1

FregeGeach Objection
Mark van Roojen
The FregeGeach objection is a problem for non-cognitivist metaethical theories
that posit a non-descriptive semantics for moral or normative domains of discourse.
Such theories typically focus in the first instance on simple indicative sentences or
utterances. They say of these sentences that their meaning is importantly different
from the meaning of similar sentences employing non-normative predicates (see
non-cognitivism; semantics, moral). Early emotivists (see emotivism), for
example, suggested that simple moral sentences expressed non-cognitive attitudes
toward their objects rather than beliefs that predicate a property of their objects.
Most early non-cognitivist accounts had little to say about more complicated
and non-indicative sentences. The FregeGeach objection (or problem), as highlighted by Peter Geach (1965) and John Searle (1962), can be seen as the problem of
supplementing the account provided for simple sentences so as to cover other,
morecomplex constructions containing moral language. And it is part of that problem that the extension should meet certain conditions of adequacy. Among these
conditions are: (1) generating a meaning for the complexes that fits with their
ordinary use to express attitudes; (2) explaining how the meanings of the more
complex constructions are a function of the meanings of their parts; and (3) doing
this in a way that preserves and explains the logical relations that intuitively
holdbetween moral sentences. Lets call these conditions (1) semantic completeness;
(2) compositionality; and (3) logical adequacy.

Semantic Completeness
Ordinarily non-cognitivists deny that the semantics of normative expressions is like
the semantics of ordinary non-normative expressions with the same surface grammar. Yet normative predicates occur in all of the same constructions in
which non-normative predicates do. Just as sentences employing ordinary, run-ofthe-mill predicates get embedded in more complex sentences, so do moral and other
normative predicates. We use sentences such as the following, and many more:
Eating meat is not wrong.
Eating meat is wrong and wasting food is wrong.
If eating meat is wrong, hes sure to do it.
Is eating meat wrong?
It is true that eating meat is wrong.
Sally believes that eating meat is wrong.

The International Encyclopedia of Ethics. Edited by Hugh LaFollette, print pages 20372046.
2013 Blackwell Publishing Ltd. Published 2013 by Blackwell Publishing Ltd.
DOI: 10.1002/ 9781444367072.wbiee313

A simple non-cognitivist theory that says, of Eating meat is wrong, that it expresses
a negative attitude toward eating meat needs supplementation to cover these locutions.
Since many of these sentences are used by people with no negative attitudes toward
eating meat, that supplementation must avoid saying that these more complex uses
express such attitudes. Furthermore, the point seems to generalize: embedded
sentences need not express the attitudes they would express in free-standing use. Any
complete non-cognitivist theory must somehow recognize this if it is to provide a
plausible account of the way speakers use these terms.

Compositionality
Speakers are able to understand the meanings of sentences they have not previously
encountered. This militates in favor of the second desideratum compositionality.
In general, competent speakers of a language can understand the meanings of novel
moral sentences upon first encountering them, so long as they already know how to
understand the sentences they embed and the connectives employed.
The problem with this, for nondescriptive semantic theories, arises because such
theories reject one way of meeting this desideratum. On standard cognitivist views,
simple sentences of subjectpredicate form, such as Fred is alive, predicate a
property of a subject. Fred contributes a person to the proposition expressed by the
sentence, and is alive contributes to it a property that of being alive. Such sentences
are true just in case the subject in fact has the property in question. From this kind
of starting place, the meanings of more complex sentences can be explained as a
function of the meanings of their constituents. Sentences negating a sentence will be
true just in case the original sentence is false. Conjunctions will be true just in case
both conjuncts are true, and so on. A person who knows what properties first-order
predicates stand for and what objects are picked out by subject terms, as well as the
rules for negation, conjunction, and so on, will be in a position to interpret novel
sentences that combine those constituents. The meanings of the complex sentences
will be a function of the constituents contributed by the terms of which they are
composed. But nondescriptivists deny that normative predicates work in this way.
So they need another way of explaining the compositionality of sentences that use
moral predicates.

