The Rule-Following Considerations
Paul A. Boghossian
Mind, New Series, Vol. 98, No. 392. (Oct., 1989), pp. 507-549.
Stable URL:
http://links.jstor.org/sici?sici=0026-4423%28198910%292%3A98%3A392%3C507%3ATRC%3E2.0.CO%3B2-T
Mind is currently published by Oxford University Press.
Your use of the JSTOR archive indicates your acceptance of JSTOR's Terms and Conditions of Use, available at
http://www.jstor.org/about/terms.html. JSTOR's Terms and Conditions of Use provides, in part, that unless you have obtained
prior permission, you may not download an entire issue of a journal or multiple copies of articles, and you may use content in
the JSTOR archive only for your personal, non-commercial use.
Please contact the publisher regarding any further use of this work. Publisher contact information may be obtained at
http://www.jstor.org/journals/oup.html.
Each copy of any part of a JSTOR transmission must contain the same copyright notice that appears on the screen or printed
page of such transmission.
The JSTOR Archive is a trusted digital repository providing for long-term preservation and access to leading academic
journals and scholarly literature from around the world. The Archive is supported by libraries, scholarly societies, publishers,
and foundations. It is an initiative of JSTOR, a not-for-profit organization with a mission to help the scholarly community take
advantage of advances in technology. For more information regarding JSTOR, please contact support@jstor.org.
http://www.jstor.org
Wed Aug 22 10:12:31 2007
The Rule-Follo wing Considerations'
PAUL A. BOGHOSSIAN
This is the Jifth of our commissioned State of the Art Series
INTRODUCTION
I. Recent years have witnessed a great resurgence of interest in the
writings of the later Wittgenstein, especially with those passagesroughly,
Philosophical Investigations ##138-242 and Remarks on the
Foundations of Mathematics, section VI-that are concerned with the topic
of rules. Much of the credit for all this excitement, unparalleled since the
heyday of Wittgenstein scholarship in the early 1g6os, must go to Saul
Kripke's Wittgenstein on Rules and Priuate ~anguage.' It is easy to explain
why.
T o begin with, the dialectic Kripke uncovered from Wittgenstein's
discussion is enormously exciting on its own terms. On Kripke's reading,
the passages on rule-following are concerned with some of the weightiest
questions ir! the theory of meaning, questions-involving the reality,
reducibility, and privacy of meaning-that occupy centre-stage in contemporary
philosophy. Furthermore, Kripke represented Wittgenstein as
defending a set of unified and extremely provocative claims concerning
these questions. And, finally, he argued for these claims with power and
clarity. The ensuing flood of articles and books on the subject of rulefollowing
was both predictable and warranted.
The present paper is the result of an invitation to survey this literature.
It could have been about exegetical matters, on what the recent discussions
have had to teach us about the historical Wittgenstein's philosophical
views. In the event, however, it is almost entirely concerned with a
retrospective assessment of the philosophical contributions. Limitations of
space dictated that a choice be made; and the philosophical assessment
seemed the more fruitful thing to doe3Despite a lot of discussion, there is
room for an improved understanding of the precise nature of Kripke's
' I am grateful to many people for helpful discussion of the issues covered in this paper, including
Mark Johnston, John Burgess, Jerry Fodor, Barry Loewer, Richard Rorty, Barry Allen, Larry Sklar,
Crispin Wright, Saul Kripke, Yeil Tennant, Steve Yablo, Kick White, and participants in various
seminars at the University of Michigan. Special thanks are due to Paul Benacerraf, Jennifer Church,
and David Velleman.
Cambridge, Harvard University Press, 1982. Henceforth, 'K'.
The main reason is that I have actually come to despair of a satisfactory interpretation of
Wittgenstein's views. I try to say why in 'The Problem of Meaning in Wittgenstein', to appear in
Meaning Scepticism, ed. K. Puhl, De Gruyter, forthcoming.
Mind, Val. 98 . 392 . October 1989 @ Oxford University Press 1989
508 Paul A. Boghossian
arguments, of their ultimate cogency, and of their relation to the wider
discussion of meaning in contemporary philosophy of mind and language.
Pulling on the thread that is Kripke's argument leads quite naturally to a
discussion of many of the most significant issues occupying philosophers
today; in that lies the main impetus behind the present essay.
I proceed as follows. In parts I and 11, I lay out the essentials of
Kripke's argument. In subsequent parts, I offer an extended critique of
the dialectic it presents, considered on its own terms and independently
of exegetical concerns. A discussion of the critical literature will be
woven in as appropriate. The moral will not be recognizably Wittgensteinian:
I shall argue that, pace Kripke's intent, the conception of meaning that
emerges is a realist, non-reductionist, and judgement-independent conception,
one which, moreover, sustains no obvious animus against private
language.
I 

KRIPKE ON MEANING AND T H E SCEPTICAL 

PROBLEM 

The sceptical problem
2 . As Kripke sees it, the burden of the rule-following considerations is that
it cannot literally be true of any symbol that it expresses some particular
concept or meaning. This is the now-famous 'sceptical conclusion' he
attributes to Wittgenstein:
[Tlhere is no fact about me that distinguishes between my meaning a definite
function by '+' . . . and my meaning nothing at alL4
How is such a radical thesis to be supported? Kripke argues, in effect, by
elimination: all the available facts potentially relevant to fixing the
meaning of a symbol in a given speaker's repertoire-facts about how the
speaker has actually used the expression, facts about how he is disposed to
use it, and facts about his qualitative mental history-are canvassed, and
found wanting. Adequate reflection on what it is for an expression to
possess a meaning ~vouldbetray, so Kripke invites us to believe, that that
fact could not be constituted by any of those.
The claim is, of course, indisputable in connection with facts about
actual use and qualitative phenomena; it is a familiar and well-assimilated
lesson of, precisely, Wittgenstein's Investigations, that neither of those
species of fact could, either in isolation or in combination, capture what it
is for a symbol to possess a meaning. Much more important and
controversial, however, is Kripke's rejection of a dispositional account of
The Rule-Following Considerations 509
meaning facts. Why are facts about how a speaker is disposed to use an
expression held to be insufficient to determine its meaning?
Kripke develops two sorts of consideration. First, the idea of meaning
something by a word is an idea with an infinitary character-if I mean plus
by '+',then there are literally no end of truths about how I ought to apply
the term, namely to just the members of this set of triples and not to
others, if I am to use it in accord with its meaning. This is not merely an
artefact of the arithmetical example; it holds for any concept. If I mean
horse by 'horse', then there are literally no end of truths about how it
would be correct for me to apply the term-to horses on Alpha Centauri,
to horses in Imperial Armenia, and so on, but not to cows or cats wherever
they may be-if I am to use it in accord with its meaning. But, Kripke
argues, the totality of my dispositions is finite, being the dispositions of a
finite being that exists for a finite time. And so, facts about dispositions
cannot capture what it is for me to mean addition by '+'.
The second objection to a dispositional theory stems from the so-called
'normativity' of meaning. This objection is somewhat harder to state, but a
rough formulation will do for now. The point is that, if I mean something
by an expression, then the potential infinity of truths that are generated as
a result are normatine truths: they are truths about how I ought to apply the
expression, if I am to apply it in accord with its meaning, not truths about
how 1,will apply it. My meaning something by an expression, it appears,
does not guarantee that I will appb it correctly; it guarantees only that
there will be a fact of the matter about whether my use of it is correct.
Now, this observation may be converted into a condition of adequacy on
theories of meaning: any proposed candidate for being the property in
virtue of which an expression has meaning must be such as to ground the
normativity of meaning-it ought to be possible to read off from any
alleged meaning-constituting property of a word, what is the correct use of
that word. And this is a requirement, Kripke maintains, that a dispositional
theory cannot pass: one cannot read off a speaker's disposition to use
an expression in a certain way what is the correct use of that expression, for
to be disposed to use an expression in a certain way implies at most that
one will, not that one should.
The contents of thought
3. But what about thoughts, intentions, and other content-bearing mental
states?How do they figure in the sceptical argument? More specifically: is
the sceptical thesis directed against them as well, or is it confined solely to
linguistic representation?
It is hard to see how a convincing meaning scepticism could be confined
purely to the linguistic domain, given the intimate relation between
thought and language. Philosophers divide, of course, on the precise
5 I o Paul A. Boghossian
nature of this relation and, in particular, on the question of priority: Do
the semantic properties of language derive from the representational
properties of thought, or is it the other way round?5Whatever the correct
answer, however, there would appear to be no plausible way to promote a
language-speczjic meaning scepticism. On the former (Gricean) picture, one
cannot threaten linguistic meaning without threatening thought content,
since it is from thought that linguistic meaning is held to derive; and on
the latter (Sellarsian) picture, one cannot threaten linguistic meaning
without thereby threatening thought content, since it is from linguistic
meaning that thought content is held to derive. Either way, content and
meaning must stand or fall together.
If a sceptical thesis about linguistic meaning is to have any prospect of
succeeding, then, it must also threaten the possibility of mental meaning
(or content). Of course, on a Sellarsian view, that result is automatic, given
a demonstration that nothing non-mental fixes linguistic meaning. But on a
Gricean view matters are not so simple. Since the Gricean holds that
linguistic items acquire their meaning from the antecedently fixed content
of mental states, an argument to the effect that nothing non-mental fixes
linguistic meaning would leave the Gricean unmoved; he needs to be given
a separate argument against the possibility of mental content. Does Kripke
see this need and does he show how it is to be met?
