment
over time:
To identify places over time requires the subject to be able to integrate the repre-
sentational contents of his successive perceptions into an integrated representa-
tion of the world around him, both near and far, past and present. I label this the
ability to engage in spatial reasoning. . . . Spatial reasoning involves the subject’s
building up a consistent representation of the world around him and of his loca-
tion in it. (Peacocke 1992, 91)
Although Peacocke does not go into details about what is to count as “a
consistent representation of the world around the subject,” it is plausible
to assume that it is rather comparable to elements (b) and (c) of what I
term the spatial awareness component, namely, that a representation of
the spatial layout of the environment registers an understanding of the
nature of space, and in particular, of the distinction between the spatial
relations that hold between places and those that hold between things.
200 Chapter 8
So it is only with (4) that the position I am developing differs from Pea-
cocke’s position. It is clear, therefore, that Peacocke’s argument presents
a challenge that has to be met: the challenge of showing how the ability
to engage in spatial reasoning is available to creatures who do not possess
the capacity for reflexive self-reference.
One way of trying to meet this challenge would be to deny the strong
conditions that Peacocke places on the capacity for place reidentification.
In particular, one might query his suggestion that place reidentification is
available only to creatures who have the capacity to construct suitably
integrated representations of their environment and to engage in spatial
reasoning in a way that necessarily involves grasp of the first-person pro-
noun. This could be done, for example, by appealing to the notion of
causally indexical comprehension developed by John Campbell (1994).2
Grasping a causally indexical notion just is grasping its implications for
one’s own actions. Examples of such causally indexical notions are that
something is too heavy for one to lift, or that something else is within
reach. As Campbell points out, grasp of causally indexical notions may
be linked with a reflective understanding of one’s own capacities and of
the relevant properties of the object. On the other hand, it need not be so
linked: he maintains that it makes sense to ascribe to creatures a grasp of
such causally indexical notions without ascribing to them any grasp of
notions that are not causally indexical (even though those causally non-
indexical notions might be essential to characterize the causally indexical
notions), because the significance of such notions is exhausted by their
implications for perception and action. And as such, causally indexical
mental states qualify as states with nonconceptual content.
In the present context, then, one might suggest that there is no need to
place the theoretical weight that Peacocke does on disengaged reflective
and reasoning abilities. Rather, we should be looking at how a given crea-
ture interacts with its surroundings, because that will be how it manifests
its grasp of the spatial properties of its environment (of the connectedness
of places, for example). On such a view, a creature could be capable of
spatial reasoning in the absence of any capacity to reflect on its interac-
tions with its environment. If this line is pursued, then it seems to provide
a way in which we can hang on to the idea of place reidentification with-
out accepting Peacocke’s claim that it implicates possession of a primitive
form of the first-person concept, precisely because it denies Peacocke’s

claim that any creature capable of place reidentification must be capable
of explicitly representing itself and its location in its surroundings.
The trouble with this suggestion, however, is that Campbell’s under-
standing of causally indexical spatial reasoning is rather orthogonal to
the account I have been developing of the essential features of experience
reflecting a nonconceptual point of view. Campbell calls upon the distinc-
tion between causal indexicality and causal nonindexicality to perform a
variety of tasks and to give concrete sense to a range of other distinctions.
One of these distinctions is that between absolute and egocentric ways of
representing space, where an absolute representation of space is one that
is not explicable in terms of an organism’s interactions with space. Camp-
bell wants to argue against the pragmatist’s denial that there are any such
absolute representations of space and to defend the view that there are
ways of thinking about space that are completely independent of one’s
engagement with it, where such engagement is construed in navigational
terms. This is all very well but, as usually happens with these “either”/
“or” distinctions, far too many discrete and rather different phenomena
get lumped together on one side of the divide. In particular, Campbell’s
notion of causal indexicality has little or nothing to say about how the
environment must be represented to make different types of navigation
possible (Campbell 1993, 25). Different types of navigation are classified
together because they all presuppose that places are physically significant
in terms of their implications for perception and action. But this glosses
over crucial distinctions, one of which is the distinction between those
forms of navigation that do presuppose an integrated representation of
the environment over time and those forms of navigation that do not. Of
course, one might think that possession of an integrated representation
of the environment over time is available only to creatures who are cap-
able of representing space absolutely, which for Campbell means that it
is available only to language-using adult humans (Campbell 1993, 25).
But this is surely an empirical claim open to empirical refutation. It is not
something that can be assumed at the beginning of enquiry.
The positions staked out by Peacocke and Campbell provide the frame-
work for an account of spatial reasoning. Such an account will show, I
hope, how the ability to engage in spatial reasoning can be available to
202 Chapter 8
creatures who do not possess the capacity for linguistic self-reference,

without weakening the notion of spatial reasoning so far that it loses
touch with the core of the notion of a nonconceptual point of view. How
can this double challenge be met? Let me return briefly to Peacocke’s char-
acterization of spatial reasoning. Recall that the crux of spatial reasoning,
as Peacocke defines it, is that it involves building up an integrated repre-
sentation of the environment over time. And it is precisely this integrated
representation of the environment that is lacking in Campbell’s con-
ception of causally indexical spatial comprehension. The contrast that
Campbell draws between causally indexical and causally nonindexical
thought is a contrast between practical engagement with the environment
and detached and reflective contemplation of the environment, and his
discussion of the contrast makes clear that (in his view, at least) the sort
of integrated spatial reasoning under consideration falls on the “detached
and reflective” side of the divide. What is needed is an account of spatial
reasoning invoking a mode of representing the environment that is richer
than what Campbell offers but yet does not involve the sophisticated
skills that Peacocke builds into his notion of spatial reasoning.
If it can be shown that it does indeed make sense to attribute to prelin-
guistic creatures an integrated representation of their environment, then
the double challenge will have been met. But how might this be done?
Everything depends, of course, on how the notion of an integrated repre-
sentation of the environment is understood, and it is unfortunate that
Peacocke does not go into much detail on this. The first step, therefore,
is to offer a more detailed account of the notion of an integrated represen-
tation of the environment over time. I shall argue in the following section
that this more detailed account is best achieved through considering what
is distinctive of behavior governed by spatial reasoning involving such a
representation—and in particular, through considering what navigational
capacities are available only through such spatial reasoning. The next sec-
tion offers an account that emphasizes four such navigational abilities,
citing experimental evidence that shows that those navigational abilities
are available to nonlinguistic creatures. In the final section I make explicit
the implications of this for primitive self-consciousness and illustrate how
such an integrated representation of the environment might emerge from
the perception of affordances (as discussed in chapter 5).
8.3 Cognitive Maps and Integrated Representations of the Environment
Readers familiar with recent work in the philosophy and psychology of

spatial representation will notice the connection between the notion of
an integrated representation of the environment over time and what psy-
chologists often call a cognitive map. Indeed, it might seem natural to
think that there is no difference between them and that the notion of
possessing an integrated representation of the environment over time is
best elucidated through the notion of a cognitive map. This seems to me
to be a serious mistake, however, for reasons that I try to bring out in
this section.
Here is a definition of what is meant by a cognitive map from one of
the most significant recent works on animal-learning theory: “A cognitive
map is a record in the central nervous system of macroscopic geometric
relations among surfaces in the environment used to plan movements
through the environment” (Gallistel 1990, 103). This conception of a
cognitive map as a record in the nervous system of geometric relations
between surfaces in the environment is not at all what this chapter is
trying to capture. What I am interested in is the spatial-awareness compo-
nent of primitive self-consciousness, and that spatial-awareness compo-
nent is a personal-level state. Records in the nervous system, however,
´
are subpersonal states. Of course, as I have argued elsewhere (Bermudez
1995b, 1995d, 1995e), it is a mistake to make a rigid distinction between
personal and subpersonal levels of explanation. It is compelling to think
that for any creature to have the sort of conscious spatial awareness
implicated in primitive self-consciousness, it must have a suitable record
in the nervous system of the geometrical relations between environmental
surfaces. It is equally compelling (at least to my mind) to think that any
adequate explanation of conscious spatial awareness must appeal to
subpersonal facts about the recording of geometric relations in the
nervous system. But the fact remains that these are two very different
things.
One of the factors that makes it tempting to equate conscious spatial
awareness with possession of a cognitive map is that many psychologists
and philosophers who place theoretical weight on cognitive maps do so
within the context of a distinction between different levels of spatial
204 Chapter 8
awareness. The most sophisticated level of spatial awareness is usually

described in terms of possession of a cognitive map. One way in which
this point is often put is that possession of a cognitive map goes hand
in hand with the possibility of identifying places on an environment-
centered (allocentric), rather than body-centered (egocentric), frame of
reference. Many have thought that the representation of space through a
cognitive map is a crucial element in the detachment from practical con-
cerns that allows us to anchor the coordinates of our spatial thought on
something other than our own bodies. One philosopher who has placed
much stress on this thought is Gareth Evans, who sees possession of a
cognitive map as one of the crucial differentia between engagement in a
spatial world and representation of a spatial world: “The network of
input-output connections which underlie the idea of an egocentric space
could never be regarded as supporting a way of representing space (even
egocentric space) if it could not be brought by the subject into coincidence
with some such larger spatial representation of the world as is constituted
by a cognitive map” (Evans 1982, 163). But what does the expression
‘cognitive map’ mean in claims like these?
One thing that seems clear is that Evans’s use of the expression ‘cogni-
tive map’ must be very different from what Gallistel takes it to mean,
namely the storage of geometric information in the nervous system. Gal-
listel makes a very plausible case that possession of a cognitive map in his
sense is not at all restricted to creatures capable of what Evans would
recognize as representation of a spatial world. Chapters 5 and 6 of The
Organization of Learning (Gallistel 1990) defend the thesis that all ani-
mals from insects upward deploy cognitive maps with the same formal
characteristics in navigating around in the environment. Gallistel argues
that the cognitive maps that control movement in animals all preserve the
same set of geometric relations within a system of earth-centered ( geocen-
tric) coordinates. These relations are metric relations, the relations stud-
ied by metric geometry, of which Euclidean geometry is the best-known
example. The distinctive feature of a metric geometry is that it preserves
all the geometric relations between the points in the coordinate system.
Gallistel’s thesis is thus in direct opposition to the widely held view that
the spatial representations of lower animals are weaker than those of
higher animals because they can record only a limited set of geometric
relations. He maintains that, although the cognitive maps of lower ani-
mals have far fewer places on them, they record the same geometrical
relations between those points as humans and other higher animals.
Moreover, he offers a uniform account of how such metric cognitive maps
are constructed in the animal kingdom:
I suggest that the surveying procedures by which animals construct their cognitive
map of their environment may be broken down into two inter-related sets of pro-
cesses, which together permit the construction of geocentric metric maps. One set
of processes constructs a metric representation in egocentric coordinates of the
relative positions of currently perceptible points, lines and surfaces—a representa-
tion of what is perceived from one’s current vantage point. The other process,
dead reckoning, provides a representation in geocentric coordinates of the van-
tage points and the angles of view (headings). Combining geocentric representa-
tions of vantage points and angles of view with egocentric representations of the
segment of the world perceived from each vantage point yields a representation
of the shape of the behaviourally relevant environment in a system of coordinates
anchored to the earth, a geocentric cognitive map. (Gallistel 1990, 106)
Dead reckoning (the process of keeping track of changes in velocity over
time) yields an earth-centered representation of vantage points and angles
of view that combines with current perceptual experience of the environ-
ment to yield an earth-centered cognitive map.
The clear implication of Gallistel’s account of animal representation is
that his sense of ‘cognitive map’ is of no use in distinguishing different
forms of spatial awareness, since it is presupposed by just about every
form of spatial awareness. How about the other senses of the expression?
Another commonly encountered way of understanding ‘cognitive map’ is
in imagistic terms, as a mental picture that the mind consults to solve
navigational problems, but this is even more clearly a nonstarter. When
we put to one side the imagistic and subpersonal understandings of the
notion of a cognitive map, we seem to be left with a notion that is rather
amorphous. Gareth Evans offers the following definition: “a representa-
tion in which the spatial relations of several distinct things are simultane-
ously represented” (1982, 151). This is not very helpful. Despite Evans’s
insistence on seeing possession of a cognitive map as the key to objective
thought about space, this definition does not distinguish between egocen-
tric and allocentric frames of reference. And it is no more helpful to be
206 Chapter 8
told by E. C. Tolman, who coined the phrase, that a cognitive map is the
mental analogue of a topographic map, for this is only as clear as the
notion of a mental analogue.
As this brief examination of the concept of a cognitive map shows, it is
a concept rather unsuited to the task of helping explain what it is to have
an integrated representation of one’s environment over time. The clearest
sense of a cognitive map is Gallistel’s construal in terms of the storage of
metric information in the nervous system, but this is not going to explain
how possession of an integrated representation of the environment over
time differs from other forms of spatial representation. None of the other
senses of ‘cognitive map’ is really clear enough to explain anything. The
conclusion to draw from this, I think, is that to make any progress on the
notion of an integrated representation of the environment over time, we
need to proceed in broadly functional or operational terms. That is, we
need to consider what forms and types of navigational behavior should
be taken as appropriate criteria for the ascription to a creature of an inte-
grated representation of the environment. I shall go on to discuss four
such types of navigational ability in the next section. Before moving on,
however, there is an important loose end to tie up.
In the previous section I argued for the importance of developing a
position midway between the cognitivist account of spatial reasoning fa-
vored by Christopher Peacocke and the theory of causally indexical spa-
tial awareness favored by John Campbell. But it might be asked at this
point if the proposal to consider what it is to possess an integrated repre-
sentation of the environment over time in broadly functional and opera-
tional terms is compatible with that stated aim. I argued that Campbell’s
notion of an awareness that is exhausted in its implications for action and
perception is not rich enough to capture the spatial-awareness component
of a nonconceptual point of view. But it is open to a defender of Camp-
bell’s position to argue that construing spatial representation in func-
tional or operational terms is simply offering up a causally indexical
account under another name. Surely, it might be argued, a functional or
operational construal of spatial representation suffers from precisely the
same defects attributed to the view that spatial representation is ex-
hausted in its implications for action and perception, simply because it
is construing spatial representation purely in terms of its behavioral

outputs.
Let me begin evaluating this objection by recalling a familiar distinction
between criterial and constitutive accounts of a particular phenomenon.
A constitutive account is intended to explain what the phenomenon in
question is, whereas a criterial account is intended to identify the condi-
tions whose satisfaction will be the basis (on causal explanatory grounds)
for a warranted claim that the phenomenon is indeed present. The first
point to make is that the operational and functional construal would have
to be construed as a constitutive account of spatial representation for the
objection to stick. But, of course, it is not being offered as a constitutive
account. It is a criterial account. The two different forms of explanation
should not be confused. The central tenet of functional /operational ex-
planation is that the best way of studying a psychological phenomenon is
in terms of the causal relations that it bears to certain inputs and certain
outputs, and it would be a caricature of this position to say that it identi-
fies the psychological phenomenon with these inputs and outputs. That,
after all, is the crux of the distinction between functionalism and crude
behaviorism. The sort of functional/operational analysis that I shall pro-
pose in the next section identifies certain forms of navigational abilities
that seem to be explicable only on the assumption that the behavior in
question is governed by an integrated representation of the environment
over time. The procedure here is, of course, inference to the best explana-
tion. As I shall try to bring out, the empirical evidence overwhelmingly
supports the view that there is a principled class of navigational abilities
for which inference to the best explanation demands the attribution of
an integrated representation of the environment over time.
8.4 Navigation Deploying an Integrated Representation of the

Environment over Time
A good deal of the most interesting work on the representation of space

has been carried out by animal-learning theorists. There is a very good
reason for this. Animal-learning theorists study nonlinguistic creatures,
and so have no access to verbal reports as a source of evidence for the
208 Chapter 8
presence of mental representations. Of necessity, therefore, they have been

forced to consider the representational abilities of animals indirectly and
to develop very sophisticated ways of evaluating the evidence for the pres-
ence of those cognitive abilities and testing to make sure that there are
no more-parsimonious explanations available that do not involve mental
representations. Similar points hold of those developmental psychologists
primarily concerned with prelinguistic infants. Here too we find a highly
sophisticated battery of experimental techniques designed to isolate pre-
cisely those behavioral responses that can be explained only intentionally.
It will be no surprise, therefore, to find that animal-learning theory and
developmental psychology will feature heavily in this section.
Let me begin by stating two basic conditions that must be satisfied be-
fore there is any possibility of ascribing to a creature integrated represen-
tations of the environment. These are necessary but not sufficient
conditions. They correspond to the four conditions that, in chapter 4, I
placed on any behavior whose explanation requires appeal to content-
bearing psychological states. Just as these four conditions mark out those
forms of behavior that can properly be described as representation-
driven, there are two conditions that must be satisfied by any form of
navigation properly describable as driven by the representation of spatial
features of the environment (which is not the same as what I am terming
an integrated representation of the environment over time).
The first of these conditions has long been recognized to be important
by animal-learning theorists. This is the condition that genuine spatial-
navigation behavior should be place-driven rather than response-driven
(or, in terms of what is by now a familiar distinction, that it be the result
of place learning rather than response learning). The first question that
must be asked of any putatively spatial behavior is whether the apparently
spatial movement is reducible to a particular sequence of motor move-
ments, as opposed to being targeted at a particular place. The point of
this condition is that it maps very neatly onto the general requirement
that no genuinely representational behavior can be explicable in stimulus-
response terms, because particular sequences of movements are para-
digms of what can be learned as responses to stimuli. The standard
stimulus-response explanation of maze learning in rats is that the animal
learns a series of responses that are reinforced by the rewards that follow
correct responses. At each point in the maze where there is more than one
possible path (the “choice points”) the rat learns to go right or go straight
ahead, where these responses can be coded in terms of the sequences of
movements that they involve, and it is these sequences of movements that
are rewarded.
The direct reinforcement of particular sequences of movements is a par-
adigm of the mechanistic explanations of behavior that stimulus-response
theory attempts to provide. This means that it will be a necessary condi-
tion on any behavior genuinely driven by an integrated representation of
the environment over time that it not be reducible to particular sequences
of movements. But what is the difference between behavior reducible to
particular sequences of movements and behavior not so reducible? A clas-
sic and elegant experiment by Tolman, Ritchie, and Kalish (1946) is a
paradigm illustration of the difference between place learning and re-
sponse learning. Tolman used a cross maze with four endpoints (north,
south, east, west—although these are labels rather than compass direc-
tions). Rats were started at north and south on alternate trials. One group
of rats were rewarded by food located at the same endpoint, say east—
the point being that the same turning response would not invariably re-
turn them to the reward. For the other group, the location of the food
reward was shifted between east and west so that, whether they started
at north or south, the same turning response would be required to obtain
the reward. This simple experiment shows very clearly the distinction be-
tween place learning and response learning. To learn to run the maze and
obtain the reward, the first group of rats (those for which the food was
always in the same place, although their starting-points differed) must
represent the reward as being at a particular place and control their move-
ments accordingly. If they merely repeated the same response, they would
only succeed in reaching the food reward on half of the trials. For the
second group, though, repeating the same turning response would invari-
ably bring them to the reward, irrespective of the starting point.
Tolman found that the first group of rats learned to run the maze much
more quickly than the second group. From this he drew conclusions about
the nature of animal learning in general, namely that it is easier for ani-
mals to code spatial information in terms of places rather than in terms of
particular sequences of movements. This general thesis—one that many
210 Chapter 8
subsequent researchers have been unwilling to accept—is not (at least for
my purposes) the most interesting aspect of Tolman’s experiment. What
is more interesting, I think, is that the distinction that his experimental
paradigm brings out offers an excellent operational illustration of a mini-
mal condition that has to be fulfilled before one can even start to enquire
whether or not a creature’s behavior is governed by an integrated repre-
sentation of its environment. The minimal condition is simply that the
behavior cannot be understood in terms of learned responses that can be
coded in terms of bodily movements rather than their distal targets.
This initial baseline condition cannot stand alone, however. There is a
second condition on the class of navigation-behaviors driven by spatial
representation of the environment. Although the contrast between re-
sponse learning and place learning is clear enough to illustrate how
behavior driven by information coded in terms of chained movement se-
quences cannot count as driven by spatial representation of the environ-
ment, it is not clear that we are in a position to affirm the converse
proposition, namely, that all behavior that resists explanation in terms
of learned responses is ipso facto driven by spatial representation of the
environment. In other words, the distinction on which the first minimal
condition rests is not exhaustive. This can be appreciated by considering
the possibility that a creature’s navigational behavior might be driven nei-
ther by coded movement responses nor by the representation of places
but instead by sensitivity to features of the environment that covary with
spatial features. Let me give an example.
Animal behaviorists puzzled by the extraordinary homing abilities of
birds, particularly of homing pigeons, have put forward several different
explanations of how a bird released into completely unfamiliar territory
hundreds of miles from its home can almost immediately set a fairly direct
course for home. On the widely accepted assumption that homing be-
havior cannot be explained in terms of any form of dead reckoning, the
birds must be reacting to stimuli that convey spatial information, and the
challenge is to identify those stimuli and how they are registered. The ex-
planations put forward fall into two broad categories (Gallistel 1990,
144–148). One set of explanations propose that the birds are registering
the angle of arrival of nonvisual stimuli propagated from their home posi-
tion, such as odors or low frequency sounds. A second class of explana-

tions appeals to the possibility of bicoordinate navigation. Bicoordinate
navigation is navigation in terms of stimuli that can be perceived from
any point on the earth’s surface and that covary with latitude and longi-
tude, such as the position of the sun and stars and/or the earth’s magnetic
field. Bicoordinate navigation on the basis of magnetic information can
illustrate the difference between sensitivity to spatial features of the envi-
ronment and sensitivity to features of the environment that covary with
spatial features. Although we are obviously dealing with navigation that
is representationally far richer than learned movement responses, in nei-
ther case do we yet have behavior driven by direct sensitivity to spatial
features of the environment. The fact that the magnetic field increases
in strength toward the poles, for example, might perhaps offer a way of
obtaining compass orientation, but it is not itself a spatial feature of the
environment. Nor are odors or low frequency sounds emanating from the
home position spatial features of that home position. If navigation behav-
ior is driven by the computation of magnetic information, then there is
no need to appeal to the distal target in explaining what is going on.
And if the behavior is explained through quite extraordinary sensitivity
to nonspatial stimuli then, even though the distal target will play a crucial
role in the explanation, those features of it involved in the explanation
will not be spatial features.
This gives us two minimal necessary (negative) conditions that must be
satisfied before there is any possibility of describing a creature’s behavior
as driven by representations of the spatial features of its environment. It
must be the case both that the behavior in question is not reducible to
coded sequences of bodily movements and that it is not driven by sensi-
tivity to features of the environment that merely covary with spatial fea-
tures. While noting that it does not seem to be true that the conjunction
of these conditions provides a sufficient condition for the existence of
behavior driven by representation of spatial features of the environment,
let me pass over the question of how exactly to specify such a sufficient
condition and turn to the further question of how one might build up
from behavior that satisfies these two minimal necessary conditions to
behavior that reflects possession of an integrated representation of the
212 Chapter 8
world. Here it seems that there are three further and more sophisticated
conditions that capture the core of this more advanced form of spatial
representation.
Possessing an integrated representation of the world involves appreciat-
ing a system of simultaneous spatial relations. This entails that, for any
given route between two places, a subject who possesses an integrated
representation of the world will appreciate that the route is not the only
possible route. In fact, it seems to be a central aspect of appreciating that
something is a route between two points that one appreciate that there are
other possible routes between those same two points. This is, of course, a
vital element in appreciating the connectivity of space. One key way of
giving practical significance to the idea that all places are connected with
each other is by appreciating that there are multiple routes between any
two places. But how are we to understand this in operational terms? What
navigational abilities would count as practical manifestations of the grasp
that any given route between two points is merely one of a set of pos-
sible routes?
In general terms, the relevant navigational ability is the capacity to
think about different routes to the same place. This navigational ability
can, of course, take many specific practical forms. One very obvious such
practical form is the capacity to navigate around obstacles. A good index
of a creature’s possession of an integrated representation of a given spatial
environment is that, should it find its customary route to a particular
place (say a source of water) blocked for some reason, it will try to navi-
gate around that obstacle. An interesting illustration of how the capacity
to think about different routes to the same place and the capacity to navi-
gate around obstacles can come together can be found in a set of experi-
ments carried out by Tolman and Honzik (1930a). Tolman and Honzik
ran rats through a maze in which three different routes of varying length
led from the starting point to a food reward. Unsurprisingly, the rats
quickly learned to follow the shortest route to the reward, and when the
shortest route was blocked, they quickly learned to use the middle-length
route. The interesting result occurred when the shortest route was
blocked at a point after the middle-length route had rejoined it. According
to Tolman and Honzik, in such a situation their rats did not attempt to
take the middle-length route but instead moved straight from the shortest
Figure 8.1
A creature that displays perspectival sensitivity will take the shortest path to
obtain food at point C. This entails that if the creature sees food at C while at
A and later moves to B, to go to C, it will follow the path BC.
route to the longest. Here the rats are representing the maze as comprising
different intersecting routes to a single point and are actively comparing
those routes to each other. They have an integrated representation over
time of at least the local environment of the maze.
As I mentioned earlier in the chapter, there are two importantly differ-
ent systems of spatial relations. There is a static and unchanging system
of spatial relations holding between places and a highly changing system
of spatial relations holding between things that are always located at a
particular place but not always at the same place. Possessing an integrated
representation of space involves grasping how these two systems of spatial
relations map onto each other and being able to bring them into harmony
for navigational purposes. Part of what makes this integrated represen-
tation so central is that all creatures capable of representing space are
themselves moving through space. This gives a second condition upon
possession of an integrated representation of the environment, namely
that any creature who can plausibly be ascribed such a representation
must be capable of reacting to the spatial properties of the objects and
stimuli it encounters in a way that is sensitive to its own changing
position.
The importance of this latter condition has been stressed by Christo-
pher Peacocke in the notion of perspectival sensitivity that he develops in
Sense and Content. Let me begin by explaining how Peacocke under-
stands the notion. Peacocke illustrates what he means with the following
example. In figure 8.1 we are to suppose that a creature at place A can
perceive an object at place B and is accustomed to obtain food at place
C. According to Peacocke, a subject who satisfies the requirement of per-
spectival sensitivity, after he has moved from A to B by a path that he
recognizes as taking him from A to B, will move directly along the line
214 Chapter 8
IIh′ B
IIh A IIl′
IIl
IIm′
IIm
C
Figure 8.2
An intentional web initially centered on place A and then recentered on place B.
A subject who wants to move from A to C but finds himself having to move from
A to B before implementing his intention will follow the route given by Pl9 if he
displays perspectival sensitivity. (From Peacocke 1983, 68. Redrawn by permis-
sion of Oxford University Press.)
BC, rather than by moving from B along a path parallel to AC, when he
desires the food that he ordinarily obtains at C. Note, moreover, that this
requirement of perspectival sensitivity is richer than the requirement men-
tioned earlier that spatial responses not be reducible to particular se-
quences of movements. Although it is true that any subject for whom the
route from A to C is coded in terms of the movements required to get from A
to C will fail to display perspectival sensitivity, the converse does not hold.
Accepting Peacocke’s illustration of perspectival sensitivity does not
mean, however, that we should accept without qualification his positive
account of what is going on when perspectival sensitivity is displayed.
The account he gives seems, in one important respect, to be too impover-
ished. Let me explain. Peacocke’s account of perspectival sensitivity is
based upon the notion of an intentional web. Consider figure 8.2. In the
figure Ph , Pl , and Pm represent experience types in which given objects
are perceptually presented, and the arrows represent the distances and
directions in which a subject would have to move were he to decide to
move toward those objects. This diagram is what Peacocke terms an in-
tentional web. It is important to realize that the experience types Ph , Pl ,
and Pm are individuated in terms of what Peacocke terms sensational
properties, that is, the nonrepresentational properties that an experience
has in virtue of what it is like to have that experience (Peacocke 1983, 5).
This provides the material for understanding his notion of perspectival
sensitivity, as follows:
A simplified general statement of the requirement of perspectival sensitivity would
be this: if the subject moves from one place to another, his intentional web must
be recentered on the place determined in normal circumstances by the change in
the sensational properties of his experience. . . . Perspectival sensitivity is literally
a matter, in actual and counterfactual circumstances, of the sensitivity of the sub-
ject’s intentional actions to variations in his perspective on the world. (Peacocke
1983, 69)
The sense in which this account of perspectival sensitivity is too impover-
ished seems to me to lie in its restriction of the notion of an intentional
web to occurrently perceived objects and routes to those objects. This
neglects the importance of spatial memory. A subject’s intentional web
must also include remembered locations and the routes to those locations.
This is so for the following reason. A perspectivally sensitive subject will
be sensitive to how the spatial properties of objects are altered as a func-
tion of his changing position. Among these spatial properties will be the
spatial relations in which occurrently perceived objects stand to objects
that have been encountered in the past and will be encountered in the
future. If these spatial relations are left out of the equation, then the prac-
tical significance of perspectival sensitivity will be severely diminished. I
propose, therefore, to take over Peacocke’s notion of perspectival sensitiv-
ity, subject to this expansion of the core notion of an intentional web.
Three examples will show how this notion of perspectival sensitivity
can be practically deployed in operational terms. Consider the following
experiment carried out by Chapuis and Varlet (1987). The experiment
was performed outdoors with Alsatian dogs. The dogs were taken on a
leash along the path ADB shown in the diagram and shown, but not al-
lowed to eat, pieces of meat at A and B (see figure 8.3). The three points
were all far enough apart to be invisible from each other.3 On the assump-
tion that the dogs, when released, would want to obtain food from both
A and B it is clear what perspectival sensitivity would demand in this
situation, namely, that if the dogs go first to A, they then proceed directly
to B along the dotted line AB, and vice versa. And this is in fact what
occurred in 96 percent of the trials. It is important to realize that what
is going on here is different from what is going on in the Tolman and
216 Chapter 8
22.6 m
40 m
24 m
30°
Figure 8.3
Perspectival sensitivity in dogs. D is the starting point, and A and B the food
points. The solid lines represent exploratory runs with the dogs on leashes, and
the dotted line the shortcut taken by the dogs when released. (From Chapuis and
Varlet 1987.)
Honzik experiments discussed above. The rats in those experiments (as

interpreted by the experimenters) are comparing three paths to a given
goal with which they are already familiar. The rats register that there are
different, though interconnected, paths to a single goal and make calcula-
tions about their relative efficiency, but this falls short of perspectival sen-
sitivity in Peacocke’s sense. What distinguishes the behavior of the dogs
is that they are using information about routes with which they are famil-
iar to devise a shortcut between two points in a way that depends on
sensitivity to their changing spatial position.
Another good example of behavior for which the demands of inference
to the best explanation seem to require the attribution of perspectival
sensitivity (behavior displaying perhaps an even greater spatial mastery)
is the performance of a young chimpanzee in the so-called traveling-
salesman combinatorial problem (Menzel 1973). The chimpanzee was
carried by an experimenter who hid food at 18 different locations, so the
chimpanzee was able to see every location and the routes between them.
When left to its own devices, the chimpanzee recovered almost all of the
food, employing an optimal path considerably more economical in terms
both of distance and of preference (when the foods were of different val-
ues). The chimp is able to construct an optimal recovery route only
because it can track spatial relations in a way that is sensitive to its chang-
ing position.
Perspectival sensitivity can clearly be demonstrated in young infants,
as is shown by a experimental paradigm employed by Acredolo, Adams,
and Goodwyn (1984). The paradigm comprises two containers sur-
rounded on three sides by transparent screens (figure 8.4). From the posi-
tion marked A in the diagram the experimental infants were shown an
object hidden in one of the two containers. Because of the glass walls,
however, they could reach it only by crawling around to the gap in the
wall on the opposite side of the apparatus (point B). The location of the
object relative to the infant is, of course, inverted as he moves around
from point A to point B (the container that was formerly on the infant’s
right is now on his left, and vice versa). The infants’ successes reported
by Acredolo, Adams, and Goodwyn clearly indicates sensitivity on the
part of the experimental infants to the spatial implications of their chang-
ing position.
218 Chapter 8
Figure 8.4
Perspectival sensitivity in infancy. The infant sees an object hidden in one of the
containers but is prevented from reaching the object by transparent glass screens.
To obtain the object, the infant must crawl around the array to B. (From Acredolo,
Adams, and Goodwyn 1984.)
Let me turn now to the third component in possessing an integrated

representation of one’s environment. As I have already mentioned several
times, the representation of space involves the capacity to represent two
different classes of spatial relations: the relatively unchanging spatial rela-
tions that hold between places and the far more dynamic relations that
hold between things, which change positions. The constraint of perspecti-
val sensitivity provides one condition on a proper grasp of how these two
different types of spatial relations interact, a condition that follows from
the fact that the perceiving subject is himself a moving thing. But this is a
further condition that needs to be extracted here. If a creature is to be
able to harmonize these two types of spatial relations, it must be able to
grasp at some level that there are two different types of spatial relations,
and this in turn seems to demand a grasp that places are distinct from the
things found at those places. This holds even when ‘thing’ is understood
weakly, so that features count as “things.” A creature that can think about
a particular spatial location only in terms of an object* or feature existing
at that location has not grasped the nature of space in the way required
to possess an integrated representation of its environment over time (any
more than a creature who can think about a particular object only in
terms of a single spatial location). There is a crucial requirement here,
namely that a creature should be able to recognize that places persist
through changes in the objects* or features located at those places. This
in turn seems to require the capacity to recognize and reidentify places
independently of objects or features located at those places.
Another set of experiments carried out on rats in mazes by Tolman
provide a good illustration of how this capacity might be practically mani-
fested. These experiments are the so-called “latent learning” experiments
(Tolman and Honzik 1930b). Two different groups of rats were run
through a maze for several days. One group were provided with rewards
at the end of the maze, while the second group went unrewarded. As one
might expect, the rats that found food at the end of the maze quickly
learned to run the maze, while the unrewarded rats just seemed to wander
aimlessly about, turning into as many blind alleys at the end of the test
period as they had at the beginning. Much more striking, though, was
what happened when a food reward was put into the maze of a rat that
had hitherto been unrewarded. The rats who previously appeared not to
have learned the correct sequence choices required to run the maze were
almost immediately able to run the maze more or less perfectly once they
encountered the food reward. Placed back at the beginning of the maze
the day after finding the reward, they took the most direct path to the end
of the maze, rather than repeat the (apparently random) movements that
had originally taken them there.
The most obvious way of interpreting these results (and certainly the
way they were interpreted by Tolman and Honzik) is to hold that the rats
were already familiar with the quickest way through the maze by the time
the food reward was introduced but simply had no reason to put their
“latent learning” into practice before encountering the food reward.
What is interesting for present purposes is how latent learning illustrates
the representation of space. Latent learning depends upon the rats regis-
tering that the place where they encounter the food is the same as the
place that they had previously visited without encountering any food. If
the rats could only think about a particular place in terms of a particular
object or feature that they find there, then latent learning would be impos-
sible. The behavior revealed in the latent-learning experiments is a good
220 Chapter 8
example of the exercise of the capacity to think about places as persisting

through changes of features or objects located at those places.
8.5 The Notion of a Nonconceptual Point of View and Primitive