Logical Adequacy
This brings us to the third desideratum to make the account consistent with the
logical relations between diverse sentences. As a general matter, sentences containing
normative predicates seem to have the same logical forms as sentences containing
non-normative predicates. Ridiculing children is fun contradicts Ridiculing
children is not fun. The same relation obtains when we replace fun with wrong.
Furthermore, conditionals embedding normative claims in their antecedents can be
used in all of the patterns of reasoning that non-normative conditionals allow.
Similar results hold for all the other sentential connectives, such as and and or.

As we noted above, descriptive semantics allows indicative sentences to represent

propositions that must be true for the sentences to be true, and these propositions
involve the properties corresponding to the predicates in the sentences. This puts
descriptivist accounts in a relatively good position to explain logic for all domains.
Sentences employing logical vocabulary such as not, or, and, if then, and so on
stand in the logical relations that they do because they express or represent
propositions that stand in analogous logical relations to one another. To take the
simplest case, the sentence Sally is tired is inconsistent with the sentence Sally is
not tired, just because its content the proposition that must be true if the sentence
is true cannot be true if the content of Sally is not tired is true. Similarly for
Sammy is corrupt and Sammy is not corrupt. Insofar as nondescriptivists deny that
simple normative sentences of the same syntactic form predicate genuine properties
and express propositions, they cannot adopt this relatively simple story for parallel
explanations in the normative domain.
Early non-cognitivist accounts of the meanings of unembedded sentences made it
hard to see how normative expressions retain a single meaning across a range of different sentences. Standard non-cognitivist theories explained the meanings ofnormative
terms in declarative sentences by telling us that they served to express a non-cognitive
attitude of disapproval or to mark that the speaker is performing a speech act, for
example condemnation. Yet, as noted above, the hypothetical antecedent in a conditional expresses no such attitude and performs no such speech act. Heres Geach:
Let us consider this piece of moral reasoning:
If doing a thing is bad, getting your little brother to do it is bad.
Tormenting the cat is bad.
Ergo, getting your little brother to torment the cat is bad.
The whole nerve of the reasoning is that bad should mean exactly the same at all four
occurrences should not, for example, shift from an evaluative to a descriptive or
conventional or inverted-commas use. But in the major premise the speaker (a father,
let us suppose) is certainly not uttering acts of condemnation: one could hardly take
him to be condemning just doing a thing. (1965: 4634)

Our problem gets its name from the fact that Geach attributes this point that
embedded claims must mean the same as their unembedded counterparts to
Frege. Non-cognitivists have heeded Geachs and Searles warning and developed
sophisticated nondescriptive accounts of normative semantics that avoid the charge
of equivocation. Still, non-cognitivists typically have to complete their account of
the whole of normative language without using many of the resources that make
semantics relatively straightforward for the cognitivists with whom they disagree.

Some Standard Non-Cognitivist Strategies

Non-cognitivists have not been without resources or ingenuity in pursuing this task.
The sections below sketch some of the various approaches that have been tried,
along with some problems encountered by each.

Imperative logic
Even before Geach and Searle highlighted the issue, R. M. Hare (1952) was at work
trying to offer a normative semantics that assimilated the logic of normative
utterances to the logic of imperatives (see prescriptivism; hare, r. m.). Hares
work was sophisticated in many ways. But its main motivating idea was pretty
intuitive. Both of the following two arguments strike us as sound, and for related
reasons:
If it is raining, take an umbrella.
It is raining.
Take an umbrella.

If it is raining, you will take an umbrella.

It is raining.
You will take an umbrella.

The example generates grounds for non-cognitivist optimism about the possibility
of an explanation of the logic of nondescriptive expressions. There is in principle no
reason why nondescriptive sentences cannot stand in logical relations to other
sentences. The mere fact that such sentences arent straightforwardly apt for truth or
falsity or that they dont represent the world as being in a certain way does not
disqualify them from having a logic. Nor is this logic unrelated to the logic of
descriptive sentences that are apt for truth and falsity. Just as natural language
includes descriptive conditionals, the imperative also occurs sometimes in
conditionals as when we ask someone to close the door if it is raining.
Still, it isnt obvious how to make this sort of proposal fully general. There seem to
be no natural English constructions embedding imperatives in the antecedents of
conditionals, nor is it obvious how to interpret such embeddings, were we to extend
natural language so as to allow them.