Colin McGinn has argued that the answer to both questions is 'no':
My third point . ..points up a real lacuna in Kripke's presentation of his paradox.
The point is that it is necessary for Kripke to apply his paradox at the level of
concepts; that is, he has to argue that the notion of possessing a determinate
concept is likewise devoid of factual foundation . ... It cannot be said, however,
that Kripke explains how this need is to be met, how this extension of the paradox
to the level of concepts is to be carried out; and brief reflection shows that the
exercise is by no means triviaL6
I think McGinn is wrong on both counts; it will be worthwhile to see why.
In fact, the suggestion that some appropriately general thought or
intention constitutes the sought after meaning-determining fact comes up
early in Kripke's presentation, before the dispositional account of meaning
is considered and found wanting:
In the United States, it is the Gricean view, that linguistic expressions acquire their semantic
properties by virtue of being used with certain intentions, beliefs, and desires, that is most influential;
whereas in Britain it appears to be the Sellarsian (Wittgensteinian?) view that thinking is a form of
internalized speaking-speech inforo interno, as Sellars likes to put it-that tends to predominate.
For the Gricean view see H. P. Grice, 'Meaning', Philosophical Review, 1957; and related papers.
See also, S. Schiffer, Meaning, Oxford, Clarendon Press, 1972. For the Sellarsian view see his
'Empiricism and the Philosophy of Minď, in his Science, Perception and Realitj~,London, Routledge
and Kegan Paul, 1963. For a debate on the priority question see 'The Chisholm-Sellars Correspondence',
in Intentionality, Mind and Language, ed. 4. Marras, Urbana, University of Illinois Press,
'972
Colin McGinn, Wittgenstein on Meaning, Oxford, Basil Blackwell, 1984, pp. 144-6.
The Rule-Following Considerations 5 I I
This set of directions, I may suppose, I explicitly gave myself at some earlier time.
It is engraved on my mind as on a slate. It is incompatible with the hypothesis that
I meant quus. It is this set of directions, not the finite list of particular additions
that I performed in the past, that justifies and determines my present response.7
And his response to it seems clear (p. 16@.The idea is that thoughts that
someone may have had concerning how he is prepared to use a certain
expression will help determine a meaning for that expression only if their
correct interpretation is presupposed. But this is equivalent to assuming,
Kripke suggests, that the sceptical challenge has been met with respect to
the expressions that jgure in those thoughts. But how was their meaning
fixed? Not by facts about their actual or counterfactual history of use, (if
the argument against a dispositional account of meaning is to be believed);
and not by facts concerning associated experiential episodes. Hence-on
the assumption that no other sort of fact is relevant to the fixation of
meaning-by nothing.
The strategy seems clear; but is it not problematic?The trouble is that it
seems to depend on the assumption that thought contents are the
properties of syntactically identifiable bearers-properties, that is, of
expressions belonging to a 'language of thoughť. And although there may
be much to recommend this view, still, does Kripke really wish to rest the
sceptical conclusion on so contestable a premiss?
Fortunately for the sceptical strategy, we will see below that, although a
contestable premiss about thought is involved, it is nothing so rich as a
language of thought hypothesis. But we will be in a position to appreciate
this properly only after we have examined McGinn's claim that, even
granted a linguistic model of thinking, it is still impossible to run a Kripkestyle
sceptical argument against thought.
The normativity of meaning
4. McGinn writes:
The issue of normativeness, the crucial issue for Kripke, has no clear content in
application to the language of thought: what does it mean to ask whether my
current employment of a word in my language of thought (i.e. the exercise of a
particular concept) is correct in the light of my earlier employment of that word?
What kind of linguistic mistake is envisaged here? . . . There is just no analogue
here for the idea of linguistic incorrectness (as opposed to thefalsity of a thought):
linguistic incorrectness (of the kind we are concerned with) is using the same word
with a different meaning from that originally intended (and doing so in ignorance
of the change), but we cannot in this way make sense of employing a concept with
a different content from that originally intended-it would just be a d@erent
concept.
The idea of mental content cannot be threatened by Kripke, McGinn
' K., pp. 15-16. Op. cit., p. 147.
512 Paul A. Boghosszan
argues, because the principal requirement by which putative reconstructions
of that notion are to be dispatched-the normativity requirement-has no
cogent application to the language of thought. The claim calls for a
somewhat more searching articulation of the normativity thesis than we have
attempted so far. In what does the normativity of meaning consist?
McGinn offers the following characterization:
The notion of normativeness Kripke wants captured is a transtemporal notion . ...
We have an account of this normativeness when we have two things: (a) an
account of what it is to mean something at a given time and (b) an account of what
it is to mean the same thing at different times-since (Kripkean) normativeness is
a matter of meaning now what one meant earlier.9
So, the later use of the expression is 'correcť, according to McGinn, if it
then expresses the same meaning as it did earlier; 'incorrecť if, without
intending to introduce a change of meaning by explicit stipulation, it
expresses a different meaning. It is in such facts as this that the
normativity of meaning is said to consist.
Supposing this were the right understanding of normativity, how would it
affect mental content scepticism? McGinn says that the problem is that we
cannot make sense of employing a concept with a different content from that
originally intended-it would just be a different concept. But although that
is certainly true, it is also irrelevant: what we need to make sense of is not
employing a concept with a different content from that originally intended,
but employing an expression in the language of thought with a different
content from that originally intended, which is a rather different matter.
As it happens, however, it is an idea that is equally problematic. The
difficulty is that we do not have the sort of access to the expressions of our
language of thought that an attribution to us of semantic intentions in
respect of them would appear to presuppose. You cannot intend that some
expression have a certain meaning unless you are able to refer to that
expression independently of its semantic properties. But we have no such
independent access to the expressions of our language of thought; we do
not, for instance, know what they look like. So we cannot have semantic
intentions in respect of them and, hence, cannot make sense of using them
correctly or incorrectly in the sense defined by McGinn.
If McGinn's understanding of normativity were the correct one, then, it
would indeed be difficult to see how it could operate at the level of thought
(though not quite for the reasons he gives). It ought to be clear, however,
that the 'normativity' requirement defined by McGinn has nothing much
to do with the concept of meaning per se and is not the requirement that
Kripke is operating with.
We may appreciate this point by observing that the requirement defined
Ibid., p. 174.
The Rzrle-Followzng Considerations 513
by McGinn could hardly act as a substantive constraint on theories of
meaning, even where these are theories solely of linguistic meaning. Any
theory of meaning that provided an account of what speakers mean by their
expressions at arbitrary times-however crazy that theory may otherwise
be-would satisfy McGinn's constraint. In particular, the main theory
alleged by Kripke to founder on the normativity requirement, would easily
pass it on McGinn's reading: since there are perfectly determinate facts
about what dispositions are associated with a given expression at a given
time-or, rather, since it is no part of Kripke's intent to deny that there
are-it is always possible to ask whether an expression has the same or a
different meaning on a dispositional theory, thus satisfying McGinn's
requirement. How to explain, then, Kripke's claim that a dispositional
theory founders precisely on the normativity requirement?
5. The answer is that the normativity requirement is not the thesis
McGinn outlines. What is it then?
Suppose the expression 'green' means green. It follows immediately that
the expression 'green' applies correctlJ1only to these things (the green ones)
and not to those (the non-greens). The fact that the expression means
something implies, that is, a whole set of normatzve truths about my
behaviour with that expression: namely, that my use of it is correct in
application to certain objects and not in application to others. This is not,
as McGinn would have it, a relation between meaning something by an
expression at one time and meaning something by it at some later time; it
is rather, a relation between meaning something by it at some time and its
use at that time.
The normativity of meaning turns out to be, in other words, simply a new
name for the familiar fact that, regardless of whether one thinks of meaning in
truth-theoretic or assertion-theoretic terms, meaningful expressions possess
conditions of correct use. (On the one construal, correctness consists in true
use, on the other, in warranted use.) Kripke's insight was to realize that this
observation may be converted into a condition of adequacy on theories of the
determination of meaning: any proposed candidate for the property in virtue
of which an expression has meaning, must be such as to ground the
'normativity' of meaning-it ought to be possible to read off from any alleged
meaning constituting property of a word, what is the correct use of that word.
It is easy to see how, on this understanding of the requirement in question, a
dispositional theory might appear to fail it: for, it would seem, one cannot read
off a disposition to use a word in a certain way what is the correct use of that
word, for to be disposed to use a word in a certain way implies at most that one
will,not that one should (one can have dispositions to use words incorrectly).1°
l o .As v e shall see belo\v, hoaever, the question \\ hether dispositional accounts of meaning really do
succumb to the normativity objection is much more complicated than this. I am not here trying to
assess the objection, but mere11 to state it.
5 14 Paul A. Boghossian
6. With this clarification of the normativity thesis in place we are finally in
a position to settle the question: can Kripke develop the same sort of
meaning-sceptical argument against a language of thought as he develops
against public language? And the answer is: clearly, yes. For: what fixes the
meaning of expressions in the language of thought? Not other thoughts, on
pain of vicious regress. Not facts about the actual tokening of such
expressions or facts about associated qualitative episodes, for familiar
reasons. And not dispositional facts about the tokening of such expressions,
for, since meaningful expressions of mentalese possess conditions of
correct use in precisely the same sense as public language expressions do,
because correctness cannot be reconstructed dispositionally. So, nothing
fixes their meaning.