Self-Consciousness
Earlier in this chapter I broke the notion of a nonconceptual point of

view down into two main components, which I termed the nonsolipsistic
component and the spatial-awareness component. This chapter has been
primarily concerned with the second of these two components. Let me
begin this section by adding the final details to this structural account of
the notion of a nonconceptual point of view.
The nonsolipsistic component (the distinction between experience and
what that experience is experience of) requires
a. grasping that an object of experience exists independently of a particu-
lar experience of it, which requires
b. grasping that a thing exists at times other than those at which it is
experienced, which requires
c. the exercise of recognitional abilities involving conscious memory,
which is most primitively manifested in
d. feature-based recognition of places.
The spatial-awareness component (putting place recognition to work for
navigational purposes) requires
a. awareness that one is navigating through the environment, which
requires
b. a degree of understanding of the nature of space, which requires
c. a grasp of the distinction between the spatial relations that hold be-
tween places and the spatial relations that hold between things, which is
manifested in navigational behavior that
d. satisfies the following two minimal conditions:
i. not being reducible to particular sequences of bodily movements,
ii. not being driven by sensitivity to features of the environment that
merely covary with spatial features of the environment;
e. and implicates the following three cognitive capacities:
i. the capacity to think about different routes to the same place,
ii. the capacity to keep track of changes in spatial relations between
things caused by one’s own movements relative to those things,
iii. the capacity to think about places independently of the objects*

or features located at those places.
A very natural question to raise when confronted with an apparently bi-
partite notion like that just put forward is, In what sense, if any, do the
two components mesh together to form a cognitive ‘natural kind’? An
unsympathetic reader might be inclined to say that the notion of a non-
conceptual point of view has been gerrymandered out of two distinct sets
of cognitive abilities that happen both to involve some notion of place
recognition and place reidentification. But this would be a serious mis-
take. To put the point briefly, the significance of the structured notion of
a nonconceptual point of view is that it allows us to understand how the
forms of first-person perceptual content discussed in chapters 5 and 6 can
be put to work to yield a genuine form of primitive self-consciousness.
This is what unites the two central components.
The notion of a nonconceptual point of view brings together the capac-
ity to register one’s distinctness from the physical environment and vari-
ous navigational capacities that manifest a degree of understanding of
the spatial nature of the physical environment. One very basic reason for
thinking that these two elements must be considered together emerges
from a point made at the beginning of the previous chapter: the richness
of the self-awareness that accompanies the capacity to distinguish the self
from the environment is directly proportionate to the richness of the
awareness of the environment from which the self is being distinguished.
So no creature can understand its own distinctness from the physical envi-
ronment without having an independent understanding of the nature of
the physical environment, and since the physical environment is essentially
spatial, this requires an understanding of the spatial nature of the physical
environment. But this cannot be the whole story. It leaves unexplained
why an understanding should be required of this particular essential fea-
ture of the physical environment. After all, it is also an essential feature
of the physical environment that it be composed of objects that have both
primary and secondary qualities, but there is no reflection of this in the
notion of a nonconceptual point of view. More is needed to understand
the significance of spatiality.
Let me step back briefly from primitive self-consciousness to consider
the account of self-identifying first-person thoughts given in Gareth
222 Chapter 8
Evans’s Varieties of Reference (1982). Evans places considerable stress on

the connection between the sophisticated form of self-consciousness that
he is considering and a grasp of the spatial nature of the world. As far as
Evans is concerned, the capacity to think genuine first-person thoughts
implicates a capacity for self-location, which he construes in terms of a
thinker’s ability to conceive of himself as identical with an element of the
objective order. Though I don’t endorse the particular gloss that Evans
puts on this, the general idea is very powerful. The relevance of spatiality
to self-consciousness comes about not merely because the world is spatial
but also because the self-conscious subject is himself a spatial element of
the world. One cannot be self-conscious without being aware that one is
a spatial element of the world, and one cannot be aware that one is a
spatial element of the world without a grasp of the spatial nature of the
world. Evans tends to stress a dependence in the opposite direction be-
tween these notions:
The 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 to the more or less stable way the world is.
The idea that there is an objective world and the idea that the subject is some-
where cannot be separated, and where he is is given by what he can perceive.
(Evans 1982, 222)
But the thrust of his work is very much that the dependence holds equally
in the opposite direction.
It seems to me that this general idea can be extrapolated and brought
to bear on the notion of a nonconceptual point of view. What binds to-
gether the two apparently discrete components of a nonconceptual point
of view is precisely the fact that a creature’s self-awareness must be aware-
ness of itself as a spatial being that acts upon and is acted upon by the
spatial world. Evans’s own gloss on how a subject’s self-awareness is
awareness of himself as a spatial being involves the subject’s mastery of a
simple theory explaining how the world makes his perceptions as they
are, with principles like ‘I perceive that such and such; such and such
holds at P; so (probably) I am at P’ and ‘I am at P; such and such does
not hold at P, so I can’t really be perceiving such and such, even though it
appears that I am’ (Evans 1982, 223). This is not very satisfactory,
though. If the claim is that the subject must explicitly hold these prin-
ciples, then it is clearly false. If, on the other hand, the claim is that these
are the principles of a theory that a self-conscious subject must tacitly
know, then the claim seems very uninformative in the absence of a spe-
cification of the precise forms of behavior that can only be explained
by the ascription of such a body of tacit knowledge. We need an account
of what it is for a subject to be correctly described as possessing such a
simple theory of perception. My own view is that some of the naviga-
tional abilities discussed in this chapter will have a central role to play in
this account. But there is no need for present purposes to make a claim
that strong. The point I wish to stress is simply that the notion of a non-
conceptual point of view as I have presented it can be viewed as capturing,
at a more primitive level, precisely the same phenomenon that Evans is
trying to capture with his notion of a simple theory of perception.
There is an important respect, though, in which Evans’s emphasis on a
simple theory of perception can be misleading. Although usefully high-
lighting the spatiality of the self-conscious subject, it focuses attention on
the passive “input” side of that spatiality, rather than the active “output”
side. A vital element in a self-conscious subject’s grasp of how he himself
is a part of the physical world is indeed derived from an understanding of
how his perceptions are a function of his location in precisely the way
that Evans brings out. But it must not be forgotten that a vital role in this
is played by the subject’s own actions and movement. Appreciating the
spatiality of the environment and one’s place in it is largely a function of
grasping one’s possibilities for action within that environment: realizing
that if one wants to return to a particular place from here one must pass
through these intermediate places, or that if there is something there that
one wants, one should take this route to obtain it. That this is something
that Evans’s account could potentially overlook emerges when one reflects
that a simple theory of perception of the form that he describes could be
possessed and deployed by a subject that only moves passively. I take it to
be an advantage of the notion sketched in this chapter that it incorporates
the dimension of action by emphasizing the practicalities of navigation.
Moreover, stressing the importance of action and movement indicates
how the notion of a nonconceptual point of view might be grounded in
the self-specifying information to be found in visual perception. I am
thinking here particularly of the concept of an affordance so central to
224 Chapter 8
Gibsonian theories of perception. One important type of self-specifying

information in the visual field is information about the possibilities for
action and reaction that the environment affords the perceiver, and I sug-
gested earlier (in chapter 5) that affordances are nonconceptual first-
person contents that serve as building blocks in the development of self-
consciousness (though there are certain qualifications that emerged in
chapter 7 that have to do with the importance of conscious memory). We
are now in a position to give this idea a little more focus (albeit program-
matically). The development of a nonconceptual point of view clearly in-
volves certain forms of reasoning, which I earlier followed Peacocke in
calling spatial reasoning. Clearly, we will not have a full understanding
of the notion of a nonconceptual point of view until we have an explana-
tion of how this reasoning can take place. What I would suggest, though,
is that Gibsonian affordances form one important class of representations
over which this reasoning takes place. The spatial reasoning involved in
developing a nonconceptual point of view upon the world is largely a
matter of calibrating different affordances into an integrated representa-
tion of the world. This idea obviously needs to be worked out in detail,
but let me make some brief comments about what might be involved in
such a process of calibration and how it relates to discussion earlier in
this chapter.
Calibrating different affordances into an integrated representation of
the world will be a function of a creature’s understanding how those af-
fordances are spatially related to each other, and the basic materials for
such a calibration will, of course, be the navigational abilities involved in
getting from one affordance to another. Let me give some examples of
what might be involved here. I will assume that when a creature is able
to navigate from one affordance to another in a way that satisfies the two
minimal conditions discussed earlier, it is appropriate to describe it as
having an understanding of the spatial relation between those affordan-
ces. The question is how an understanding of individual spatial relations
of that sort becomes an integrated representation of the environment.
One very basic point here is that the spatial relations holding between
affordances have the properties of symmetry and transitivity. Correspond-
ingly, calibrating affordances involves recognizing such symmetry and
transitivity. Let me term this a grasp of affordance symmetry and af-
fordance transitivity. If a particular route R leads from affordance A to

affordance A9, then a creature who has grasped affordance symmetry will
recognize that reversing the route will take him from affordance A9 to
affordance A. By the same token, a creature who has grasped affordance
transitivity will recognize that if route R leads from affordance A to af-
fordance A9 and route R9 leads from affordance A9 to affordance A0, then
R followed by R9 will take him from A to A0, and so on for any combina-
tion of connected routes.
At a more sophisticated level, consider the following. Calibrating af-
fordances into an integrated representation of the world will involve the
capacity to distinguish affordances from the places at which those af-
fordances hold, and vice versa. A grasp of what I would term affordance
identities is one fundamental component of this capacity. Grasping af-
fordance identities is grasping that a given place can have different af-
fordances over time and, correlatively, that a different affordance does
not necessarily mean a different place. Understanding affordance identi-
ties will involve, for example, understanding that the spatial relations
holding between a given affordance A at a particular place P and other
affordances A9, A0, . . . at other places P9, P0, . . . will continue to hold
even if A intermittently ceases to hold at P and is replaced by A*. More
concretely, suppose that an animal lives near a watering hole, from which
it is safe to drink at all times of day except early morning and dusk when
it is visited by predators. Here we have a place P (the watering hole) with
two different affordances: A (water) and A* (serious danger). A creature
who has grasped the identity of A and A* will understand that the spatial
relations that hold between A and other affordances A9, A0, . . . at P9, P0,
. . . hold equally between A* and those other affordances. For example,
suppose the creature is familiar with the route from a particular tree to
the watering hole and usually visits the watering hole after visiting the
tree. If the creature has grasped the affordance identity of A and A*, it
will know not to take that route at dawn or dusk.
Suppose that a creature moves towards an integrated representation of
its environment by applying its understanding of affordance symmetry,
affordance transitivity, and affordance identity to its understanding of
the spatial relations holding between individual affordances. When does
the process end? When would it be right to say that the creature has an
226 Chapter 8
integrated representation of its environment? Let me define the number

of affordances with which a creature is familiar as its affordance space,
and let a section of an affordance space be connected if and only if the
creature in question recognizes that there is a route from any given af-
fordance to any other given affordance. It seems clear that no creature
can possess an integrated representation of its environment unless its af-
fordance space is connected, or at least close to being connected. More-
over, repeated application of affordance symmetry, affordance transitivity,
and affordance identity could yield just such a connected affordance
space. This is not to say, of course, that a connected affordance space
is sufficient for an integrated representation of the environment. As dis-
cussed in the previous section, there are conditions on navigational be-
havior that any creature possessing an integrated representation of its
environment must satisfy, and it is possible that a creature might fail to
satisfy those conditions despite possessing a connected affordance space.
After all, a connected affordance space does not automatically generate
the capacity to keep track of those changes in spatial relations between
objects caused by one’s own movements relative to those objects, for ex-
ample. Nor will every connected affordance space be able to serve as a
basis for further types of spatial reasoning. The London Underground
map is an example of a connected affordance space, but not one that
would enable a navigator to devise shortcuts not marked on the map. At
the very least, a connected affordance space must be sensitive to the dis-
tance and direction between affordances.
Nonetheless, once a creature has a connected affordance space, or
something close to it, it is relatively straightforward to see how it might
progress toward an integrated representation of its environment by ac-
quiring the further capacities discussed in the previous section. It is plau-
sible that the capacity for thinking about different routes to the same
place could emerge naturally out of a fully connected affordance space,
particularly one in which there is a range of affordance identities. Like-
wise for the capacity to think about places independently of the objects
or features located at those places.
The concept of an affordance, therefore, is the key to seeing how my
discussion of spatial reasoning and of integrated representations of the
environment meets the Acquisition Constraint in the manner outlined in
chapter 1. The constraint demands, in effect, that any learned cognitive

ability be constructible out of more primitive abilities already in existence.
As emerged in chapter 5, there are good reasons to think that the percep-
tion of affordances is innate. And so if, as I have suggested, the perception
of affordances is the key to the acquisition of an integrated spatial repre-
sentation of the environment via the recognition of affordance symmet-
ries, affordance transitivities, and affordance identities, then it is perfectly
conceivable that the capacities implicated in an integrated representation
of the world could emerge nonmysteriously from innate abilities.
9
Psychological Self-Awareness: Self and Others
Let me take stock. Chapter 5 showed how the very structure of perceptual
experience can be a source of nonconceptual first-person contents
through the pick-up of self-specifying information. In chapter 6, I argued
that somatic proprioception can also be a source of first-person contents,
and that those first-person contents incorporate an awareness of the body
as a spatially extended and bounded object that is distinctive in virtue of
its responsiveness to the will. The next two chapters explored how these
basic building blocks could be developed into a more sophisticated form
of self-awareness that I termed a nonconceptual point of view. The key
elements of a nonconceptual point of view are, first, a more sophisticated
registering of the distinction between self and environment than is avail-
able in either perceptual experience or somatic proprioception and, sec-
ond, a capacity for spatial reasoning that brings with it an awareness of
oneself as moving within, acting upon, and being acted upon by, the spa-
tial environment.
These forms of self-awareness so far discussed are predominantly phy-
sical and bodily. What we have been dealing with is largely awareness
of the embodied self as a bearer of physical properties. There has so far
been little discussion of the explicitly psychological dimension of self-
awareness. But awareness of the embodied self as a bearer of psychologi-
cal properties is an equally significant strand in self-awareness and, more-
over, of central importance in solving the paradox of self-consciousness.
It will be the subject of this chapter.
230 Chapter 9
9.1 The Symmetry Thesis: An Unsuccessful Defence
An initial question that arises in thinking about psychological self-

awareness is the relation between psychological self-awareness and aware-
ness of other minds. As a way of broaching the issue, let me start with
the relation between psychological self-awareness and awareness of other
minds in subjects who are capable of fully conceptual thought and lin-
guistic mastery. As I have already mentioned, psychological predicates
have a constant sense, whether they are applied to oneself or to others. It
is not the case that psychological predicates have a first-person sense, on
the one hand, and a second- and third-person sense, on the other. This
is prima facie evidence for a commonality between psychological self-
awareness and awareness of other minds, at least at the conceptual level.
Let me offer the following as a way of capturing this prima facie common-
ality more generally:
The Symmetry Thesis A subject’s psychological self-awareness is consti-
tutively linked to his awareness of other minds.
The Symmetry Thesis will be the subject of this section and the next. I
will consider two arguments in support of the Symmetry Thesis, only one
of which will turn out ultimately to be sound. First, though, let me discuss
the Symmetry Thesis a little further.
The Symmetry Thesis has both a weak reading and a strong reading,
which it is useful to distinguish. The weak reading is that there can be no
psychological self-awareness without awareness of other minds. I am tak-
ing awareness of other minds to involve the capacity to discriminate at
least some psychological-state types in other subjects of experience. This
means that the Symmetry Thesis is directly antithetical to any view on
which psychological self-awareness is conceptually distinct from aware-
ness of other minds. Let me put the point another way. On the weak
reading, the Symmetry Thesis implies that to the extent that psychological
awareness in general is a matter of applying psychological predicates, it
is impossible for a subject’s repertoire of psychological predicates to be
composed solely of predicates that have only a first-person application.1
The same point can be put in a way that does not make it a point about
language mastery. If we define a psychological representation as a mental
representation that enables a subject to discriminate psychological-state
Psychological Self-Awareness: Self and Others 231
types, the weak reading of the Symmetry Thesis maintains that not all
of a subject’s psychological representations can be applicable only to the
first person.
Thus, the Symmetry Thesis, as weakly construed, is incompatible with
the notion that psychological self-awareness is independent of awareness
of other minds, in, for example, the way that seems to be implied by those
versions of the argument from analogy that hold that our capacity to
ascribe psychological properties to others is derived by analogy and ex-
tension from our logically independent capacity to ascribe psychological
properties to ourselves. On most versions of the argument from anal-
ogy, there is no constitutive connection between first-person and third-
person ascriptions of psychological properties. The analogy approach to
awareness of other minds generally appears within the context of a debate
predicated on the assumption that it makes perfectly good sense for a
psychologically self-aware subject not to be aware that there are any other
minds at all. Consider the following well-known passage from John Stu-
art Mill:
By what evidence do I know, or by what considerations am I led to believe, that
there exist other sentient creatures; that the walking and speaking figures which
I see and hear, have sensations and thoughts, or in other words, possess Minds?
. . . I conclude it from certain things, which my experience of my own states of
feeling proves to me to be marks of it. . . . I am conscious in myself of a series of
facts connected by a uniform sequence, of which the beginning is modifications
of my body, the middle is feelings, and the end is outward demeanour. In the case
of other human beings I have the evidence of my senses for the first and last links
of the series, but not for the intermediate link. . . . Experience, therefore, obliges
me to conclude that there must be an intermediate link. (1889, 243)
The implication here is clearly that apprehending the connections be-
tween the three links in the sequence is independent of the identifying the
presence of two of those links in other subjects and could take place in
its absence. Even on its weak construal, the Symmetry Thesis rules this
possibility out.
A second and stronger reading of the Symmetry Thesis would be that
no subject can discriminate in himself psychological-state types that he
cannot discriminate in other subjects. This is stronger than the first impli-
cation, which leaves open the possibility that, despite the impossibility
of a psychologically self-aware subject being completely unaware of the
existence of other minds, that subject’s understanding of other minds
232 Chapter 9
might be impoverished relative to his psychological self-understanding.

This point can be put in a way that matches the previous point: to the
extent that psychological awareness is a matter of ascribing psychological
predicates, whether to oneself or to others, there can be no psychologi-
cal predicates which have only a first-person application. Alternatively,
and more broadly, whatever psychological representations a subject has,
none of those psychological representations can have only a first-person
application.
It is important to note, however, that the Symmetry Thesis does not
have any implications for the obvious fact that our ways of finding out
about our own psychological properties can differ fundamentally from
our ways of finding out about the psychological properties of other
people. Typically, we find out about many of our psychological properties
without any need to observe our behavior, whereas observation of be-
havior is often required to find out about other people’s psychological
properties. No theory that denied this obvious fact could be true. The
Symmetry Thesis, though, says no more than that the set of capacities
that we put to work in finding out whether or not we instantiate any of
a given set of psychological properties cannot exist in the absence of a
comparable set of capacities to determine whether other people instanti-
ate any of that same set of psychological properties. It does not require
that it be the same capacities performing both tasks, or that those capaci-
ties be structurally similar in any way.
Let me turn now to the question of why one might think that the Sym-
metry Thesis is true. I have already suggested one prima facie piece of
evidence, which is the constancy of sense of psychological predicates
across first-person, second-person and third-person applications. But this
is no more than prima facie evidence. The fact that where a psychological
predicate can be applied in first-, second-, and third-person ways, it means
the same in each application does not entail either that if a subject has
some psychological predicates, he must have at least some with a third-
person application (the weak reading) or that none of those psychological
predicates can have only a first-person application (the strong reading).
To move beyond prima facie evidence, the first argument in support of
the Symmetry Thesis that I want to consider draws on a widely accepted
claim about the nature of concept mastery. Although the argument’s most
articulate development has come from Gareth Evans, its origins seem to
lie in some terse remarks of Strawson’s. Here is how Strawson puts the
conclusion of the argument: “It is a necessary condition of one’s ascribing
states of consciousness, experiences, to oneself in the way one does, that
one should also ascribe them, or be prepared to ascribe them, to others
who are not oneself” (Strawson 1959, 99). This seems close to what I
have termed the strong reading of the Symmetry Thesis, namely that no
subject can discriminate in himself psychological-state types that he can-
not discriminate in others. Note, however, that this argument, if sound
(which I shall argue that it is not), would prove the Symmetry Thesis in a
manner that might well turn out to be too restricted. Because it rests cru-
cially upon a particular view of the requirements for concept mastery,
it is restricted to the discrimination of psychological-state types at the
conceptual level, unless it can be shown that the requirements in question
are not exclusive to conceptual representation.
This following sentence from Strawson conveys the essentials of Straw-
son’s and Evans’s position: “The main point here is a purely logical one:
the idea of a predicate is correlative with that of a range of distinguishable
individuals of which the predicate can be significantly, though not neces-
sarily truly affirmed” (Strawson 1959, 99, n. 1). From this requirement
on predicates in general it follows that psychological predicates must have
a range of distinguishable individuals to which they can be applied. And
this seems to lead to the conclusion that no psychological predicate can
have only a first-person application. Evans has developed the same point
into what he terms the Generality Constraint:
What we have from Strawson’s observation is that any thought which we can
interpret as having the content that a is F involves the exercise of an ability—
knowledge of what it is for something to be F—which can be exercised in indefi-
nitely many distinct thoughts, and would be exercised in, for instance, the thought
that b is F. And this of course implies the existence of a corresponding kind of
ability, the ability to think of a particular object. (Evans 1982, 103)
We thus see the thought that a is F lying at the intersection of two series of
thoughts: on the one hand, the series of thoughts that a is F, that b is F, that c is
F, . . . and, on the other hand, the series of thoughts that a is F, that a is G, that a
is H. (Evans 1982, 104, n. 1)
Evans’s Generality Constraint imposes a stronger requirement than the
comments from Strawson that gave rise to it. Evans imposes a require-
234 Chapter 9
ment of general recombinability, suggesting that mastery of the concept F

requires being able to think that x is F for arbitrary x, whereas Strawson
is requiring simply the ability to think that x is F for a range of objects.
This difference is not relevant to the following, and for the sake of sim-
plicity, I shall refer to both Evans and Strawson as defending forms of the
Generality Constraint.
Evans’s Generality Constraint has the advantage of precision, and it is
clear how it might be put to work in support of the Symmetry Thesis. Let
me offer a schematic argument:
1. Self-ascription of psychological states in subject-predicate form is the
paradigm form of psychological self-awareness.
2. Self-ascription of psychological states requires conceptual mastery of
both subject and predicate.
3. The Generality Constraint applied to (2) requires the ability to general-
ize both the subject and the predicate.
4. Generalizing the predicate involves the ability to apply that predicate
to arbitrary distinguishable individuals.
5. Since the predicate is a psychological predicate, these arbitrary distin-
guishable individuals will include other psychological subjects.
From (5), the Symmetry Thesis follows straightforwardly. Note, more-
over, that it is the strong reading of the Symmetry Thesis that is sup-
ported. The conclusion (5) implies that no psychological predicate can
have only a first-person application.
In evaluating this argument, we need to distinguish clearly between the
following two questions:
(A) If an individual is aware of the existence of other psychological
subjects, does the Generality Constraint require that any individual prop-
erly credited with mastery of a given psychological predicate be able to
apply that psychological predicate to those other subjects?
(B) If an individual is not aware of the existence of other psychological
subjects, does this mean that he cannot satisfy the Generality
Constraint?
These two questions have very different implications for the Symmetry
Thesis. An affirmative answer to the first question leaves open the possi-
bility that would be denied by an affirmative answer to the second ques-
tion, namely that an individual unaware of the existence of other

psychological subjects might nonetheless have genuine mastery of a psy-
chological predicate. No defence of the Symmetry Thesis can allow this
to be a genuine possibility. Since the weight of the argument falls on the
Generality Constraint, the Generality Constraint must yield an affirma-
tive answer not just to (A) but also to (B). Unless this is so, stage (5) in
the Strawson and Evans argument will not go through.
But it is not clear that the Generality Constraint does yield an affirma-
tive answer to (B). There are ways in which the Generality Constraint
could be satisfied by an individual unaware of the existence of other psy-
chological subjects. Everything here hinges on how we understand the
concept of an arbitrary distinguishable individual. Recall that we are
dealing here with truth-evaluable subject-predicate sentences. The Gener-
ality Constraint does not prescribe (nor is there any reason why it should)
that thoughts that would show that the constraint was satisfied by a given
subject-predicate thought must have the same truth value as the original
thought. From this one might conclude that there is no need for the range
of distinguishable individuals to include other psychological subjects.
Why is the Generality Constraint not satisfied by a subject’s capacity to
entertain the thought that a given psychological predicate does not apply
to any one of a range of inanimate objects?
It might be objected that this is self-defeating because, simply in virtue
of entertaining the thought that an inanimate object does not have a given
psychological property, one is (mistakenly) treating it as a psychological
subject. This cannot be right, however. A minimal requirement for some-
thing to be a psychological subject is surely that at least some psychologi-
cal predicates must be true of it. It is not clear to me that to entertain the
thought that a given psychological predicate does not apply to a given
inanimate object is somehow to commit a category mistake. How, after
all, can it be a category mistake if it is a true thought?
In any case, however, there are other grounds on which one can object
to the Strawson and Evans argument. The basic thought motivating ac-
ceptance of the Generality Constraint is that there is a deep incoherence
in the idea of a predicate that is applicable only on a single occasion: to
a given individual at a given time. Predicates must be multiply instanti-
able. That, it seems to me, is the force of Strawson’s claim that “the idea
236 Chapter 9
of a predicate is correlative with that of a range of distinguishable individ-

uals of which the predicate can be significantly, though not necessarily
truly affirmed.” 2 It is unfortunate, however, that both Evans and Strawson
take a very synchronic view of the satisfaction conditions for their respec-
tive constraints. Had they taken a more diachronic perspective, they
would have recognized that a subject himself can, over time, provide a
range of different occasions on which a predicate may be significantly
affirmed. To appreciate the significance of this, recall some comments that
Strawson makes in the footnote from which I have already quoted: “A
necessary condition of one’s ascribing predicates of a certain class to one
individual, i.e. oneself, is that one should be prepared, or ready, on appro-
priate occasions, to ascribe them to other individuals, and hence that one
should have a conception of what those appropriate occasions for ascrib-
ing them would be” (Strawson 1959, 99, n. 1).3 Strawson’s view is that
ascribing predicates to other individuals, or being prepared so to ascribe
them, implicates a conception of the appropriate ascription conditions for
those predicates. Presumably, the argument is completed by noting that
possessing such a conception is a necessary condition of being able to
apply the predicate at all. I can see no reason, however, why a conception of
the appropriate ascription conditions is not equally implicated by ascrib-
ing, or being prepared to ascribe, predicates to oneself at different times.
Combining this with the earlier point provides a compelling counter-
weight to the move from (4) to (5) in the Strawson and Evans argument.
Although it is (arguably) the case that mastery of a given psychological
predicate requires the capacity to apply it to arbitrary distinguishable in-
dividuals, it does not follow that these arbitrary distinguishable individu-
als must include other psychological subjects. In such a case, therefore,
the Generality Constraint cannot be used to argue that psychological
predicates must have second- and third-person uses, as well as first-person
uses. Without this, though, the Strawson and Evans argument cannot pro-
vide a satisfactory defence of the Symmetry Thesis. Evans writes, “No
judgment will have the content of a psychological self-ascription, unless
the judger can be regarded as ascribing to himself a property which he
can conceive as being satisfied by a being not necessarily himself—a state
of affairs which he will have to conceive as involving a persisting subject
of experience” (Evans 1982, 232). To maintain this, however, we need
more than the Generality Constraint. In the next section I offer what I
take to be a more powerful defence of the Symmetry Thesis. In one respect
this defence will be wider in scope than the Strawson and Evans argu-
ment, because it will not be restricted to conceptual thought. In another
respect, though, it will be narrower in scope, because it will not support
a strong reading of the Symmetry Thesis.
9.2 The Symmetry Thesis: A Neo-Lockean Defence
The argument in support of the Symmetry Thesis that I will offer in this
section is much briefer than the unsuccessful argument from the General-
ity Constraint just examined. It starts not from putative requirements on
concept mastery and psychological self-ascription but from a point about
self-awareness that has already been discussed. This is the point that self-
awareness has a fundamentally contrastive dimension. The contrastive di-
mension of self-awareness had an important part to play in developing
the notion of a nonconceptual point of view. At the core of the notion of
a nonconceptual point of view is the capacity to distinguish self from the
environment, and the richness of the self-awareness that accompanies this
capacity is directly proportionate to the richness of the awareness of the
environment from which the self is being distinguished. A similar thought
is the key to seeing why the Symmetry Thesis is true.
Let me begin, though, by bringing out in a little more detail what I
mean by the contrastive dimension of self-awareness. As I have empha-
sized, an important element in self-awareness is a subject’s capacity to
distinguish himself from the environment and its contents. I will use the
phrase ‘distinguishing self-awareness’ as a shorthand for this. The discus-
sion of somatic proprioception in chapter 6 provided a clear example of
distinguishing self-awareness. I showed how an awareness of the body as
a spatially extended object distinctive in virtue of its responsiveness to
the will was part of the content of somatic proprioception. As I noted,
however, this awareness is only a limited form of distinguishing self-
awareness, yielding little more than a basic distinction between self and
nonself. The reason for this awareness being so restricted is simply
that somatic proprioception does not implicate (though, of course, it can
be accompanied by) a particularly rich conception of what the self is
238 Chapter 9
opposed to. The conception of the environment emerging from somatic

proprioception need not be anything richer than what is not responsive
to the will, and this yields little more than a conception of basic numerical
distinctness. Clearly, if a richer conception of the environment were
in play, then the distinguishing self-awareness in somatic proprioception
would be correspondingly richer. For example, if a subject were able to
conceive of his environment as containing spatially extended objects, then
he would be capable of distinguishing self-awareness as a spatially ex-
tended object that is distinctive within the class of spatially extended ob-
jects. The significance of possession of a nonconceptual point of view is
precisely that it provides a richer awareness of the environment, thus mak-
ing for a correspondingly richer degree of distinguishing self-awareness.
There is, I conjecture, a general principle to be extracted here. Distin-
guishing self-awareness implies tacit reference to what might be termed a
contrast space. I have distinguishing self-awareness of myself as f only to
the extent that I can distinguish myself from other things that are f. These
other things that are f form the contrast space. This principle about
contrast spaces can be seen as an extension of a familiar neo-Lockean
point about identity (Mackie 1976, chap. 5; Wiggins 1980; Ayers 1991,
vol. 2).
The neo-Lockean point about identity is that questions about identity
over time can only be posed and answered relative to a given categoriza-
tion. It does not make sense to ask whether something that exists at a
particular time is or is not identical with something that exists at a later
time. What we have to ask is whether this thing is the same x as that
thing, where ‘x’ is a sortal representation that picks out a category or
kind. A sortal representation, as I propose to use the term, is a mental
representation, not necessarily linguistic, that classifies things as members
of a given kind. The kind in question may, but need not, be a natural kind.
What distinguishes a sortal representation from an ordinary descriptive
or functional representation is that it is associated with more or less deter-
minate criteria of application and identity, from which it follows that
once we have a given sortal representation in play, we will know what
criteria of identity to look for. Of course, distinctness being the opposite
of identity, it follows that we cannot ask simply whether this is or is not
distinct from that. What we must ask is whether this is a distinct x from
that, where ‘x’ picks out a category or type of object.
Let me term the Lockean thesis the restricted thesis of relative identity
over time. To hold the restricted thesis of relative identity is to hold that
questions of identity over time are always relative to a given sortal repre-
sentation. This needs to be kept separate from the unrestricted thesis of
relative identity over time, according to which two individuals x and y
may be identical with respect to one sortal representation but distinct
with respect to another. Let me illustrate this with the very relevant ex-
ample of personal identity. Locke’s view (which I am not endorsing) is
that a single block of flesh and bone can embody two different countable
things: a living human being (relative to the sortal representation man)
and a conscious rational subject capable of memory (relative to the sortal
representation person). The man and the person are different things that
happen most of the time to be physically coextensive. This is a form of
restricted relative identity. According to the unrestricted theory of rela-
tive identity, on the other hand, the block of flesh and bone is identical
with the man and also identical with the person, although the man is
not identical with the person. Two good reasons for not endorsing the
unrestricted thesis are, first, that it comes into conflict with the principle
(Leibniz’s Law) that if x and y are identical, then they share all their prop-
erties and, second, that it means abandoning the transitivity of identity.
Fortunately, the unrestricted thesis is not entailed by the restricted thesis.4
To reach the unrestricted thesis, one needs to add to the restricted thesis
the premise that an individual can be a member of more than one kind,
where the kinds in question are not subordinate to one another.
Locke himself believed that questions about identity and distinctness
(or diversity, as he put it) were problematic only when considered over
time. Recent work on personal identity has not followed him in this. Con-
sideration of split-brain patients in particular has suggested that identity
and distinctness at a time can also be very problematic.5 For example, one
(admittedly rather controversial) way of considering split-brain patients
would be to hold that at any given time after the severing of the cor-
pus callosum, a block of flesh and bone can embody more than one con-
scious rational subject, despite embodying only one living human being.
Whether or not this is a correct description of split-brain patients, it is
a conceivable possibility that theories of personal identity need to take
account of by incorporating some of sort of means to answer questions
of identity at a time. Presumably, this will require applying criteria of
240 Chapter 9
identity in accordance with a version of the restricted thesis of relative