Minimalism, syntax, and discipline

Minimalist or deflationist theories about truth and truth-aptness have had some
currency in recent years, and some proponents have suggested that they might help
with the FregeGeach problem. Deflationist theories of truth suggest that truth is
not a robust or a substantial property, and that truth predications are not to be
understood, at least in the first instance, as predicating a substantial property of the
objects to which they are applied (see minimalism about truth, ethics and).
Rather, a truth predicate helps us to say things of a sort we say anyway, but it allows
us to generalize or expand the range of this ability. One kind of minimalist might say
that we understand what the truth predicate does when we understand that sentences
of the form X is true are appropriately assertable just when X would be assertable,
and that such sentences say no more than what X would have said, were we to assert
it. Such predicates are useful because they allow us to say things like Everything that
Fred said is true without having to repeat everything that Fred said, or even to know
everything that Fred said.
Some theorists also favor minimal accounts of the difference between truth-apt discourse and discourse that does not admit of either truth or falsity. On one such view,

all it takes for an atomic sentence to be capable of truth or falsity is to be in the indicative mood and to have a standard use to perform a recognizable speech act. Very often
the two views are held together, and some non-cognitivists have proposed that the
resulting view can help with the FregeGeach problem (Horwich 1993; Stoljar 1993).
It is pretty easy to see how minimalism about truth will yield a meaning for one
sort of complex normative construction, namely constructions predicating truth of
a normative sentence or claim. Those sentences say just what an assertion of the
free-standing embedded claim would say. A sentence such as It is true that lying is
wrong says no more and no less than the sentence Lying is wrong. We can add
this rule to any theory of the meaning of atomic normative sentences to give a
compositional account of such sentences.
Some minimalists think that we can go further than that, by putting minimalism
about truth-aptness to work alongside deflationism about truth. Given that atomic
normative sentences are in the indicative mood and that non-cognitivists typically
tell us the speech acts they are conventionally used to perform, they are minimally
truth-apt and we can generate a minimal truth condition for them. For example,
Lying is wrong is true just in case lying is wrong. And, the proposal goes, from here
we can use a truth-functional account of the meanings of various logical connectives
to explain the meanings of more complex expressions.
There are some reasons for skepticism, however. Not every sort of speech act
not even one with a highly regulated character and employing a specific form
of words to be accomplished seems to be of the right sort to give an enlightening story about the meaning of a complex that disjoins words of that form. Nor
do things seem to get much better if the form of words is syntactically in the
indicative mood. Suppose we started to greet one another by using hello in the
predicate position, so that I could greet Sally with Sally is Hello, and that this
convention became universally adopted. It is not at all obvious how that would
help us understand the meaning of Either Sally is Hello, or Jim is Lazy (Dreier
1996).

Expressivism
Most recent non-cognitivist theories have been expressivist. The basic expressivist
idea is to satisfy the completeness constraint by pairing each grammatical naturallanguage sentence with an attitude that provides its meaning. Compositionality can
then be achieved if the attitudes expressed by complex sentences are a predictable
function of the attitudes expressed by free-standing occurrences of the atomic
sentences of which the complexes are composed. The hope is that the attitudes
expressed by such sentences already stand in certain logical relations to one another.
In other words, expressivists contend that certain attitudes, however expressed, are
logically incompatible with certain other attitudes. Perhaps, for example, approval of
murder is incompatible with disapproval of murder. If that is correct, expressivists
hope to explain the logic of certain linguistic expressions as falling out of the logical
relations among the attitudes that those expressions express.