Indeed, we are also now in a position to see, as promised, that nothing
so rich as a language of thought hypothesis is strictly needed. A language
of thought model is composed out of two theses: (a) that thinking the
thought that p involves tokening an item-a representation-that means
that p; and (b) that the representation whose tokening is so involved
possesses a combinatorial syntactic and semantic structure. In other
words, according to a language of thought hypothesis, thought contents
are the semantic properties of syntactically and semantically structured
bearers. But it should be quite clear that nothing in the sceptical argument
depends on the assumption of structure: even if the representation were to
possess no internal syntax, we could still ask, in proper Kripkean fashion,
what its correctness conditions are and in virtue of what they are
determined.
It would appear, however, that the sceptical argumenťs strategy does
presuppose that content properties have some sort of bearer (even if not
necessarily a structured one). For, otherwise, there will be no natural way
to formulate a dispositional theory of thought content, and no natural way
to bring the normativity requirement to bear against it. There has to be
something-a state, event, or particular, it need not matter which-whose
disposition to get tokened under certain circumstances constitutes, on a
dispositional theory, its possession of a certain content. And although this
commitment is, I suppose, strictly speaking contestable, it is also very
natural and plausible. After all, contents do not figure in a mental life
except as subtended by a particular mode-belief, desire, judgement,
wish-and, hence, are naturally understood as the properties of the states
or events that instantiate those modes.
And so we see that the sceptical argument must, can, and does (in
intent, anyway) include mental content within the scope of the scepticism
it aims to promote.l l
'' Since nothing will hang on it, and since it will ease exposition, I shall henceforth write as if a
language of thought hypothesis were true.
The Rule-Following Considerations j I j
The constitutive nature of the sceptical problem
7. Having a meaning is essentially a matter of possessing a correctness
condition. And the sceptical challenge is to explain how anything could
possess that.
Notice, by the way, that I have stated the sceptical problem about
meaning without once mentioning Kripke's notorious sceptic. That
character, as everyone knows, proceeds by inviting his interlocutor to
defend a claim about what he previously meant by the expression '+'.The
interlocutor innocently assumes himself to have meant addition; but the
sceptic challenges him to prove that the concept in question was not in fact
quaddition, where quaddition is just like addition, except for a singularity
at a point not previously encountered in the interlocutor's arithmetical
practice.
It may seem, then, that the sceptical problem I have described could not
be Kripke's. For Kripke's problem appears to be essentially epistemological
in character-it concerns a speaker's ability to defend a particular meaning
ascription; whereas the problem I have outlined is constitutive, not
epistemological-its topic is the possibility of meaning, not our knowledge
of it.
In fact, however, the two problems are the same; Kripke merely
chooses to present the constitutive problem in an epistemological guise.
Epistemological scepticism about a given class of judgements is the view
that our actual cognitive capacities are incapable of delivering justified
opinions concerning judgements in that class. Kripke's sceptic is not after
a thesis of that sort. This is evident from the fact that his interlocutor, in
being challenged to justify his claim that he meant addition by '+', is
permitted complete and omniscient access to all the facts about his
previous behavioural, mental, and physical history; he is not restricted to
the sort of knowledge that an ordinary creature, equipped with ordinary
cognitive powers, would be expected to possess.'2 Kripke's sceptical
scenario is, thus, completely unsuited to promoting an epistemological
scepticism. What it is suited for is the promotion of a constitutive
scepticism. For if his sceptic is able to show that, even with the benefit of
access to all the relevant facts, his interlocutor is still unable to justify
any particular claim about what he meant, that would leave us no choice
but to conclude that there are no facts about meaning.13
Pace many of Kripke's readers, then, the problem is not-not even in
l 2 McGinn's failure to note this leads him to wonder how the constitutive and epistemological
aspects of Kripke's discussion are related, 'for the epistemological claim is clearly distinct from the
metaphysical claim' (op. cit., p. 149).
l 3 This point is made verj- nicely by Crispin Wright in his 'Kripke's Account of the .Jrgument
.Against Private Language', Journal of Ph~losophj~,1984, pp. 761-2. Wright, however, discerns another
sort of epistemological dimension to the sceptical problem. I s ill discuss that below.
516 Paul A. Boghossian
part-pistemological scepticism about meaning.14 But, of course, one
may agree that the problem is constitutive in character, and yet believe it
to have an epistemological dimension. According to Crispin Wright, for
example, Kripke is not interested in the mere possibility of correctness
conditions; he is interested in the possibility of correctness conditions that
may be, at least in one's own case, known non-inferentially.' The problem
is essentially constitutive in character; but acceptable answers to it are to
be subject to an epistemic constraint.
I do not wish to argue about this at length. It does seem to me that, once
we have corrected for the distortions induced by the dialogic setting, there
ought not to be any residual temptation to think that epistemological
considerations are playing a critical role in Kripke's argument. In any case,
whatever intention Kripke may have had, the considerations he adduces
on behalf of the sceptical conclusion appear to owe nothing to epistemological
constraints and can be stated without their help.16 That, anyway, is
how I shall present them.
The 'rule-following' considerations?
8. It would not be inappropriate to wonder at this point what all this has to
do with the topic of rule-following? Where, precisely, is the connection
between the concepts of meaning and content, on the one hand, and the
concept of following a rule, on the other, forged? I shall argue that, in an
important sense, the answer is 'nowhere', and hence that 'the rulefollowing
considerations' is, strictly speaking, a misnomer for the discussion
on offer.
Many writers seem to assume that the connection is straightforward;
they may be represented as reasoning as follows. Expressions come to have
correctness conditions as a result of people following rules in respect of
them; hence, exploring the possibility of correctness is tantamount to
exploring the possibility of rule-following.
But, at least on the ordinary understanding of the concept of following a
rule, it cannot be true of all expressions-in particular, it cannot be true of
mental expressions-that they come to have correctness conditions as a
result of people following rules in respect of them. The point is that the
ordinary concept of following a rule-as opposed to that of merely
l4 For example, McGinn, op. cit., pp. 140-jo; G. Baker and P. Hacker, 'On Misunderstanding
Wittgenstein: Kripke's Private Language Argumenť, Synthise, 1984, pp. 409-10. Neil Tennant has
complained that Kripke's sceptic does not ultimately supply a convincing bent-rule reinterpretation of
his interlocutor's words. See his 'Against Kripkean Scepticism', forthcoming. Tennant may well be
right about this. But here again, I think, the perception that this affects the force of the sceptical
problem about meaning is a result of taking the dialogic setting too seriously. The constitutive problem
about meaning-hob1 could there so much as be a correctness condition-can be stated quite forcefully
~vithoutthe actual provision of a convincing global reinterpretation of a person's words.
l 5 See op. cit., pp. jjz-j. l 6 With one relatively minor exception to be noted below.
The Rule-Following Considerations 517
conforming to one-is the concept of an intentional act: it involves the
intentional attempt to bring one's behaviour in line with the dictates of
some grasped rule. Crispin Wright has decribed this intuitive conception
very clearly:
Correctly applying a rule to a new case will, it is natural to think, t) pically involve
a double success: it is necessary both to apprehend relevant features of the
presented situation and to know what, in the light of those apprehended features,
will fit or fail to fit the rule. Correctly castling in the course of a game of chess, for
instance, will depend both on apprehension of the configuration of chessmen at
the time of the move, and on a knowledge of whether that configuration (and the
previous course of the game) permits castling at that point.17
As such, however, the ordinary concept of following a rule is the concept of
an act among whose causal antecendents lie contentful mental states; consequently,
it is a concept that presupposes the idea of a correctness condition,
not one that can, in full generality, help explain it. Since it makes essential
play with the idea of a propositional attitude, which in turn makes essential
play with the idea of content, rule-following in this sense presupposes that
mental expressions have conditions of correct application. On pain of regress,
then, it cannot be true that mental expressions themselves acquire meaning
as a result of anyone following rules in respect of them.
What Kripke's discussion is concerned with is the possibility of
correctness; so long as we keep that clearly in mind, talk of 'rule-following'
is harmless. Simon Blackburn has captured this perspective very well:
I intend no particular theoretical implications by talking of rules here. The topic is
that there is such a thing as the correct and incorrect application of a term, and to
say that there is such a thing is no more than to say that there is truth and falsity. I
shall talk indifferently of there being correctness and incorrectness, of words being
rule-governed, and of their obeying principles of application. Whatever this is, it
is the fact that distinguishes the production of a term from mere noise, and turns
utterance into assertion-into the making of judgment.18
I1 

THE SCEPTICAL SOLUTION 

A non-factualist conception of meaning
9. Having established to his satisfaction that no word could have the
property of expressing a certain meaning, Kripke turns to asking how this
l 7 Crispin Wright: '\Iiittgenstein's Rule-Follo~vingConsiderations and the Central Project of
Theoretical Linguistics', in Rejections on Chottzsky, ed. '4. George, Osford, Basil Blackvell, 1989,
P. 255,
Simon Blackburn, 'The Individual Strikes Back', S,>~nth?se,1984, p p 281-2. M y onl!- disagreement
with this passage concerns its identification of correctness conditions with truth conditions.
Truth conditions are simply one species of a correctness condition; proof conditions or justification
conditions supply further instances.