identity holding at a time as well as over time. Corresponding to the re-
stricted thesis of relative identity at a time will be a thesis of relative dis-
tinctness at a time, according to which questions of distinctness at a time
are always relative to a given sortal representation. If this were not so,
then because x and y are identical if and only if they are not distinct, it
would be possible to answer questions about identity without reference
to a given sortal representation, contrary to the thesis of relative identity.
The relevance of this to distinguishing self-awareness is as follows. Dis-
tinguishing self-awareness is a general label for the various ways in which
a subject can recognize his distinctness from the environment and its con-
tents. According to the restricted thesis of relative distinctness, questions
about distinctness are relative to a given sortal representation. But what
is involved in answering a question about distinctness relative to a given
sortal? The key here, it seems to me, is recognizing that questions of the
form ‘Are x and y distinct with respect to sortal z?’ should really be re-
phrased in the form ‘Can we count more than one instance of sortal z
here?’ 6 Once this is recognized, the Symmetry Thesis falls out soon
enough. Assume that sortal z is a psychological sortal. If a subject is to
be capable of recognizing that he is distinct from his environment with
respect to sortal z, he must be capable of posing and answering questions
of the form ‘Can more than one instance of sortal z be counted here?’
The capacity to pose such questions presupposes an understanding that
sortal z is at least potentially multiply instantiated (that there are, or at
least could be, other instances of sortal z), while the capacity to answer
them presupposes an ability to distinguish and count instances of sortal
z. Putting these together yields the conclusion that, insofar as sortal z is
to provide a form of distinguishing self-awareness, it cannot have solely a
first-person application. Distinguishing self-awareness requires a contrast
space, as a function of what it is to recognize distinctness.
Although the neo-Lockean argument does support the Symmetry The-
sis, it supports only the weak reading. The choice, it will be remembered,
was between construing the Symmetry Thesis as holding either that if a
subject has a range of psychological representations, he must have at least
some with a third-person application (the weak reading) or that none of
those psychological representations can have only a first-person applica-
tion (the strong reading). The defence I have offered of the Symmetry
Thesis is based on the requirements of distinguishing self-awareness. It

follows, therefore, that a constitutive link between first- and third-person
application has been shown to hold only for those psychological represen-
tations implicated in distinguishing self-awareness. I have not argued (and
do not believe) that the Symmetry Thesis holds in its strong form.
The psychological representations implicated in distinguishing self-
awareness are, of course, those psychological representations that define
the category (sort or kind) of psychological subjects. So the upshot of the
neo-Lockean defence of the Symmetry Thesis is that the psychological
representations that define the category of psychological subjects nec-
essarily have first-, second-, and third-person applications. In the next
section I will have more to say about what those psychological representa-
tions actually are.
Let me turn now to an issue that has been hovering in the background
throughout this and the previous section. To what extent are the argu-
ments I have discussed in support of the Symmetry Thesis applicable at
the nonconceptual level to non-language-using creatures? The (unsuccess-
ful) argument from the Generality Constraint is closely tied to the putative
requirements of concept mastery, and hence is not straightforwardly ap-
plicable at nonconceptual levels of representation. But the same is not
true of the just-offered neo-Lockean argument from distinguishing self-
awareness. The point of the argument is that a subject’s distinguishing
self-awareness depends on his capacity to discriminate and count individ-
ual instances of a given category. I have put this point in terms of a sub-
ject’s mastery of sortal representations, deliberately leaving open the
possibility that the categorization in question might be independent of
language mastery.7 Of course, as in earlier chapters, work is needed to
show that the categories in question can be understood and applied at
the nonconceptual level, and this task will occupy the final sections of
this chapter.
9.3 The Core Notion of a Psychological Subject
The existence of a constitutive link between psychological self-awareness

and awareness of other minds implies that the best place to look for prim-
itive forms of psychological self-awareness is in social interactions. A sub-
ject’s recognition that he is distinct from the environment in virtue of
242 Chapter 9
being a psychological subject must take place against the background of

a contrast space that includes other psychological subjects. This has a
clear implication for the sort of evidence that will settle the question of
whether psychological self-awareness can exist in a form that is noncon-
ceptual and independent of language mastery. What we are looking for
are social interactions involving prelinguistic or nonlinguistic subjects for
which the best explanation involves ascribing the appropriate form of
distinguishing self-awareness to a nonlinguistic or prelinguistic subject.
Clearly, the first step must be to clarify what exactly the appropriate form
of distinguishing self-awareness involves. That will be the subject of this
section.
A subject has distinguishing self-awareness to the extent that he is able
to distinguish himself from the environment and its contents. He has dis-
tinguishing psychological self-awareness to the extent that he is able to
distinguish himself as a psychological subject within a contrast space of
other psychological subjects. What does this require? The discussion in
the previous section has shown how this question is to be answered. Dis-
tinguishing self-awareness is relative to a given sortal categorization, and
psychological self-awareness is relative to the sortal category of psycho-
logical subjects. We need to turn our attention to the sortal category of
psychological subjects. What are the criteria of identity and application
associated with the sortal category of psychological subjects?
The first point to make is that the category of a psychological subject
is what one might term a complex sortal category, analyzable in terms of
more basic categories. There are several relevant criteria of identity and
application for psychological subjects, and each of these criteria of iden-
tity pick out a further psychological category. Each of these psychological
categories is itself independently analyzable. This is one of the reasons
why distinguishing psychological self-awareness is a matter of degree. A
subject can master some of the relevant criteria of identity and applica-
tion without mastering others. A subject in such a position will have a
restricted range of psychological categories in terms of which he can dis-
tinguish himself from the social environment. The more of these psycho-
logical categories a subject acquires, the closer he will move toward
distinguishing awareness of himself as a psychological subject.
The category of a psychological subject is vague, in the following stan-
dard sense. There are borderline cases of individuals for whom (or which)
it is not possible to give a determinate answer to the question of whether

or not they qualify as psychological subjects. This is an unsurprising con-
sequence of the point made in the previous paragraph about the category
of psychological subjects being a complex sortal category. Given the com-
plexity and diversity of the criteria of identity and application, the differ-
ent combinations in which they may appear, and the fact that it will not
always be straightforward to determine whether a given criterion actually
applies, it is unsurprising that there does not appear to be a set of neces-
sary and sufficient conditions that will settle in any given case whether
one is dealing with a psychological subject or not. Nor should this be a
matter of great concern, since it is not clear that there are any philosophi-
cally interesting kinds or sorts that are completely determinate. As far
as analyzing distinguishing psychological self-awareness is concerned, the
vagueness of the category of a psychological subject means simply that
we should concern ourselves with trying to identify the core of the notion
of a psychological subject. By the core of the notion of a psychological
subject, I mean a set of basic psychological categories whose instantiation
by a given individual collectively provides strong prima facie evidence that
the individual in question is a psychological subject, where the instantia-
tion of those categories counts as strong prima facie evidence for the pres-
ence of f if and only if a judgement based on such evidence would count
as warranted. One might define a warranted judgement in this context as
a judgement that would often settle a debate as to the presence of f and
would at the minimum be taken very seriously in such a debate.8
The account of the notion of a self-aware psychological subject I am
looking for will take the following form:
(PS1) Self-aware psychological subjects are aware of themselves as x, y,
z, etc.
The schematic letters correspond to the basic psychological categories
that collectively count as prima facie evidence for the presence of a psy-
chological subject. Of course, there are indefinitely many psychological
categories whose instantiation provides strong prima facie evidence for
the presence of a psychological subject. The category of individuals ca-
pable of autobiographical memory seems to qualify, for example. This
poses an obvious problem. How is one to identify what I am terming the
core of the notion of a psychological subject? How is one to distinguish
244 Chapter 9
the basic from the nonbasic among all the psychological categories whose
instantiation provides strong prima facie evidence for the presence of a
psychological subject?
The solution to this puzzle is to distinguish between those psychologi-
cal categories that one cannot hold to be instantiated without assuming
that they are instantiated in a psychological subject and those psycho-
logical categories that can be ascribed to an individual without thereby
identifying that individual as a psychological subject. To ascribe auto-
biographical memories to an individual is ipso facto to identify that indi-
vidual as a psychological subject. So the inference from an individual’s
possession of autobiographical memories to that individual’s being a psy-
chological subject seems somewhat analytic (as would be the inference
from an individual’s being a milkman to his being a man). On the other
hand, one can decide that a creature is sentient (that is to say, capable of
feeling pleasure and pain) without ipso facto placing it in the category of
psychological subjects. Sentience is not a sufficient condition for psycho-
logical-subjecthood. So a creature’s sentience can count as one of a range
of facts that collectively count as strong prima facie evidence for psycho-
logical subjecthood. I thus assume that the psychological categories that
feature in a completed version of (PS1) will not individually count as suf-
ficient conditions for psychological subjecthood.
Perhaps the most obvious candidate for inclusion in a completed ver-
sion of (PS1) that satisfies the requirement just noted will be the psycho-
logical category of perceivers of the world. This follows straightforwardly
from the fact that perceiving the world in at least one modality is a ne-
cessary condition for being ascribed any psychological properties at all,
and only creatures to which psychological properties can be ascribed can
count as psychological subjects. So, to be aware of oneself as a psycholog-
ical subject must involve being aware of oneself as a perceiver. It is helpful
at this point, I think, to advert back to Gareth Evans’ idea of a simple
theory of perception, briefly discussed toward the end of the previous
chapter. Here is what he says:
Any thinker who has an idea of an objective spatial world—an idea of a world
of objects and phenomena which can be perceived but which are not dependent
upon being perceived for their existence—must be able to think of his perception
of the world as being simultaneously due to his position in the world, and to the
condition of the world at that position. The 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 to the
more or less stable way the world is. (Evans 1982, 222)
The significance of a subject’s mastery of the sort of simple theory of
perception that Evans outlines does not lie solely in what it betokens for
his command of the objectivity of the spatial environment.9 It is also
highly relevant to his psychological self-awareness. To have mastered a
simple theory of perception is quite simply to be aware of oneself as a
perceiver of the environment.
Perception, obviously, is necessary but not sufficient for being a psy-
chological subject. The category of psychological subjects is much nar-
rower than the category of perceivers, and the category of psychologically
self-aware subjects is correspondingly much narrower than the category
of self-aware perceivers. It is a shortcoming in Evans’s discussion of simple
theories of perception that he gives the impression that mastery of such a
theory is somehow the key to self-awareness. The truth of the matter is
that a subject’s awareness of himself as a perceiver of the environment
is just one of the several strands in psychological self-awareness. John
Campbell’s recent book (1994), which follows Evans in many respects,
develops Evans’s position at this crucial point. Campbell stresses that we
need to think in terms, not just of a theory of perception, but of a joint
theory of perception and action. This broaches what I take to be the sec-
ond strand in the notion of a psychological subject. Psychological subjects
are agents who intentionally act upon the world because of their percep-
tions and desires. Correspondingly, to be aware of oneself as a psychologi-
cal subject is to be aware of oneself as an agent.10 The psychological
category of agents itself has several strands, one of which emerges in this
passage from John Campbell: “This theory explains our perceptions as
the joint upshot of the way things are in the world and the way things are
with us, and it explains the effects of our actions as the joint consequences
of our bodily movements and the way things were around us to begin
with” (Campbell 1994, 217). As it stands, however, this is rather incom-
plete. A subject’s actions are not just the joint consequences of bodily
movements and the layout of the environment. A subject’s bodily move-
ments are what they are because of his intentions, and an action is suc-
cessful or unsuccessful to the extent that those intentions are satisfied.
Lack of success can be due to the particular body movements made not
246 Chapter 9
being sufficiently skillful or to their being blocked in some way or to the

environment not actually being the way it was perceived and expected to
be. A subject will be aware of himself as an agent to the extent that he is
capable of recognizing and distinguishing these various factors and possi-
bilities. Again, we are dealing with something that is a matter of degree
and that will have an unavoidably vague dimension.
Let me move on to a third strand in the notion of a psychological sub-
ject. Psychological subjects do not simply perceive the world and act upon
it. There are psychological reactions to the world that are not exhausted
by perception and action (although such reactions do, of course, have
implications for perception and action). I refer here to moods, emotions,
feelings of happiness and unhappiness, which I shall collectively term the
reactive psychological states. It is hard to envisage any recognizable psy-
chological life in the absence of such phenomena. Any sentient creature
is capable of the subjective valences of pleasure and pain that qualify as
the most basic forms of psychological phenomena of this type, and noth-
ing that is not sentient could count as a psychological subject. As an argu-
ment for this, consider the following. A necessary condition of being a
psychological subject is that a creature should be capable of acting inten-
tionally. Intentional action is constitutively motivated by desires and be-
liefs (or protodesires and protobeliefs), and no nonsentient creature can
have a desire. Desires are not themselves reactive states, but it is hard to
see how desires could arise in the absence of reactive psychological states.
It follows from this that any individual who is to count as a self-aware
psychological subject must be aware of himself as having reactive psycho-
logical states, or as bearing reactive psychological attitudes to the world.
In addition to this a priori argument for the thesis that psychological
subjects are essentially capable of reactive psychological states, a further
set of reasons emerges when one reflects on the functions that emotions
and other reactive psychological states serve in maintaining the organism.
Some of these functions are physiological. For example, emotional states
play a vital role in eliciting autonomic and endocrine responses. The emo-
tion of fear generates changes in heart rate and release of adrenalin. Some
byproducts of reactions are psychological, such as the role that emotions
play in the storing and triggering of episodic memories. Others are social,
the role of emotions in promoting social bonding being an obvious ex-
ample. It is certainly arguable that no creature lacking these functions

would have the behavioral flexibility and cognitive sophistication to count
as a psychological subject. As Edmund Rolls (1990, 1995) has stressed,
emotional states provide a computationally simple way of integrating sen-
sory input with motor output. The sensory system need transmit only
the valence to the appropriate action-generating systems, rather than a
full representation of the state of affairs that gave rise to the valence. The
emotion of fear is a basic and clear example. There are obvious adaptive
advantages in avoidance and flight responses being triggered as quickly
as possible. Combining these functional considerations with the a priori
suggestions yields a compelling case for seeing the capacity for reactive
attitudes as essential to psychological subjects.
So the central psychological categories in terms of which the category
of psychological subjects should be understood are the categories of per-
ceivers, agents, and bearers of reactive psychological states. I offer the
following as the core of the notion of a self-aware psychological subject:
(PS2) Psychological subjects with a perspective on the world are aware
of themselves as perceivers, as agents, and as having reactive psycholog-
ical states.
As mentioned earlier it is relatively unimportant whether these turn out
to be jointly sufficient or even severally necessary. All of the concepts in-
volved here are complex and vague in ways that make it extremely un-
likely that they will each always have determinate criteria of identity. My
claim is simply that when we encounter a subject aware of himself as a
perceiver, as an agent, and as having psychological states, we have strong
prima facie evidence that we are dealing with a self-aware psychological
subject.
9.4 The Emergence of Psychological Self-Awareness in Social

Interactions
Let me now bring the various strands of this chapter to bear on the issue
of psychological self-awareness. According to the weak version of the
Symmetry Thesis that has been defended, there is a constitutive link be-
tween psychological self-awareness and awareness of other minds. This
248 Chapter 9
link holds because a subject’s recognition that he is distinct from the envi-
ronment in virtue of being a psychological subject depends on his ability
to identify himself as a psychological subject within a contrast space of
other psychological subjects. This self-identification as a psychological
subject will take place relative to the set of categories that define the core
of the concept of a psychological subject. So a suitably self-aware subject
will be capable of distinguishing himself as a perceiver within a contrast
space of perceivers, as an agent within a contrast space of agents, and as
a bearer of reactive attitudes within a contrast space of other bearers of
reactive attitudes. It follows from this that the best place to look for prim-
itive forms of psychological self-awareness is in social interactions. This
offers a clear way to proceed that will settle the question of whether psy-
chological self-awareness can exist in a nonconceptual form that is inde-
pendent of language mastery. In line with the methodology I have been
following throughout the book, what would settle the matter would be
social interactions involving prelinguistic or nonlinguistic subjects for
which inference to the best explanation requires ascribing the appropriate
form of distinguishing self-awareness to a nonlinguistic or prelinguistic
subject.11
Here, as at various points in earlier chapters, we will need to look at
empirical work if we are properly to identify and to understand the poten-
tial explananda. No doubt there are various areas of ethology and psy-
chology that might be highly relevant here. What I want to concentrate
on, however, is an impressive body of developmental results concerning
the interactions between infants and their parents during the first year of
the infants’ lives. There is a concensus of opinion among researchers from
a range of different traditions that at about the age of 9 months human
infants undergo a social-cognitive revolution. This is a revolution that
takes place in several dimensions. One significant point that has been
stressed, particularly by Piaget and his followers, is that at 9 months in-
fants become capable of new ways of acting upon objects. At about 9
months (the beginning of Piaget’s stage IV) infants start to search for hid-
den objects and are able to solve the problem of reaching an object placed
out of reach on a cloth that is itself within reach: by pulling the cloth
toward them until they can grasp the object (Piaget 1954). This develop-
ment in the perception and representation of objects is not, however,
directly relevant to psychological self-awareness, although it will turn out

to be indirectly relevant. What I want to focus on is the social dimension
of the 9-month revolution. Here is how Colwyn Trevarthen, one of the
leading workers in the area, describes the social transition that takes
place: “The most important feature of the new behaviour at 9 months is
its systematically combining interests of the infant in the physical, pri-
vately known reality near him, and his acts of communication addressed
to persons. A deliberately sought sharing of experiences about events and
things is achieved for the first time” (Trevarthen and Hubley 1978, 184).
What I will be proposing is that this new behavior emerging in 9-month-
old infants manifests distinguishing self-awareness relative to the three
categories at the core of the notion of a psychological subject, in other
words, that what we see at 9 months is the emergence of psychological
self-awareness in infancy. Let me start, though, by sketching the broad
contours of the transition in infant development, and in particular the
background against which it takes place.
Infants are social beings from their very earliest days (Trevarthen
1993). An important illustration of this is provided by the work on neo-
natal imitation cited and briefly discussed in chapter 5. Meltzoff and
Moore (1983) found that infants with a mean age of 32 hours (including
one as young as 42 minutes) were capable of imitating gestures of mouth
opening and tongue protrusion performed by an investigator. As I
´
sketched in chapter 5 (see also Bermudez 1996), neonatal imitation
behavior reveals that infants have a sophisticated social awareness of
physical commonalities between themselves and human adults. Infant im-
itation of facial gestures presupposes a recognition on the part of the in-
fants that the gestures they see in front of them are gestures that they
themselves can make (although they cannot see themselves making them).
This recognition in turn must rest on an awareness of having a common
physical structure with the experimenter, and with other human beings in
general. Given that the capacity for imitation behavior seems to set in
immediately after birth, it is natural to conclude that this offers one re-
spect in which the human infant is born as a social being.
There is, of course, a range of further empirical evidence supporting
the thesis that the infant universe is social from the very start of life. One
body of evidence comes from the primitive discriminatory abilities that
250 Chapter 9
infants display. Discrimination of voices is a good example. Infants as

young as 3-days-old seem to prefer their own mother’s voice over the voice
of another infant’s mother (DeCasper and Fifer 1980). There is also evi-
dence that slightly older infants (2-weeks-old) were more likely to stop
crying on hearing their mother’s voice than when they heard the sound of
a female stranger (Bremner 1988, 157). One possible explanation for this
sensitivity to the maternal voice is that it is acquired while the infant is
still in the womb. No such explanation is available, however, for some
of the striking results that have emerged with respect to young infants’
abilities for discriminating faces. Particularly interesting are the results
obtained by Field, Woodson, Greenburg, and Cohen (1982) showing that
infants with a mean age of 45 hours revealed a clear preference for the
face of their mother rather than the face of a stranger. The same study
also showed that the same infants, after being repeatedly presented with
their mother’s face, became habituated to it and would eventually look
longer at a stranger’s face.
This sensitivity on the part of young infants to perceptually discrimina-
ble features of other individuals is matched by a sensitivity to the rather
more subtle matter of other people’s emotional states. Again, this is some-
thing that starts more or less from birth, as is indicated by the phenome-
non often termed empathic arousal, where infants begin to cry when they
hear another infant cry (Sagi and Hoffman 1976). It is worth adverting
also to a phenomenon (noted in chapter 5) that was discovered while the
experiments revealing empathic arousal were being replicated. Neonates
are capable of discriminating their own crying from the crying of other
infants (as evidenced by the fact that their distress crying significantly
decreases when a recording of themselves crying replaces a recording of
another infant crying). As they grow older, infants’ capacities for discrim-
inating emotional states become more sophisticated. The 3-month-old
infants tested by Kuchuk, Vibbert, and Bornstein (1986) showed a prefer-
ence for a face with a smiling expression over a face with a neutral expres-
sion, and this preference increased with the intensity of the smile.
These different forms of infant social sensitivity manifest themselves
in surprisingly complex forms of interaction between infants and their
caregivers. By the time they are 2-months-old, infants engage in extended
and coordinated protodialogues or protoconversations involving recipro-
cal vocalization, with each partner taking their turn and adapting their
behavior to fit in with the other (Bateson 1975, Trevarthen 1993). An
excellent description of a typical episode of protoconversation, worth
quoting in full, has been provided by Colwyn Trevarthen:
The start of communication is marked by orienting of the baby. Babies 6 weeks
or older focus on the mother’s face and express concentrated interest by stilling of
movement and a momentary pause in breathing. The infant’s interest as a whole
conscious being is indicated by the coordination and directedness of this behav-
iour, which aims all modalities to gain information about the mother’s presence
and expressions. Hands and feet move and clasp the mother’s body or her sup-
porting hand, the head turns to face her, eyes fix on her eyes or mouth, and ears
hold and track her voice.
The next, and crucial, phase is signaled by the infant’s making a “statement of
feeling” in the form of a movement of the body, a change in hand gesture away
from clasping the mother, a smile or a pout, a pleasure sound or a fretful cry. The
mother, if she is alert and attentive, reacts in a complementary way. A positive,
happy expression of smiling and cooing causes her to make a happy imitation,
often complementing or praising the baby in a laughing way, and then the two of
them join in a synchronised display that leads the infant to perform a more serious
utterance that has a remarkably precocious form.
This infant utterance is the behaviour, in that context of interpersonal coordi-
nation and sharing of feelings, which justifies the term protoconversation. It looks
and sounds as though the infant, in replying to the mother, is offering a message
or statement about something it knows and wants to tell. Mothers respond and
speak to these bursts of expression as if the infant were really saying something
intelligible and propositional that merits a spoken acknowledgment. (1993,
130–132)
An experimental paradigm developed by Murray and Trevarthen (1985)
makes a persuasive case for why this sort of behavior is properly described
as a primitive form of conversation behavior, rather than simply as the
mother accommodating herself to the random vocalizations of her child.
Murray and Trevarthen used closed-circuit television to set up remote in-
teractions between 6-to-8-week-old infants and their mothers, with each
partner seeing a life-size full-face image of the other. Even in this artificial
medium there ensued a fairly normal protoconversation of the sort de-
scribed above (barring the bodily contact, of course). The experimenters
then broke the closed-circuit link and, after a short while, began replaying
to the infants videotapes of their mother filmed a few minutes earlier
during the real-time interaction. The infants showed real distress at this,
despite the fact that the videotapes showed precisely the same images of
252 Chapter 9
the mother that had recently given them so much pleasure. The explana-
tion of the infants’ distress is, of course, that they miss the contingency
of their mothers’ utterances and expressions on their own utterances and
expressions. They see what is clearly half a dialogue that appears to be
directed at them without any sense of participation in that dialogue—
without the mutual accommodation, adjustment, and responses that de-
fine normal protoconversations.
There is a distinction to be made between thin and thick interpretations
of these mother-infant protoconversations and the discriminative abilities
that they presuppose and implicate. On the thick reading (which is
strongly suggested by Trevarthen’s own interpretation of his results), these
mother-infant interactions reveal genuine intentional communication
involving attempts on the part of both partners to translate their own
feelings to the other participant. On this interpretation, the infants’ be-
havior is not only purposeful but also displays recognition that the moth-
ers’ responses are equally purposeful. The infants are aware of what they
are trying to do and aware of what their mothers are trying to do. Their
regulation of the protoconversation is directed toward bringing what they
are trying to do in accord with what their mothers are trying to do. Con-
sider Trevarthen’s summary of infant development during the first year
of life:
We have shown that from birth infants have no trouble in detecting and inter-
acting discriminately and optionally with the mental states of other persons. Very
soon after birth, they can enter into a dynamic exchange of mental states that has
a conversational potentially intention-and-knowledge-sharing organisation and
motivation. The emotional and purposeful quality of these interchanges of mo-
tives undergoes rapid differentiation in the games of ensuing months. It becomes
more elaborate, more quickly reactive, and more directive in relation to the re-
sponses of the other and is protracted into longer narratives of feeling. We say the
infant is developing a more assertive, more conscious self, but we mean that the
infant’s experience of being a performer in the eyes of the other is gaining in
power, presence and pleasure. (1993, 161)
This passage makes very clear that if the thick reading is right, explaining
what is going on in mother-infant interactions will require ascribing to
the young infant a form of distinguishing self-awareness. As defined in
the previous section, distinguishing self-awareness involves a recognition
of oneself as a perceiver, an agent, and a bearer of reactive attitudes
against a contrast space of other perceivers, agents, and bearers of reactive

attitudes. It can only make sense to speak of the infant’s experience of
being a performer in the eyes of the other if the infant is aware of himself
as an agent and of his mother as a perceiver. Similarly, talk of intentional
interchanges and narratives of feeling presuppose that the infant is aware
of himself and his mother as bearers of reactive attitudes.
It seems to me, however, that the view that the components of dis-
tinguishing psychological self-awareness are all present in the protocon-
versations that start in the second month of life glosses over the very
important distinction between thinking about something in a particular
way and responding differentially to something that has a particular fea-
ture. The fact that the infant regulates his behavior in conformity with
his mother’s responses does not license the conclusion that he is thinking
about either himself or his mother in any particular way. It does not li-
cense ascribing such thoughts to him because those thoughts are not nec-
essary to explain how he behaves. As has been emphasised in earlier
chapters, it is the principle of inference to the best and most parsimonious
explanation that we must use to regulate content ascriptions to nonlin-
guistic and prelinguistic creatures, and this principle does not require us
to credit protoconversational infants with distinguishing psychological
self-awareness. Simpler explanations are available.12
One such possible explanation might employ a version of Piaget’s no-
tion of secondary circular reactions (although Piaget sees these as charac-
teristic of his stage IV, which covers the fifth to ninth months of the first
year and hence start rather later than protoconversational behavior). Pia-
get’s secondary circular reactions involve a primitive distinction between
means and ends, and the stage of development that he characterizes in
terms of them is one in which infants manipulate objects with a view to
bringing about sights that they find interesting and pleasurable.13 There
are all sorts of reasons why an infant might derive pleasure from proto-
conversational interactions. The infant might derive pleasure from noting
the physical contingency between his own bodily movements and the
movements of his mother.14 Or he might derive pleasure from noting (vi-
sually, proprioceptively, or both) the match between his own bodily move-
ments and his mother’s movements. Or, of course, he might simply find
the whole process intrinsically interesting. Note, moreover, that this line
254 Chapter 9
of explanation is not necessarily incompatible with an intentional expla-

nation of the infant’s behavior (in the sense discussed in chapter 4). It is
incompatible only with explanations that describe protoconversational
infants as engaged in intentional communication.
Another, and perhaps in this context more plausible, explanation of
what is going on in mother-infant protoconversations would take more
seriously the idea that protoconversations involve the evocation and regu-
lation of feelings, while denying that this involves the sort of communica-
tive acts postulated by the thick reading. The key to this second possible
explanation is the distinction between, on the one hand, having evoked
in one a particular feeling and evoking that feeling in another individual
and, on the other, communicating to another individual the fact that one
has a particular feeling. Only the second of these implicates the thick
interpretation of protoconversations. Evoking a feeling in another person
does not require thinking about a person in a way that identifies them as
a person who has feelings (or, by extension, as a bearer of reactive atti-
tudes). And this is compatible with the behavior that evokes the feeling
in the other person being intentional—a point easily missed. Suppose, for
example, that I evoke a feeling in you because I have learned that this will
result in the same feeling being evoked in me in a heightened form. This
evocation could be intentional simply under the description of my in-
tending to bring it about that I experience the same feeling in a heightened
form. It can be the case that feelings are transferred and shared intention-
ally in a way that affects the feelings of each participant without it being
the case that the infant is intentionally communicating the fact that he
has a certain feeling to an individual whom he identifies as someone ca-
pable of having feelings herself. Of course, there is little reason to think
that anything like this goes on in normally developed adult subjects (al-
though it is possible that this offers one way of understanding what seem
to be social interactions in autistic and psychopathic subjects). Nonethe-
less, it offers a plausible way of fleshing out the thin reading of what is
going on in mother-infant protoconversations.
The plausibility of the thin interpretation of social interactions in early
infancy emerges even more clearly when we consider the developments
that take place at about 9 months of age, because at this stage infants
start to behave in ways that compel one to explain them using the catego-
ries deployed by the thick interpretation. In most general terms, what

happens at 9 months is a fundamental change in the nature of the infant’s
interactions with people and objects. I have noted that in the first 9
months of life infants engage in surprisingly complicated social interac-
tions with their caregivers. The same period is also one of keen explora-
tion and manipulation of physical objects. The consensus of opinion
among workers in the area is that what is distinctive of infant behavior
and cognition in the first 9 months of life is an inability to integrate these
two different types of interaction. These young infants can interact with
other people in the protoconversational manner described, and they can
explore and manipulate objects, but they can only do one task at a time.
What happens at 9 months is that infants suddenly become capable, not
just of undertaking both types of interaction at the same time, but also
of coordinating them in a way that creates a fundamentally new type of
interaction. This new type of interaction is triadic rather than dyadic.
Infants become capable of employing their interactions with people in
their interactions with objects, and vice versa. Figure 9.1, drawn from
Tomasello 1993, illustrates the general structure of the transition. In the
remainder of this chapter I will examine three of the new social interac-
tions made possible by this transition from dyadic to triadic interactions.
What I will argue to be the best explanation of these interactions involves
ascribing to the infants involved the distinguishing psychological self-
awareness discussed above. To each of the three types of interaction there
corresponds one strand of the core notion of a psychological subject iden-
tified in the previous section.
The first type of social interaction has been well studied by develop-
mental psychologists as the phenomenon of joint selective visual atten-
tion. Joint visual attention is the simplest form of triadic interaction, and
the basis of the other two that I shall consider. Joint visual attention oc-
curs when infants attend to objects as a function of where they perceive
another individual’s gaze to be directed and, conversely, when infants di-
rect another individual’s gaze to an object in which they are interested.
Let me briefly review the respective evidence for these two aspects of joint
visual attention.
I begin with the evidence for the infant’s ability to look to where they
perceive another individual’s gaze to be directed. Scaife and Bruner (1975)
256 Chapter 9
Before 9 months of age
Infant Adult Infant Adult
(or anywhere else)
Object Object
(a) Infant engagement with object prior (b) Infant engagement with person prior to
to 9 months (adult passive onlooking) 9 months (object not part of interaction)
After 9 months of age
Infant Adult Infant Adult
Object Object
(c) Joint attention after 9 months (d) Self-perception after 9 months
Figure 9.1
Infants’ engagements with objects and persons before and after 9 months of age.
Solid arrows indicate perception and dashed arrows a comprehension of another’s
perspective. (From Tomasello 1993, 177.)
set up an experiment in which infants of varying ages were confronted

with an experimenter (not a caregiver) who established eye contact with
them and then looked away. The aim was to find the age at which infants
followed the movement of the experimenter’s gaze at a level above chance.
The group of infants aged 8 months and over performed considerably
better than chance—a particularly significant result in view of the infant’s
lack of familiarity with the experimenter and the artificiality of the para-
digm. Murphy and Messner (1977) showed that 9-month-old infants are
capable of recognizing maternal pointing as a cue for directing their atten-
tion. Although their ability at that age to follow the invisible line from
the pointing hand to the target object is subject to certain limitations (par-
ticularly when there is a large angle of separation between the pointing
limb and the target object and when there are distracting features), the
infants were clearly able to register that their attention is being directed
toward the target object. There is evidence also that infants, after follow-
ing the direction of the point, look back at the person pointing for feed-
back as to whether they have indeed targeted the right object. The
implications of these and other findings (Bruner 1975) are summed up by
Jerome Bruner in these terms:
What has been mastered at this first stage is a procedure for homing in on the
attentional locus of another. It is a disclosure and discovery routine and not a
naming procedure. It is highly generative within the limited world inhabited by
the infant in the sense that it is not limited to specific kinds of objects. It has,
moreover, equipped the child with a technique for transcending egocentrism, for
insofar as he can appreciate another’s line of regard and decipher their marking
intentions, he has plainly achieved a basis for what Piaget has called decentration,
using a coordinate system for the world other than the one of which he is the
centre. (Bruner 1977, 276)
The converse aspect of joint visual attention, the crucial embedding of
object-focused attention in social contexts, begins at about the age of 6
months when infants begin to switch their gaze back and forth between
object and caregiver (Newson and Newson 1975). These gestures develop
during the last third of the first year into attempts on the part of the infant
to establish an object as the focus of joint attention (Leung and Rheinhold
1981). Daniel Stern provides an eloquent summary of how infants can
take the initiative in generating states of joint visual attention:
Infants begin to point at about nine months of age, though they do so less fre-
quently than mothers do. When they do, their gaze alternates between the target
and the mother’s face, as when she is pointing to see if she has joined in to share
the attentional focus. It seems reasonable to assume that, even prior to pointing,
the infant’s beginning capacity to move about, to crawl or cruise, is crucial in
discovering alternative perspectives as is necessary for joint attention. In moving
about, the infant continually alters the perspective held on some unknown sta-
tionary sight. Perhaps this initial acceptance of serially different perspectives is a
necessary precursor to the more generic “realization” that others can be using a
different coordinate system from the infant’s own. (Stern 1985, 129–130)
It is interesting to consider the conclusions that Stern draws from the
development of these two aspects of joint visual attention:
These observations lead one to infer that by nine months infants have some sense
that they can have a particular attentional focus, that their mother can also have
a particular attentional focus, that these two mental states can be similar or not,
and that if they are not they can be brought into alignment and shared. Inter-
attentionality becomes a reality. (Stern 1985, 130)
258 Chapter 9
Stern’s conclusion, in essence, is that joint visual attention illustrates that