Different theorists have proposed different analyses of this broad sort, each
differing from the next with respect to the attitudes involved, and also with respect
to the nature of the sort of inconsistency that is logically relevant.
Higher order attitude expressivism One expressivist strategy has been to generate a compositional semantics for the relevant domains by suggesting that certain
logical connectives help to form sentences that express higher order attitudes
toward the attitudes expressed by the claims that the resulting complex embeds.
Simon Blackburn (1984) proposes that conditionals express an attitude of approval
or disapproval toward conjunctions of attitudes. For example, If lying is wrong,
telling your little brother to lie is wrong, when sincerely uttered, might express
disapproval of the following combination: disapproving of lying without also disapproving of telling ones brother to lie. If thats correct, there would then be a kind
of incoherence in holding that attitude while at the same time disapproving of
lying but not disapproving of telling ones brother to lie. That incoherence, then,
can be used to explain the inconsistency of sentences that express the attitudes in
question. First, we explain that attitudes are inconsistent when they display this
sort of incoherence. Second, we explain the inconsistency of sentences as falling
out of their expressing attitudes that are themselves inconsistent in this sense.
Similar accounts would have to be given of other sentential connectives to make
the story fully general.
Various critics have argued that the proposal is inadequate by virtue of conflating
one kind of incoherence (perhaps pragmatic incoherence) with another kind the
kind involved in genuine logical inconsistency. It has been known for a long time
that not every set of incoherent attitudes involves accepting beliefs with inconsistent
contents. There is something incoherent and self-defeating about accepting and
asserting sentences like G. E. Moores example, It is raining and I dont believe that
it is. But the claim it makes is not logically inconsistent. It could be true. So the
inconsistency is something other than true logical inconsistency. One way to
pursuethis argument is by showing that higher order attitude accounts are apt to
overgenerate instances of putative logical inconsistency. They will consider
inconsistent any attitudes that display the requisite incoherence and any sentences
that express those same attitudes. Yet, these critics maintain, not all such sentences
or attitudes involve logical inconsistency (van Roojen 1996).
Committing attitude expressivism Simon Blackburn has coined the term quasirealism for the research program of explaining the realist-seeming nature of moral
discourse by using only the materials to which non-cognitivists are entitled (see quasirealism). Fellow travelers worthy of the name include Allan Gibbard, Terry Horgan,
and Mark Timmons. Partly in response to worries about FregeGeach, several of these
theorists have postulated something of the order of a new attitude type, not easily
expressible (if at all) by using non-normative vocabulary. Different theorists have proposed different versions of the idea. Allan Gibbard at one point worked with normacceptance, whereas more recently he has suggested that the state of mind in question

is a kind of planning attitude. Horgan and Timmons (2006) postulate ought-beliefs,

which are propositional attitudes of a sort distinct from the ordinary non-normative
beliefs that they dub is-beliefs. These proposals all postulate that the target attitudes
commit a person to reasoning in certain ways if she reasons with them at all, and that
certain combinations of such attitudes generate genuinely inconsistent commitments.
Importantly, these theorists can be very strict about the kind of inconsistency in question, so that Moore-type incoherence does not count as logical inconsistency.
The proposals are diverse, and much of their payoff for FregeGeach turns on
relatively technical features of the resulting accounts. So the following discussion
will oversimplify to some extent and still be difficult going. Allan Gibbards (2003)
proposal is the most developed. He suggests that we can capture both ordinary
descriptive beliefs and normative (or planning) states by determining the set of
states of mind with which they disagree. The attitudes expressed by simple normative
and descriptive sentences can thus be represented by the set of such judgments that
they rule out. Sentences expressing those judgments will be inconsistent with one
another just when the judgments they express are similarly inconsistent.
Gibbard exploits this idea to construct a formal system for tracking and explaining the
logical relations between simple and more complex judgments and the sentences that
express them. He uses possible worlds to represent belief contents, and detailed plans to
represent the normative attitude of being committed to a plan. A descriptive belief rules
out all the possible worlds in which it is false, but no plans. So the attitudes it rules out
can be represented by the set of pairs pairing each of those ruled-out worlds with every
possible plan. A purely normative judgment rules out all the plans inconsistent with the
plan it commits one to, but no beliefs. So the attitudes it rules out can be represented by
the set of pairs of worlds and plans where the plan-member of each pair is inconsistent
with the plan it commits one to. More complex judgments are composed of combinations
of such simple cognitive and normative judgments. Conjunctive judgments are
inconsistent with any attitude that a component judgment rules out, so they rule out the
set of world/plan pairs ruled out by any conjunct. Disjunctive judgments rule out any
attitude ruled out by all the conjuncts, so they can be represented as ruling out the
corresponding world/plan pairs. Similar accounts are given of the other logical
connectives. The payoff is that we can use these rules to capture the logical relations
between judgments of all kinds normative, descriptive, and mixed. From this we get an
isomorphic story about the logical relations between sentences.
The negation problem Unwin (2002) raises a general worry about extant
quasi-realist proposals of this sort and their treatment of negation: at some level
they help themselves to that which they should be trying to explain. When I
disagree with an ordinary descriptive belief of yours, I do so by believing something something that is inconsistent with what you believe. Ordinary beliefs are
inconsistent just when their contents are. But extant non-cognitivist proposals
dont generate contents that are complex enough to allow this sort of explanation
one that generates disagreement or inconsistency just when we hold an attitude
of a single type towardcontents that are inconsistent (Schroeder 2008). Instead,