518 Paul A. Boghossian
conclusion is to be accommodated. The question is urgent, in his view,
because the conclusion threatens to be not merely shocking but paradoxical.
The trouble is that we would ordinarily take a remark to the effect that
there could not be any such thing as the fact that I mean something by the
'+' sign, to entail that there is nothing I could mean by the use of that
sign. Applied quite generally, across all signs and all people, the claim
becomes the seemingly paradoxical and self-refuting thesis that no one
could mean anything by their use of linguistic expressions.
A scepticism about meaning facts would appear to be, then, prima facie
anyway, an unstable position. Sustaining it requires showing that what it
asserts does not ultimately lapse into a form of pragmatic incoherence.
What is called for, in other words, is a rehabilitation of our ordinary
practice of attributing content to our thoughts and utterances, which
nevertheless conserves the sceptical thesis that there are no facts for such
attributions to answer to. That is what the 'sceptical solution' is designed
to do. It is alleged to have the following startling consequence: the idea of a
language whose meanings are constituted solely out of an individuaľs
speaker's properties, considered 'completely in isolation from any wider
community to which he may belong', is incoherent.19
The sceptical solution has two parts that are usefully distinguished. The
first consists in the suggestion that we replace the notion of truth
conditions, in our intuitive picture of sentence meaning, by that of
assertibility conditions. The second consists in a description of the
assertibility conditions for meaning-attributing sentences, in the course of
which it is argued that it is essential to such sentences that their
assertibility conditions advert to the actions or dispositions of a commu-
nity.
The adjustment recommended in the first part is supposed to help
because
if we suppose that facts or truth conditions are of the essence of meaningful
assertion, it will follow from the skeptical conclusion that assertions that anyone
ever means anything are meaningless. On the other hand, if we apply to these
assertions the tests suggested . . . no such conclusion follows. All that is needed to
legitimize assertions that someone means something is that there be roughly
specifiable circumstances under which they are legitimately assertible, and that
the game of asserting them has a role in our lives. No supposition that 'facts
corresponď to those assertions is needed."
l 9 Following Goldfarb, we may call this the concept of a 'solitary language'. See his 'Kripke on
Wittgenstein on Rules',Journal of Philosophy, 1985. Goldfarb goes on to say that the idea of a solitary
language is more general than that of a Wittgensteinian 'private language', for the latter essentially
involves the idea of necessary unintelligtbility to another. It is hard to assess this, because it is hard to
know how to interpret 'necessary unintelligibility'. Surely it cannot mean: a language to whose
predicates no two people could attach the same descriptive conditions. And it is not clear what it is to
mean, if not that. For useful discussion see C. Wright, 'Does Philosophtcal Investigations 1.258-60
Suggest a Cogent Argument Against Private Language?', in Subject, Thought and Context, ed. P. Pettit
and J. McDowell, Oxford, Clarendon Press, 1986. 20 K., pp. 77-8.
The Rule-Following Considerations j I9
The proposed account is, in effect, a global non-factualism: sentence
significance is construed quite generally in assertion-theoretic terms and
no invidious distinction is drawn between the sort of significance possessed
by meaning-attributing sentences and that possessed by sentences of other
types.
The argument against solitary language
10. The argument against 'solitary language' emerges, according to
Kripke, from the observation that, so long as a speaker is considered in
isolation we can assign no assertibility conditions to judgements to the
effect that he has misapplied a symbol in his repertoire:
[I]f we confine ourselves to looking at one person alone, this is as far as we can go
. . .. There are no circumstances under which n-e can say that, even if he inclines
to say '12j1,he should have said 'j', or vice-cersa . . .. Under what circumstances
can he be wrong? No one else by looking at his mind or behavior alone can say
something like, 'He is wrong if he does not accord with his own intention'; the
whole point of the skeptical argument was that there are no facts about him in
virtue of which he accords with his intentions or not.21
The possibility of error, however, is essential to our ordinary concept of
meaning, and can only be accommodated if we widen our gaze and take
into consideration the interaction between our imagined rule-follower and
a linguistic community. Were we to do so, Kripke continues, we could
introduce assertibility conditions for judgements about error in terms of
the agreement, or lack of it, between a given speaker's propensities in the
use of a term and the community's. Since, however, this would appear to
be the only way to give substance to the correlative notions of error and
correctness, no one considered wholly in isolation from other speakers
could be said to mean anything. And so a solitary language is impossible.
Let us turn now to an assessment of the various central aspects of
Kripke's argument.
I11 

ASSESSMENT O F T H E ARGUMENT AGAIKST 

SOLITARY LAKGUAGE 

Constitutive accounts and solitary language
I I . Kripke is very clear about the limited, wholly descriptive nature of the
sceptical solution, at least in his 'officiaľ explications of the view:
We have to see under what circumstances attributions of meaning are made and
what role these attributions play in our lives. Following Wittgenstein's exhortaK.,
p. 88
520 Paul A. Boghossian
tion not to think but to look, we will not reason a priori about the role such
statements ought to play; rather we will find out what circumstances actually
license such assertions and what role this license actually plays. It is important to
realize that we are not looking for necessary and sufficient conditions (truth
conditions) for following a rule, or an analysis of what such rule-following
'consists in'. Indeed such conditions would constitute a 'straighť solution to the
skeptical problem, and have been rejected."
It is important to see that the counselled modesty-we will not reason a
priori about the role such statements ought to play-is compulsory. The
assertibility conditions may not be understood to provide the content (or
truth conditions) of the meaning-attributing sentences, on pain of falling
prey to the accepted sceptical considerations. (That is why the solution on
offer has to be sceptical: it has already been conceded that nothing could
cogently amount to the fact that a meaning sentence reports). It would
appear to follow from this, however, that the sceptical solution can do no
more than record the conditions under which speakers in fact consider the
attribution of a certain concept warranted and the endorsement of a
particular response appropriate. The Wittgensteinian exhortation 'not to
think but to look' is not merely (as it may be) good advice; the modesty it
counsels is enforced by the fact that truth conditions for these sentences
has been jettisoned. For how, in the absence of a conception of the truth
conditions of meaning attributing sentences, could the project of providing
an account of their assertion conditions aspire to anything more than
descriptive adequacy? Were we equipped with an account of their truth
conditions, of course, we might be able to reason a priori about what their
assertion conditions ought to be and, hence, potentially, to revise the
conditions for assertion actually accepted for them. But without the benefit
of such an account there is no scope for a more ambitious project: a
descriptively adequate account of the actual assertion conditions for such
sentences is the most one may cogently aim for.
If this is correct, however, we ought to be puzzled about how the
sceptical solution is going to deliver a conclusion against solitary language
of the requisite modal force: namely, that there could not be such a
language. For even if it were true that our actual assertibility conditions for
meaning-attributing sentences advert to the dispositions of a community,
the most that would license saying is that our language is not solitary. And
this would be a lot less than the result we were promised: namely, that any
possible language has to be communal.
Communal assertibility conditions?
12.Putting this worry to one side, let us ask whether it is in fact true that,
if we accept the sceptical conclusion, we cannot introduce substantive
The Rule-Following Considerations 52 I
assertibility conditions for meaning-attributions that do not advert to the
dispositions of a community of speakers? It appears, on the contrary, that
not only can we introduce such conditions, but have actually done so.23
Consider the following:
(A) It is warranted to assert of Jones that he means addition by ' +',
provided he has responded with the sum in reply to most
arithmetical queries posed thus far.
As a description of our practice, (A) is, of course, quite rough: room has to
be made for the importance of systematic deviations, the greater importance
attaching to simple cases, and many other such factors. But all these
refinements may be safely ignored for the purpose of raising the following
critical question: what in the sceptical conclusion rules out attributions of
form (A)? It had better rule them out, of course, if the argument against
solitary language is to be sustained, for (A)adverts to no one other than the
individual. But as Goldfarb points out, there appears to be nothing in the
sceptical conclusion that will rule it It can hardly be objected that
the interpretation of 'sum' is being presupposed in the statement of the
condition, for the sceptical solution is not meant to be a straight solution to
the problem about meaning; as Kripke himself says, in fending off a
similar imagined objection to his own account of the assertibility condi-
tions:
What Wittgenstein is doing is describing the utility in our lives of a certain
practise. Necessarily he must give this description in our own language. As in the
case of any such use of our language, a participant in another form of life might
apply various terms in the description (such as 'agreemenť) in a non-standard
'quus-like' way. . . . This cannot be an objection to Wittgenstein's solution unless
he is to be prohibited from any use of language at
Nor is there any problem in the assumption that it is a genuinely factual
matter what any two numbers sum to; as Kripke himself repeatedly
emphasizes, the sceptical argument does not threaten the existence of
mathematical facts. But how, then, is (A) to be ruled out, and the argument
against solitary language preserved?
13. Could it perphaps be argued that (A) is permissible though parasitic on
the communal assertibility conditions Kripke outlines? As a matter of fact,
just the opposite seems true.26
Kripke's communitalian account of meaning-attributions runs as fol-
lows:
Smith will judge Jones to mean addition by 'plus' only if he judges that Jones's
answers to particular addition problems agree with those he is inclined to give . . ..
2 3 This sort of rejoinder is can\-assed both in Goldfarb, op. cit., and in McGinn, op. cit.