the infant possesses what I have termed distinguishing psychological self-
awareness relative to the category of perceivers of the environment. Let
me reconstruct the reasoning that might lead to this conclusion. Recall
that we are operating under the constraints of inference to the best expla-
nation, with parsimony taken as a necessary condition upon the best ex-
planation. The explananda in one form of joint visual attention are, first,
the fact that the infant’s attention tracks the mother’s attention and, sec-
ond, the fact that the infant seeks feedback from the person pointing.
Both of these explananda seem to demand ascribing to the infant a recog-
nition that the mother (or the experimenter) perceives the world from a
perspective that is different from the infant’s own but that the infant can
learn about by adjusting the focus of his own attention. Unless the infants
do have this recognition, there will be no reason for them to extend their
gaze from the pointing finger to the object to which the finger is pointing
and certainly no reason for them to look back to the mother for feedback.
Built into this recognition on the part of the infant is an awareness of
himself as a perceiver, in particular, as a perceiver who is trying to perceive
what his mother perceives.
The explanatory requirement to assume that the infant is aware of him-
self as a perceiver is even clearer in the second form of joint visual atten-
tion, where the explanandum is the infant’s trying to direct another
individual’s gaze to an object in which they are interested. This is a means-
end activity. The infant is trying to bring about the desired end that his
mother look at the object or event in which he is interested. This, of
course, implicates a grasp that his mother is a perceiver. That the infant
also has a concomitant grasp that he himself is a perceiver emerges more
clearly in the means he chooses to bring this end about. In essence what
the infant tries to do is to make the mother recognize that he, as a per-
ceiver, is looking at a particular object, with the eventual aim that her
recognition that this is what he is trying to do will cause the mother to
look in the same direction.
It is interesting now, and will be helpful in the next chapter, to outline
some representative first-person contents that seem to be implicated in
the best description and explanation of what is going on in joint visual
attention. The first type of joint visual attention (where the infant attends
to an object as a function of where he sees the other’s gaze directed) seems

to require attributing to the infant contents like the following:
(1) The infant recognizes, “Mother wants me to look where she is
looking.”
Or perhaps:
(2) The infant recognizes, “Mother is trying to get me to look where
she is looking.”
In the second type of joint visual attention the direction of fit is reversed:
infants are trying to direct the mother’s attention to an object they are
looking at. Here inference to the best explanation seems to push one to-
ward something like the following psychological attributions. First is a
desire:
(3) The infant desires, “Mother will look where I am looking.”
But the desire is efficacious in explaining behavior only when combined
with an instrumental protobelief along the following lines:
(4) The infant recognizes, “Mother will look where I am looking if I
look back and forth from her to it.”
Of course, particular explanations in particular contexts may well depart
from these paradigms. But it does seem that central cases of joint visual
attention will need to be explained through the attributions of intentions
and protobeliefs like these.
Joint visual attention is, of course, a form of collaborative activity, and
it is possible to argue that the infant’s efforts in bringing about a redirec-
tion of the focus of the mother’s attention itself presupposes a degree of
distinguishing psychological self-awareness as an agent. The argument
might run, for example, like the following. The infant’s intention is that
the mother’s recognition that he is trying to bring it about that she look
where he is looking, cause her to look where he is looking. If, like many
philosophers of action (e.g., O’Shaughnessy [1980]), one takes the pres-
ence of a trying to betoken the presence of an action, then the content of
the infant’s intention seems to include a recognition of his own agency. I
do not want to put too much weight on this argument, however, as there
is excellent evidence for ascribing to infants a more full-blooded sense
of agency at about the same age. This is evidence from the games and
260 Chapter 9
collaborative activities that infants engage in with their caregivers (what

some workers in the area call coordinated joint engagement). As before,
I’ll start by briefly reviewing the data and then move on to explain their
implications.
Much of the most interesting work in this area has taken the form of
longitudinal studies that trace the development of a single infant over
time. The data that I want to focus on comes from Trevarthen and Hu-
bley’s (1978) work with an infant called Tracey. They placed Tracey and
her mother interacting and playing in a darkened and comfortably fur-
nished room and filmed them unobtrusively from an adjacent room. A
total of 32 visits during Tracey’s first year provided a detailed perspective
on Tracey’s development. The room was equipped with toys, and the ex-
perimenters were particularly interested in the development of different
forms of play. They noted an important transition between 9 and 10
months:
At 40 weeks Tracey’s mother became an acknowledged participant in actions.
Tracy repeatedly looked up at her mother’s face when receiving an object, pausing
as if to acknowledge receipt. She also looked up to her mother at breaks in her
play, giving the indication of willingness to share experiences as she had never
done before. Tracey pulled the cart in by the string, watching it move remote
from her hand. She accepted many changes among coloured beads by her mother,
pausing in her manipulation to look at what was shown to her. She followed when
her mother pointed to a bead while speaking, and calmly accepted removal of an
object without loss of interest in the shared play. At one point she gently moved
her mother’s hand aside so she could get to beads beneath it. When her mother
showed her how to make the wheels of the inverted trolley turn and squeak, Tra-
cey watched closely and touched the wheels. When her mother eagerly said “Pull
it” Tracey made a move to draw the trolley towards her, but failed because the
string was not taut, at the same time, expecting success, she looked up and smiled
eagerly at her mother. This was clearly a learned anticipation of the pleasure they
usually shared when she did the trick of pulling in the trolley correctly. (Trevar-
then and Hubley 1978, 201–204)
This is a clear illustration of what was earlier termed a triadic interaction,
in which infants employ their interactions with people in their interac-
tions with objects, and vice versa. In the next few weeks the games Tracey
and her mother played became more sophisticated:
At both 45 and 47 weeks a large transformation in the balance of Tracey’s com-
munications with her mother was completed, and the effect on her mother was
very great. For the first time Tracey gave a play object happily to the mother when
asked to do so. They played with one of Tracey’s toys, a rattle with a clear plastic
globe, and a ball inside it. The globe could be unscrewed from the handle to take
out or put in the ball. When the ball was offered to Tracey she promptly put it in
the globe, and when her mother assembled the two parts she eagerly shook it. . . .
Then Tracey’s mother rolled a cloth ball to her saying, “Ready, steady, go!” Tracey
caught it in two hands and grinned delightedly banging the table with both hands
and the ball, and looking up at her mother’s face. As her mother held her hand
out palm up to receive the object saying, “Where’s the ball?”, Tracey hesitated a
moment and, distracted by the sound of someone entering the room to the side
away from her mother, turned to hold out the ball to the visitor. Her mother
continued, “No, here: over this side.” Tracey looked at her mother’s hand, quickly
reached to give her mother the ball, looked to her face and smiled. “Thank you!”
said the mother, and Tracey gave a triumphant vocalization and hit the table.
(Trevarthen and Hubley 1978, 204–205)
Other longitudinal studies (Bakeman and Adamson 1984) have found
similar patterns in infant development.
As Trevarthen and Hubley’s descriptions make clear, coordinated joint
engagement clearly involves (and indeed presupposes) joint visual atten-
tion. Therefore, coordinated joint engagement requires distinguishing
psychological self-awareness relative to the category of perceivers. What
is distinctive about it, however, is that it also implicates distinguishing
psychological self-awareness relative to the category of agents. The plea-
sure that Tracey takes in the various games is not just pleasure at her own
agency (in the way that, for example, many infants show pleasure in the
simple ability to bring about changes in the world like moving a mobile)
but pleasure at successfully carrying out an intention—a form of pleasure
possible only for creatures who are aware of themselves as agents. When
the intention that is successfully carried out is a joint intention (as it is in
several of the games described), the pleasure shared with the other partici-
pant reflects an awareness that they too are agents.
As we did with joint visual attention, let us look at the details of some
of the contents that appear to be implicated in the best explanation of
coordinated joint engagement. If we are trying to explain why infants take
such pleasure when they succeed in various tasks or games, the natural
explanation would involve something like the following attribution:
(5) The infant recognizes, “I have succeeded in what I intended to do.”
Similarly, when the task or game is a cooperative one, the content of the
intention is first-person plural rather than first-person singular:
262 Chapter 9
(6) The infant recognizes, “We have succeeded in what we are trying
to do.”
Of course, in cooperative games (like the game that Trevarthen reports
Tracey and her mother playing with the cloth ball) a vital part of what is
going on appears to be each party’s recognition of the other’s intentions.
Thus the Tracey manifests intentions with the following content:
(7) The infant recognizes, “Mother wants me to give her the ball.”
It is clear from these specimen contents (again, all first-personal) how
coordinated joint engagement implicates distinguishing self-awareness
relative to the category of agents. All three contents involve the infant’s
awareness of herself as an agent, and (6) and (7) combine this with an
awareness of the mother’s agency.
So the well-documented phenomena of joint visual attention and coor-
dinated joint engagement in infants during the last quarter of the first
year provide excellent examples of social interactions for which inference
to the best and most parsimonious explanation requires ascribing to the
infants in question distinguishing psychological self-awareness relative
to the psychological categories of perceiver and agent. Of the three
strands of distinguishing self-awareness identified earlier in the chapter,
this leaves only the third undiscussed: awareness of oneself as a bearer of
reactive attitudes relative to a contrast space of other bearers of reactive
attitudes. Once again, the crucial period of infant development that starts
at around 9 months of age presents a striking example of social inter-
actions for which the best explanation requires attributing to the infants
involved precisely this type of distinguishing self-awareness. The crucial
stage is the emergence of what developmental psychologists call social
referencing.
The essence of social referencing is the regulation of one’s own behavior
by investigating and being guided by the emotional reactions of others to
a particular situation. In its typical manifestation in infancy, the infant
will come across a puzzling or perhaps intimidating situation and will
look toward his mother for guidance. His subsequent behavior will be
influenced by his perception of her emotional reaction to the situation.
Like coordinated joint engagement, social referencing presupposes joint
visual attention: not only do both mother and infant have to be attending
to the same features of the situation for social referencing to take place,
but the infant has to be confident that he has successfully captured his
mother’s attention if he is to trust her reactions. Social referencing, there-
fore, clearly implicates distinguishing self-awareness as a perceiver. It is
richer than joint visual attention and coordinated joint engagement be-
cause it opens up the emotional and reactive dimension of self-awareness.
Like joint attention and joint engagement, however, the self-awareness
emerges in the context of cooperative and interactive behavior into which
the infant enters to cope with a world of unfamiliar and often alarming
objects and people.
I will cite only one study to illustrate the phenomenon of social refer-
encing: Klinnert et al. 1983. It is particularly interesting because it em-
ploys the experimental paradigm of the visual cliff cited in chapter 5 as
an important example of how infants are capable of picking up self-
specifying visual information almost immediately after birth. Recall that
the visual cliff is a table made of clear glass and divided into two halves.
On one of the two halves a chequered pattern is placed immediately be-
low the surface of the glass to create an appearance of opacity, while a
similar chequered pattern is placed at a variable distance below the sur-
face of the other half to create the appearance of a sudden drop-off. What
Klinnert et al. discovered is that 12-month-old infants clearly employ
social referencing when they reach the drop-off point as they try to cross
the tabletop to their mother. They look down at the deep side and then
across to their mother. The mother was in each case instructed to adopt
a predetermined facial expression (either fear or happiness). The results
clearly showed that the infants registered and responded to the mother’s
emotional reaction to the situation. Of the 19 infants whose mothers
smiled, 14 crossed the deep side, while of the 17 infants whose mothers
showed fear, none crossed the deep side to the mother.
What conclusions can be drawn from social referencing? The expla-
nanda are that the infants look towards their mothers for guidance when
they encounter a surprising or alarming event like the visual cliff and that
the mother’s facial expression is highly effective in determining their sub-
sequent behavior. Explaining the first of these clearly requires attributing
to the infant a comprehension that his mother reacts to things and events
in the same way that he does, that she also finds events surprising or
264 Chapter 9
alarming. As mentioned earlier, this presupposes, but is not exhausted by,

the infant’s comprehension that both he and his mother are perceivers of
the world. Following the mother’s line of sight or the invisible line that
leads from her pointing fingertip to an object does nothing more than
direct the infant’s attention to a particular object or event. It does not tell
the infant anything about the object. Social referencing, in contrast, is a
way of finding out about things in the world. It is not too fanciful, I think,
to draw a linguistic analogy: whereas joint visual attention is a primitive
analogue of the linguistic act of reference, social referencing is a primitive
analogue of the linguistic act of predication. This means that the effective-
ness of social referencing depends upon the infant’s recognizing that his
mother’s emotional reactions are a guide to the nature of the puzzling
object or event. This is what explains his willingness to tailor his own
emotional reactions to the the emotional reactions that he perceives in his
mother, which in turn presupposes an awareness that both he and his
mother are bearers of reactive attitudes. This, of course, is just distin-
guishing self-awareness relative to the category of bearers of reactive
attitudes.
Once more, it is worth making explicit specimens of the general type
of contents that might feature in descriptions and explanations of social
referencing. The infant on the visual cliff looks towards his mother for
guidance. Why? The natural explanation is that the infant has an instru-
mental protobelief to the effect that his mother’s reactions will provide an
indication of whether the situation really is as bad as it appears:
(8) The infant recognizes, “Mother is like me and reacts in the same
sort of way to events as I do.”
Something slightly more complex is required to translate this into ac-
tion—capturable perhaps by the following pair of contents:
(9) The infant recognizes, “The expression on her face means that she
herself would not go forward.”
(10) The infant intends, “I will do what she would do.”
Again, the combination of (8), (9), and (10) clearly implicates the pres-
ence of distinguishing psychological self-awareness relative to the cate-
gory of bearers of reactive attitudes.
9.5 Conclusion
Let me draw the strands of this chapter together. I have been focusing
on psychological self-awareness. It emerged early on in the chapter that
psychological self-awareness needs to be viewed in the context of what I
termed the Symmetry Thesis, namely that a subject’s psychological self-
awareness is constitutively linked with his awareness of other minds. The
neo-Lockean defence of the Symmetry Thesis that I offered led to the
thought that the Symmetry Thesis holds for self-awareness relative to
the psychological categories that define and form the contrast space for
the complex category of psychological subjects. These categories are
those of perceivers, agents, and bearers of reactive attitudes.
The central question tackled in this chapter was whether it is legitimate
to speak of nonconceptual psychological self-awareness in the way that
it has been seen to be legitimate to speak of nonconceptual bodily self-
awareness in previous chapters. As in previous chapters, I suggested that
an answer to this question needed to be constrained by the requirements
of inference to the best and most parsimonious explanation. What is
needed to settle the question is clear evidence of forms of behavior for
which inference to the best explanation requires attributing to non-
linguistic or prelinguistic creatures precisely such psychological self-
awareness. In the final section of the chapter I showed how such evidence
was to be found in the complex social interactions in which human in-
fants partake during and after the last quarter of the first year. Joint visual
attention exemplifies distinguishing psychological self-awareness relative
to the category of perceivers. Coordinated joint engagement exemplifies
distinguishing psychological self-awareness relative to the category of
agents. And social referencing exemplifies distinguishing psychological
self-awareness relative to the category of bearers of reactive attitudes.
Thus, with respect to these three core elements in the notion of a psycho-
logical subject, the notion of nonconceptual psychological self-awareness
is indeed legitimate. It remains, however, to see how this can be deployed
to solve the paradox of self-conscious. For this I turn to the next and
final chapter.
10
Solving the Paradox of Self-Consciousness
The preceding chapters have provided the materials with which to offer a
solution to the paradox of self-consciousness identified in chapter 1. In
this final chapter I will draw together the various strands of the argument
and show how they illuminate the paradox of self-consciousness. The first
step, though, must be to recapitulate the paradox and the various moves
that have been made towards a satisfactory solution to the paradox.
10.1 A Recapitulation
The paradox of self-consciousness as identified in chapter 1 can be pre-

sented as an inconsistency between the following six propositions:
1. The only way to analyze self-consciousness is by analyzing the capac-
ity to think ‘I’-thoughts.
pronoun.
4. Linguistic mastery of the first-person pronoun requires the capacity to
think ‘I’-thoughts.
5. A noncircular analysis of self-consciousness is possible.
6. The capacity to think ‘I’-thoughts meets the Acquisition Constraint
(in the paradigmatic way defined in chapter 1).
It is propositions (5) and (6) that create the paradox. If the Thought-
Language Principle is correct, then it does indeed follow that the only way
to analyze the capacity to think ‘I’-thoughts is through an analysis of the
268 Chapter 10
conditions of mastery of the first-person pronoun. But there are two kinds
of circularity created by the proposal (as per (1)) that this is the key to
providing an analysis of the capacity for self-conscious thought. The first
type of circularity, which I termed explanatory circularity, arises because
the capacity for self-conscious thought is presupposed in a satisfactory
account of mastery of the first-person pronoun. The second type of circu-
larity (capacity circularity) arises because the putative interdependence
between self-conscious thought and linguistic mastery of the first-person
pronoun rules out the possibility of explaining how the capacity either
for self-conscious thought or for linguistic mastery of the first-person
pronoun arises in the normal course of human development. Neither ca-
pacity is innate, and yet each presupposes the other in a way that seems
to imply that neither can be acquired unless the other capacity is already
in place.
Clearly, defusing the paradox of self-consciousness requires removing
at least one of the propositions (1) to (6) so as to leave the propositions
that remain consistent. The proposal that emerged from considering the
strengths and shortcomings of the functionalist theory of self-reference in
chapter 2 is that it is the Thought-Language Principle that must be re-
jected. Rejecting the Thought-Language Principle means rejecting the
theory of content that underlies it, what I termed the classical conception
of content. The classical conception of content has two key components.
The first component is the Conceptual-Requirement Principle, that the
ascriptions of content to an individual are constrained by the concepts
that the individual possesses. The second component is the Priority Prin-
ciple, to the effect that there is a constitutive connection between concep-
tual abilities and correlated linguistic abilities such that conceptual
abilities cannot be possessed by nonlinguistic creatures. These two com-
ponents jointly make up the Thought-Language Principle, and rejecting
either one of them will be sufficient to reject the Thought-Language Prin-
ciple. I proposed retaining the Priority Principle and rejecting the prin-
ciple that ascriptions of content are constrained by concept possession.
This makes possible a theory of nonconceptual content, according to
which states with representational content can be properly ascribed to
individuals without those individuals necessarily possessing the concepts
required to specify how those states represent the world.
Solving the Paradox of Self-Consciousness 269
Of course, rejecting the Thought-Language Principle and the classical

conception of content is useless unless it is shown that rejecting the
Thought-Language Principle does not affect the truth values of the propo-
sitions that remain. The general strategy that I proposed here involves
distinguishing between those forms of full-fledged self-consciousness that
presuppose mastery of the first-person concept and linguistic mastery of
the first-person pronoun and those forms of primitive or nonconceptual
self-consciousness that do not presuppose any such linguistic or concep-
tual mastery. The operative idea was that if the self-consciousness presup-
posed by linguistic mastery of the first-person pronoun could be
construed as nonconceptual rather than full-fledged, then the two forms
of circularity that create the paradox of self-consciousness could be neu-
tralized. A noncircular analysis of full-fledged self-consciousness in terms
of linguistic mastery of the first-person pronoun would be available if
the ‘I’-thoughts presupposed by such linguistic mastery turned out to be
instances not of full-fledged self-consciousness but instead of nonconcep-
tual self-consciousness. This would also provide the key to showing how
linguistic mastery of the first-person pronoun meets the Acquisition Con-
straint. On the assumption that linguistic mastery of the first-person
pronoun can be analyzed satisfactorily in terms of nonconceptual
self-consciousness, the Acquisition Constraint will be satisfied if it
can be shown that there could be a plausible developmental progression
from the cognitive skills and abilities that normal human infants have
available to them at birth via the relevant forms of nonconceptual self-
consciousness to linguistic mastery of the first-person pronoun.
If a solution to the paradox of self-consciousness along these lines is
to be independently plausible, there are several conditions that must be
satisfied. First, the rejection of the classical conception of content must
be independently motivated. The paradox of self-consciousness cannot
be the only reason for rejecting the classical conception of content if the
appearance of a purely ad hoc move is to be avoided. A second condition
is that the theory of nonconceptual content be subject to general con-
straints on the ascription of states with representational content. This in
turn requires developing a general account of the marks that characterize
content-bearing states in general and limiting ascriptions of nonconcep-
tual content to states in which those marks are clearly present. Moreover,
270 Chapter 10
some ways of rejecting the classical conception of content to allow the

existence of nonconceptual contents are nonetheless incompatible with
the proposed solution to the paradox. It might be argued, for example,
that states with nonconceptual content can be ascribed only to creatures
that have a certain basic conceptual and linguistic mastery. A third condi-
tion, therefore, is that the conception of nonconceptual content in play
be one that permits representational states with nonconceptual content
to be ascribed to creatures that have no linguistic (and hence conceptual)
abilities at all, that is, that recognizes the possibility of autonomous non-
conceptual contents.
In chapter 3, I undertook to show that these conditions are indeed sat-
isfied. I began by discussing arguments from the theory of perceptual
experience in support of the general thesis that not all contents can be
conceptual. Those arguments do not, however, support a notion of auton-
omous nonconceptual content. Accordingly, I then argued that evidence
from experiments studying dishabituation in object perception in early
infancy shows that the perceptual experience of infants does genuinely
represent the world in a content-bearing way. Hence, since these are in-
fants who manifestly lack any linguistic or conceptual abilities, the ways
in which they represent the world can only be captured through a notion
of autonomous nonconceptual content. In chapter 4, after having thus
legitimated the notion of nonconceptual content, I offered an account
of the essential marks of content-bearing states that is neutral between
conceptual and nonconceptual content and thus provides a principled
way of identifying when it is appropriate to ascribe states with noncon-
ceptual content.
Nonetheless, a simple rejection of the classical conception is not suffi-
cient, since there are many philosophers who would be prepared to coun-
tenance the possibility of nonconceptual content without accepting that
there might be nonconceptual first-person contents, or ‘I’-thoughts. Thus,
to use the theory of nonconceptual content to solve the paradox of self-
consciousness, I must independently motivate the possibility of noncon-
ceptual first-person contents, and hence the possibility of nonconceptual
self-consciousness. This is a more substantial task, occupying chapters 5
through 9. The methodology I adopted rested on the first of the marks of
content identified in chapter 3, namely that content-bearing states serve
to explain behavior in situations where the connections between sensory

input and behavioral output cannot be plotted in a lawlike manner. I
pointed out in chapter 2 ( pace the functionalist theory of self-reference)
that not every instance of intentional behavior where there are no such
lawlike connections between sensory input and behavioral output needs
to be explained by attributing to the creature in question representational
states with first-person contents. Nonetheless, many such instances of in-
tentional behavior do need to be explained in this way. This offers a way
of establishing the legitimacy of nonconceptual first-person contents.
What would satisfactorily demonstrate the legitimacy of nonconceptual
first-person contents would be the existence of forms of behavior in pre-
linguistic or nonlinguistic creatures for which inference to the best under-
standing or explanation (which in this context includes inference to the
most parsimonious understanding or explanation) demands the ascrip-
tion of states with nonconceptual first-person contents.
I applied this methodological principle in four separate domains, each
of which yielded a different type of nonconceptual self-consciousness. In
chapter 5, drawing on J. J. Gibson’s account of perception, I developed
the view that the pick-up of self-specifying information in exteroceptive
perception is a source of nonconceptual first-person contents from the
beginning of life. This was argued partly on the basis of an analysis of the
structure of the visual field and of visual experience in general and partly
as a form of inference to the best explanation from experimental work
on the sensory capacities of human infants. The second domain where
I looked for evidence of nonconceptual self-consciousness was somatic
proprioception (the various forms of bodily awareness). This occupied
chapter 6. I suggested that somatic proprioception was best understood
as a form of perception of the body. The best explanation of the phenome-
nology of bodily self-perception took the form of an account in which
somatic proprioception emerged as a distinctive form of awareness of the
material self as a spatially extended and bounded physical object that is
distinctive in being responsive to the will.
The nonconceptual first-person contents implicated in somatic propri-
oception and the pick-up of self-specifying information in the structure
of exteroceptive perception provide very primitive forms of nonconcep-
tual self-consciousness, albeit forms that can plausibly be viewed as in
272 Chapter 10
place from birth or shortly afterward. That such nonconceptual first-

person contents exist is a necessary condition of there being a solution to
the paradox of self-consciousness, because they are an essential part of
the case against the central assumption generating the paradox (namely,
that it is incoherent to ascribe thoughts with a first-person content to
creatures who lack linguistic mastery of the first-person pronoun). None-
theless, they are far from sufficient. A solution to the paradox requires
showing how we can get from these primitive forms of self-consciousness
to the full-fledged self-consciousness that comes with linguistic mastery
of the first-person pronoun. This progression will have to be both logical
(in a way that will solve the problem of explanatory circularity) and onto-
genetic (in a way that will solve the problem of capacity circularity).
Clearly, this requires that there be forms of self-consciousness that, while
still counting as nonconceptual, are nonetheless more developed than
those yielded by somatic proprioception and the structure of extero-
ceptive perception, and moreover that it be comprehensible how these
more developed forms of nonconceptual self-consciousness should have
emerged out of basic nonconceptual self-consciousness.
The dimension along which forms of self-consciousness must be com-
pared is the richness of the conception of the self that they provide. None-
theless, a crucial element in any form of self-consciousness is how it
enables the self-conscious subject to distinguish between self and environ-
ment—what many developmental psychologists term self-world dualism.
In this sense, self-consciousness is essentially a contrastive notion. One
implication of this is that a proper understanding of the richness of the
conception of the self that a given form of self-consciousness provides
requires that we take into account the richness of the conception of the
environment with which it is associated. In the case of both somatic pro-
prioception and the pick-up of self-specifying information in exterocep-
tive perception, there is a relatively impoverished conception of the self
associated with a comparably impoverished conception of the environ-
ment. One prominent limitation is that both are synchronic rather than
diachronic. The distinction between self and environment that they offer
is a distinction that is effective at a time but not over time. The contrast
between propriospecific and exterospecific invariants in visual percep-
tion, for example, provides a way for a creature to distinguish between
itself and the world at any given moment, but this is not the same as a
conception of oneself as an enduring thing distinguishable over time from
an environment that also endures over time.
To capture this diachronic form of self-world dualism, I introduced the
notion of a nonconceptual point of view. Having a nonconceptual point
of view on the world involves taking a particular route through the envi-
ronment in such a way that one’s perception of the world is informed by
an awareness that one is taking such a route. This diachronic awareness
that one is taking a particular route through the environment turned out
to involve two principal components: a nonsolipsistic component and a
spatial awareness component. The nonsolipsistic component is a subject’s
capacity to draw a distinction between his experiences and what those
experiences are experiences of, and hence his ability to grasp that what is
experienced exists at times other than those at which it is experienced.
This requires the exercise of recognitional abilities involving conscious
memory and can be most primitively manifested in the feature-based rec-
ognition of places. I glossed the spatial awareness component of a non-
conceptual point of view in terms of possession of an integrated
representation of the environment over time. That a creature possesses
such an integrated representation of the environment is manifested in
three central cognitive/navigational capacities: a capacity to think about
different routes to the same place, a capacity to keep track of changes in
spatial relations between objects caused by one’s own movements relative
to those objects, and a capacity to think about places independently of
the objects or features located at those places. I cited evidence from both
ethology and developmental psychology indicating that these central cog-
nitive/navigational capacities are present in both nonlinguistic and prelin-
guistic creatures. Again, my argumentative strategy was inference to the
best explanation.
What is significant about the notion of a nonconceptual point of view
is that it manifests an awareness of the self as a spatial element moving
within, acting upon, and being acted upon by the spatial environment.
This is far richer than anything available through either somatic proprio-
ception or the self-specifying information available in exteroceptive
perception. Nonetheless, like these very primitive forms of self-conscious-
ness, a nonconceptual point of view is largely awareness of the material
self as a bearer of physical properties. This limitation raises the question
274 Chapter 10
of whether there can be a similarly nonconceptual awareness of the mate-

rial self as a bearer of psychological properties. This question was the
subject of chapter 9. In approaching the possibility of nonconceptual psy-
chological self-consciousness, I was guided by the thought that there
might be a constitutive link between a subject’s psychological self-
consciousness and his awareness of other minds (the Symmetry Thesis).
In the first part of the chapter I defended a weak version of the Symmetry
Thesis, according to which there are some psychological categories that
a subject cannot apply to himself without also being able to apply them
to other psychological subjects. These psychological categories form the
core of the notion of a psychological subject. This is so because a subject’s
recognition that he is distinct from the environment in virtue of being a
psychological subject depends on his ability to identify himself as a psy-
chological subject within a contrast space of other psychological subjects,
and this self-identification as a psychological subject takes place relative
to a set of categories that collectively define the core of the concept of a
psychological subject.
There appear to be three central psychological categories defining the
core of the concept of a psychological subject: the category of perceivers,
the category of agents, and the category of bearers of reactive attitudes.
This, in conjunction with the Symmetry Thesis, offers a clear way of an-
swering the question of whether psychological self-consciousness can ex-
ist in a form that is nonconceptual and independent of linguistic mastery.
In line with the methodology I adopted previously in this book, what
would settle the matter would be social interactions involving prelinguis-
tic or nonlinguistic subjects for which inference to the best explanation
requires assuming that those subjects are applying the relevant psycholog-
ical categories to themselves and to others. In the final section of the chap-
ter I looked at various experimental paradigms of research on the social
cognition of infants and showed that there are compelling grounds for
attributing to prelinguistic infants in the final quarter of the first year
distinguishing psychological self-consciousness relative to the three key
categories.
The situation at the moment, therefore, is that four different forms of
primitive nonconceptual self-consciousness have been identified and
shown to be psychologically real. With this we have the materials for a
solution to the paradox of self-consciousness. The next and final stage is

to show how these materials can be used to circumvent the paradox of
self-consciousness.
10.2 Nonconceptual Self-Consciousness and Explanatory Circularity
What is responsible for explanatory circularity is the fact that an ade-

quate explanation of what linguistic mastery of the first-person pronoun
consists in involves ascribing to the person who has mastered the first-
person pronoun certain first-person thoughts that, by the Thought-
Language Principle, are themselves comprehensible only by ascribing to
the language-user mastery of the first-person pronoun. The first stage in
resolving this aspect of the paradox of self-consciousness is to get clear
on precisely what those first-person thoughts are.
The initial argument in chapter 1 was that a grasp of the token-reflexive
rule governing the reference of the first-person pronoun cannot be suffi-
cient to explain mastery of the first-person pronoun. A minimal require-
ment upon proper and genuine employment of the first-person pronoun
is that the person who employs a token of the first-person pronoun so as
to secure genuine and proper self-reference should know that he is the
producer of the token in question. This minimal requirement must be
reflected in an account of mastery of the first-person pronoun. A natural
question at this point is whether this minimal requirement is also (in con-
junction, of course, with a suitable grasp of the token-reflexive rule) suf-
ficient for linguistic mastery of the first-person pronoun. More precisely,
given that an account of what it is to have mastery of the first-person
pronoun will need to specify the conditions holding on genuine and
proper employment of the first-person pronoun, does such an account
need to specify any conditions other than mastery of the token-reflexive
rule and a knowledge that one is the producer of the relevant token of ‘I’?
Brief reflection shows that these conditions cannot be sufficient. There
are straightforward counterexamples in which a language-user produces
a token of ‘I’ in full understanding of the token-reflexive rule, whereby
every token of ‘I’ refers to the person who produces it, and in the full
knowledge that he himself is the producer of that token, while nonethe-
less failing to refer to himself. One such counterexample is provided by
276 Chapter 10
the use of language in fiction. An actor uttering a soliliquy satisfies both