Unwin argues, non-cognitivists are forced to postulate at least two distinct attitude
types, one of which is expressed by atomic normative sentences and the other of
which is expressed by their negations.
Recall that the expressivist program is to explain the logical properties of
sentences as falling out of the logical properties of the attitudes they express. So,
for each sentence whose logical properties we aim to explain, there must be an
attitude that corresponds to it. Take the sentence Lying is wrong. We can
correlate it with a state of mind that it expresses, one we normally call believing
that lying is wrong, but which non-cognitivists will want to treat as importantly
different from non-normative beliefs. Its negation, Lying is not wrong, should
then express an attitude that contradicts it. It is the attitude we normally attribute
when we say that a person believes that lying is not wrong. If the normative
content of the sentence is contributed by the predicate wrong, and if the way that
predicate contributes the content is by expressing a certain practical attitude
toward the act type contributed by the rest of the sentence, it looks as if the
attitude cant be of the same sort as the un-negated judgment. Suppose we want
to figure out which of Sallys possible states of mind would be expressed by Lying
is not wrong, and suppose we think that sentences employing wrong express
attitudes of a type well call disapproval. Then Lying is wrong will express the
attitude Sally has when she disapproves of lying. What is her state of mind when
she accepts its negation? It looks like we have the following possibilities:
Sally disapproves of not lying.
Sally does not disapprove of lying.
But the first doesnt seem to be what we want, because Sally can think neither Lying
is wrong nor Not lying is wrong consistent with accepting the negation of Lying
is wrong. And the second is compatible with having no view about lying whatsoever. (Note: nothing here really turns on the fact that we have chosen disapproval as
our attitude type. Analogous problems seem to arise whenever we try to make do
with just one attitude type and an unstructured attitudinal semantics for normative
terms.)
Expressivists can and do try to generate an extensionally adequate story by
postulating a distinct attitude type, which one expresses when one accepts the
negation of a normative claim (Horgan and Timmons 2006). This seems to give
up on the compositionality desideratum, at least when it is generalized to negations of negations and beyond. It is fair to ask, of each distinct attitude type, why
it stands in the relations of inconsistency in which it stands. Schroeder (2008)
presses this objection especially clearly. He also suggests that the problem is not
confined to the treatment of negation, because other connectives raise related
problems. He goes on to suggest a way forward for expressivism to allow
normative terms to contribute a normative attitude plus something more, such as
a property or a relation. The additional complexity can then be used to provide
an additional place for negation to operate. Placing this suggestion within a

compositional semantics for the whole of a language requires the acceptance of

some surprising claims, about purely descriptive semantics as well as about
normative semantics.