24 Ibid. 2 5 K., p. 146.
26 This is argued in McGinn, op cit., pp. 185--7, from which this point is derived.
522  Paul A. Boghossian
If Jones consistently fails to give responses in agreement . . . with Smith's, Smith
will judge that he does not mean addition by 'plus'. Even if Jones did mean it in
the past, the present deviation will justify Smith in judging that he has lapsed.27
According to this account, then, I will judge that Jones means addition by
'plus' only if Jones uses 'plus' enough times in the same way I am inclined
to use it. As a rough description of our practice, and many important
refinements aside, this seems acceptable enough. One of the refinements
that is called for, however, exposes the fact that Kripke's communitarian
conditions are parasitic on the solitary conditions, and not the other way
round.
It would be absurd for me, under conditions where I had good reason to
believe that I had become prone to making arithmetical mistakes-perhaps
owing to intoxication or senility or whatever-to insist on agreement
with me as a precondition for crediting Jones with mastery of the concept
of addition. And this would appear to show that, at a minimum, Kripke's
communitarian account must be modified to read:
(B)  It is warranted to assert of Jones that he means addition by '+',
provided he agrees with my responses to arithmetical queries,
under conditions mhere I have been a reliable computer of sums.
But this modification would seem immediately to reveal that the reference
to 'my own responses' is idle, and that the basic assertion condition I
accept is just (A):
It is warranted to assert of Jones that he means addition by '+',
provided he has responded with the sum in reply to most arithmetical
queries posed thus far.
It would appear, in other words, that the acceptability of the communitarian
conditions is strongly parasitic on the acceptability of the solitary ones,
and not the other way around.
In sum: both because it is difficult (impossible?)to generate constitutive
results out of non-constitutive accounts, and because our actual assertibility
conditions for meaning ascriptions appear not to be communitarian, I
conclude that the sceptical solution does not yield a convincing argument
against solitary language.
IV 

IRREALIST CONCEPTIONS O F MEANING 

14.The argument against solitary language was supposed to flow from the
adjusted understanding of sentence significance forced by the sceptical
conclusion. The sceptical conclusion has it that it cannot literally be true of
The Rule-Following Considerations 523
any symbol that it expresses a particular meaning: there is no appropriate
fact for a meaning-attributing sentence to report. The sceptical solution's
recommendation is that we blunt the force of this result by refusing to
think of sentence significance in terms of possession of truth conditions,
or a capacity to state facts. We should think of it, rather, in terms
of possession of assertibility conditions. But is this solution forced?
Are there not, perhaps, other ways of accommodating the sceptical
conclusion?
The solution on offer is bound to strike one as an overreaction, at least at
first blush, in two possible respects. First, in that it opts for a form of nonfactualism,
as opposed to an error theory; and second, in that the
recommended non-factualism is global, rather than restricted solely to the
region of discourse-meaning talk-that is directly affected by the
sceptical result it seeks to accommodate.
Semantically speaking, the most conservative reaction to the news that
nothing has the property of being a witch is not to adopt a non-factualist
conception of witch talk, it is to offer an error conception of such talk. An
error conception of a given region of discourse conserves the region's
semantical appearances-predicates are still understood to express
properties, declarative sentences to possess truth conditions; the ontological
discovery is taken to exhibit-merely-the systematic falsity of the
region's (positive, atomic) sentences.28
Could not the moral of the sceptical argument be understood to consist
in an error conception of meaning discourse? It could not, for an error
conception of such discourse, in contrast with error conceptions of other
regions, is of doubtful coherence. The view in question would consist in
the claim that all meaning-attributions are false:
(I) For any S: rS means that pl is false.
But the disquotational properties of the truth predicate guarantee that (I)
entails
(2) For any S : rS1 has no meaning.
(I) implies, that is, that no sentence whatever possesses a meaning. Since,
however, a sentence cannot be false unless it is meaningful to begin with,
this in turn implies that (I) cannot be true: for what (I) says is that some
sentences-namely meaning-attributing sentences-are false.29
See John Mackie, Ethics: Inventing Right nnd Wrong, London, Penguin, 1977, for such a
conception of moral discourse.
29 An error conception of meaning has been advocated by Paul Churchland; see his 'Eliminative
Materialism and the Propositional Attitudes', Journal of Phzlosophy, 1981. This argument is elaborated
and defended in my 'The Status of Contenť, Philosophical Revzew, forthcoming April 1990.
524 Paul A. Boghossian
So it appears that Kripke was right to avoid an error conception of
meaning discourse. But does his non-factualist conception fare any
better?
15. The canonical formulation of a non-factualist view-and the one that
Kripke himself favours-has it that some targeted declarative sentence is
not genuinely truth-conditional. A non-factualism about meaning, consists,
that is, in the view that
(3) For any S,p: rS means that p? is not truth-conditional
As I noted above, however, the projectivism recommended by the
sceptical solution is intended to apply globally: it is not confined solely to
meaning-attributing sentences. Thus,
(4) For any S : rS? is not truth conditional.
Why does Kripke adopt so extreme a view? Why does he not suggest
merely that we abandon a truth-conditional model for semantic discourse,
while preserving it, as seems natural, for at least some regions of the rest of
language? Kripke does not say. But it may be that he glimpsed that the
global character of the projectivism is in factforced in the present case.30
For consider a non-factualism solely about meaning-the view that, since
there is no such property as a worďs meaning something, and hence no
such fact, no meaning-attributing sentence can be truth-conditional. Since
the truth-condition of any sentence S is (in part, anyway) a function of its
meaning, a non-factualism about meaning will enjoin a non-factualism
about truth-conditions: what truth-condition Spossesses could hardly be a
factual matter if that in virtue of which it has a particular truth-condition
is not itself a factual matter. And so we have it that (3) entails:
(5) For all S,p: rS has truth-condition p? is not truth-conditional.
However, since, courtesy of the disquotational properties of the truth
predicate, a sentence of the form rS has truth-condition pl is true if and
only if S has truth-condition p, and since (5) has it that rS has truthcondition
pl is never simply true, it follows that
(4) For any S : rS1 is not truth-conditional
just as predicted.
It is, then, a fascinating consequence of a non-factualism about
meaning, that it entails a global non-factualism; in this respect, if no other,
a non-factualism about meaning distinguishes itself from a similar thesis
30 Somewhat different arguments are given for this both in Crispin Wrighťs 'Kripke's Accounť,
loc. cit., pp. 769-70 and in my 'Meaning, Content and Rules', in Part I of my Ph.D. Dissertation
Essays on Meaning and BelieJ; Princeton, 1986.
The Rule-Following Considerations 525
about any other subject matter. Crispin Wright has suggested that it also
renders it irremediably problematic:
it is doubtful that it is coherent to suppose that projectivist views could be applied
quite globally. For, however exactly the distinction be drawn between fact-stating
and non-fact-stating discourse, the projectivist will presumably want it to come by
way of a discocerjt that certain statements fail to qualify for the former class; a
statement of the conclusion of the skeptical argument, for instance, is not itselfto
be projective.31
It is hard not to sympathize with Wrighťs suggestion that there must be
something unstable about a projectivist thesis that is itself within the scope
of the projectivism it recommends. But it is also not entirely clear to me in
what the instability consists. T o be sure, a global projectivism would have
to admit that it is no more than assertible that no sentence possesses a truth
condition. But what is wrong with that? If there is an instability here, it is
not a transparent one.
16. In fact, however, I do believe that a non-factualism about meaning is
unstable, but not because of its global character. Rather, the reasons have
to do with the clash between what you have to suppose about truth in
order to frame a non-factualist thesis about anything, and what you have to
suppose about truth as a result of accepting a non-factualism about
meaning. I have developed the argument for this in some detail elsewhere;32
here I have space only to sketch its outlines.
Consider a non-factualist thesis about, say, the good:
(7) All sentences of the form rx is goodl are not truth-conditional.
The point that needs to be kept in focus is that the sentence of which truth
conditions are being denied is a significant declarative sentence. For this
fact immediately implies that the concept of truth in terms of which the
non-factualist thesis is framed cannot be the dejationary concept that A. J.
Ayer succinctly described as follows:
. . . to say that p is true is simply a way of asserting p. . . . The traditional
conception of truth as a 'real quality' or a 'real relation' is due, like most
philosophical mistakes, to a failure to analyze sentences correctly . . .. There are
sentences in which the word 'truth' seems to stand for something real . . . [but]
our analysis has shown that the word 'truth' does not stand for anything.33
If the concept of truth were, as Ayer claims in this passage, merely the
concept of a device for semantic ascent, and not the concept of some
genuine property-some 'real relation'-that a sentence (or thought) may
enjoy, then non-factualism is nowhere a coherent option. For on a
deflationary understanding of truth, a sentence will be truth-conditional
31 Ibid., p. 770, 3 2 In 'The Status of Contenť, loc. cit.
3 3 A. J. Ayer, Langucige, Truth ntzd Logic, Nev York, Doler, 1952, p. 89.
526 Paul A. Boghossian
provided only that it is apt for semantic ascent; and it will be apt for
semantic ascent provided only that it is a significant, declarative sentence.
But it is constitutive of a non-factualist thesis precisely that it denies, of
some targeted, significant, declarative sentence, that it is truth-conditional.
It follows, therefore, that a non-factualism about any subject matter
presupposes a conception of truth richer than the deflationary: it is
committed to holding that the predicate 'true' stands for some sort of
language-independent property, eligibility for which will not be certified
purely by the fact that a sentence is declarative and significant. Otherwise,
there will be no understanding its claim that a significant sentence,
declarative in form, fails tc possess truth-conditions.