conditions but does not refer to himself. Fiction is a difficult case, how-
ever, since it is arguable that no genuine and proper reference takes place
in fictional discourse. Fortunately, there are two further, straightforward
counterexamples that do not depend on any peculiarities of fictional situ-
ations. When I dictate a letter to a secretary who writes down at the end
of the letter the words ‘I look forward to hearing from you’, she does so
fully comprehending both the token-reflexive rule that the relevant token
of ‘I’ refers to whoever utters or inscribes it and the fact that she is writing
down that token. But she is obviously not referring to herself, nor, of
course, does she take herself to be referring to herself. The same is true
of the simultaneous interpreter who translates my words into French
without placing them in oratio obliqua. These are all cases where some-
one knowingly produces a token of ‘I’ (or ‘je’) in full comprehension of
the token-reflexive rule, and yet it is clear that they neither refer to them-
selves nor take themselves to be referring to themselves.
Evidently, then, more is required for proper use of the first-person pro-
noun than knowingly producing a token of ‘I’ in full comprehension of
the token-reflexive rule. A natural suggestion here would be that what is
missing in the examples just given is an intention on the part of the lan-
guage users to use the token of ‘I’ that they knowingly produce to refer
to themselves. Perhaps the reason that neither the secretary nor the inter-
preter take themselves to be referring to themselves is that they know that
they do not intend to refer to themselves. There is an important difference
between knowingly producing a token of ‘I’ that would, in certain circum-
stances, refer to oneself and producing a token of ‘I’ knowing that one is
intending to refer to oneself. An ostensibly competing explanation would
be that the token-reflexive rule does not apply to either the secretary or
the translator because neither of them produces the respective token of ‘I’
in the relevant sense of ‘produces’. Indeed, it is highly plausible that there
is more to producing a token of ‘I’ than simply uttering or inscribing it.
Yet in spelling out what the production of a token of ‘I’ involves over and
above uttering it or inscribing it, it is natural to advert to the intention to
utter or inscribe that token of ‘I’ with the intention to refer to oneself.
Let me start by trying to get a proper understanding of what is involved
in the phenomenon of intentional self-reference by means of the first-
person pronoun. What is the content of the relevant intention? The best
place to start is with Grice’s communication-theoretic account of mean-
ing (1957), on the entirely reasonable assumption that an account of the
intention to refer linguistically to oneself can be derived only from an
account of communicative intent in general. As is well-known, the general
form of Grice’s attempt to use the notion of communicative intent to illu-
minate the notion of meaning is this: that A means something by x is to
be understood in terms of A’s intention that the utterance of x should
produce some effect in an audience by means of the recognition of this
intention.1 Although this general form has been somewhat refined in re-
sponse to counterexamples (see particularly Grice 1969), we can safely
ignore these complexities. What is immediately relevant is a general bifur-
cation that Grice introduces between two different types of responses that
might be communicatively intended—what might be termed the informa-
tional and the practical. When an informative message is intended, the
response that an utterer intends to produce in his audience is a belief or
some other kindred cognitive state. When a practical message is intended,
the response that an utterer intends to produce is an action of some form
or other. Let me schematize these. Communicatively intending an infor-
mational message takes (roughly) the following form:
An utterer u utters x to mean p if and only if u utters x intending
a. that some audience a should come to believe that p,
b. that a should be aware of intention (a),
c. that the awareness mentioned in (b) should be part of a’s reason for
believing p.
Communicatively intending a practical response takes this form:
An utterer u utters x to mean that an audience a should do f if and only
if u utters x intending
a. that a should do f,
c. that the awareness mentioned in (b) should be part of a’s reason for
doing f.
Other philosophers of language in the communication-theoretic tradition
(Armstrong 1971, Bennett 1976) have followed Grice in this bifurcation in
the central notion of intending to bring about a response in the audience.
278 Chapter 10
It is not clear from the work of Grice and his successors whether they
intend their account of communicative intent to be part of a statement of
the necessary and sufficient conditions that must be satisfied on any given
occasion of language use, so that whenever a token sentence x is used to
mean that p, or to bring about the response f, it must be accompanied
by the appropriate communicative intent. Many philosophers would find
such a claim psychologically implausible. Let me offer such philosophers
a weaker construal of the communicative-intent theory. One way of read-
ing the theory on which it would not have this sort of psychological im-
plausibility would be as an account that operates at the level of sentence
types rather than sentence tokens. A weak reading of the claim that the
meaning of sentence type X is given by a complex communicative intent
of the sort just indicated is that no speaker could be credited with a mean-
ingful use of a token of type X unless he learned how to employ tokens
of type X through appreciating the complex communicative intention
given by the theory. For the purposes of this chapter I shall stick with the
Gricean position as formulated for expository convenience. For what I
have to say, however, the weaker reading is perfectly sufficient.
The two-branched analysis of communicative intent in terms of in-
tended informational and practical messages is relatively straightforward
at the level of the sentence. There is a simple and easily identifiable dis-
tinction between assertoric or declarative sentences, which it is plausible
to analyze along the first branch as intended to generate beliefs in the
audience, and imperative or injunctive sentences, which it is equally plau-
sible to analyze along the second branch as intended to bring about some
act or other on the part of the audience. The particular use to which I
wish to put the theoretical notion of communicative intent is in illuminat-
ing intentional self-reference, and there are two good reasons why this
bifurcated theory of communicative intent cannot be unproblematically
applied here.
The first reason is a general one. At a level of analysis below the sen-
tence we are dealing with particular acts like reference and predication,
which do not lend themselves to this clean bifurcation. Let me bring this
out with respect to the act of reference. If a clear bifurcation between
informational and practical communicative intent is to be maintained be-
low the level of the sentence, then reference ought to be construed in
terms of informational communicative intent, so that the response that

an utterer intends to produce in his audience is a belief. Let us assume,
for the sake of simplicity, that we are dealing with the simple subject-
predicate sentence ‘That is green’ uttered in the presence of a green
Granny Smith apple to which I am pointing and that I intend to produce
in my audience (Jones) the belief that the apple is green. How are we to
understand the reference of the demonstrative pronoun ‘that’? Straight-
forwardly adapting the informational account would yield the thought
that my utterance of the demonstrative pronoun (accompanied by a suit-
able ostensive gesture) is intended to produce in Jones the belief that the
object to which I am pointing is the subject of my sentence through what
Bennett helpfully terms the Gricean mechanism (that is, through Jones’s
recognition of my intention to produce in him that belief). But it is surely
not necessary either for Jones to have any such belief or for this to be
what I am trying to do, for me successfully to communicate to him the
thought that the apple is green. Nor is it necessary for me to intend to
produce in him any beliefs about the apple being the subject of my sen-
tence. I do, of course, intend to produce in Jones a belief about the apple,
namely that it is green. As part of that project I refer demonstratively to
the apple. And the apple is indeed the subject of my sentence. But none
of this adds up to my intending Jones to form a belief about the apple
being what my sentence is about. I shall return to this shortly.
The second reason that the theory of communicative intent is problem-
atic at this level is a psychological one highly pertinent to the matter in
hand. It is relatively uncontroversial among developmental psychologists
that a full-fledged concept of belief emerges relatively late in cognitive
development, and certainly long after the emergence of even relatively so-
phisticated language use (including, for that matter, competent use of the
first-person pronoun). The evidence for the late emergence of the concept
of belief is the well-known false-belief task, a beautifully simple paradigm
designed to identify whether or not young children can deploy the idea
of a mental state representing the world to be a way other than the way
it is. In the false-belief task, children are presented with the following
situation. A character, Maxi, places some chocolate in a cupboard and
then leaves the room. While he is away, the chocolate is moved to another
location. The child is asked where Maxi will look for the chocolate when
280 Chapter 10
he returns. In order for the child to give the correct answer, he must un-
derstand that Maxi still represents the chocolate as being where in fact it
is not, namely where Maxi originally left it. And that means that the child
must be able to contrast his perception of the situation as it is with Maxi’s
beliefs about the situation as it is no longer. The experimental evidence is
that children consistently fail on this task until well into their fourth year
(Wimmer and Perner 1983; see Perner 1991 for further discussion). Now,
although there is, of course, much to be said about the false-belief task,
the central point for present purposes is that no subject who fails the
false-belief task can be ascribed the concept of belief, since possessing the
concept of belief requires comprehension of the possibility that another
subject represents the world in a way other than it is.2 This has the
straightforward consequence that the concept of belief cannot feature in
any communicative intention that is to be ascribed to infant language
users younger than three years of age. So, on the assumption that the
notion of communicative intent is indeed applicable here, it must at least
be possible for there to be an appropriate informational communicative
intention that does not take the concept of belief to be central.
To return, then, to the first point, what is the communicative intention
governing my demonstrative reference to the apple, since it is not an inten-
tion that Jones should come to believe (via the Gricean mechanism) that
the apple is the subject of my sentence? The simplest and most straight-
forward suggestion would be that I intend the demonstrative reference to
the apple to draw Jones’s attention to the apple. Of course, this is hardly
sufficient. I could call Jones’s attention to the apple by biting into it or by
placing it on my son’s head and shooting an arrow at it. These would not
be ways of drawing attention to the apple that would count as instances
of demonstrative reference, the reason being, of course, that they can be
effective without proceeding through Jones’s recognition of my intention.
Genuine demonstrative reference occurs when I intend Jones’s attention
to be drawn to the apple by means of the Gricean mechanism, namely that
my utterance of the demonstrative pronoun accompanied by a suitable
ostensive gesture should identify the apple to Jones in virtue of his recog-
nition that such is my intention.
This proposal brings out what is wrong with the bifurcation between
informational and practical communicative intent at the subsentential
level. When I intend to draw Jones’s attention to the apple, I do not intend
simply to produce beliefs in Jones (as the purely informational account of
communicative intent at the level of the sentence would maintain). Nor
do I intend simply to get Jones to do something. What I intend is that
Jones should do something (redirect his attention) in a way that will put
him in informational contact with the apple. This is, moreover, fully in
line with the second point. When I intend that Jones’s attention should
be drawn via the Gricean mechanism to the apple, I do not intend Jones
to form any beliefs about the subject of my sentence. Nor is it necessary
for Jones to form any such beliefs. All that is required is that his attention
should in fact be drawn to the apple through the Gricean mechanism in
a way that will lead him correctly to realize that the apple is what I am
describing as green.
Of course, these observations fall a long way short of a general commu-
nication-theoretic account of reference. There are all sorts of other condi-
tions that might arguably be needed. It might be thought necessary, for
example, that linguistic reference take place as part of the utterance of a
complete sentence. Something also needs to be said (and will be shortly)
about the linguistic meaning of the word through which reference is ef-
fected. I take it, however, that from what’s been said we can derive at least
the bare bones of a satisfactory account of the communicative intent that
governs linguistic reference. Such an account will proceed along some-
thing like the following lines:
An utterer u utters x to refer to an object o if and only if u utters x
intending
a. that some audience a should have their attention drawn to o,
c. that the awareness mentioned in (b) should be part of the explanation
for a’s attention being drawn to o.
This general outline gives us sufficient tools to return to the problem of
self-reference.
Recall what led us to communication-theoretic accounts of reference.
I originally raised the question of whether it is sufficient for genuine lin-
guistic self-reference that a speaker should knowingly produce a token of
‘I’ in full knowledge of the token-reflexive rule governing the reference of
‘I’. I showed that this is not sufficient and suggested that what would be
282 Chapter 10
sufficient would be producing a token of ‘I’ in full knowledge of the

token-reflexive rule and with the intention to refer to oneself. This is what
led me to investigate the communicative intent that governs linguistic ref-
erence—on the two assumptions, first, that intentional self-reference will
be a special case of intentional reference and, second, that the best way
to understand intentional reference is in terms of the communicative in-
tent that governs acts of reference. The next step must be to apply this to
the first-person pronoun.
A straightforward application of the sketch just given of the communi-
cative intent governing the act of reference yields something like the
following:
An utterer u utters ‘I’ to refer to himself* if and only if u utters ‘I’ in full
comprehension of the token-reflexive rule that tokens of ‘I’ refer to their
producer and with the tripartite intention
a. that some audience a should have their attention drawn to u,
b. that a should be aware of u’s intention that a’s attention should be
drawn to u,
for a’s attention being drawn to u.
For familiar reasons, this cannot be satisfactory. An utterer’s intentional
self-reference does not require that he intend to draw attention to himself
in the way just mentioned. Utterer u can refer to himself (more accurately,
himself*) without intending to draw attention to u. This, after all, is the
problem that initially led to the paradox of self-consciousness. It will not
be of any use to replace ‘u’ with ‘the utterer of this token’ in each of the
three parts of the tripartite intention. Although that would undoubtedly
secure successful self-reference in cases where, for example, u is unaware
that he is u, it is surely only in the most unusual cases that people intend
to refer to themselves as producers of tokens of ‘I’. This is sufficient to
prevent it from featuring in a statement of the communicative intent gov-
erning paradigm cases of intentional self-reference. It is, of course, rea-
soning along precisely these lines that brought us to the paradox of self-
consciousness in chapter 1, via the conclusion that any adequate account
of linguistic mastery of the first-person pronoun would have to include
certain first-person thoughts among its conditions. We can now see the
form that such an account will have to take:
a. that some audience a should have their attention drawn to him*,
b. that a should be aware of his* intention that a’s attention should be
drawn to him*,
for a’s attention being drawn to him*.
Each of the three clauses of the tripartite intention is a first person
thought, in virtue of the presence in each of them of the indirect reflexive
pronoun he*.
If this account of linguistic mastery of the first-person pronoun is to
escape the explanatory circularity that partly creates the paradox of self-
consciousness, then, for reasons that have already been discussed in this
chapter, each of the three clauses in the statement of the tripartite com-
municative intention must be formulable in terms of nonconceptual first-
person thoughts. It is at this point that the exploration of nonconceptual
first-person thoughts in the main body of this book pays dividends, for
we now have the materials to provide such a formulation. That is, the
methodology of inference to the best and most parsimonious explanation
has shown how we can be warranted in identifying and explaining behav-
ior by means of nonconceptual first-person contents of precisely the sort
that would provide a noncircular account of linguistic mastery of the first-
person pronoun.
Clause (a) in the tripartite intention is that the utterer should utter a
token of ‘I’ with the intention that some audience should have their atten-
tion drawn to himself*. There are two key components here. The first
component is that the utterer should intend to draw another’s attention
to something. That this is possible at the nonconceptual level is clearly
shown by my discussion of joint selective visual attention in the previous
chapter. The best explanation of joint selective visual attention involving
young infants is that they are acting on the intention to direct another
individual’s attention to an object in which they are interested. Of course,
in the cases of joint selective visual attention I examined, the objects in
question are not the infants themselves. Accordingly, the second compo-
nent of clause (a) of the communicative intention is that the utterer should
284 Chapter 10
be aware of himself* as a possible object of another’s attention. This is

largely a matter of physical self-consciousness. The materials here are pro-
vided by proprioceptive self-consciousness and the various forms of
bodily self-consciousness implicated in possession of a nonconceptual
point of view. Proprioceptive self-consciousness in particular provides a
conception of the body as a spatially extended and bounded physical ob-
ject that is unique in its responsiveness to the will. One way, therefore, in
which clause (a) of the complex tripartite intention might be satisfied at a
nonconceptual level would be through an intention to draw an audience’s
attention to the material self as revealed in proprioceptive self-
consciousness. Another way would be through an intention to draw an
audience’s attention to the material self as a spatial element moving
within, acting upon, and being acted upon by, the spatial environment—
the conception of the self that is implicated in possession of a nonconcep-
tual point of view.
In clause (b) the requirement is that the utterer of ‘I’ should intend that
his audience recognize his* intention to draw its attention to him*. This
is a reflexive awareness of the intention in the first clause. The real issue
that it raises is one about how iterated psychological states can feature in
the content of intentions. In the first clause we have what can usefully be
described as an uniterated intention. An uniterated intention is one in
which a psychological state does not feature within the scope of the ‘that’
clause specifying the content of an intention or any other content-bearing
state, or where a psychological state does so feature, that psychological
state does not have a further psychological state featuring in the ‘that’
clause that specifies its content. The first clause in the intention is an unit-
erated intention in the second of these two senses because the psychologi-
cal state that features in the content of the intention does not take a
further psychological state as its object. A first-order iteration occurs
when a further psychological state is embedded within the context cre-
ated by the psychological state featuring within the scope of the ‘that’
clause specifying the intention. A second-order iteration occurs when
there is a third psychological state embedded within the context created
by that further psychological state. Generally, an iteration to the order n
occurs when there is an (n 2 1)th order psychological state embedded
within the scope of the ‘that’ clause specifying the content of a given psy-
chological state. Clearly, in clause (b) specifying the communicative intent
governing the use of ‘I’, we have a second-order iteration. There are three
psychological states embedded within the scope of the ‘that’ clause. What
complicates the matter further is that one of those psychological states is
a first-person state. Moreover, it is a first-person psychological state that
has a further psychological state within the scope of the ‘that’ clause spec-
ifying its content. This poses an obvious challenge. Are there resources at
the nonconceptual level for thinking an intention of this complexity?
There are two distinguishable questions here. The first question is
whether there are resources at the nonconceptual level for thinking a
thought whose content contains a second-order iteration. The second
question is whether there are resources at the nonconceptual level for em-
bedding a first-person psychological state within a first- or higher-order
iteration. Let me take them in reverse order. The discussion of psychologi-
cal self-consciousness in the previous chapter revealed several examples
of first-person psychological states embedded within first- or higher-order
iterations. A canonical example comes from joint visual attention, where
inference to the best explanation seems to require attributing to infants
contents like the following:
(1) The infant recognizes, “Mother wants me to look where she is
looking.”
This is a first-person content embedded in a first-order iteration. The
structure emerges more clearly when (1) is reformulated less idiomatically
as follows:
(2) The infant recognizes, “Mother wants that I look where she is
looking.”
Of course, this is not simply a peculiarity of joint visual attention. This
sort of embedding of first-person contents in first-order iterations occurs
whenever there is recognition of another’s intention that one should do
something. Recognitional states like these play a crucial role in the coop-
erative games and projects that are so important in infancy after the last
quarter of the first year. An important source of infants’ pleasure and
enjoyment is their recognition that they have successfully performed what
286 Chapter 10
their mothers intended them to. And this implicates an embedding of a

first-person content within a first-order iteration.
The fact remains, however, that the descriptions and explanations given
of the various examples of joint visual attention, coordinated joint en-
gagement, and social referencing that I have considered do not provide
any examples of second or higher-order iterations. They are all forms of
cognition and action for which inference to the best explanation can rest
satisfied with first-order iterations. Nonetheless, it is possible to describe
a consistent and indeed familiar situation for which inference to the best
explanation does appear to require attributing a second-order iteration,
and indeed, a second-order iteration that embeds a first-person content.
This can be done by varying slightly one of the examples of coordinated
joint engagement already discussed. Trevarthen and Hubley’s longitudinal
study of Tracey describes a game with a trolley:
When her mother showed her how to make the wheels of the inverted trolley turn
and squeak, Tracey watched closely and touched the wheels. When her mother
eagerly said “Pull it” Tracey made a move to draw the trolley towards her, but
failed because the string was not taut, at the same time, expecting success, she
looked up and smiled eagerly at her mother. (Trevarthen and Hubley 1978, 204)
Explaining Tracey’s pleasure in this simple game requires attributing to
her the following two true protobeliefs:
(3) Tracey recognizes, “I have succeeded in what I set out to do.”
(4) Tracey recognizes, “Mother recognizes that I have succeeded in
what I set out to do.”
I am assuming that Tracey’s expectant smile toward her mother antici-
pates a reciprocating smile. Suppose, however, another situation in which
things are similar except that no such reciprocating smile appears on the
mother’s face. Tracey* (as I shall call her) is consequently distressed. One
way of explaining this state of affairs would be in terms of Tracey*’s rec-
ognizing her mother’s failure to recognize her intention that her mother
should recognize her success. This in turn requires attributing to Tracey*
the intention to get her mother to take pleasure when she succeeds in
carrying out her intention:
(5) Tracey* intends, “Mother is pleased when she recognizes that I
have succeeded in what I set out to do.”
This is a second-order iteration, and moreover a second-order iteration

that embeds a first-person content.
It should not be thought that describing such a possible scenario is
simply stipulating a solution to the problem, because the possibility of
the scenario as described presupposes the coherence of ascribing second-
order iterations at the nonconceptual level. Certainly, there would be
room for considerable disquiet if the only motivation for believing in the
possibility of the scenario was to solve the problem at hand. But this is
surely not the case. Not only is the sort of interaction between infant and
mother described readily conceivable and indeed familiar, but the expla-
nation proposed can be tested. Suppose, for example, that one is wonder-
ing whether Tracey*’s distress ought not to be explained more simply in
terms of her being unhappy that her mother had failed to smile when a
certain effect was attained. This would, of course, require no more than
the following uniterated intention:
(6) Tracey* intends, “Mother is pleased when such and such occurs.”
Now it seems to me that what is at issue between these two interpreta-
tions can be empirically tested, because they each implicate different
counterfactuals. Suppose that the relevant effect is the moving of the trol-
ley. If the explanation implicating the uniterated intention (6) is correct,
then one would expect Tracey* to be distressed whenever the trolley
moves and her mother does not smile or otherwise show pleasure. If, on
the other hand, the explanation implicating the second-order iteration in
intention (5) is correct, then one would expect Tracey* to show distress
only in those situations where her mother fails to show pleasure and the
movement of the trolley is contingent upon Tracey*’s pulling it. There is
nothing puzzling about the thought that this counterfactual might be true
rather than the previous one, and hence there must be a genuine and psy-
chologically real distinction between (5) and (6).
I take it, therefore, that clause (b) in the tripartite communicative intent
governing the deployment of the first-person pronoun is fully specifiable
at the nonconceptual level. That leaves us simply with clause (c) in the
tripartite intention, the clause specifying the so-called Gricean mecha-
nism that the intention specified in clause (b) should be part of the expla-
nation for the success of the intention specified in clause (a). Here again
are the conditions we are working with:
288 Chapter 10
a. that some audience a should have their attention drawn to him*,
b. that a should be aware of his* intention that a’s attention should be
drawn to him*,
for a’s attention being drawn to him*.
The first point to make here is that the iteration involved in clause (c) is
not at a higher level than the iteration involved in clause (b). It is not that
(a) is embedded in (b). Rather, the content of the intention in (c) is that
there be a certain relation between (a) and (b). What is this relation? In
Gricean accounts of meaning, the relation is often glossed in terms of the
awareness mentioned in (b) being a reason for the success of the intention
in (a). This does not seem appropriate, however, since the notion of a
reason is not operative at the level of nonconceptual content.3 What is
needed is a nonconceptual analog of the reason-giving interpretation of
the Gricean mechanism. Some form of causal interpretation is the obvi-
ous alternative. Suppose that we reformulate (c) as follows:
c9. that the awareness in (b) should bring it about that a’s attention be
drawn to him*
Now since (b) has already been explained and since the intention that a’s
attention be drawn to him* is the content of (a), which has also already
been explained, the only new element to be explained here is the causal
relation of bringing-it-about-that and how it can form part of an inten-
tion linking (a) and (b). There is a reason for referring to the causal rela-
tion of bringing-of-about-that, rather than to the notion of cause. The
concept of cause brings with it a degree of theory that does not seem to
be required to manipulate the Gricean mechanism. For example, the con-
cept of cause has clear modal elements with respect to the necessary con-
nection between cause and effect and the concomitant impossibility of
the cause not being followed by the relevant effect. These modal elements
are not required for the Gricean mechanism to be understood and em-
ployed. Consequently, the concept of cause need not feature in the content
of the communicative intention defining mastery of the first-person pro-
noun. The notion of bringing-it-about-that is intended to indicate a
causal relation that can be understood without a full-fledged understand-

ing of full-fledged causation.4
The causal relation of bringing-it-about-that is integral to the notion
of a nonconceptual point of view and to the self-consciousness that it
implicates. Possession of a nonconceptual point of view involves an
awareness of the self as acting upon and being acted upon by the spatial
environment, and this, of course, depends crucially on understanding
both the general notion of one’s bringing it about that a certain effect
occurs (without which intentional action would be impossible) and the
general notion of things being brought about in virtue of, for example,
interactions between objects (without which it would be impossible to
make predictions about the future other than those based on the most
simple correlations of features). Of course, there is a distinction to be
made between physical causation and psychological causation. The causal
relation of bringing-it-about-that implicated in a nonconceptual point of
view is a physical notion, while the type of bringing-it-about-that involved
in understanding the operation of the Gricean mechanism is clearly psy-
chological. So if the intention in (c9) is to be available at the noncon-
ceptual level, there must be warrant for attributing a grasp of the
psychological bringing-it-about-that relation to prelinguistic creatures.
Not every form of psychological bringing-it-about-that will be suffi-
cient to underwrite comprehension of the Gricean mechanism. Awareness
of oneself as an agent presumably involves a comprehension that one’s
intentions can be effective in bringing about changes in the world, and
this is obviously a form of psychological bringing-it-about-that. Nonethe-
less, it is important to distinguish between what might be termed mind-
to-world bringing-it-about-that and mind-to-mind bringing-it-about-
that. It is the second of these that is required. Comprehension of the
Gricean mechanism requires understanding that one’s manifest intentions
can be efficacious in bringing about changes in the mental states of others.
Such an understanding is, of course, available at the nonconceptual level,
as emerged during my discussion of psychological self-consciousness in
the previous chapter. Both coordinated joint engagement and joint visual
attention depend on both participants comprehending the possibility of
mind-to-mind bringing-it-about-that. Here is one example of a proto-
belief that might feature in the explanation of canonical instances of joint
visual attention:
290 Chapter 10
(7) The infant recognises, “Mother will look where I am looking if I

look back and forth from her to it.”
This instrumental protobelief is a recognition of mind-to-mind bringing-
it-about-that. The infant recognizes that his looking back and forth from
his mother to the object will bring it about that she looks where he is
looking.
Once the comprehension of mind-to-mind bringing-it-about-that at the
nonconceptual level is granted, the legitimacy of clause (c9) in the commu-
nicative intention governing use of the first-person pronoun swiftly fol-
lows. Clause (c9) comprises the following intention:
c9. that the awareness in (b) should bring it about that a’s attention be
drawn to him*
This can be broken down into three elements—specifically, a relation of
mind-to-mind bringing-it-about-that between a’s awareness of the utter-
er’s intention that a’s attention be drawn to him* and a’s awareness actu-
ally being drawn to him. All three elements have now been accommodated
at the nonconceptual level, so there is no obstacle to the thought that their
combination in (c9) can be accommodated at the nonconceptual level.
Is this sufficient to solve the problem of explanatory circularity that is
one of the two planks of the paradox of self-consciousness? The problem
of explanatory circularity derives from the fact that the capacity for self-
conscious thought is presupposed in a satisfactory account of mastery of
the first-person pronoun. This appears to render circular the project of
elucidating the capacity for self-conscious thought through mastery
of the first-person pronoun. We are now in a position, however, to see
why the circularity here is apparent rather than real. A communication-
theoretic account of the requirements on mastery of the first-person pro-
noun does indeed involve, in addition to mastery of the token-reflexive
rule, three intentions with first-person contents. A real and vicious circu-
larity would arise only if the first-person thoughts implicated in mastery
of the first-person pronoun themselves implicate mastery of the first-
person pronoun. What has emerged, however, is that these first-person
thoughts are of a kind that can be nonconceptual in a way that makes
them logically independent of mastery of the first-person pronoun. They
can be used to explain the requirements on linguistic mastery of the first-
person pronoun without themselves presupposing linguistic mastery of

the first person. I take it that this is sufficient to solve the problem of
explanatory circularity. It also holds considerable promise for solving the
problem of capacity circularity, which I take up in the next section of
the book.
10.3 Solving the Problem of Capacity Circularity
The problem of capacity circularity arose in chapter 1 in the context of a

constraint on the psychological reality of cognitive abilities that I termed
the Acquisition Constraint:
The Acquisition Constraint If a given cognitive capacity is psychologi-
cally real, then there must be an explanation of how it is possible for an
individual in the normal course of human development to acquire that
cognitive capacity.
The Acquisition Constraint is not a very strong constraint. It does not
require that any account be provided of how the cognitive ability in ques-
tion is actually acquired. Nor in fact does it require that any such account
be epistemically available. What it does provide is a negative test. If, for
any designated cognitive ability, there are good reasons to think that it
cannot be acquired in the normal course of human development, then its
psychological reality is cast in doubt.
Of course, the force of the Acquisition Constraint depends upon how
the notion of acquisition is to be understood. After all, one way in which
the Acquisition Constraint could be satisfied for a given cognitive capacity
would be if it turned out that the emergence of that cognitive capacity
was due to the maturation of an innate module. One might be forgiven
for feeling that this makes the Acquisition Constraint too weak a tool.
With this apparently catch-all way of potentially satisfying the Acquisi-
tion Constraint, how could anything fail to meet it? But what makes mat-
ters slightly simpler here is that it is often part of our understanding of a
given psychological capacity that it should have been acquired in a certain
way. So when we ask whether a given cognitive capacity is psychologically
real, the specification of that cognitive ability will often include an outline
conception of how it is acquired. This is certainly the case with the mul-
292 Chapter 10
tiple cognitive capacities implicated in language mastery. We understand

these cognitive capacities to have been acquired through a process of
learning, a process that typically starts in the second year of life and con-
tinues for several years afterwards. This conception of how language is
acquired is so deeply embedded in our understanding of language that
the question of how the various cognitive capacities implicated in lan-
guage use meet the Acquisition Constraint is the question of how they are
learned. And should it turn out that one or more of these capacities can-
not be learned then, although it might of course meet the Acquisition
Constraint in some other way, it seems fair to say that the capacity as we
understand it is not psychologically real.
In chapter 1, I gave the following more-detailed account of how, for the
purposes of this book, the Acquisition Constraint was canonically to be
understood. Every individual has an innate set of cognitive capacities that
he possesses at birth. Let me call that set S0. At any given time t after
birth, an individual will have a particular set of cognitive capacities. Let
me call this set St. Now consider a given cognitive capacity c putatively in
St. Suppose that for any time t 2 n the following two conditions are satis-
fied. First, it is conceivable how c could have emerged from capacities
present in St2n. Second, it is conceivable how the capacities present in St2n
could have emerged from the capacities present in S0. By it being conceiv-
able that one capacity could emerge from a given set of capacities, I mean
that it is intelligible that the individual in question could deploy the cogni-
tive capacities it already has to acquire the new capacity. If these condi-
tions are satisfied, then we have a paradigm case of learning.
If mastery of the first-person pronoun is to meet the Acquisition Con-
straint (at least on this canonical construal), then clearly there must be
some time t when St includes the capacity for linguistic mastery of the
first-person pronoun, and a corresponding time t 2 n when St2n does not
include that capacity but includes other capacities on the basis of which
it is intelligible that an individual could acquire the capacity for linguistic
mastery of the first-person pronoun. This creates an apparent circularity,
because any such St2n will have to contain the capacity to think thoughts
with first-person contents, and it is natural to think that the capacity to
think thoughts with first-person contents cannot exist in the absence of
the capacity for linguistic mastery of the first-person pronoun.
Can the problem of capacity circularity be dealt with in a way that

shows how the Acquisition Constraint can be canonically met by linguis-
tic mastery of the first-person pronoun? The solution to this problem is
similar in general form to the solution to the problem of explanatory cir-
cularity discussed in the previous section. The apparent vicious circularity
is removed by showing that the first-person contents implicated in the
complex communicative intention governing mastery of the first-person
pronoun can be understood at the nonconceptual level, and hence inde-
pendently of linguistic mastery of the first-person pronoun. In the terms
used above, although it is true that the capacities present in St2n do indeed
include the capacity to think certain thoughts with first-person contents,
this does not mean that the capacities in St2n must also include linguistic
mastery of the first-person pronoun, because the crucial thoughts in-
volved in learning how to use the first-person pronoun can be noncon-
ceptual and hence independent of language mastery. This emerges
particularly clearly if the neo-Gricean specification of the communicative
intent governing the correct use of the first-person pronoun is read, as I
suggested in section 10.2, as offering conditions on learning the proper
use of the token-reflexive rule. If that suggestion is accepted, then we
have, first, a clear specification of a set of first-person thoughts that must
be grasped by anybody who successfully learns the first-person pronoun
and, second, an illustration of how those first-person thoughts can be
nonconceptual. This is sufficient, I think, to remove the threat posed by
capacity circularity, because it dispels the idea that the first-person
thoughts required for the acquisition of mastery of the first-person pro-
noun could not exist unless mastery of the first-person pronoun was al-
ready in play.
I have made various suggestions about how the sophisticated but none-
theless nonconceptual first-person contents required to explain mastery
of the first-person pronoun emerge from the primitive first-person con-
tents in an individual’s innate endowment. Some of these suggestions are
best viewed as selective rational reconstructions of how this process might
take place. One example would be the discussion at the end of chapter 8
of how an integrated representation of the spatial environment might
emerge from the propriospecific information present in the structure of
visual perception. I suggested there that the calibration of what Gibson
294 Chapter 10
calls affordances through recognition of affordance symmetries, af-

fordance transitivities, and affordance identities could bridge the gap
between the perception of individual spatial relations and the construc-
tion of an integrated representation of the environment. I certainly do
not claim that the rational reconstructions I have offered are complete or
anything near it. Nor do I claim that the various suggestions I have made
about how certain cognitive abilities might conceivably emerge from oth-
ers must feature in a true account of this emergence. What I do claim to
have done, however, is to have dispelled worries that there are principled
reasons for thinking that no such account will be forthcoming.
10.4 The Way Forward
The principal aim of this book has been to offer a solution to the paradox
of self-consciousness by illuminating the various different forms of non-
conceptual self-consciousness that are both logically and ontogenetically
more primitive than the higher forms of self-consciousness that are more
usually the focus of philosophical debate. What philosophers tend to
think of as the unitary phenomenon of self-consciousness has a complex
and multilayered structure, which is all too often neglected not least
because so much of it lies concealed beneath the surface of language. I
have tried to compensate by exposing the nonconceptual foundations
on which the structure rests. A full account of the structure of self-
consciousness, however, will need to illuminate those higher, conceptual
forms of self-consciousness to which I have devoted little attention in this
book. Let me end this book with a few programmatic comments about
the form that I believe such an account will take and about how it might
emerge from some of the points that I have made in this book.
In chapter 1, I considered what I termed the deflationary theory of self-
consciousness. The starting point of the deflationary theory is the thought
that an explanation of everything that is distinctive about self-con-
sciousness will emerge out of an account of what it is for a subject to be
capable of thinking about himself. The conceptual forms of self-
consciousness—specimens of which are the abilities to entertain autobio-
graphical memories, to make plans for the future, and to formulate
second-order desires (desires about desires that one should have)—can all
be understood as involving propositional attitudes to first-person concep-

tual contents. Deflationary theorists are rightly impressed by the thought
that all of the conceptual abilities deployed in such first-person contents
are, with a single exception, also deployable with the same sense in con-
tents that are not first-personal. Since illumination of the distinctive fea-
tures of self-consciousness will not emerge from considering those
conceptual abilities that can feature indifferently in contents that are first-
personal and contents that are not, the deflationary theorist is driven to
the thought that an explanation of mastery of the first-person concept is
all that is required for a theoretical elucidation of the conceptual forms
of self-conscious thought.
If the deflationary theory is correct, then it might be natural to think
that an account of mastery of the first-person pronoun (and hence of
what it is to possess the first-person concept) based upon the discussion
in this book could come close to providing a complete philosophical eluci-
dation of self-consciousness. Sadly, however, this would be premature.
The deflationary theory rests on insights that have an important role to
play in the development of a complete account of the higher, conceptual
forms of self-consciousness, but it would be a grave error to think that
such an account will be so easily forthcoming.
Let me grant the first part of the deflationary position, namely that an
account of the higher forms of self-consciousness will be an account of
the varieties of self-conscious thought at the conceptual level. It is natural
to agree with the deflationary theory that such an account will have two
components. The first component will elucidate the contents of first-
person thoughts at the conceptual level, while the second component will
elucidate the nature of the complex propositional attitudes (such as mem-
ory and desire) that a thinker can bear to those first-person thoughts. The
deflationary theory takes its stand on a clear demarcation between these
two components, viewing the second component as the business not of
the theory of self-consciousness but of a general theory of propositional
attitudes and the first component as exhausted by an account of what it
is to possess the first-person concept in conjunction with a general ac-
count of concepts.
It is certainly true that there are no propositional attitudes that can be
taken only to first-person contents. Autobiographical memory, for ex-
296 Chapter 10
ample, is best understood not as a distinctively first-person propositional