Hybrid theories and fictionalism

Some theorists partial to non-cognitivism have attempted to employ descriptivist
resources to handle the FregeGeach problem while at the same time suggesting that
non-cognitivism is right rather than wrong. According to such views, expressivists
are right to think that moral terms function to express a non-cognitive attitude. But
they need not deny that these statements, in a context, have descriptive content. One
version of this idea is adopted by fictionalists, who suggest that indicative moral
sentences have a descriptive semantics even while they are normally used by speakers
to express attitudes other than belief in their contents. If the sentences have just the
descriptive content they seem to have, there may be no special problem with
explaining how they stand in relations of implication to one another (Kalderon
2005; see fictionalism, moral).
Another way is to combine descriptive and nondescriptive elements in the
semantics (see hybrid theories of moral statements). For instance, a moral
sentence might express disapproval of a property, which at the same time it predicates
of an object. On some such views, the identity of the predicated property is a function
of the non-cognitive attitude that the sentence also expresses (Ridge 2006). On other
such views, moral terms function more like epithets with constant descriptive
content (Boisvert 2008). Advocates of these theories try then to use the postulated
descriptive content to explain the logical relations that hold among such utterances.
The extent to which these theories succeed is a matter of some controversy, and the
terrain is too complex to be surveyed here. Schroeder (2009) offers an extensive
overview of various hybrid theories and of the potential problems for each.
See also: ayer, a. j.; emotivism; fictionalism, moral; hare, r. m.; hybrid
theories of moral statements; minimalism about truth, ethics and;
non-cognitivism; prescriptivism; quasi-realism; semantics, moral
REFERENCES
Blackburn, Simon 1984. Spreading the Word. Oxford: Clarendon.
Boisvert, Dan 2008. Expressive-Assertivism, Pacific Philosophical Quarterly, vol. 89,
pp. 169203.
Dreier, James 1996. Expressivist Embeddings and Minimal Truth, Philosophical Studies,
vol. 83, pp. 2951.
Geach, Peter 1965. Assertion, in The Philosophical Review, vol. 74, pp. 44965.
Gibbard, Allan 2003. Thinking How to Live. Cambridge, MA: Harvard University Press.
Hare, R. M. 1952. The Language of Morals. Oxford: Clarendon.
Horgan, Terrance, and Mark Timmons 2006. Cognitivist Expressivism, in Horgan and
Timmons (eds.), Metaethics After Moore. Oxford: Oxford University Press, pp. 25598.

10
Horwich, Paul 1993. Gibbards Theory of Norms, Philosophy and Public Affairs, vol. 22,
pp. 6779.
Kalderon, Mark Ely 2005. Moral Fictionalism. Oxford: Oxford University Press.
Ridge, Mike 2006. Ecumenical Expressivism: Finessing Frege, Ethics, vol. 116, pp. 30236.
Schroeder, Mark 2008. Being For: Evaluating the Semantic Program of Expressivism. Oxford:
Oxford University Press.
Schroeder, Mark 2009. Hybrid Expressivism: Virtues and Vices, Ethics, vol. 119,
pp. 257309.
Searle, John 1962. Meaning and Speech Acts, Philosophical Review, vol. 71, pp. 42332.
Stoljar, Daniel 1993. Emotivism and Truth Conditions, Philosophical Studies, vol. 70,
pp. 81101.
Unwin, Nicholas 2002. Norms and Negation: A Problem for Gibbards Logic, Philosophical
Quarterly, vol. 49, pp. 6075.
van Roojen, Mark 1996. Expressivism and Irrationality, Philosophical Review, vol. 105,
pp. 31155.

FURTHER READINGS
Blackburn, Simon 1993. Essays in Quasi-Realism. New York: Oxford University Press.
Blackburn, Simon 1998. Ruling Passions. Oxford: Clarendon Press.
Dreier, James 2006. Negation for Expressivists: A Collection of Problems with a Suggestion
for their Solution, in Shafer-Landau (ed.), Oxford Studies in Metaethics, vol. 1. Oxford:
Oxford University Press, pp. 21734.
Geach, Peter T. 1960, Ascriptivism, Philosophical Review, vol. 69, pp. 2215.
Gibbard, Allan 1990. Wise Choices, Apt Feelings. Cambridge, MA: Harvard University Press.
Hale, Bob 1993. Can There Be a Logic of Attitudes? in John Haldane and Crispin Wright
(eds.), Reality, Representation and Projection. Oxford: Oxford University Press, pp. 33763.
Schroeder, Mark 2008. What Is the FregeGeach Problem? Philosophy Compass, vol. 3. At
http://onlinelibrary.wiley.com/doi/10.1111/j.1747-9991.2008.00155.x/abstract.
Searle, John 1969. Speech Acts. Cambridge: Cambridge University Press.