So we have it that any non-factualist thesis presupposes that truth is, as
I shall henceforth put it, robust. But, now, notice that judgements about
whether an object possesses a robust property could hardly fail to be
factual. If P is some genuinely robust property, then it is hard to see how
there could fail to be a fact of the matter about whether an object has P. It
does not matter if P is subjective or otherwise dependent upon our
responses. So long as it is a genuine, language-independent property,
judgements about it will have to be factual, will have to be possessed of
robust truth-conditions. In particular, if truth is a robust property, then
judgements about a sentence's truth value must themselves be factual. But
we saw earlier-see ( 5 ) above-that a non-factualist thesis about meaning
implies that judgements about a sentence's truth cannot be factual:
whether a certain sentence is true cannot be a factual matter if its meaning
is not. And this exposes the contradiction we have been stalking: a nonfactualism
about meaning implies both that truth is robust and that it is
not.
17. It is hard to do justice to the issues involved within the confines of the
present essay.34 I do hope, however, that the preceding discussion has
succeeded in sowing some doubts about the cogency of irrealist conceptions
of meaning-whether in the form of a non-factualism about
meaning, as in the sceptical solution, or an error theory, as suggested, for
instance, by Churchland.
The uncompromising strength of the claim is bound to arouse suspicion.
Irrealist conceptions of other domains may not be particularly
appealing or plausible, but they are not incoherent. Why should matters
stand differently with meaning discourse?
The source of the asymmetry is actually not that hard to track dow~:.It
consists in the fact that error and non-factualist theories about any subject
matter presuppose certain claims about truth and truth-conditions, that an
error or non-factualist conception directed precisely at our talk of meaning
itself ends up denying. Not surprisingly the ensuing result is unstable.
34 Again, for a more detailed treatment see 'The Status of Contenť, loc. cit.
The Rule-Following Considerations 527
Thus, an error thesis about any subject matter presupposes that the
target sentences are truth-conditional. But an error thesis directed
precisely at our talk about meaning entails the denial of that presupposition.
Thus, also, a non-factualism about any subject matter presupposes a
robust conception of truth. But a non-factualism directed precisely at our
talk about meaning entails the denial of that presupposition.
if these considerations are correct, then, they would show that the
sceptical conclusion cannot be sustained: there appears to be no stable way
of accol-ilmodating the claim that there are no truths about meaning.
Something must be wrong, therefore, with the argument that appeared to
lead us to it. What could it be?
V
REDUCTIVE ACCOUNTS O F MEANING
18. The sceptical argument has been faulted on a number of grounds, the
most important being:
That its arguments against dispositional accounts of meaning do not
work.
That it neglects to consider all the available naturalistic facts.
That its conclusion depends on an unargued reductionism.
The first two objections issue from a naturalistic perspective: they claim
that the sceptical argument fails to establish its thesis, even granted a
restriction to naturalistic facts and properties. The final objection concedes
the failure of naturalism, but charges that the sceptical argument is
powerless against an appropriately anti-reductionist construal of meaning.
In this part I shall examine the naturalistic objections, and in the next the
anti-reductionist suggestion.
I should say at the outset, however, that I see no merit to objections of
the second kind and will not discuss them in any detail here. All the
suggestions that I have seen to the effect that Kripke ignores various viable
reduction bases for meaning facts seem to me to rest on misunderstanding.
Colin McGinn, for example, claims that Kripke neglects to consider the
possibility that possession of a concept might consist in possession of a
certain sort of capacity. Capacities, McGinn explains, are distinct from
dispositions and are better suited to meet the normativity c o n ~ t r a i n t . ~ ~
This rests on the misunderstanding of normativity outlined above. Warren
Goldfarb charges that Kripke neglects to consider causal/informational
accounts of the determination of meaning.36This derives from a failure to
see that, in all essential respects, a causal theory of meaning is simply one
species of a dispositional theory of meaning, an account that is, of course,
35 See McGinn, op. cit., pp. 168-74. 36 See Goldfarb, op. cit., n. 13
528 Paul A. Boghossian
extensively discussed by Kripke. It is unfortunate that this connection is
obscured in Kripke's discussion. Because Kripke illustrates the sceptical
problem through the use of an arithezetzcal example, he tends, understandably,
to focus on conceptual role versions of a dispositional account
of meaning, rather than on causal/informational versions. This has given
rise to the impression that his discussion of dispositionalism does not
cover causal theories. But the impression is misleading. For the root form
of a causal/informational theory may be given by the following basic
formula:
0 means (property) P by predicate S iff (it is a counterfactual
supporting generalization that) 0 is disposed to apply S to P.
Dispositions and meaning: Jinitude
19.The single most important strand in the sceptical argument consists in
the considerations against dispositional theories of meaning. It would be
hard to exaggerate the importance of such theories for contemporary
philosophy of mind and semantics: as I have just indicated, the most
influential contemporary theories of content-determination-'informationaľ
theories and 'conceptual-role' theories-are both forms of a
dispositional account.37In my discussion I shall tend to concentrate, for
the sake of concreteness, on informational theories of the content of mental
symbols; but the issues that arise are general and apply to any dispositional
theory whatever.
The root form of an information-style dispositional theory is this:
My mental symbol 'horse' expresses whatever property I am disposed
to apply it to.
Kripke's first objection amounts, in effect, to suggesting that there will
always be a serious indeterminacy in what my dispositions are, thus
rendering dispositional properties an inappropriate reduction base for
meaning properties. For, Kripke argues, if it is indeed the property horse
that I am disposed to apply the term to, then I should be disposed to apply
it to all horses, including horses so far away and so far in the past that it
would be nonsense to suppose I could ever get into causal contact with
them. Otherwise, what is to say that my disposition is not a disposition to
apply the term to the property nearbv horse, or some such? But no one can
3 7 For correlational theories see: F. Dretske, Knowledge and the Flow of Information, Cambridge,
MIT Press, 1981;D. Stampc, 'Towards a Causal Theory o f Linguistic Representation', Midwest
Studies in Philosophy, vol. 2, Minneapolis, University of Minnesota Press, 1977; Jerry Fodor,
Psychosemantics, Cambridge, MIT Press, 1987. For conceptual role theories see: H. Field, 'Logic,
Meaning and Conceptual Role',3ournal ofPhilosophy, 1977;Ned Block, 'Advertisement for a Semantics
for Psychology', Midwest Studies tn Phtlosophy, vol. 10, Minneapolis, University of Minnesota Press,
1986.
The Rule-Following Conszderattons 529
have a disposition to call all horses 'horse', for no one can have a
disposition with respect to inaccessible objects.
The argument does not convince. Of course, the counterfactual
If I were now to go to Alpha Centauri, I would call the horses there
'horse',
is false. If I were now to go to Alpha Centauri, I probably would not be in
any position to call anything by any name, for I would probably die before
I got there. But that by itself need not pose an insuperable obstacle to
ascribing the disposition to me. All dispositional properties are such that
their exercise-the holding of the relevant counterfactual truth-is
contingent on the absence of interfering conditions, or equivalently, on the
presence of ideal conditions. And it certainly seems conceivable that a
suitable idealization of my biological properties will render the counterfactual
about my behaviour on Alpha Centauri true. Kripke considers such a
response and complains:
But how can we have any confidence in this? How in the world can I tell what
would happen if my brain were stuffed with extra brain matter? . . . Surely such
speculation should be left to science fiction writers and futurologists.38
If the point is supposed to be, however, that one can have no reason for
accepting a generalization defined over ideal conditions unless one knows
exactly which counterfactuals would be true if the ideal conditions
obtained, then, as Jerry Fodor has pointed out, it seems completely
~ n a c c e p t a b l e . ~ ~For example, no one can claim to know all of what would
be true if molecules and containers actually satisfied the conditions over
which the ideal gas laws are defined; but that does not prevent us from
claiming to know that, if there were ideal gases, their volume would vary
inversely with the pressure on them. Similarly, no one can claim to know
all of what would be true if I were so modified as to survive a trip to Alpha
Centauri; but that need not prevent us from claiming to know that, if I
were to survive such a trip, I would call the horses there 'horse'.40
Still, it is one thing to dispel an objection to a thesis, it is another to
prove the thesis true. And we are certainly in no position now to show that
we do have infinitary dispositions. The trouble is that not ez9ery true
counterfactual of the form.
If conditions were ideal, then, if C, S would do A
can be used to attribute to S the disposition to do A in C. For example,
one can hardly credit a tortoise with the ability to overtake a hare, by
pointing out that if conditions were ideal for the tortoise-if, for example,
38 K., p. 27. 

39 Scc ' A Theor!- of Contenť, Part 11, p. I j (manuscript). 

40 For a related criticism of Kripke on this score see Blackburn, op. cit. 

530 Paul A. Boghossian
it were much bigger and faster-then it would overtake it. Obviously, only
certain idealizations are permissible; and also obviously, we do not now
know which idealizations those are. The set of permissible counterfactuals
is constrained by criteria of which we currently lack a systematic account.
In the absence of such an account, we cannot be completely confident that
ascriptions of infinitary dispositions are acceptable, because we cannot be
completely confident that the idealized counterfactuals needed to support
such ascriptions are licit. But I think it is fair to say that the burden of
proof here lies squarely on Kripke's shoulders: it is up to him to show that
the relevant idealizations would be of the impermissible variety. And this
he has not done.