attitude but rather in terms of a distinctive set of first-person contents to
which the attitude of remembering can be taken. By the same token,
second-order desires are best understood in terms of a distinctive set of
first-person contents to which the attitude of desire can be taken. None-
theless, in both of these cases there are crucial facts, indeed defining facts,
about the distinctive set of first-person contents to which the relevant atti-
tude can be taken that cannot be captured by the deflationary theory.
Part of what makes something an autobiographical memory is that it
stands in certain inferential relations to other autobiographical memories
and to other first-person contents. Similarly, second-order desires have a
distinctive role to play in practical reasoning that is also capturable in
terms of the inferential relations in which they stand and the dispositions
(both cognitive and behavioral) to which they give rise. More generally,
each of the higher, conceptual forms of self-consciousness is partially in-
dividuated by a different set of inferential patterns and dispositions to
behavior that govern any first-person content to which that attitude is
taken. Now, as suggested earlier, these inferential patterns and behavioral
dispositions must be described at the level of content rather than the level
of attitude. What makes memories autobiographical memories, or desires
second-order desires, is simply that they are memories or desires with
distinctive first-person contents. Of course, for the very reasons at the
heart of the deflationary theory, it is at the level of the first-person concept
that we will need to explain what this distinctiveness amounts to and how
it emerges. This is quite compatible with the deflationary theorist’s central
tenet that an account of the first-person concept is the key to explaining
the conceptual forms of self-consciousness. On the other hand, it seems
to be clearly incompatible with the deflationary theorist’s proposal for
implementing that central tenet, namely that an account of the first-
person concept will be derivable from an account of linguistic mastery of
the first-person pronoun. There are no facts about linguistic mastery of
the first-person pronoun that will determine or explain what might, fol-
lowing David Kaplan, be termed the cognitive dynamics of the first-
person concept.
The way forward for a theory of self-consciousness, it seems to me, is,
first, to chart the characteristic features individuating the various distinct
conceptual forms of self-consciousness in a way that will provide a taxon-

omy of self-consciousness and then, second, to show how these character-
istic features can be determined at the level of content. What I hope is
now clear is that these higher forms of self-consciousness emerge from a
rich foundation of nonconceptual first-person thought, which I have tried
to expose and clarify in this book. It is my firm conviction that these
different forms of nonconceptual first-person thought hold the key, not
just to an eventual account of how mastery of the first-person pronoun
meets the Acquisition Constraint, but to a proper understanding of the
complex phenomenon of self-consciousness.
Notes
Chapter 1
1. Pears 1988, vol. 2, chapter 10, is an illuminating discussion of the background

to Wittgenstein’s thinking about the self in The Blue and Brown Books.
2. This is still not quite right. A further technical problem is noted in Evans 1982,
pp. 189–191, but need not detain us here.
3. A more involved argument for a similar conclusion can be found in Noonan
1979.
4. Compare Shoemaker 1968, 9–10. He amends his view in 1970, nn. 3 and 5.
His claim in the latter work is that one can make self-reference immune to error
through misidentification with names and definite descriptions, provided that in
using them the speaker intends to refer to himself. This is a valid point, but a
circumscribed one, since it seems that intending to refer to oneself via a name or
definite description can only be done if certain further thoughts are presupposed,
and those further thoughts will be first-personal and involve identifying oneself
as the referent of that name or definite description.
5. What I am calling the deflationary account of self-consciousness has not been
adopted as such in the literature. It is, however, suggested in Shoemaker 1968,
subject to the qualifications in n. 4 above.
6. It will be clear from what follows that these problems are not peculiar to the
deflationary theory. On the contrary, as I will bring out at the end of the chapter,
they fall naturally out of a set of assumptions that are intuitively highly plausible.
I am focusing on the deflationary theory here because the paradox of self-
consciousness is best appreciated within the confines of a theory of self-
consciousness and the deflationary theory is relatively straightforward to ex-
pound, as well as having a certain intuitive plausibility. Although this book will
offer a solution to the paradox of self-consciousness, this solution is not intended
to rehabilitate the deflationary theory. I shall return to the deflationary theory in
section 10.4.
7. It is natural to relativize specifications of the token-reflexive rule to the present
tense. This clearly holds for linguistic utterances of ‘I’. It does not hold for written
300 Notes to Pages 14–22
inscriptions, however. Adequately specifying the temporal dimension of self-

reference will doubtless require some ingenuity. For present purposes, however,
the various rough-and-ready formulations I offer will suffice.
8. For a stimulating discussion of how such an explanation might be derived, see
Altham 1979. The thought that the token-reflexive rule cannot give the meaning
of ‘I’ as a result of the interdependence between the indirect reflexive pronoun
and the first-person pronoun has been emphasised by Anscombe (1975, 47–48).
Anscombe puts this to work in support of her eccentric claim that the first-person
pronoun is not a referring expression. As far as this claim is concerned, I am fully
in agreement with the careful and perceptive criticisms to be found in Taschek
1985. It should be noted that Taschek has no quarrel with Anscombe’s point
about the circularity of the token-reflexive rule as an explanation of the semanti-
cal role of ‘I’ (1985, 640). I do quarrel with it, however. See pp. 16–17.
9. This is the formulation favoured by Campbell (1994, 102). I have discussed
Campbell’s use of this rule in Bermu´dez 1995a.
10. Compare Nozick 1981, 79–84.
11. All that my argument requires at this point is that such knowledge should be
a necessary condition for mastery of the semantics of the first-person pronoun.
In chapter 10, where I give a full account of how such mastery is to be understood,
it will emerge that it is not a sufficient condition.
12. See Peacocke 1979 for a discussion of several such local holisms. Grice and
Strawson (1956) defend what they see as a local holism in the theory of meaning
against what is effectively a charge of explanatory circularity in Quine 1951. Hur-
ley, forthcoming, is a sustained examination of the local holism formed by percep-
tion and action.
13. I discuss the question of how philosophy and psychology might profitably
interact in Bermu´dez 1995d and criticize some of the grounds for which philoso-
´
phers try to keep them apart in Bermudez 1995b.
14. The sense of ‘module’ here is importantly different from that pioneered in
Fodor 1983. Among their other defining characteristics, Fodor’s modules operate
at the input-output level, are informationally encapsulated, and are domain-
specific. None of these features hold of the modules proposed to explain how
language mastery and understanding of other minds meet the acquisition con-
straint. See Karmiloff-Smith 1992 and Segal 1996 for further discussion.
15. The linguistic arguments in favor of the nativist theory are reviewed in Pinker
1979. For a recent and philosophically driven presentation of Chomsky’s view,
see Chomsky 1995. For further philosophical discussion, see the essays in
George 1989.
16. The debate was opened in the context of chimpanzees in Premack and Wood-
ruff 1978. The idea that the psychological abilities that emerge between the ages
of 3 and 4 years should be explained in terms of a theory-of-mind module
(ToMM) was suggested in Leslie 1987 and has recently been vigourously argued
in Baron-Cohen 1993. Baron-Cohen also suggests a possible ontogenesis for
Notes to Pages 34–61 301
ToMM in terms of more primitive modules, such as the eye-direction-detector

module (EDDM) and the shared-attention module (SAM). Karmiloff-Smith 1992
provides a sustained critique of such extreme nativist theories, both in the area of
children’s theory of mind and in the area of language acquisition.
Chapter 2
1. The account I favor does not actually speak of the contents of belief as com-
posed of concepts. Rather, it maintains that belief contents (and the contents of
propositional attitudes in general) are conceptual, where a conceptual content is
one that cannot be ascribed to a creature unless that creature possesses the con-
cepts involved in specifying that content. See chapter 3.
2. Perry 1979 provides the classic discussion of the relation between first-person
thoughts and action.
3. And indeed, the account that I shall give below is a broadly functional account.
4. As I shall stress in chapter 4, there are different degrees of structure associated
with different types of representations. I shall defend there the view that all genu-
ine representation of the world involves a minimal degree of structure. This needs
to be kept firmly distinguished, however, from the view that all genuine represen-
tations involve the sort of structure to be found in natural languages. It is certainly
not true that all genuine representations permit the sort of universal and general-
ized recombinability characteristic of natural language. That view would be hard
to maintain in conjunction with a rejection of the thought-language principle.
Chapter 3
1. The original impetus came from Evans 1982. It is important to note, though,
that Evans’s understanding of nonconceptual content is distinctive. He holds that
it applies to perceptual-information states, which do not become conscious per-
ceptual experiences until they are engaged with a concept-applying and reasoning
system (1982, 226–227).
2. This view doesn’t follow from the preceding, since I have said nothing to rule
out the possibility that perceptual beliefs might have nonconceptual contents.
3. I take it that there are comparable normative dimensions for the propositional
attitudes other than belief.
4. This is what I tried to capture by including the Conceptual-Requirement Prin-
ciple as part of the classical view of content.
5. There seem, however, to be possible versions of developmental explanation
that are compatible with denying the Autonomy Principle. For example, it could
be argued that, although in general nonconceptual and conceptual contents form
a holism, there are certain conceptual contents that develop out of nonconcep-
tual contents.
6. This is Peacocke’s own position. I have criticised it in Bermu´dez 1994, and he

replies in Peacocke 1994.
7. I shall return to this in the final section of this chapter.
8. This is stressed in Peacocke 1992.
9. This is certainly Spelke’s own view: “The earliest developing conceptions of
physical objects are the most central conceptions guiding mature object percep-
tion and physical reasoning. For adults, such conceptions are overlaid by a wealth
of knowledge about the appearances and the behavior of particular kinds of ob-
jects. Even this more specific and limited knowledge, however, reflects the core
knowledge from which it grew” (Spelke and Van de Walle 1993, 157).
10. It is worth mentioning in this context that, although developmental psychol-
ogists seem to find unproblematic the conclusion that infants show surprise in the
drawbridge experiments because it is impossible for the screen to pass through
the object, this is at best questionable. On the inferential interpretation, the sur-
prise would be due to the infant’s identifying an object and inferring from their
knowledge of how objects behave that what they see is impossible. But why say
this rather than that the infants show surprise simply because there are no prece-
dents for what they see? Certainly, more argument is needed than simply pointing
out that what the infants see is really impossible.
11. I am grateful to Jim Russell for pressing me here.
12. Some qualifications are required here. It is known from studies of adult sub-
jects, for example, that perceptual predictions of object transformations and in-
teractions tend to be based on kinematic, rather than dynamic, principles and
information (Cooper and Munger 1993). This has obvious computational advan-
tages and clearly shows that ordinary object perception need not always involve
the perceptual pick-up of the full range of object properties. But that is no objec-
tion to the claim that perception of objects depends on perceptual sensitivity to
object properties. Perceptual sensitivity to object properties needs to be under-
stood at the level of competence rather than performance.
13. The experiments on infant expectations in collision events reported in Bail-
largeon 1995 are an excellent example of how detailed work might be carried out
on infant sensitivity to object properties.
14. This would be consistent with the general view of the importance of agency
in mental development outlined in Russell 1996, and is of course a modern ver-
sion of a broadly Piagetian position.
15. The inference to an apelike (pongid) neural organization is based on the loca-
tion of the lunate sulcus, on the assumption that the massively expanded posterior
parietal cortex characteristic of the human brain pushed the lunate sulcus back-
wards and downwards to the position it occupies in the human brain.
16. It has occasionally been suggested in the literature that Neanderthal homi-
nids had a completely modern superlaryngeal vocal tract. Liebermann (1991)
compellingly argues that this would have meant that Neanderthal’s larynx was in
his chest.
17. The crucial claim is, of course, that the emergence of language did not sig-
nificantly precede the descent of the larynx. This presupposes an equation of lan-
guage with speech. Some authors have objected to this, suggesting that a gestural
language might have preceded spoken language (Corballis 1991), or that Homo
erectus had a protolanguage comparable to that of a 2-year-old child (Parker and
Gibson 1979). Both these suggestions remain at the level of conjecture, however.
18. Farah (1990) distinguishes between two types of simultagnosia, which she
terms dorsal and ventral (after the location of the relevant lesion site). Although
ventral simultagnosics cannot recognize multiple objects, they can nonetheless see
multiple objects (unlike dorsal simultagnosics).
19. Farah (1990) questions whether patients with associative object agnosias re-
ally do have intact visual perception, noting that they often fail more sophisticated
visual tests. Nonetheless, they generally seem to have relatively unimpaired visual
perception in comparison with apperceptive agnosics, which seems enough to
suggest a significant distinction between the two groups of disorders.
20. See Shallice 1988, section 8.4, for an illuminating discussion of the connec-
tions between Marr’s theory and relevant neuropsychological findings.
21. This is not to deny that the 3D sketch is itself a content-bearing state. See
Bermu´dez 1995e for further discussion.
Chapter 4
1. The minimal account is an account of what it is in general for a state to have

content. This is a different level of theorizing from specific theories of content. A
possible-worlds account of content is an example of such a specific theory, as is
Peacocke’s own theory of scenario content. The general theory explains what the
specific theories have in common.
2. The importance of an exceptionless connection for the theory of information
is stressed in Dretske 1981, chap. 3. He gives several arguments for refusing to
accept conditional probabilities of less than 1. The most important are the follow-
ing. First, doing otherwise would admit the possibility that a signal could carry
the information that s is G and the information that s is F, but not the information
that s is G and F (because the probability of s being F and G might be below the
acceptable value of conditional probability). Second, the flow of information is
transitive and that this transitivity would be lost if the value of conditional proba-
bility were less than one.
3. It is this that led Dretske substantially to modify the theory proposed in Dret-
ske 1981. See Dretske 1990 and chapter 3 of Dretske 1992.
4. This is central to J. J. Gibson’s notion of the direct perception of affordances,
which will be discussed in the next chapter.
5. Although the number of approaches to the magazines for rats in the omis-
sion schedule eventually fell below the level for the control group, for which the
instrumental contingency was maintained, it remained significantly higher than

the level for a second control group, for which food and tone were unpaired.
6. This is noted in Peacocke 1983, 61.
7. In fact, this is precisely the form that Bennett adopts for the teleological laws
that he thinks explain much animal behavior. The form of a teleological law for
Bennett is this:
(x)(F )(t)((Rx & x registers that F /Gx at time t) → Fx at time t 1 d)
Rx is to be read as the claim that suitable enabling conditions are satisfied (includ-
ing the animal’s having the satisfaction of G as a goal); F /Gx as the claim that x
is so structured that the truth of Fx at a later time (t 1 d) is causally sufficient for
Gx at a still later time.
8. The core of Smolensky’s proposal is that compositional structure can be cap-
tured within a network if the network represents particular “fillers” (e.g., a noun)
occupying particular roles (e.g., the role of a grammatical object). But there are
other ways in which PDP researchers have suggested that the compositionality
requirement might be satisfied without collapsing into CTC. Ramsey (1992), for
example, describes a class of connectionist networks in which the different syn-
tactic roles of particular atomic units are encoded via slightly different patterns
of activation. These different patterns of activation legitimate different types of
molecular combination.
Chapter 5
1. Although Gibson himself displayed little interest in the matter, there are impor-
tant processing questions about the relation between retinal flow and flow in the
optic array. See Harris, Freeman, and Willis 1992 for a framework within which
these questions might be addressed.
2. An interesting illustration of the explanatory powers of the theory of invari-
ants comes with the traditional problem of size constancy. Depending on one’s
distance from an object, the solid visual angle that it subtends will vary. So why
does the object not seem smaller or larger, depending on one’s distance from it?
There is a perceptual invariant that might well explain why objects are perceived
as being of a constant size, irrespective of their distance from the perceiver. The
relevant invariant is the ratio of an object’s height to the distance between its base
and the horizon. This ratio remains constant irrespective of the object’s distance
from the perceiver.
3. This is, of course, one of the ways in which Gibson himself describes it.
4. The efference copy is a perception of the motor command as it leaves the mo-
tor apparatus. The efference copy is usually distinguished from what is known as
the corollary discharge (Gallistel 1980, chap. 7). ‘Efference copy’ and ‘corollary
discharge’ both refer to the same signal, but there are two different theories about
the role that the signal plays in controlling action, and in particular in allowing
the organism to distinguish between exafferent signals (due to the motion of the
environment) from reafferent signals (due to the organism’s own motion). Ac-
cording to corollary-discharge theory, the copy of the motor signal directly inhib-
its the action of any functional unit (such as the optokinetic unit) that would
compensate for reafference by blocking the relevant action (as would an un-
checked optokinetic reaction). In efference-copy theory, on the other hand, a copy
of the motor signal is passed on by the relevant functional unit (e.g., the optoki-
netic unit) as if it were a command from that unit. This allows the efference copy
and the reafference to cancel each other out. As Gallistel notes, however, the dif-
ference between the two theories emerges only when reafference is abnormal.
5. For interesting and provocative discussion of some of the ontological implica-
tions of ecological optics, see Reed 1987.
6. See also Rock 1975, 124–129.
7. See further the experimental evidence reviewed in the next section.
8. It is important to recognize the differences between misperception and percep-
tual illusions. In addition to the familiar examples of visual illusions, like the
waterfall illusion, there are well-documented cases of artificially induced disorders
of bodily perception in which subjects report experiencing limbs in anatomically
and physically impossible positions (Lackner 1988). These are fundamentally dif-
ferent from the misperception of affordances, because there is no way in which
they involve the pick-up of misinformation. In neither the proprioceptive cases
nor the visual cases is there a direct pick-up of information from the optical flow.
The perceptual systems are confronted with artificially induced sensory input that
they are forced to make sense of in any way they can.
9. The source for this is, of course, Perry 1979. For an opposed view, see Milli-
kan 1990.
10. I am referring here to the level of autonomous nonconceptual content, in the
sense brought out in chapter 3.
11. I will discuss more complicated forms of instrumental protobeliefs in chapter
7. These emerge with the advent of memory.
12. See also Neisser 1991, Butterworth and Hicks 1990, and the papers collected
in Neisser 1993.
13. Comprehensive reviews will be found in the essays collected in Neisser 1993
and Rochat 1995.
14. I have discussed the philosophical significance of neonatal imitation from the
viewpoint of applied ethics in my 1996 essay.
15. Through the course of their development there are interesting variations in
the extent to which infants are susceptible to discrepant visual information. Expe-
rience in sitting and crawling decreases susceptibility. See Butterworth and Hicks
1990 for further discussion.
16. This is not to rule out, though, the possibility that the first-person contents
of perceptual experience can be understood in terms of the states of visual-
processing subsystems in the brain, and that the states of such subsystems might
be representational and have content (Bermu´dez 1995e). My suggestion is simply
that these subpersonal representational states, despite having contents, will not
have first-person contents.
Chapter 6
1. I am following closely the general introduction to Bermu´dez, Marcel, and

Eilan 1995.
2. Such reasons are discussed by Anscombe (1962), who develops a position ulti-
mately derived from Wittgenstein (see Budd 1989).
3. In this and the following I will drop the explicit restriction to conscious
somatic proprioception. Unless there is any indication to the contrary, I will
use ‘somatic proprioception’ to refer only to the conscious modes of somatic
proprioception.
4. Although Shoemaker himself does not explicitly consider somatic propriocep-
tion, he does explicitly argue for the Humean conclusion that the self cannot be
the object of any form of inner perception, as in the object-perception model.
5. When Shoemaker continues his discussion of the object constraint on pp. 256–
257, he modulates into what I am calling the multiple-objects constraint by defin-
ing a notion of ‘object awareness’ in such a way that it must be possible for one
to bear it to a range of different objects.
6. It is important to distinguish exploratory or active touch from passive touch.
It is well known that the tactile discrimination of differently textured surfaces is
greatly increased when the finger moves relative to the surface, but it does not
seem to matter whether the movement is carried out by the subject or not. For
more complicated discrimination tasks, such as the discrimination of three-
dimensional shapes and the determination of their structural and material proper-
ties, however, active movement is essential (Gordon 1978, Lederman and Klat-
zky 1987).
7. The physiology of exploratory or active touch has been well researched, at
least with respect to the hand (Gordon 1978). Active touch depends on the joint
processing of proprioceptive and exteroceptive information. Proprioceptive infor-
mation about the movement of the hand is required so that predictable informa-
tion stemming from the subject’s own movement can be separated out from the
novel information about the contours of the object.
8. Of course, this is subject to the proviso that the behavior really is intentional.
´
For further discussion, see Meltzoff 1993 and Bermudez 1996.
9. It is not yet known whether the process of coordinate transformation proceeds
in a hierarchical fashion, with each representation actually computed in increas-
ing order of complexity, or whether there is a single highly distributed body-
centered representation that can integrate signals with different coordinate
frames.
10. Moreover, as shown by the movement errors discussed in Ghez et al. 1995,
visual information about initial limb position (and not just the position of the
target) has an essential role to play in controlling action. Both visual and proprio-
ceptive information are integrated at every stage of visually guided reaching.
11. For further discussion, see the papers in part 1 of Eilan, McCarthy, and
Brewer 1993.
12. Of course, it is possible to imagine complicated scenarios in which one’s pro-
prioceptive system is hooked up to another’s body so that misidentification is pos-
sible. This is explored in Armstrong 1984 and Cassam 1995. As Cassam notes, if
this really is a coherent possibility, then there is no logical necessity about immu-
nity to error. But it is not clear why one would expect logical necessity for some-
thing that is ultimately determined by contingent facts about the human body.
The sense of necessity relevant here is nomological necessity (i.e., truth in all pos-
sible worlds with the same laws of nature as this world), and so we will have to
demonstrate that any proposed counterexample to the modal claim is nomologi-
cally possible rather than merely logically possible. But how is one to decide
whether being hooked up to somebody else’s body so as to feel their sensations
really is nomologically possible? How would one convince those who are sure
that it is not? In any case, though, and this is the second point, it is far from clear
that the modal claim really is so important. If Armstrong-type scenarios are in-
deed nomologically possible, and were it ever to come about that they are more
common than the sort of somatic proprioception under consideration, in which
there is de facto only one object of awareness, then clearly we would need to
develop both a new theory of somatic proprioception and a new theory of self-
awareness. But there seems little reason why that should influence how we decide
the question of whether somatic proprioception, in the form that we are familiar
with, can be a form of self-awareness, also in the form that we are familiar with.
Let me weaken the claim about immunity to accommodate Armstrong-style sce-
narios as follows: necessarily, in a body with unsupplemented somatic proprio-
ceptive information systems, there is only one object of somatic proprioception,
which is the body itself. This is quite enough to be getting on with.
13. Something like this conflation appears in Evans 1982, 237.
14. Let me stress, though, that my way of spelling out this connection is rather
different from Shoemaker’s or Evans’s. See n. 16 below.
15. The only counter to this would be via the claim that psychological properties
somehow manifest themselves to introspection in an explicitly first-person way,
so that it is impossible to introspect in the manner that (3) attempts to capture.
Something like this view is defended in Chisholm 1969.
16. This is a good place to distinguish my use of the notion of immunity to error
through misidentification as a criterion of self-awareness from Evans’s. Evans
treats the first-person pronoun on the model of a perceptual demonstrative. He
tries to show that demonstrative self-reference achieved with ‘I’ depends on epis-
temological links with one’s embodied self that are comparable to the percep-
tual links that make possible demonstrative reference to perceived objects. The
epistemological links are provided by information channels, one of which is what

I have been terming somatic proprioception. Evans identifies the relevant informa-
tion channels as all and only those channels that give rise to judgments immune
to error through misidentification relative to the first-person pronoun. There are
certain features of Evans’s position that sharply distinguish it from mine, how-
ever. The first is that he denies that genuine self-awareness is available in the ab-
sence of mastery of the first-person pronoun. Moreover, and this is perhaps part
of the explanation for this denial, he thinks that states of the somatic propriocep-
tive system (and indeed information-carrying states generally) do not qualify as
experiences until conceptual and reasoning abilities are brought to bear upon
them. Against this I am arguing that somatic proprioception (as opposed to judg-
ments based on proprioceptive information channels) is itself a source of self-
awareness. As will be made clear in section 6.5, I find it perfectly intelligible, and
indeed necessary, to talk of proprioceptive experiences in the absence of con-
ceptual and reasoning abilities, and it is these experiences (not the first-person
judgments that might or might not be based on them) that are the sources of
self-awareness.
17. Ramachandran suggests that the referral of sensations should be explained
through what he terms the “remapping hypothesis,” namely that the area in the
somatosensory cortex originally responsible for processing sensory information
from the amputated limb becomes reorganized and sensitive to sensory informa-
tion from parts of the body originally processed by adjacent areas of the somato-
sensory cortex. It is well known that adjacent areas in the somatosensory cortex
do not represent adjacent body parts. This explains why there are well-defined
distributions of points on the body surface that yield referred sensations in ampu-
tated limbs.
18. See the essays in Prigatano and Schacter 1991 for further discussion.
19. See the papers cited in n. 11 for further discussion.
20. A form of this proposal is to be found in Peacocke 1992, chap. 3.
21. Nothing hangs on the vocabulary of pains, or other bodily sensations, ex-
isting at particular places. I am not assuming that there are such things as pains
that are the objects of somatic proprioception, as will become clear in my account
of the descriptive aspect of proprioceptive content.
22. I shall henceforth use ‘moveable’ in this limited sense. The sense in which the
torso is immoveable is similarly restricted. Of course, I can move my torso by
leaning forward or by turning at the waist, but the range of movement is very
limited.
23. Strictly speaking, of course, one of these specifications will be sufficient to
fix the location. This doesn’t mean, however, that only one specification need fea-
ture in the frame of reference. Compare the situation of twenty ships moving in
convoy flanked by four tugs, which keep the same position relative to each other.
The fact that the location of any given ship can be fixed by its distance and direc-
tion from any one of the tugs is compatible with all four tugs’ constituting the
frame of reference.
24. Certain documented phenomena prima facie seem hard to reconcile with the
description of the phenomenology of bodily experience just made, and hence with
the account of the spatial content of proprioception that I have been offering.
Some subjects have reported feeling sensations in space or even in the bodies of
´ ´
others (Shapiro, Fink, and Bender 1952; Bekesy 1967). See Martin 1993 for an
interesting discussion. Can phenomena like this be accommodated in terms of A
locations and B locations? Clearly, they cannot have A locations, in the sense of
being experienced as being located in identifiable body parts (unlike, for example,
pains that are felt in phantom limbs). But they might plausibly be ascribed B
locations. As Martin notes, exosomesthetic sensations tend to be felt in exten-
sions of the body, albeit indeterminate ones. This suggests that it might make
sense to attribute to them a B location based upon their coordinates relative to the
nearest hinges. Of course, this would be radically different from the ascriptions of
ordinary B locations, but this might be seen as reflecting the bizarre phenomenol-
ogy of the situation.
25. Let me stress that these are not intended to be constitutive accounts. The
proposal is not that actions, or even dispositions to act, constitute proprioceptive
content in that the correctness conditions are given by the success or failure of the
relevant actions. That would have the unacceptable consequence that states with
proprioceptive content would in principle be unavailable to a paraplegic. The rep-
resentational states that cause the relevant actions are what have correctness con-
ditions, and those correctness conditions are bodily events. The actions provide
illustrations of how one might recognize when and if the correctness conditions
are satisfied.
26. In fact, they may even be available before birth. See the evidence discussed in
Gallagher 1996.
Chapter 7
1. Note, moreover, that ‘object*’ qualifies as a count term rather than a mass
term.
2. The point is that certain terms that we employ as count terms to individuate
particular objects can have a more primitive deployment as mass terms. As Camp-
bell puts the point, “There may be a use of ‘tiger’ as a mass term which is prior
to its use as a count noun. This use of ‘Tiger!’ would be merely a response to the
presence of tigerhood, by someone who might be quite incapable of making the
distinction between one tiger and two being present, or having the idea of its
being the same tiger again as was here previously” (1993, 65).
3. This primitive form of place recognition falls significantly short of the capacity
to identify places in a way that would reflect a more general understanding of
space. In the next chapter I discuss how primitive place recognition might be built
up into this more sophisticated way of thinking about space.
4. This is just one aspect of a more general distinction between conscious and
nonconscious information processing that can be observed in various neuro-
psychological disorders, such as blindsight and prosopagnosia, as well as in

associative-priming experiments on normal patients. A general survey of the neu-
ropsychological issues will be found in Schacter, McAndrews, and Moscovitch
1988. For associative priming, see Marcel 1983.
5. Campbell himself makes such a distinction (1994, 236).
6. For some recent discussions of psychological theories of concept acquisition,
and in particular the turn away from prototype theories, see the essays collected
in Neisser 1987.
7. Compare Price 1953, chap. 2. It is also, I think, Kant’s view that primitive
recognition is at the root of all cognition. This seems to me to be the upshot of
the discussion of the threefold synthesis in the transcendental deduction of the
categories (see Kant 1781/1964, A98–A110).
8. Bennett does not offer an argument in support of this assumption, but it could
be defended along the following lines. Only if the two cognitive abilities could be
separately manifested would it be possible to come up with a determinate answer
to the question of which particular ability is implicated in a given situation.
9. There are some difficulties in disentangling Gibson’s defensible points from his
untenable diatribes against information-processing accounts of perception (Fodor
and Pylyshyn 1981). It is clear that what Gibson notes about “direct perception”
does not mean that there is no processing at all going on, although that appeared
to be his own view. Some explanation needs to be given at the neurophysiological
and other subpersonal levels of how the organism can be sensitive to perceptual
invariants. Useful suggestions for how Gibson’s insights can be accommodated
within a more traditional information-processing theory will be found in McCar-
thy 1993.
10. This would be a weak sense of what has come to be known in the literature
as phenomenal consciousness, as opposed to access consciousness. See the Intro-
duction to Davies and Humphreys 1993 for an explanation of the distinction.
11. This is one of the points at which ontogeny and phylogeny come apart, since
there is a pressing need to give an explanation of why and, more importantly, how
the capacity to feel sensations should have emerged during evolution.
Chapter 8
1. See Peacocke 1992, 90–92. This argument is offered in opposition to the Au-
tonomy Principle, discussed in chapter 3 above. My 1994 essay argues that this
argument cannot be satisfactorily deployed against the Autonomy Principle. Pea-
cocke replies to this argument in 1994.
2. There is further discussion of Campbell’s book in my 1995a and 1997 essays.
3. Note that this means that Peacocke’s original notion of perspectival sensitivity
would not be applicable here.
Chapter 9
1. I am describing predicates as having a first-person application when the sub-

ject using them is capable of understanding that they can apply to himself. Simi-
larly with second- and third-person predicates. By extension, therefore, a
predicate will have only a first-person application when the subject using it is
capable only of understanding that it applies to himself.
2. This is very relevant to some of the points made against Evans’s development
of the Generality Constraint in an acute paper by Charles Travis (1994). Travis
objects that satisfying Evans’s Generality Constraint for a and F is not a necessary
condition of being able to think the thought that a is F, because thinking the
thought that a is F requires a degree of context sensitivity that a subject may not
have for the thought that b is F. This point is well taken. Nonetheless, it doesn’t
affect the more general point about the unacceptability of uniquely instantiable
predicates that inspires Evans and Strawson.
3. A source of confusion is that although Strawson makes these points in dis-
cussing an objection to his own position, they are claims that he endorses. The
passage continues “but not necessarily that one should do so on any occasion”—
it is this continuation that Strawson has qualms about.
4. See Wiggins 1980 for an extended discussion and Mackie 1976, 160–161, for
a more concise one.
5. Nagel 1971 has been very influential in this respect.
6. I am assuming, for simplicity’s sake, that x and y are both perceptually present
to the subject.
7. Of course, if (as I shall argue) the relevant categories can be apprehended and
applied by nonlinguistic and prelinguistic creatures, then the weak and strong
readings of the Symmetry Thesis will need to be reformulated.
8. Note that I am not claiming that the instantiation of any one of these psycho-
logical categories on its own provides strong prima facie evidence for the presence
of a psychological subject. Nor am I ruling out the possibility that a subset of
these psychological categories might provide strong prima facie evidence.
9. See pp. 221–223.
10. It is purely for convenience of exposition that I am treating perception and
action as separate strands. It seems clear that at every level of analysis, from the
neuronal upward, perception and action are inextricably linked. See the essays in
Hurley, 1998, for a sustained defense of the noninstrumental interdependence of
perception and action. She develops a picture of man and other animals as
complex-, dynamic-feedback systems structured by feedback loops that run from
action to perception, as well as from perception to action.
11. I am assuming here the soundness of the argument provided in section 3.3
for the Priority Principle, so that evidence of distinguishing psychological self-
awareness in nonlinguistic or prelinguistic creatures will ipso facto count as
evidence for the possibility of distinguishing psychological self-awareness at the

nonconceptual level.
12. I should note at this point that the current account replaces what I said about
infant’s awareness of their own agency in Bermu´dez 1995c.
13. I now think that an interpretation along these lines is most appropriate for
the experimental data discussed in my 1995c paper.
14. In chapter 4 and Bermu´dez 1996, I suggested that neonatal imitation behav-
ior involved a pick-up of self-specifying information, and hence representations
with first-person contents. Nothing that I said implies that those first-person con-
tents implicated a form of psychological self-awareness. On the contrary, the
properties picked up are purely physical.
Chapter 10
1. In the following I will use ‘utterance’ in a broad sense on which inscriptions