Dispositions and meaning: normativity
20. Few aspects of Kripke's argument have been more widely misunderstood
than his discussion of the 'normativity' of meaning and his
associated criticism of dispositional theories. This is unsurprising given
the difficulty and delicacy of the issues involved. In what sense is meaning
a normative notion? Kripke writes:
Suppose I do mean addition by '+'.What is the relation of this supposition to the
question how I will respond to the problem '68 +57'? The dispositionalist gives a
descriptive account of this relation: if' +' meant addition, then I will answer '125'.
But this is not the right account of the relation, which is normative, not
descriptive. The point is not that, if I meant addition by '+', I will answer ' I Z ~ ' ,
but that, if I intend to accord with my past meaning of' +',I should answer '125'.
Computational error, finiteness of my capacity, and other disturbing factors may
lead me not to be disposed to respond as I should, but if so, I have not acted in
accordance with my intentions. The relation of meaning and intention to future
action is normative, not de~criptive.~'
The fact that I mean something by an expression, Kripke says, implies
truths about how I ought to use that expression, truths about how it would
be correct for me to use it. This much, of course, is incontestable. The fact
that 'horse' means horse implies that 'horse' is correctly applied to all and
only horses: the notion of the extension of an expression just is the notion
of what it is correct to apply the expression to. It is also true that to say
that a given expression has a given extension is not to make any sort of
simple descriptive remark about it. In particular, of course, it is not to say
that, as a matter of fact, the expression will be applied only to those things
which are in its extension. Kripke seems to think, however, that these
observations by themselves ought to be enough to show that no dispositional
theory of meaning can work. And here matters are not so
straightforward.
The Rule-Followzng Conszderuttons 531
Let us begin with the very crude dispositional theory mentioned above:
'horse' means whatever property I am disposed to apply it to. This is a
hopeless theory, of course, but the reasons are instructive. There are two
of them, and they are closely related. The first difficulty is that the theory
is bound to get the extension of 'horse' wrong. Suppose I mean horse by it.
Then, presumably, I have a disposition to call horses 'horse'. But it will
also be true that there are certain circumstances-sufficiently dark
nights-and certain cows-sufficiently horsey looking ones-such that, I
am disposed, under those circumstances, to call those cows 'horse' too.
Intuitively, this is a disposition to make a mistake, that is, to apply the
expression to something not in its extension. But our crude dispositional
theory, given that it identifies the property I mean by an expression with the
property I am disposed to apply the expression to, lacks the resources by
which to effect the requisite distinction between correct and incorrect
dispositions. If what I mean by an expression is identified with whatever I
am disposed to apply the expression to, then everything I am disposed to
apply the expression to is, ipsofacto, in the extension of that expression.
But this leads to the unacceptable conclusion that 'horse' does not express
the property horse but rather the disjunctive property horse or cow.
There is a related conceptual difficulty. Any theory which, like the
crude dispositional theory currently under consideration, simply equates
how it would be correct for me to use a certain expression with how I am
disposed to use it, would have ruled out, as a matter of definition, the very
possibility of error. And as Wittgenstein was fond of remarking, if the idea
of correctness is to make sense at all, then it cannot be that whatever seems
right to me is (by definition) right.
One would have thought these points too crucial to miss; but it is
surprising how little they are appreciated. In a recent, comprehensive
treatment of conceptual role theories, Ned Block has written
of a choice that must be made by [conceptual role semantics] theorists, one that
has had no discussion (as far as I know): namely, should conceptual role be
understood in ideal or normative terms, or should it be tied to what people
actually do? . . . I prefer not to comment on this matter . . . because I'm not sure
what to say . . .42
This ought to seem odd. If conceptual role is supposed to determine
meaning, then there can be no question, on pain of falling prey to Kripke's
objection, of identifying an expression's conceptual role with a subjecťs
actual dispositions with respect to that expression.
21.The objections from normativity show, then, that no dispositional
theory that assumes the simple form of identifying the property I mean by
'horse' with the property I am disposed to call 'horse', can hope to succeed.
But what if a dispositional theory did not assume this simple form? What
42 iicd Block, op. cit., p. 631.
532 Paul A. Boghossian
if, instead of identifying what I mean by 'horse' with the entire range of my
dispositions in respect of 'horse', it identified it only with certain select
dispositions. Provided the theory specified a principle of selection that
picked out only the extension-tracking dispositions; and provided also that
it specified that principle in terms that did not presuppose the notion of
meaning or extension, would it not then be true that the objections from
normativity had been disarmed?
Let us try to put matters a little more precisely. If a dispositional theory
is to have any prospect of succeeding, it must select from among the
dispositions I have for 'horse', those dispositions which are meaningdetermining.
In other words, it must characterize, in non-intentional and
non-semantic terms, a property M such that: possession of M is necessary
and sufficient for being a disposition to apply an expression in accord with
its correctness condition^.^^ Given such a property, however, could we not
then safely equate meaning something by an expression with: the set of
dispositions with respect to that expression that possess M ? For, since
dispositions with that property will be guaranteed to be dispositions to
apply the expression correctly, both of the objections from normativity
canvassed so far would appear to have been met. There will be no fear that
the equation will issue in false verdicts about what the expression means.
And, since it is only M-dispositions that are guaranteed to be correct, it
will no longer follow that whatever seems right is right: those dispositions
not possessing M will not be dispositions to apply the expression to what it
means and will be free, therefore, to constitute dispositions to apply the
expression falsely.
At this point two questions arise. First, is there really such a property
M ? And, second, supposing there were, is there really no more to
capturing the normativity of meaning than specifying such a property?
Now, Kripke is clearly sceptical about the existence of an appropriate
M-property. I will consider that question below. But more than this,
Kripke seems to think that even if there were a suitably selected
disposition that captured the extension of an expression accurately, that
disposition could still not be identified with the fact of meaning, because it
still remains true that the concept of a disposition is descriptive whereas
the concept of meaning is not. In other words, according to Kripke, even if
there were a dispositional predicate that logically covaried with a meaning
predicate, the one fact could still not be identified with the other, for they
are facts of distinct sorts. A number of writers have been inclined to follow
him in this. Simon Blackburn, for instance, has written:
43 It is occasionally suggested that it would be enough if possession of M were sufficientfor the
disposition's correctness. But that is not right. If only sufficiency were required we would not know,
simply by virtue of a definition of M, the expression's meaning. For although we would know what
properties were definitely part of the expression's meaning we would not know if we had them all. And
so we would not have even a sufficient condition for the expression's possessing a given meaning.
The Rule-Followzng Considerations 533
I share Kripke's view that whatever dispositions we succeed in identify-ing they
could at most give us standards for selection of a function which we mean. They
couldn't provide us with an account of what it is to be faithful to a previous rule.
It is just that, unlike Kripke, I do not think dispositions are inadequate to the task
of providing standards. Indeed, I think they must be.44
Blackburn here is explicitly envisaging the successful, substantive specification
of dispositions that mirror the extensions of expressions correctly.
But he cites the normative character of facts about meaning as grounds for
denying a dispositional reduction. But what precisely has been left over,
once the extensions have been specified correctly?
One might have a thought like this. A proper reduction of the meaning of
an expression would not merely specify its extension correctly, it would also
reveal that what it is specifying is an extension-namely, a correctness
condition. And this is what a dispositional theory cannot do. There might be
dispositions that logically covary with the extensions of expressions; so that
one could read off the dispositions in question the expressions' correctness
conditions. But the dispositional fact does not amount to the meaning fact,
because it never follows from the mere attribution of any disposition,
however selectively specified, that there are facts concerning correct use;
whereas this does follow from the attribution of an extension. T o be told
that 'horse' means horse implies that a speaker ought to be motivated to
apply the expression only to horses; whereas to be told, for instance, that
there are certain select circumstances under which a speaker is disposed to
apply the expression only to horses, seems to carry no such implication.
It is not clear that this is in general true. Perhaps the M-dispositions are
those dispositions that a person would have when his cognitive mechanisms
are in a certain state; and perhaps it can be non-question-beggingly
certified that that state corresponds to a state of the proper functioning of
those mechanisms. If so, it is conceivable that that would amount to a noncircular
specification of how the person would ideally respond, as
compared with how he actually responds; and, hence, that it would suffice
for capturing the normative force of an ascription of meaning.
There is clearly no way to settle the matter in advance of the
consideration of particular dispositional proposals. What we are in a
position to do, however, is state conditions on an adequate dispositional
theory. First, any such theory must specify, without presupposing any
semantic or intentional materials, property M. This would ensure the
theory's extensional correctness. Second, it must show how possession of
an M-disposition could amount to something that deserves to be called a
correctness condition, something we would be inherently motivated to
44 'The Individual Strikes Back', loc. cit., pp. 289-91. Similar concessions are made by Wright in
his 'Kripke's Accounť, loc. cit., pp. 771-2; and by John McDowell, 'Wittgenstein on Following a
Rule', Synthdse, 1984, p. 329.
534 Paul A.Boghossian 

satisfy. This would ensure the intensional equivalence of the two
properties in question, thus paving the way for an outright reduction of
meaning to dispositions.
What property might M be? There are, in effect, two sorts of proposal:
one, long associated with Wittgenstein himself, seeks to specify M by
exploiting the notion of a community; the other, of more recent
provenance, attempts to define M in terms of the notion of an optimality
condition. I shall begin with the communitarian account.