on pieces of paper and computer disks count as utterances. By the same token, I
shall use ‘utterer’ in such a way that one can qualify as an utterer with one’s vocal
cords cut.
2. I take it that failing the false-belief task is not prima facie evidence that the
subject cannot comprehend the possibility of himself representing the world in a
way other than it is.
3. See chapter 3.
4. I will hyphenate ‘bringing-it-about-that’ as a reminder that it is functioning as
a sort of technical term.
References
Aaronson, D., and Rieber, R. W. (eds.). 1975. Developmental Psycholinguistics

and Communication Disorders. New York: New York Academy of Sciences.
Acredolo, L. P., Adams, A., and Goodwyn, S. W. 1984. “The Role of Self-
Produced Movement and Visual Tracking in Infant Spatial Orientation.” Journal
of Experimental Child Psychology 38: 312–327.
Altham, J. E. J. 1979. “Indirect Reflexives and Indirect Speech.” In Diamond and
Teichmann 1979.
Andersen, R. A. 1995. “Coordinate Transformations and Motor Planning in Pos-
terior Parietal Cortex.” In Gazzaniga 1995.
Anscombe, G. E. M. 1962. “On Sensations of Position.” Analysis 22: 55–58.
Anscombe, G. E. M. 1975. “The First Person.” In Guttenplan 1975.
Armstrong, D. M. 1968. A Materialist Theory of the Mind. London: Routledge
and Kegan Paul.
Armstrong, D. M. 1971. “Meaning and Communication.” Philosophical Review
80: 427–447.
Armstrong, D. M. 1984. “Consciousness and Causality.” In Armstrong and Mal-
colm 1984.
Armstrong, D. M., and Malcolm, N. 1984. Consciousness and Causality. Oxford:
Basil Blackwell.
Ayers, M. 1991. Locke: Epistemology and Ontology. London: Routledge.
Baillargeon, R. 1987. “Object Permanence in 3.5- and 4.5-Month-Old Infants.”
Developmental Psychology 23: 655–664.
Baillargeon, R. 1995. “Physical Reasoning in Infancy.” In Gazzaniga 1995.
Bakeman, R., and Adamson, L. B. 1984. “Coordinating Attention to People and
Objects in Mother-Infant and Peer-Infant Interaction.” Child Development 55:
1278–1289.
Baldwin, T. R. 1995. “Objectivity, Causality, and Agency.” In Bermu´dez, Marcel,
and Eilan 1995.
314 References
Baron-Cohen, S. 1993. “From Attention-Goal Psychology to Belief-Desire Psy-

chology: The Development of a Theory of Mind and Its Dysfunction.” In Baron-
Cohen, Tager-Flusberg, and Cohen 1993.
Baron-Cohen, S., Tager-Flusberg, H., and Cohen, D. J. (eds.). 1993. Understand-
ing Other Minds: Perspectives from Autism. Oxford: Oxford University Press.
Barwise, J., and Perry, J. 1981. “Situations and Attitudes.” Journal of Philosophy
78: 669–691.
Bateson, M. C. 1975. “Mother-Infant Exchanges: The Epigenesis of Conversa-
tional Interaction.” In Aaronson and Rieber 1975.
´ ´
Bekesy, G. von. 1967. Sensory Inhibition. Princeton: Princeton University Press.
Bennett, J. 1964. Rationality. London: Routledge and Kegan Paul.
Bennett, J. 1976. Linguistic Behaviour. Cambridge: Cambridge University Press.
Bermu´dez, J. L. 1994. “Peacocke’s Argument against the Autonomy of Noncon-
ceptual Content.” Mind and Language 9: 402–418.
Bermu´dez, J. L. 1995a. “Aspects of the Self.” Inquiry 38: 489–501.
Bermu´dez, J. L. 1995b. “Syntax, Semantics, and Levels of Explanation.” Philo-
sophical Quarterly 45: 361–367.
Bermu´dez, J. L. 1995c. “Ecological Perception and the Notion of a Nonconcep-
´
tual Point of View.” In Bermudez, Marcel, and Eilan 1995.
´
Bermudez, J. L. 1995d. “Transcendental Arguments and Psychology.” Metaphi-
losophy 26: 379–401.
´
Bermudez, J. L. 1995e. “Nonconceptual Content: From Perceptual Experience to
Subpersonal Computational States.” Mind and Language 10: 333–369.
´
Bermudez, J. L. 1996. “The Moral Significance of Birth.” Ethics 106: 378–403.
´
Bermudez, J. L. 1997. “Practical Understanding vs Reflective Understanding.”
Philosophy and Phenomenological Research 57: 635–641.
Bermu´dez, J. L., Marcel, A. J., and Eilan, N. (eds.). 1995. The Body and the Self.
Cambridge: MIT Press.
Boden, M. (ed.). 1990. The Philosophy of Artificial Intelligence. Oxford: Oxford
University Press.
Bower, T. G. R. 1972. “Object Perception in Infants.” Perception 1: 15–30.
Bremner, J. G. 1988. Infancy. Oxford: Basil Blackwell.
Brewer, B. 1993. “The Integration of Spatial Vision and Action.” In Eilan, McCar-
thy, and Brewer 1993.
Brown, J. W. 1988. The Life of the Mind. Hillsdale, N.J.: Laurence Erlbaum.
Bruner, J. S. 1975. “The Ontogenesis of Speech Acts.” Journal of Child Language
2: 1–19.
Bruner, J. S. 1977. “Early Social Interaction and Language Acquisition.” In Schaf-
fer 1977.
References 315
Budd, M. 1989. Wittgenstein’s Philosophy of Psychology. London: Routledge.

Butterworth, G. E., and Hicks, L. 1977. “Visual Proprioception and Postural Sta-
bility in Infancy: A Developmental Study.” Perception 6: 255–262.
Butterworth, G. E., and Hicks, L. 1990. “Self-Perception in Infancy.” In Cicchetti
and Beeghly 1990.
Campbell, J. 1993. “The Role of Physical Objects in Spatial Thinking.” In Eilan,
McCarthy, and Brewer 1993.
Campbell, J. 1994. Past, Space, and Self. Cambridge: MIT Press.
Carey, S. 1982. “Semantic Development: The State of the Art.” In Wanner and
Gleitman 1982.
Carey, S., and Gelman, R. (eds.). 1991. The Epigenesis of Mind: Essays on Biol-
ogy and Cognition. Hillsdale, N.J.: Lawrence Erlbaum.
Carruthers, P., and Smith, P. K. (eds.). 1996. Theories of Theories of Mind. Cam-
bridge: Cambridge University Press.
Cassam, Q. (ed.). 1994. Self-Knowledge. Oxford Readings in Philosophy. Oxford:
Oxford University Press.
´
Cassam, Q. 1995. “Introspection and Bodily Self-Ascription.” In Bermudez, Mar-
cel, and Eilan 1995.
˜
Castaneda, H.-N. 1966. “ ‘He’: A Study in the Logic of Self-Consciousness.”
Ratio 8: 130–157.
˜
Castaneda, H.-N. 1969. “On the Phenomeno-logic of the I.” In Cassam 1994.
Chapuis, N., and Varlet, C. 1987. “Shortcut by Dogs in Natural Surroundings.”
Quarterly Journal of Experimental Psychology 39: 49–64.
Chisholm, R. M. 1969. “On the Observability of the Self.” Philosophy and Phe-
nomenological Research 30: 7–21.
Chomsky, N. 1981. Lectures on Government and Binding Theory. Dordrecht:
Foris Publications.
Chomsky, N. 1995. “Language and Nature.” Mind 104: 1–61.
Cicchetti, D., and Beeghly, M. (eds.). 1990. The Self in Transition: Infancy to
Childhood. Chicago: University of Chicago Press.
Clark, A. 1989. Microcognition. Cambridge: MIT Press.
Cockburn, D. (ed.). 1991. Human Beings. Cambridge: Cambridge University
Press.
Cole, J., and Paillard, J. 1995. “Living without Touch and Peripheral Information
about Body Position and Movement: Studies with Deafferented Subjects.” In Ber-
´
mudez, Marcel, and Eilan 1995.
Cooper, L. A., and Munger, M. P. 1993. “Extrapolating and Remembering Posi-
tions along Cognitive Trajectories: Uses and Limitations of Analogies to Physical
Motion.” In Eilan, McCarthy, and Brewer 1993.
316 References
Corballis, M. 1991. The Lopsided Ape: Evolution of the Generative Mind. Cam-
bridge: Cambridge University Press.
Costall, A., and Still, A. (eds.). 1987. Cognitive Psychology in Question. Brigh-
ton: Harvester Press.
Crane, T. 1988. “The Waterfall Illusion.” Analysis 48: 142–147.
Crane, T. (ed.). 1992. The Contents of Experience. Cambridge: Cambridge Uni-
versity Press.
Cussins, A. 1990. “The Connectionist Construction of Concepts.” In Boden
1990.
Davidson, D. 1975. “Thought and Talk.” In Guttenplan 1975.
Davies, M., and Humphreys, G. (eds.). 1993. Consciousness. Oxford: Basil
Blackwell.
Davis, R. T., and Fitts, S. S. 1976. “Memory and Coding Processes in Discrimina-
tion Learning.” In Medin, Roberts, and Davis 1976.
Davis, S. (ed.). 1992. Connectionism: Theory and Practice. New York: Oxford
University Press.
DeCasper, A. J., and Fifer, W. 1980. “Of Human Bonding: Newborns Prefer
Their Mothers’ Voices.” Science 208: 1174–1176.
Dennett, D. 1981. Brainstorms. Brighton: Harvester Press.
Diamond, A. 1991. “Neuropsychological Insights into the Meaning of Object
Concept Development.” In Carey and Gelman 1991.
Diamond, C., and Teichmann, J. (eds.). 1979. Intention and Intentionality: Es-
says in Honour of G. E. M. Anscombe. Brighton: Harvester Press.
Dickinson, A. 1980. Contemporary Animal Learning Theory. Cambridge: Cam-
bridge University Press.
Dickinson, A. 1988. “Intentionality in Animal Conditioning.” In Weiskrantz
1988.
Donald, M. 1991. Origins of the Modern Mind: Three Stages in the Evolution of
Culture and Cognition. Cambridge: Harvard University Press.
Dretske, F. 1981. Knowledge and the Flow of Information. Cambridge: MIT
Press.
Dretske, F. 1990. “Misrepresentation.” In Lycan 1990.
Dretske, F. 1992. Explaining Behavior: Reasons in a World of Causes. Cam-
bridge: MIT Press.
Dummett, M. 1978. Truth and Other Enigmas. London: Duckworth.
Dummett, M. 1993. The Origins of Analytical Philosophy. London: Duckworth.
´
Eilan, N., Marcel, A. J., and Bermudez, J. L. 1995. “Self-Consciousness and the
´
Body: An Interdisciplinary Introduction.” In Bermudez, Marcel, and Eilan 1995.
Eilan, N., McCarthy, R., and Brewer, M. W. (eds.). 1993. Spatial Representation:
Problems in Philosophy and Psychology. Oxford: Basil Blackwell.
References 317
Evans, G. 1982. The Varieties of Reference. Oxford: Clarendon Press.

Evans, G. 1985. “Molyneux’s Question.” In G. Evans, Collected Papers. Oxford:
Clarendon Press.
Farah, M. J. 1990. Visual Agnosia: Disorders of Object Recognition and What
They Tell Us about Normal Vision. Cambridge: MIT Press.
Field, J. 1976. “Relation of Young Infants’ Reaching Behaviour to Stimulus Dis-
tance and Solidity.” Developmental Psychology 12: 444–448.
Field, T. M., Woodson, R., Greenburg, R., and Cohen, D. 1982. “Discrimination
and Imitation of Facial Expression by Neonates.” Science 218: 179–181.
Fodor, J. A. 1975. The Language of Thought. New York: Crowell.
Fodor, J. A. 1983. The Modularity of Mind. Cambridge: MIT Press.
Fodor, J. A. 1986. “Why Paramecia Don’t Have Mental Representations.” In
French, Uehling, and Wettstein 1986.
Fodor, J. A. 1987. Psychosemantics: The Problem of Meaning in the Philosophy
of Mind. Cambridge: MIT Press.
Fodor, J. A., and Pylyshyn, Z. W. 1981. “How Direct Is Visual Perception? Some
Reflections on Gibson’s Ecological Approach.” Cognition 6:136–196.
Fodor, J. A., and Pylyshyn, Z. W. 1988. “Connectionism and Cognitive Architec-
ture: A Critical Analysis.” Cognition 28: 3–71.
French, P. A., Uehling, T. E., and Wettstein, H. K. (eds.). 1986. Studies in the Phi-
losophy of Mind. Midwest Studies in Philosophy, no. 10. Minneapolis: University
of Minnesota Press.
Gaffan, D. 1977. “Monkeys’ Recognition of Complex Pictures and the Effects of
Fornix Transection.” Quarterly Journal of Experimental Psychology 29:
505–514.
Gallagher, S. 1996. “The Moral Significance of Primitive Self-Consciousness: A
´
Response to Bermudez.” Ethics 107: 129–140.
Gallistel, C. R. 1980. The Organization of Action. Hillsdale, N.J.: Laurence
Erlbaum.
Gallistel, C. R. 1990. The Organization of Learning. Cambridge: MIT Press.
Gazzaniga, M. S. 1992. Nature’s Mind. Harmondsworth: Penguin.
Gazzaniga, M. S. (ed.). 1995. The Cognitive Neurosciences. Cambridge: MIT
Press.
George, A. (ed.). 1989. Reflections on Chomsky. Oxford: Basil Blackwell.
Ghez, C., Gordon, J., Ghilardi, M. F., and Sainburg, R. 1995. “Contributions of
Vision and Proprioception to Accuracy in Limb Movements.” In Gazzaniga 1995.
Gibson, E. J. 1969. Principles of Perceptual Learning and Development. New
York: Appleton-Century-Crofts.
Gibson, J. J. 1979. The Ecological Approach to Visual Perception. Boston:
Houghton Mifflin.
318 References
Glisky, E. L., Schacter, D. L., and Tulving, E. 1986. “Computer-Learning by

Memory-Impaired Patients: Acquisition and Retention of Complex Knowledge.”
Neuropsychologia 24: 313–328.
Gordon, G. (ed.). 1978. Active Touch. Oxford: Pergamon Press.
Grice, H. P. 1957. “Meaning.” Philosophical Review 66: 377–388. Page refer-
ences to Grice 1989.
Grice, H. P. 1969. “Utterer’s Meaning and Intentions.” Philosophical Review 78:
147–177. Page references to Grice 1989.
Grice, H. P. 1989. Studies in the Way of Words. Cambridge: Harvard University
Press.
Grice, H. P., and Strawson, P. F. 1956. “In Defence of a Dogma.” Philosophical
Review 65: 141–158.
Guttenplan, S. (ed.). 1975. Mind and Language. Oxford: Clarendon Press.
Haegman, L. 1993. Introduction to Government and Binding Theory. Oxford:
Basil Blackwell.
Hampshire, S. 1959. Thought and Action. London: Chatto and Windus.
Harris, M., Freeman, T., and Williams, G. 1992. “Surface Layout from Retinal
Flow.” In Humphreys 1992.
Holland, D. C. 1979. “Differential Effects of Omission Contingencies on Various
Components of Pavlovian Appetitive Responding in Rats.” Journal of Experimen-
tal Psychology: Animal Behavior Processes 5: 65–78.
Horgan, T., and Tienson, J. 1992. “Structured Representations in Connectionist
Systems.” In Davis 1992.
Hume, D. 1739–1740/1978. A Treatise of Human Nature. Edited by L. A. Selby-
Bigge and revised by P. H. Nidditch. Oxford: Clarendon Press.
Humphreys, G. W. (ed.). 1992. Understanding Vision. Oxford: Basil Blackwell.
Hurley, S. 1998. Consciousness and Action. Cambridge: Harvard University
Press.
Kant, I. 1781/1964. Critique of Pure Reason. Translated by Norman Kemp
Smith. London: Macmillan.
Karmiloff-Smith, A. 1992. Beyond Modularity: A Developmental Perspective on
Developmental Psychology. Cambridge: MIT Press.
Kellman, P. J., and Spelke, E. S. 1983. “Perception of Partly Occluded Objects in
Infancy.” Cognitive Psychology 15: 483–524.
Kitcher, P. 1990. Kant’s Rational Psychology. Oxford: Clarendon Press.
Klinnert, M. D., Campos, J. J., Sorce, J. F., Emde, R. N., and Svejda, M. 1983.
“Emotions as Behaviour Regulators: Social Referencing in Infancy.” In Plutchik
and Kellerman 1983.
Kuchuk, A., Vibbert, M., and Bornstein, M. H. 1986. “The Perception of Smiling
and Its Correlates in Three-Month-Old Infants.” Child Development 57:
1054–1061.
References 319
Lackner, J. R. 1988. “Some Proprioceptive Influences on the Perceptual Represen-

tation of Body Shape and Orientation.” Brain 111: 281–297.
Lederman, S. J., and Klatzky, R. L. 1987. “Hand Movements: A Window into
Haptic Object Recognition.” Cognitive Psychology 19: 342–368.
Lee, D. N., and Aronson, E. 1974. “Visual Proprioceptive Control of Standing in
Human Infants.” Perception and Psychophysics 15: 529–532.
Lehrer, K. 1974. Knowledge. Oxford: Clarendon Press.
Leslie, A. M. 1987. “Pretense and Representation: The Origins of ‘Theory of
Mind’.” Psychological Review 94: 412–426.
Leung, E., and Rheinhold, H. 1981. “Development of Pointing as a Social Ges-
ture.” Developmental Psychology 17: 215–220.
Lieberman, P. 1991. Uniquely Human: The Evolution of Speech, Thought, and
Selfless Behavior. Cambridge: Harvard University Press.
Lishman, J. R., and Lee, D. N. 1973. “The Autonomy of Visual Kinaesthetics.”
Perception 2: 287–294.
Lock, A. (ed.). 1978. Action, Gesture, Symbol: The Emergence of Language. Lon-
don: Academic Press.
Loewer, B., and Rey, G. (eds.). 1991. Meaning in Mind: Fodor and His Critics.
Oxford: Basil Blackwell.
Lycan, W. G. (ed.). 1990. Mind and Cognition: A Reader. Cambridge, Mass.:
Basil Blackwell.
Mackie, J. L. 1976. Problems from Locke. Oxford: Clarendon Press.
MacWhinney, B. (ed.). 1987. Mechanism of Language Acquisition. Hillsdale,
N.J.: Lawrence Erlbaum.
Marcel, A. J. 1983. “Conscious and Unconscious Perception: Experiments on
Visual Masking and Word Recognition.” Cognitive Psychology 15: 197–237.
Marr, D. 1982. Vision. San Francisco: W. H. Freeman and Sons.
Marr, D., and Nishihara, H. K. 1978. “Representation and Recognition of the
Spatial Organisation of Three-Dimensional Shapes.” Proceedings of the Royal
Society of London B204: 269–294.
Martin, G. B., and Clark, R. D. 1982. “Distress Crying in Neonates: Species and
Peer Specificity.” Developmental Psychology 18: 3–9.
Martin, C. B., and Deutscher, M. 1966. “Remembering.” Philosophical Review
75: 161–196.
Martin, M. G. F. 1992. “Sight and Touch.” In Crane 1992.
Martin, M. G. F. 1993. “Sense Modalities and Spatial Properties.” In Eilan, Mc-
Carthy, and Brewer 1993.
´
Martin, M. G. F. 1995. “Bodily Awareness: A Sense of Ownership.” In Bermudez,
Marcel, and Eilan 1995.
McCarthy, R. A. 1993. “Assembling Routines and Addressing Representations:
320 References
An Alternative Conceptualization of ‘What’ and ‘Where’ in the Human Brain.”

In Eilan, McCarthy, and Brewer 1993.
McDowell, J. 1994. Mind and World. Cambridge: Harvard University Press.
McGinn, C. 1989. Mental Content. Oxford: Basil Blackwell.
Medin, D. L., Roberts, W. A., and Davis, R. T. (eds.). 1976. Processes of Animal
Memory. Hillsdale, N.J.: Lawrence Erlbaum.
Mellor, D. H. 1988–1989. “I and Now.” Proceedings of the Aristotelian Society
89: 79–94. Page references to Mellor 1991.
Mellor, D. H. 1991. Matters of Metaphysics. Cambridge: Cambridge University
Press.
Meltzoff, A. N. 1993. “Molyneux’s Babies: Cross-Modal Perception, Imitation,
and the Mind of the Preverbal Infant.” In Eilan, McCarthy, and Brewer 1993.
Meltzoff, A. N., and Moore, M. K. 1977. “Imitation of Facial and Manual Ges-
tures by Human Neomnates.” Science 198: 75–78.
Meltzoff, A. N., and Moore, M. K. 1983. “Newborn Infants Imitate Adult Facial
Gestures.” Child Development 54: 702–709.
Melzack, R. 1992. “Phantom Limbs.” Scientific American 266: 90–96.
Menzel, E. W. 1973. “Chimpanzee Spatial Memory Organization.” Science 182:
943–945.
Mill, J. S. 1889. An Examination of Sir William Hamilton’s Philosophy. Sixth
edition. London: Longmans.
ˆ
Millikan, R. G. 1990. “The Myth of the Essential Indexical.” Nous 24: 723–734.
Murphy, C. M., and Messner, D. J. 1977. “Mothers, Infants, and Pointing: A
Study of a Gesture.” In Schaffer 1977.
Murray, L., and Trevarthen, C. 1985. “Emotional Regulation of Interactions be-
tween Two-Month-Olds and Their Mothers.” In T. M. Field and N. A. Fox (eds.),
Social Perception in Infancy. Norwood, N.J.: Ablex.
Nagel, T. 1971. “Brain Bisection and the Unity of Consciousness.” Synthese 20:
396–413.
Neisser, U. 1976. Cognition and Reality. New York: W. H. Freeman.
Neisser, U. (ed.). 1987. Concepts and Conceptual Development. Cambridge:
Cambridge University Press.
Neisser, U. 1988. “Five Kinds of Self-Knowledge.” Philosophical Psychology 1:
35–59.
Neisser, U. 1991. “Two Perceptually Given Aspects of the Self and Their Develop-
ment.” Developmental Review 11: 197–209.
Neisser, U. (ed.). 1993. The Perceived Self. Cambridge: Cambridge University
Press.
Neumann, O., and Prinz, W. (eds.). 1990. Relationships between Perception and
Action. Berlin: Springer Verlag.
References 321
Newson, J., and Newson, E. 1975. “Intersubjectivity and the Transmission of

Culture.” Bulletin of the British Psychological Society 28: 437–445.
Noonan, H. 1979. “Identity and the First Person.” In Diamond and Teichmann
1979.
Nozick, R. 1981. Philosophical Explanations. Oxford: Clarendon Press.
O’Brien, L. 1994. “Anscombe and the Self-Reference Rule.” Analysis 54: 277–
281.
O’Shaugnessy, B. 1980. The Will. 2 vols. Cambridge: Cambridge University Press.
O’Shaugnessy, B. 1989. “The Sense of Touch.” Australasian Journal of Philoso-
phy 67: 37–58.
Parker, S. T., and Gibson, K. P. 1979. “A Developmental Model for the Evolution
of Language and Intelligence in Early Hominids.” Behavioral and Brain Sciences
2: 367–408.
Peacocke, C. 1979. Holistic Explanation: Action, Space, Interpretation. Oxford:
Clarendon Press.
Peacocke, C. 1983. Sense and Content. Oxford: Clarendon Press.
Peacocke, C. 1986. “Analogue Content.” Proceedings of the Aristotelian Society,
suppl. vol. 60: 1–17.
Peacocke, C. 1989. Transcendental Arguments in the Theory of Content. Inaugu-
ral Lecture. Oxford: Oxford University Press.
Peacocke, C. 1992. A Study of Concepts. Cambridge: MIT Press.
Peacocke, C. 1994. “Nonconceptual Content: Kinds, Rationales, and Relations.”
Mind and Language 9: 419–430.
Pears, D. 1975. “Russell’s Theories of Memory.” In D. Pears, Questions in the
Philosophy of Mind. London: Duckworth.
Pears, D. 1988. The False Prison. 2 vols. Oxford: Clarendon Press.
Perner, J. 1991. Understanding the Representational Mind. Cambridge: MIT
Press.
ˆ
Perry, J. 1979. “The Problem of the Essential Indexical.” Nous 13: 3–21.
Piaget, J. 1954. The Construction of Reality in the Child. Translated by M. Cook.
New York: Basic Books.
Piaget, J., and Inhelder, B. 1969. The Psychology of the Child. New York: Basic
Books.
Piatelli-Palmarini, M. 1989. “Evolution, Selection, and Cognition: From ‘Learn-
ing’ to Parameter Setting in Biology and the Study of Language.” Cognition 31:
1–44.
Pinker, S. 1979. “Formal Models of Language Learning.” Cognition 1: 217–
283.
Pinker, S. 1987. “The Bootstrapping Problem in Language Acquisition.” In Mac-
Whinney 1987.
322 References
Pinker, S., and Bloom, P. 1990. “Natural Language and Natural Selection.” Be-
havioral and Brain Science 13: 707–784.
Pitcher, G. 1970. “Pain Perception.” Philosophical Review 79: 368–393.
Plutchik, R., and Kellerman, H. (eds.). 1983. Emotion: Theory, Research, Experi-
ence. Boston: Academic Press.
Pope, M. J. 1984. “Visual Proprioception in Infant Postural Development.” Ph.D.
thesis, University of Southampton.
Premack, D., and Woodruff, G. 1978. “Does the Chimpanzee Have a Theory of
Mind?” Behavioral and Brain Science 4: 515–526.
Price, H. H. 1953. Thinking and Experience. London: Hutchinson University
Library.
Prigatano, G., and Schacter, D. (eds.). 1991. Awareness of Deficit after Brain
Injury. Oxford: Oxford University Press.
Prinz, W. 1990. “A Common Coding Approach to Perception and Action.” In
Neumann and Prinz 1990.
Quine, W. V. O. 1951. “Two Dogmas of Empiricism.” Philosophical Review 60:
20–43. Page references to Quine 1953.
Quine, W. V. O. 1953. From a Logical Point of View. Harvard: Harvard Univer-
sity Press.
Quine, W. V. O. 1960. Word and Object. Cambridge: MIT Press.
Quinton, A. 1973. The Nature of Things. London: Routledge and Kegan Paul.
Ramachandran, V. S. 1994. “Phantom Limbs, Neglect Syndromes, Repressed
Memories, and Freudian Psychology.” International Review of Neurobiology
37: 291–333.
Ramsey, W. 1992. “Connectionism and the Philosophy of Mental Representa-
tion.” In Davis 1992.
Reed, E. S. 1987. “Why Do Things Look As They Do? The Implications of J. J.
Gibson’s The Ecological Approach to Visual Perception.” In Costall and Still
1987.
Rochat, P. 1995. The Self in Infancy: Theory and Research. Amsterdam: Elsevier.
Rock, I. 1975. An Introduction to Perception. New York: Macmillan.
Rolls, E. T. 1990. “A Theory of Emotion, and Its Application to Understanding
the Neural Basis of Emotion.” Cognition and Emotion 4: 161–190.
Rolls, E. T. 1995. “A Theory of Emotion and Consciousness, and Its Application
to Understanding the Neural Basis of Emotion.” In Gazzaniga 1995.
Russell, B. 1921. The Analysis of Mind. London: Allen and Unwin.
Russell, J. 1980. “Action from Knowledge and Conditioned Behaviour. Part One:
The Stratification of Behaviour.” Behaviorism 8: 87–98.
Russell, J. 1996. Agency: Its Role in Mental Development. Hove: Lawrence
Erlbaum.
References 323
Sagi, A., and Hoffman, M. 1976. “Empathic Distress in the Newborn.” Develop-
mental Psychology 12: 175–176.
Scaife, M., and Bruner, J. S. 1975. “The Capacity for Joint Visual Attention in
the Infant.” Nature 253: 265–266.
Schacter, D. L., McAndrews, M. P., and Moscovitch, M. 1988. “Access to Con-
sciousness: Dissociations between Implicit and Explicit Knowledge in Neuropsy-
chological Syndromes.” In Weiskrantz 1988.
Schaffer, H. R. (ed.). 1977. Studies in Mother-Infant Interaction. Proceedings of
the Loch Lomond Symposium, Ross Priory, University of Strathclyde, September
1975. New York: Academic Press.
Schopenhauer, A. 1844/1966. The World as Will and Representation. Translated
by E. J. F. Payne. 2 vols. New York: Dover Publications.
Segal, G. 1996. “The Modularity of Theory of Mind.” In Carruthers and Smith
1996.
Shapiro, M. F., Fink, M., and Bender, M. B. 1952. “Exosomesthesia or Displace-
ment of Cutaneous Sensation onto Extrapersonal Space.” AMA Archives of Neu-
rology and Psychiatry 68: 481–490.
Shoemaker, S. 1963. Self-Knowledge and Self-Identity. Ithaca: Cornell Univer-
sity Press.
Shoemaker, S. 1968. “Self-Reference and Self-Awareness.” Journal of Philosophy
65: 555–567. Page references to Shoemaker 1984.
Shoemaker, S. 1970. “Persons and Their Pasts.” American Philosophical Quar-
terly 7: 269–285. Page references to Shoemaker 1984.
Shoemaker, S. 1984. Identity, Cause, and Mind. Cambridge: Cambridge Univer-
sity Press.
Shoemaker, S. 1986. “Introspection and the Self.” In French, Uehling, and
Wettstein 1986.
Shoemaker, S. 1994. “Self-Knowledge and ‘Inner Sense’.” Philosophy and Phe-
nomenological Research 54: 249–314.
Simner, M. L. 1971. “Newborn’s Response to the Cry of Another Infant.” Devel-
opmental Psychology 5: 136–150.
Slater, A., Morrison, V., Somers, M., Mattock, A., and Taylor, D. 1990. “New-
born and Older Infants’ Perceptions of Partly Occluded Objects.” Infant Behavior
and Development 13: 33–49.
Smolensky, P. 1991. “Connectionism, Constituency, and the Language of
Thought.” In Loewer and Rey 1991.
Snowdon, P. F. 1991. “Personal Identity and Brain Transplants.” In Cockburn
1991.
´
Snowdon, P. F. 1995. “Persons, Animals, and Bodies.” In Bermudez, Marcel, and
Eilan 1995.
324 References
Spelke, E. S. 1988. “The Origins of Thought and Language.” In Weiskrantz 1988.

Spelke, E. S. 1990. “Principles of Object Perception.” Cognitive Science 14:
29–56.
Spelke, E. S., and Van de Walle, G. A. 1993. “Perceiving and Reasoning about
Objects: Insights from Infants.” In Eilan, McCarthy, and Brewer 1993.
Stern, D. 1985. The Interpersonal World of the Infant. New York: Basic Books.
Strawson, P. F. 1959. Individuals: An Essay in Descriptive Metaphysics. Lon-
don: Methuen.
Strawson, P. F. 1966. The Bounds of Sense. London: Methuen.
Strawson, P. F. 1994. “The First Person—and Others.” In Cassam 1994.
Taschek, W. W. 1985. “Referring to Oneself.” Canadian Journal of Philosophy
15: 629–652.
Taylor, C. 1964. The Explanation of Behaviour. London: Routledge and Kegan
Paul.
Tinbergen, N. 1951. The Study of Instinct. Oxford: Oxford University Press.
Tobias, P. V. 1987. “The Brain of Homo habilis: A New Level of Organization in
Cerebral Evolution.” Journal of Human Evolution 16: 741–765.
Tolman, E. C., and Honzik, C. H. 1930a. “ ‘Insight’ in Rats.” University of Cali-
fornia Publications in Psychology 4: 215–232.
Tolman, E. C., and Honzik, C. H. 1930b. “Introduction and Removal of Reward,
and Maze Learning in Rats.” University of California Publications in Psychology
4: 257–275.
Tolman, E. C., Ritchie, B. F., Kalish, D. 1946. “Studies in Spatial Learning. II:
Place Learning vs Response Learning.” Journal of Experimental Psychology 36:
221–229.
Tomasello, M. 1993. “On the Interpersonal Origins of Self-Concept.” In Neisser
1993.
Travis, C. 1994. “On Constraints of Generality.” Proceedings of the Aristotelian
Society 94: 165–188.
Trevarthen, C. 1993. “The Self Born in Intersubjectivity.” In Neisser 1993.
Trevarthen, C., and Hubley, P. 1978. “Secondary Intersubjectivity: Confidence,
Confiding, and Acts of Meaning in the First Year.” In Lock 1978.
Ts’o, D. Y., and Roe, A. W. 1995. “Functional Compartments in Visual Cortex:
Segregation and Interaction.” In Gazzaniga 1995.
Van Gulick, R. 1990. “Functionalism, Information, and Content.” In Lycan 1990.
Von Hofsten, C. 1982. “Foundations for Perceptual Development.” Advances in
Infancy Research 2: 241–261.
Walker, Ralph C. S. 1978. Kant. London: Routledge and Kegan Paul.
Wanner, E., and Gleitman, L. R. (eds.). 1982. Language Acquisition: The State
of the Art. Cambridge: Cambridge University Press.
References 325
Weiskrantz, L. (ed.). 1988. Thought without Language. Oxford: Clarendon Press.

Wiggins, D. 1980. Sameness and Substance. Oxford: Basil Blackwell.
Wimmer, H., and Perner, J. 1983. “Beliefs about Beliefs: Representation and Con-
straining Function of Wrong Beliefs in Young Children’s Understanding of Decep-
tion.” Cognition 13: 103–128.
Wittgenstein, L. 1921/1961. Tractatus Logico-philosophicus. Translated by D. F.
Pears and B. F. McGuinness. London: Routledge and Kegan Paul.
Wittgenstein, L. 1958. Preliminary Studies for the “Philosophical Investiga-
tions,” Generally Known as the Blue and Brown Books. Oxford: Basil Black-
well.
Woodruff, G., and Premack, D. 1979. “Intentional Communication in the Chim-
panzee.” Cognition 7: 333–362.
Index
Acquisition Constraint, 19–21, 23, Aristotle, 139

24, 44–45, 129, 162, 189, 191, Armstrong, D. M., 54, 277, 307
226, 267, 269, 291–293, 297 (n. 12)
and first-person pronoun, 20, Aronson, E., 111, 128
292–294 Associative agnosias, 79–81
introduced, 19 Australopithecus, 77
Acredolo, L. P., 217, 218 Autonomy Principle, 46, 58–82, 85,
Adams, A., 217, 218 97, 270, 301 (n. 5), 310 (n. 1)
Adamson, L. B., 261 and infant cognition, 62–76
Affordances, 112–114, 117, 118, 127, and phylogeny of language, 76–79
223–227, 293–294. See also Ecolog- stated, 61
ical optics and visual agnosias, 79–81
affordance identities, 225–226, 293 Ayers, M., 141, 238
affordance space, 226–227
affordance symmetry, 224–226, 293 Baillargeon, R., 64–66, 68, 302
affordance transitivity, 225–226, 293 (n. 13)
nonextensionality of, 121–122 Bakeman, R., 261
Alexia, 80 Baron-Cohen, S., 300–301 (n. 16)
A location vs. B location in proprio- Barwise, J., 9
ception, 154–158 Basic inductive generalizations,
Altham, J. E. J., 300 (n. 8) 183–186
Amnesia, 174–175, 176 Bateson, M. C., 251
Andersen, R. A., 141 Behavior
Animal learning theory, 62, 207 emergence of intentional, 192
Anosagnosia, 151–152 intentional vs. nonintentional, expla-
A, not-B error or stage 4 error nation of, 36–38, 85–91, 115–118,
(Piaget), 70, 75 246
Anscombe, G. E. M., 16–17, 300 narrowly vs. broadly construed, 86
(n. 8), 306 (n. 2) stimulus-response explanations of,
Apperceptive agnosias, 79–81 88–89, 208–210
´ ´
Argument from analogy, 231 Bekesy, G. von, 309 (n. 24)
328 Index
Belief. See also Content; Representa- on causally indexical comprehension,

tional states; Subjective beliefs 200–202, 206–207
conditions on ascription of, 33–39 on temporal orientation toward the
emergence of concept of, 279–280 past, 178–180
functionalist view of, 29 Capacity circularity, 18–25, 28, 47,
necessarily conceptual, 34, 52–54, 49, 58, 60, 62, 268
301 (n. 1) and Acquisition Constraint, 19–20
normativity of, 52–54 and developmental explanation, 60,
utility conditions of, 29, 34, 37–39 62
Belief-desire psychology, 36–37, 85 distinguished from explanatory circu-
Bender, M. B., 309 (n. 24) larity, 18
Bennett, J., 88–89, 184–185, 277, and innatism, 21–24
279, 304 (n. 7), 310 (n. 8) introduced, 18
Bermu´dez, J. L., 132, 249, 300 (nn. 9, and nonconceptual self-
13), 302 (n. 6), 303 (n. 21), 305– consciousness, 291–294
306 (n. 16), 306 (nn. 1, 8), 310 solution to problem of, 291–294
(nn. 1, 2), 312 (nn. 12, 13, 14) Carey, S., 70
Bicoordinate navigation, 211 Cassam, Q., 143–144, 307 (n. 12)
˜
Blindsight, 79 Castaneda, H.-N., 4, 14, 148
Bloom, P., 79 Causally indexical comprehension
Bodily awareness, 131–162. See also (Campbell), 200–202, 206–207
Proprioception, somatic Causation, concept of, 288
of body as responsive to will, 150, Chapuis, N., 215–216
161, 229, 237, 271, 284 Chisholm, R., 307 (n. 15)
of felt boundaries of body, 149–150, Choice-reaction-time tasks, 140
161, 229, 271, 284 Chomsky, N., 22, 300 (n. 15)
structure of, 152–158 Clark, A., 93
Bodily self. See Bodily awareness Clark, R. D., 124
Body-relative space, 142, 145, 204 Cognitive dynamics (Kaplan), 296
frame of reference of, 152–158 Cognitive map, 203–207
Bornstein, M. H., 250 and levels of self-awareness, 203–204
Bower, T. G. R., 127 as record in nervous system,
Bremner, J. G., 250 203–206
Brewer, B., 140, 151, 158, 307 (n. 11) Cohen, D., 250
Bringing-it-about-that, 288–290, 312 Cohesion, principle of (Spelke), 63–
(n. 4) 64, 74
Broca’s area, 77–78 Color perception, 55–57, 96
Brown, J. W., 76, 80 Common sensibles, 139
Bruner, J. S., 255, 257 Communicative intent, 277–291
Budd, M., 306 (n. 2) of first-person pronoun, 282–283,
Butterworth, G. E., 128, 305 (nn. 12, 293
15) informational vs. practical, 277
and linguistic reference, 280–281
Campbell, J., 172, 198, 245, 300 sentential vs. subsentential, 278–281
(n. 9), 310 (chap. 7, n. 5; chap. 8, Computational theory of cognition
n. 2) (CTC), 93–94, 304 (n. 8)
Index 329
Computational theory of vision Coordinated joint engagement in in-

(Marr), 81–82 fants, 260–262, 263, 265, 285–286,
Concepts 289
acquisition of, 22, 60–62, 69–71, Corballis, M., 77, 78, 303 (n. 17)
74–76 Corollary discharge, 304–305 (n. 4)
conditions on ascription of, 34 Count terms, 173, 309 (n. 2)
evolution of, 60–62 Criterial vs. constitutive accounts of
inferential role of, 67–68, 69–70 content, 84, 207
normativity of, 52–54 Cussins, A., 60
relation of, to language mastery, 41–
42, 68, 70–71 (see also Priority Davies, M., 310 (n. 10)
Principle) Davis, R. T., 186
Conceptual-Requirement Principle, Dead reckoning, 205, 210
41, 42, 43, 44, 45, 268, 301 (n. 4) DeCasper, A. J., 250
stated, 41 Deflationary account of self-
Conditioning, 36–37, 87–89, 119 consciousness, 9–13, 14–21, 31, 32,
classical/Pavlovian, 36, 89 294–296, 299 (nn. 5, 6)
instrumental/operant, 36, 89, 190 and paradox of self-consciousness,
Consciousness, 30, 189–192, 309–310 14–21
(n. 4), 310 (n. 10) stated, 10–11
Constancy-of-sense thesis, 12, 13, 24– Deutscher, M., 175, 179–180
25, 230, 232 Developmental explanation, 60–62,
Contact, principle of (Spelke), 64, 74 74
Content. See also Representational and Autonomy Principle, 61–62, 301
states; Thought (n. 5)
classical view of, 28, 39, 41–43, 45, Developmental psychology, 62–76,
49, 269–270 149, 172, 208, 248
and concept possession, 41–44 Diamond, A., 75
(see also Nonconceptual content) Dickinson, A., 36, 62, 87, 190
conditions on ascribing, 33–39, 47, Dishabituation experiments, 47, 64–
83–94, 269 66, 68, 72, 75
constitutive vs. criterial accounts of, Distinguishing self-awareness,
84–85 237–238
demonstrative, 38, 55–57 psychological, 240–241, 242, 249,
indexical, 38 252, 255, 261
and information, 84–85 and relative identity, 240
nonconceptual (see Nonconceptual Distress crying in neonates, 124–125
content) Donald, M., 77, 78
of perceptual experience (see Percep- Drawbridge experiment (Baillargeon),
tual content) 64–66, 68, 70
Russellian, 52 Dretske, F., 84, 303 (nn. 2, 3)
Continuity, principle of (Spelke), 64, Dummett, M., 12, 41, 42
66, 74
Contrast space, 238, 242, 248, 265, Ecological optics, 103–129, 165. See
274 also Affordances; Gibson, J. J.
Cooper, L. A., 302 (n. 12) content of, 115–123
330 Index
Ecological optics (continued) Field, J., 127

and information-processing ac- Field, T. M., 126, 250
counts, 310 (n. 9) Fifer, W., 250
memory in, 187–188 Fink, M., 309 (n. 24)
perceptual invariants in, 108–109 First-person contents, 5–12, 27–39,
and self-perception in infancy, 44–47, 49. See also First-person pro-
123–128 noun; Nonconceptual first-person
self-specifying information in, 103– contents; Nonconceptual psycholog-
114, 223–224 ical self-awareness; Subjective beliefs
Ecological self, 124, 131. See also Eco- functionalist theory of, 27–39, 42
logical optics; Nonconceptual first- identification component of, as
person contents basic, 7
Efference copy, 111, 304–305 (n. 4) identification component vs. predica-
Egocentric space. See Body-relative tion component of, 5–6, 10
space without linguistic self-reference,
Eilan, N., 132, 306 (n. 1), 307 (n. 11) 27–39 (see also Nonconceptual
Emotions, 246–247 first-person contents)
Empathic arousal in infants, 250 role of, in explanation of action, 34–
Evans, G., 5, 6, 50, 59, 92, 139, 144, 35, 116–117, 148
299 (n. 2), 301 (n. 1), 307 (nn. 13, First-person-motivation thesis, 35,
14, 16) 37–39
on cognitive maps, 204–205 First-person perspective, entry into,
on first-person thoughts, 221–223 123–124, 128
on Generality Constraint, 233–237, First-person pronoun, 3, 9–12, 14–
311 (n. 2) 18, 21, 22, 24–25, 27–29, 31, 32,
on immunity to error through mis- 44–46, 49, 132, 146, 198–199,
identification, 307–308 (n. 16) 267–268
on simple theory of perception, 222– and Acquisition Constraint, 20,
223, 244–245 292–294
Exosomesthesia, 309 (n. 24) communicative intent governing use
Explanation, two senses of, 19 of, 282–291, 293
Explanatory circularity, 14–16, 47, conditions on proper use of,
49, 58, 60, 268, 283, 290–291 275–276
distinguished from capacity circular- functionalist account of mastery of,
ity, 18 31
introduced, 14–16 guaranteed reference of, 9–10
and nonconceptual self- noncircular account of mastery of,
consciousness, 275–291 283–291
and possession-conditions explana- as perceptual demonstrative (Evans),
tion, 60 307–308 (n. 16)
solution to problem of, 275–291 not a referring expression (Ans-
combe), 16–17, 300 (n. 8)
False-belief task, 279–280, 312 (n. 2) role of nonconceptual content in mas-
Farah, M., 80, 303 (nn. 18, 19) tery of, 45, 47, 283–291
Features and feature-placing thought, semantics of, 9–12 (see also Token-
38–39, 172–174, 195 reflexive rule)
Index 331
semantics vs. pragmatics of, 17 Hinges in proprioceptive content,

used as object vs. used as subject, 155–157, 161
5–6, 11 defined, 157
Fitts, S. S., 186 Hoffman, M., 250
Focal vs. peripheral awareness, Holland, D. C., 87–88
138–139 Homo erectus, 78
Fodor, J. A., 77, 86, 89, 93, 300 Homo habilis, 78
(n. 14), 310 (n. 9) Homo sapiens, 77–78
Freeman, T., 304 (n. 1) Honzik, C. H., 212, 217, 219
Functionalist account of first-person Hubley, P., 249, 260–261, 287
thought, 27–39, 42, 46 Hume, D., 104
and mastery of first-person pronoun, Humphreys, G. W., 310 (n. 10)
31 Hurley, S. L., 300 (n. 12), 311 (n. 10)
and self-reference, 27–30, 271
Identification constraint (Shoemaker),
Gaffan, D., 185–186 136–137
Gallagher, S., 309 (n. 26) Identity. See also Relative identity
Gallistel, C. R., 203–206, 210, 304– neo-Lockean conception of, 238–240
305 (n. 4) of persons, 239
Gazzaniga, M., 22 Imitation by neonates, 125–127, 140,
Generality Constraint (Evans), 59, 92, 249, 312 (n. 14)
233–237, 241, 311 (n. 1) Immunity to error through misidenti-
George, A., 300 (n. 15) fication, 6–12, 143–145
Ghez, C., 307 (n. 10) defined, 7
Gibson, E., 113–114 and guaranteed reference of first-
Gibson, J. J., 46, 103–114, 131, 165, person pronoun, 9–10
187–188, 190, 224, 271, 303 (n. 4), and identification component of first-
304 (nn. 1, 3), 310 (n. 9) person contents, 11–12
Gibson, K. P., 303 (n. 17) and justification, 6–7
Glisky, E. L., 175, 176 not logically necessary, 307 (n. 12)
Goodwyn, S. W., 217, 218 of perceptual demonstrative judge-
Gordon, G., 137, 306 (nn. 6, 7) ments, 143
Government-binding (GB) paradigm Shoemaker’s definition of, 6
(Chomsky), 22 of somatic proprioception, 143–145,
Greenburg, R., 250 147, 148
Grice, H. P., 277–278, 300 (n. 12) Indexical beliefs, 117. See also Subjec-
Gricean mechanisms, 279–281, 287– tive beliefs
289. See also Communicative intent Indirect reflexive pronoun (‘he*’), 4,
14–17, 22, 28, 45, 49
Haegman, L., 22 introduced, 4
Haptic exploration, 133, 137–138, Infant cognition, 62–76, 100, 248–
152, 306 (nn. 6, 7) 262, 271, 274
Harris, M., 304 (n. 1) and protoconversations, 250–254
He*, 4, 14–17, 22, 28, 45, 49 revolution in, at 9 months, 248–249,
introduced, 4 254–255
Hicks, L., 128, 305 (nn. 12, 15) triadic vs. dyadic, 255, 260
332 Index
Inference to best explanation, 46, 100, visual, 110–112, 114, 115, 116, 117,
207, 217, 248, 253, 259, 265, 271, 122, 128
273 Klatzky, R. L., 306 (n. 6)
Innate endowment, 19, 22, 292–293 Klinnert, M. D., 263
Innate releasing mechanisms, 36 Kuchuk, A., 250
Innatism, 21–24
Integrated representation of the envi- Lackner, J. R., 305 (n. 8)
ronment, 199–207, 273 Language
basic conditions on, 208–211 generativity of, 92
and cognitive maps, 203–207 phylogenesis of, 76–79
construction of, 224–227, 294 Language learning
and navigational abilities, 207–220 and capacity circularity, 20
need for operational understanding innatist theory of, 23–24
of, 206 of syntax vs. semantics, 23–24
place-driven vs. response-driven, Latent-learning experiments, 219
208–210 Lederman, S. J., 306 (n. 6)
and sensitivity to genuine spatial fea- Lee, D. N., 111, 128
tures, 210–211 Leibniz’s Law, 239
and two systems of spatial relations, Leslie, A., 300 (n. 16)
213–217 Leung, E., 257
Integrated sensory field (Ayers), 141– Lieberman, P., 78, 302 (n. 16)
142, 145 Lishman, J. R., 111
Intentional web (Peacocke), 214–215 Local holism (Peacocke), 18, 300
Introspection, 135, 146 (n. 12)
Inverting lenses, 142 Locke, J., 239
Iterated psychological states, 284–287 Looming, 109, 122, 128
‘I’-thoughts, 2–12, 24–25. See also
First-person contents; First-person Mackie, J. L., 238, 311 (n. 4)
pronoun; Nonconceptual first- Marcel, A. J., 132, 306 (n. 1), 310
person contents; Nonconceptual (n. 4)
psychological self-awareness Marr, D., 76, 80, 108, 303 (n. 20)
defined, 4 Martin, C. B., 175, 179–180
and genuine self-consciousness, 2–4 Martin, G. B., 124
oratio obliqua vs. oratio recta, 3–4, Martin, M. G. F., 137, 139, 158, 309
14–15 (n. 24)
and self-reference, 2–4 Mass terms, 173, 309 (n. 2)
Maze learning, 208–209, 219
Joint selective visual attention in in- McAndrews, M. P., 174, 310 (n. 4)
fants, 255–259, 263, 265, 283, 285, McCarthy, R. A., 307 (n. 11)
289 McDowell, J., 54–58
Mellor, D. H., 29–39, 42
Kalish, D., 209 Meltzoff, A. N., 125–126, 249, 306
Kant, I., 165–167, 198, 310 (n. 7) (n. 8)
Kaplan, D., 296 Melzack, R., 151
Karmiloff-Smith, A., 300 (n. 14), 301 Memory
(n. 16) autobiographical, 243–244, 294,
Kinesthesis, 134, 140, 150, 168 295–296
Index 333
autobiographical vs. episodic, 181 autonomous, 58–82, 270–274 (see

causal theory of, 175–176, 179–180 also Autonomy Principle)
conscious, 174–180, 185–186, 193, and computational theory of vision,
194, 273 81–82
and ecological optics, 187–188 conditions on ascribing, 83–94, 115–
Menzel, E. W., 217 123, 253
Messner, D. J., 256 defined, 52
Mill, J. S., 23 and developmental explanation,
Millikan, R. G., 305 (n. 9) 60–62
Misrepresentation, 84–85, 88, 89–90, different applications of, 58–62
116 and infant cognition, 62–76 (see also
misperception vs. perceptual illu- Infant cognition)
sions, 305 (n. 8) nonextensionality of, 119–122
Modality specificity, 139–142 and phylogeny of language, 76–79
Moore, M. K., 125–126, 249 and possession-conditions explana-
Moscovitch, M., 174, 310 (n. 4) tion, 58–59
Moving-room experiments, 111–112, of somatic proprioception, 151–162
116 and visual agnosias, 79–81
Multiple-objects constraint (Shoe- of visual perception, 103–114
maker), 136–137, 142–143, 144 Nonconceptual first-person contents,
Munger, M. P., 302 (n. 12) 44–47, 49, 97, 103–104, 114, 163,
Murphy, C. M., 256 167, 188, 193, 224, 229, 271–274
Murray, L., 251 conditions on ascribing, 115–123
in coordinated joint engagement,
Nagel, T., 311 (n. 5) 261–262, 285–286
Nativism (innatism), 21–24 in ecological optics, 103–114
Navigation, 196–198, 201–202. See embedded within iterated psychologi-
also Integrated representation of the cal states, 285–287
environment; Nonconceptual point introduced, 49
of view; Space; Spatial awareness in joint visual attention, 258–259,
as evidence for integrated represen- 283, 285
tation of environment, 207–220 in navigational abilities, 46
around obstacles, 212–213 role of, in explanation of action,
and two systems of spatial relations, 117–123
213–217 role of, in intentional self-reference,
Neisser, U., 124, 188, 305 (nn. 12, 283–291
13), 310 (n. 6) in social contexts, 46
Neuroanatomy, 77 in social referencing, 264
Newson, E., 257 in somatic proprioception, 131–162
Newson, J., 257 Nonconceptual point of view, 168–
Nishihara, H. K., 81 192, 193–227, 229, 237, 238
Nonconceptual content, 43–47, 50– construction of, 224–227
82, 268–270. See also Noncon- defined, 168
ceptual first-person contents; diachronic nature of, 169–170,
Nonconceptual point of view; 272–273
Nonconceptual psychological and ecological optics, 168–188
self-awareness and memory, 174–180
334 Index
Nonconceptual point of view (continued) functionalist solution to, 28, 31–32

nonsolipsistic component of, 197– solution to, 272, 275–294
198, 220, 273 stated, 24
and spatial awareness, 193–198 Parallel distributed processing (PDP),
spatial-awareness component of, 93, 304 (n. 8)
198, 199, 220–221, 273 Parietal cortex, 77–78, 140
Nonconceptual psychological self- Parker, S. T., 303 (n. 17)
awareness, 248–265, 274 Peacocke, C., 8, 18, 50, 51, 52, 53,
in coordinated joint engagement, 59–61, 84, 224, 300 (n. 12), 302
261–262 (nn. 6, 8), 304 (n. 6), 308 (n. 20),
in joint visual attention, 258–259 310 (nn. 1, 3)
in social referencing, 264 on definition of nonconceptual con-
Nonconceptual recognition, 182–184 tent, 52
and basic inductive generalizations, and minimal account of concept pos-
183–186 session, 84
Nonsolipsistic consciousness (Straw- and perspectival sensitivity, 213–215
son), 167–170, 178, 193–194, 196. on place reidentification, 198–202
See also Nonconceptual point of and possession-conditions explana-
view tion, 59
Noonan, H., 299 (n. 3) and scenario/protopropositional con-
Nozick, R., 300 (n. 10) tent, 95–101, 303 (n. 1)
Pears, D. F., 183, 299 (n. 1)
Object*, 100–101, 172, 174, 177, Perception. See also Perceptual
309 (n. 1) content
defined, 100 cross-modal nature of, 139–142
Object, concept of, 67–68, 69–71, 99– models of (Shoemaker), 135–136
100, 171–172 sensational properties of, 214–215
Object constraint (Shoemaker), 136– simple theory of (Evans), 222–223,
137, 145 244–245
Object permanence, 62–76, 97, 103, Perceptual content, 46, 50–58. See
302 (n. 10) also Nonconceptual content
Object properties, 72–74, 100–101, analogue nature of (Peacocke), 51
302 (n. 12) epistemic theory of (Armstrong), 54
Object-recognition system, 80 fineness of grain of, 50, 54–56, 96
O’Brien, L., 17 in infancy, 62–76
Omission schedule, 87–88, 303–304 vs. perceptual belief, 51
(n. 5) Perceptual space, 56
Optic flow, 109–112 Perner, J., 280
O’Shaughnessy, B., 137, 151, 259 Perry, J., 9, 148, 301 (n. 2), 305 (n. 9)
Perspectival sensitivity (Peacocke),
Paleoneurology, 77 213–217
Paradox of self-consciousness, 1, 16, Phantom limb, 151, 308 (n. 17)
27–28, 43–45, 47–48, 49, 58, 123, Piaget, J., 62, 66, 70, 75, 125, 248,
163, 229, 265, 267, 272, 282, 294. 253
See also Capacity circularity; Ex- Piatelli-Palmarini, M., 78
planatory circularity Pinker, S., 23, 79, 300 (n. 15)
Index 335
Place reidentification/recognition, proprioceived body vs. physical body,

171–174, 189, 193. See also Inte- 151–152
grated representation of the environ- spatial content of, 154–158
ment; Navigation; Nonconceptual types of information in, 132–134
point of view; Spatial awareness Prosopagnosia, 80, 81
and conscious memory, 177–179 Protobeliefs, 118–119, 120, 122, 161,
and first-person concept, 198–200 259, 286
types of, 194–195 and basic inductive generalizations,
Point of view, 164–168. See also Non- 184
conceptual point of view instrumental vs. perceptual, 118–119
Pope, M. J., 128 Protoconversations with infants,
Possession-conditions explanation 250–254
(Peacocke), 58–59, 61 thick vs. thin interpretations of,
P* predicates (Shoemaker), 8 252–254
Premack, D., 62, 300 (n. 16) Protodesires, 118
Price, H. H., 310 (n. 7) Protopropositional content (Pea-
Prigatano, G., 308 (n. 18) cocke), 96–101
Primitively compelling inference (Pea- Psychological self-awareness, 229–
cocke), 8 265. See also Distinguishing self-
Principle of cohesion (Spelke), 63–64, awareness; Nonconceptual psycho-
74 logical self-awareness
Principle of contact (Spelke), 64, 74 and awareness of other minds,
Principle of continuity (Spelke), 64, 230–241
66, 74 criteria for, 242
Principle of solidity (Spelke), 64–65, emergence of, in infancy, 247–264
74 in social interactions, 241–247
Prinz, W., 140 Psychological subject
Priority Principle, 42–44, 61, 68, 199, as agent, 245–246
268, 311–312 (n. 11) as bearer of reactive attitudes,
stated, 42 246–247
Propositional content, 3, 83–85 core notion of, 241–247, 274
Proprioception, somatic, 46, 131–162, as perceiver, 244–245
163, 167, 188, 193, 194, 229, 237, Psychologism, 20
271, 284 Pylyshyn, Z., 93, 310 (n. 9)
A location vs. B location, 154–158
conscious vs. nonconscious, Quasi-indicator. See Indirect reflexive
133–134 pronoun (‘he*’)
content of, 151–161 Quine, W. V. O., 173, 183, 300
correctness conditions of, 159–161, (n. 12)
309 (n. 25) Quinton, A., 172
descriptive content of, 158–161
as form of perception, 135–145 Ramachandran, V. S., 151, 308
as form of self-consciousness, (n. 17)
145–151 Ramsey, W., 304 (n. 8)
mediate vs. immediate, 133–134 Reaching behavior in infants, 127
and neonatal imitation, 125 Reed, E. S., 305 (n. 5)
336 Index
Registration (Bennett), 88–89 broad vs. narrow, 160–161

Reinhold, H., 257 as a cognitive state, 1
Relative identity contrastive nature of, 164, 221, 237–
and distinguishing self-awareness, 238, 272
240–241 core requirements on, 148–151
restricted vs. unrestricted, 238–239 gradations in, 163–164, 187, 189,
Remapping hypothesis, 308 (n. 17) 193
Representational states narrow vs. broad, 149–151
cognitive integration of, 90–91, 94 ontogenesis of, 18–19, 291–294
holism of, 90–91 primitive forms of, 76, 103, 128–
as intermediaries between sensory in- 129, 131–132, 161, 269
put and behavioral output, 86–90, and self-ascription of predicates, 12,
94, 115–116, 270–271 13
structure of, 38–39, 91–94, 301 and spatial awareness, 221–223
(n. 4) Self-reference, 2–4, 14–18, 27–30,
Ritchie, B. F., 209 202. See also First-person contents;
Rock, I., 305 (n. 6) First-person pronoun
Roe, A. W., 80 conditions on, 275–276, 281
Rolls, E. T., 247 functionalist account of, 27–30, 271
Russell, B., 183 identification-free, 7–8
Russell, J., 36, 75, 190, 302 (nn. 11, and immunity to error, 9–12, 299
14) (n. 4)
intentional, 282–283
Sagi, A., 250 as production of a given token (‘I’),
Scaife, M., 255 15–16
Scenario content (Peacocke), 95–101 reflexive vs. accidental, 2–4, 32–33,
Schacter, D. L., 174, 175, 308 (n. 18), 34–35, 147–148, 200
310 (n. 4) without self-knowledge, 27–30
Schopenhauer, A., 104, 105, 106, Self-world dualism, 149–151,
107, 108, 109 163–164
Secondary circular reactions (Piaget), Semantic-bootstrapping hypothesis
253 (Pinker), 23
Segal, G., 300 (n. 14) Shallice, T., 303 (n. 20)
Self Shapiro, M. F., 309 (n. 24)
in content of perception, 103–108 Shoemaker, S., 6, 8, 135–137, 143,
embodied, 131–162 (see also Proprio- 144, 299 (nn. 4, 5), 306 (nn. 4, 5),
ception, somatic) 307 (n. 14)
introspection of, 104, 136 Simner, M. L., 124
Self-consciousness. See also Simple argument about somatic pro-
Deflationary account of self- prioception, 134–135, 145
consciousness; First-person con- Simultagnosia, 79, 303 (n. 18)
tents; Nonconceptual first-person Size constancy, 304 (n. 2)
contents; Nonconceptual psycholog- Slater, A., 100
ical self-awareness; Paradox of self- Smolensky, P., 93–94, 304 (n. 8)
consciousness; Proprioception, Social referencing in infants, 262–264,
somatic 265
Index 337
Solid angles in vision, 106–107, 152 second-order, 30

Solidity, principle of (Spelke), 64–65, self-reference in, 30
74 utility conditions of, 29, 34, 37–39
Somatic proprioception. See Proprio- Symmetry Thesis, 230–241, 265, 274,
ception, somatic 311 (n. 7)
Somatoparaphrenic delusions, distinction between weak reading
151–152 and strong reading of, 230–232
Space. See also Integrated representa- and Generality Constraint, 233–237
tion of the environment; Naviga- neo-Lockean defence of, 237–241,
tion; Nonconceptual point of view; 265
Spatial awareness and relative identity, 240
absolute vs. egocentric representation strong reading of, 233
of, 201 weak reading of, 240–241, 247
allocentric vs. egocentric representa-
tion of, 204, 205 Taschek, W. W., 300 (n. 8)
connectivity of, 212 Taylor, C., 85, 120–121
two types of spatial relations, 194– Temporal orientation toward the past,
195, 197, 213–217, 220 178–180
understanding, 194–195, 196–197 Tensor-product representations, 94
Spatial-attention system, 80 Thought. See also Content; Represen-
Spatial awareness tational states
and nonconceptual point of view, analysis of, through linguistic expres-
193–198 sion (Dummett), 12–13, 40–41, 43
and self-consciousness, 198, communicability of (Dummett),
221–223 12–41
understanding places as places, relation of, to language, 12–13,
218–220 39–48
Spatial reasoning, 200–202. See also structured vs. unstructured, 38–39
Integrated representation of the Thought-Language Principle, 12–13,
environment 24–25, 39–45, 267–269, 275, 301
Spelke, E., 63–66, 302 (n. 9) (n. 4)
on concept of objects, 68 stated, 13
Split-brain patients, 239–240 Tobias, P. V., 78
Stage 4 error or A, not-B error Token-reflexive rule, 14–18, 275–276,
(Piaget), 70, 75 281–282, 290
Stern, D., 257–258 different versions of, 14–16
Stimulus-response (S-R). See Behavior and intentional self-reference, 276
Strawson, P. F., 9, 165–168, 169, 171, temporal dimension of, 299–300
172, 233–237, 300 (n. 12), 311 (n. 7)
(nn. 2, 3) Tolman, E. C., 206, 209–210, 212,
Subjective beliefs (Mellor), 29–31, 34, 215, 219
37–39, 117. See also First-person Tomasello, M., 255, 256
contents; Nonconceptual first- Touch, phenomenology of, 137–139.
person contents See also Haptic exploration
conscious, 29, 30 Transducers, 89
defined, 29 Transformational grammar, 22
338 Index
Traveling-salesman combinatorial
problem, 217
Travis, C., 311 (n. 1)
Trevarthen, C., 249, 251–252, 260–
261, 286
Ts’o, D. Y., 80
Tulving, E., 175
Universality-of-belief thesis, 35–37, 46
Valences, 113
Van de Walle, G. A., 63, 302 (n. 9)
Van Gulick, R., 90
Varlet, C., 215–216
Vestibular system, 133, 141, 150
Vibbert, M., 250
Visual agnosias, 79–81, 303 (n. 19)
Visual-cliff experiments, 113–114,
116, 263, 264
Visual kinesthesis. See Kinesthesis
Von Hofsten, C., 127
Wernicke’s area, 77
Wiggins, D., 238, 311 (n. 4)
Williams, G., 304 (n. 1)
Wimmer, H., 280
Wittgenstein, L., 5–6, 7, 104, 105, 106,
107, 108, 109, 299 (n. 1), 306 (n. 2)
Woodruff, G., 62, 300 (n. 16)
Woodson, R., 250

The Paradox of Self-Consciousness - José Bermúdez

Uploaded by

Copyright:

Available Formats

The Paradox of Self-Consciousness - José Bermúdez

Uploaded by

Document Information

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

The Paradox of Self-Consciousness - José Bermúdez

Uploaded by

Copyright:

Available Formats

Preface

The phenomenon of self-consciousness provides one of the enduring

The perspective from which this book is written is philosophical natu-

This book was conceived and written during my tenure (1993–1996) of

From Perceptual Experience to Subpersonal Computational States”

It is very natural to think of self-consciousness as a cognitive state or,

Confronted with the range of putatively self-conscious cognitive states

oneself. So it is natural to conclude that ‘I’-thoughts involve a distinctive

J. L. B. will be the next person to get a parking ticket in central Cam-

makes all those cognitive states modes of self-consciousness is the fact

1.2 Two Types of First-Person Content

The class of first-person contents is not a homogenous class. There is an

Definition To say that a statement “a is f” is subject to error through

This leaves us with a class of self-ascriptions that are identification-free.

1.3 The First-Person Pronoun and a Deflationary Account of

Since we cannot explain what immunity to error through misidentifica-

‘I’ refers. Nonetheless, this is not a counterexample to the suggested view,

component of such judgements that mastery of the semantics of the first-

We communicate thoughts by means of language because we have an implicit

1.4 Explanatory Circularity and Capacity Circularity

According to claim 3a, once we have explained what it is to master the

call that I characterized a first-person content as one that can be specified

person pronoun) will end up being viciously circular because mastery of

explanans. This is explanatory circularity. It is important to note, how-

capacities it already has to acquire the new capacity. If those conditions

cept can be acquired. The Acquisition Constraint has nothing to say

1.5 Capacity Circularity: An Innatist Solution?

It might be thought that capacity circularity as thus described both

to these questions will involve an answer to the question of how it is

1.6 The Paradox of Self-Consciousness

A paradox is an unacceptable conclusion or set of conclusions reachable

One possible reaction to the paradox of self-consciousness as outlined in

In the first section of this chapter I examine the functionalist account

2.1 The Functionalist Account of Self-Reference

The circularity problems at the root of the paradox of self-consciousness

pronoun is at the core of the functionalist theory of self-reference and

What secures self-reference in belief b( p) is the contiguity of cause and

If we adopt the functionalist theory, the paradox of self-consciousness

Let me start by mentioning one possible line of objection to the general

propriate desire(s) in the appropriate circumstances and provided there

consider them in reverse order. The universality-of-belief thesis that every

2.2 Rejecting the Classical View of Content

remembered, was one of the inconsistent set of propositions that together

adverted to by Dummett directly yield the conclusion that there cannot

a direct function of its degree of language mastery. There is, of course,

2.3 The Outline of a Solution

This distinction between conceptual and nonconceptual modes of

contents to creatures who have mastered the semantics of the first-person

If proposition (2) is rejected, then consistency will indeed be restored.

I undertake the first of these tasks in chapters 3 and 4. There I outline

What gives bite to the paradox of self-consciousness as developed in chap-

be an attack on the classical theory. In the remainder of this chapter I will

3.1 Conceptual and Nonconceptual Content: The Richness of

A dominant view among philosophers is that the ways in which a creature

how they represent. They are distinguished according to whether, in speci-

doxastic and nondoxastic onto the distinction between conceptual and

inferences that it licenses to other propositional-attitude contents, where

uated in terms of the normative dimensions of their inferential role, it

This general strategy is not, of course, confined to colors. Theorists of

it to have a component that is nondoxastic but nonetheless representa-