The communitarian account
22. The idea that correctness consists in agreement with one's fellows has a
distinguished history in the study of Wittgenstein. Even before the current
concern with a 'rule-following problem', many commentators-whether
rightly or wrongly-identified communitarianism as a central thesis of the
later writings. As a response to the problem about meaning, it found its
most sustained treatment in Wrighťs Wittgenstein on the Foundations of
~ a t h e m a t i c s . ~ ~Which of the many dispositions a speaker may have with
respect to a given expression determine its meaning? Or, equivalently,
which of the many dispositions a speaker may have with respect to an
expression are dispositions to use it correctly? Wrighťs communitarian
account furnishes the following answer:
. . . it is a community of assent which supplies the essential background against
which alone it makes sense to think of individuals' responses as correct or
incorrect . . .. None of us can unilaterally make sense of correct employment of
language save by reference to the authority of communal assent on the matter;
and for the community itself there is no authority, so no standard to meet.46
It is important to understand that, according to the proposal on offer, the
correct application of a term is determined by the totality of the
community's actual dispositions in respect of that term. The theory does
not attempt, in specifying the communal dispositions that are to serve as
the constitutive arbiters of correctness, to select from among the community's
actual dispositions a privileged subset. There is a reason for this.
Communitarianism is a response to the perceived inability to define a
distinction, at the level of the individual, between correct and incorrect
dispositions. The suggestion that correctness consists in agreement with
the dispositions of one's community is designed to meet this need. The
proposal will not serve its purpose, however, if the problem at the level
of the individual is now merely to be replayed at the level of the
45 Cambridge, Harvard University Press, 1980.(His more recent writings suggest that Wright no
longer holds this view.) See also Christopher Peacocke, 'Reply: Rule-Following: The Nature of
Wittgenstein's Arguments', in Wittgenstein on Following a Rule, ed. Holtzman and Leich, London,
Routledge and Kegan Paul, 1981.
46 Ibid., p p 219-20.
The Rule-Followzng Constderntzons 535
community. A communitarian does not want it to be a further question
whether a given actual communal disposition is itself correct. The
proposal must be understood, therefore, as offering the folowing characterization
of M: M is the property of agreeing with the actual dispositions of
the community.
How does the proposal fare with respect to the outlined adequacy
conditions on dispositional theories?
Consider first the 'intensionaľ requirement, that possession of the
favoured M-property appear intuitively to resemble possession of a
correctness condition. Does communal consensus command the sort of
response characteristic of truth?
A number of critics have complained against communitarianism that
communal consensus is simply not the same property as truth, that there is
no incoherence in the suggestion that all the members of a linguistic
community have gone collectively, but non-collusively, off-track in the
application of a given predicate.47 This is, of course, undeniable. But the
communitarian is not best read as offering an analysis of the ordinary
notion of truth, but a displacement of it. His thought is that the emaciated
notion of truth !-ielded b!- communitarianism is the best we can hope to
expect in light of the rule-following considerations. The crucial question,
then, is not whether communitarianism captures our ordinar!. notion of
truth, for it quite clearlj- does not; it is, rather, whether communitarianism
offers any concept deserving of that name.
This is a large question on which I do not propose to spend a lot of
time.48 Although there are subtle questions about how much of logic will
be recoverable from such a view, and whether it can be suitablj- nonreductivelj-
articulated (can 'non-collusive agreemenť be defined without
the use of intentional materials?), I am prepared to grant, for the sake of
argument, that the proposal does not fare all that badly in connection with
the 'intensionaľ requirement. Non-collusive communal agreement on a
judgement does usually provide one with some sort of reason for
embracing the judgement (even if, unlike truth, not with a decisive one); it
thus mimics to some degree the sort of response that is essential to truth.
Where communitarianism fails, it seems to me, is not so much here as with
the extensional requirement.
Consider the term 'horse'. What dispositions do I have in respect of
this expression? T o be sure, I have a disposition to apply it to horses.
But I also have a disposition, on sufficientlj- dark nights, to applj- it to
deceptively horse!- looking cows. Intuitivelj-, the facts are clear. 'Horse'
means horse and my disposition to applj- it to cows on dark nights is
47 See Blackburn, op. cit.
48 For a more extensi\-e discussion see my Essays on Meaning and Belief, loc. cit.; see also
Blackburn, Spreading the Word, Oxford, Oxford Uni\-ersity Press, 1984, pp. Xzff.
536 Paul A. Boghosslnn
mistaken. The problem is to come up n-ith a theorj- that delivers this
result systematically and in purelj- dispositional terms. The communitarian's
idea is that the correct dispositions are constitutivelj- those which
agree n-ith the communitj-'s. What, then, are the cornmunit!-'s dispositions
likely to be?
The cornmunit!-, I submit, however exactlj- specified, is bound to
exhibit preciselj- the same dualitj- of dispositions that I do: it too will be
disposed to call both horses and deceptively horsey looking co\\-s on dark
nights 'horse'. After all, if I can be taken in bj-a deceptively horse!- looking
con- on a dark night, \\-hat is to prevent 17,000 people just like me from
being taken in by the same, admittedlj- effective, impostor? The point is
that manj- of the mistakes we make are sj~s~emittic:they arise because of the
presence of features-bad lighting, effective disguises, and so forth-that
have a generalizable and predictable effect on creatures n-ith similar
cognitive endow-ments. (This is presumablj- what makes 'magicians'
possible.) But, then, an!- of mj- dispositions that are in this sense
sj-stematicallj-mistaken, are bound to be duplicated at the level of the
communitj-. The communitarian, however, cannot call them mistakes, for
thej- are the communitj-'s dispositions. He must insist, then, firm
conviction to the contrarl- notwithstanding, that 'horse' means not horse
but, rather, horse or cow.
The problem, of course, is general. There are countless possible
impostors under countless possible conditions; and there is nothing special
about the term 'horse'. The upshot n-ould appear to be that, according to
communitarianism, none of our predicates have the extensions we take
them to hare, but mean something n-ildlj-disjunctik-e instead. IVhich is to
saj- that communitarianism is bound to issue in false verdicts about the
meanings of most expressions, thus failing the first requirement on an
adequate dispositional theory.
It seems to me that n-ehave no option but to reject a pure communitarianism.
If we are to have any prospect of identifj-ing the extensions of our
expressions correctly, it will simplj- not do to identify truth with
communal consensus. Even from among the cornmunit)llsdispositions, n-e
have to select those which ma! be considered meaning-determining, if \T-e
are to have a plausible theorj- of meaning. IVhich is to say that we are still
lacking what communitarianism was supposed to provide: the specification
of a propertj- M such that, possession of rVI b!- a disposition is necessar!and
sufficient for that disposition's correctness.
Of course, once we have abandoned communitarianism, we lack any
motive for defining M over communal dispositions; nothing-at least
nothing ob\-ious-tells against defining .If directlj- over an indicidz~nľs
dispositions. IVhich is preciselj- the way the voluminous literature on this
topic approaches the problem and to a discussion of which I no\{- propose
to turn.
Optimal u'zsposztlo~zs
23. The literature supplies what is, in effect, a set of variations on a basic
theme: M is the propert!. of: being a disposition to apply (an expression) in
a scrtain !11pc o f s i t z t a ~ i o n . ~ ~The idea behind such proposals is that there is
a certain set of circumstances-call them 'optimalit!- conditions1-under
which subjects are, for one or another reason, incapable of mistaken
judgements; hence, we ma!- equate \\-hat they mean b!- a given (mental)
expression with, the properties the!- are disposed to applj- the expression
to, under optimal conditions. Different proposals provide different characterizations
of the conditions that are supposed to be optimal in this sense.
Fred Dretskc, for example, holds that optimal conditions are the conditions
under n-hich the meaning of the expression \\-as first acquired. X
number of other n-riters subscribe to some form or other of a ~cleologi~al
proposal: optimalit!- conditions are those conditions-defined b!- evolutionary
biologj--under which our cogniti\-e mechanisms are functioning
just as the!- are supposed to."
Now, Kripke is very short with such possible elaborations of a dispositional
theor!-. He brieflj-considers the suggestion that we attempt to define idealized
dispositions and says that 'a little experimentation \\-ill reveal the futility of
such an efforť.'' But, surelj-, this underestimates the complexit!- of the
problems involved and fails to do justice to the influence that such proposals
currentlj-exert. IVhat Kripke needs, if his rejection of dispositional accomts is
to succeed, but does not really provide, is a set of principled considerations
against the existence of non-semanticall!-, non-intentionall!. specifiable optimalit!-
conditions. What I would like to do in the remainder of this section is
to begin to sketch an argument for that conclusion. Several specific problems
for specific versions of an optimalit!- theory have received discussion in the
literature." Here, honever, I n-ant to attempt an argument with a more
general sn-eep:I want to argue that n-ehave reason to believe that there could
not be naturalisticall!- specifiable conditions under which a subject \\-ill be
disposed to appl>-an expression only to what it means; and, hence, that no
attempt at specif!-ing such conditions can hope to s~cceed.'~
'' There is one exception to this generalization: JerrJ- Fodor's recent propoial has it that S's
meaning-determining are those that ser\e as an 'as!mmetric dependence baie' fix S's other
dispositions. See his '..\ 'ľhcor! of Contenť. Part 11. fhrthcoming. In 'Saturali~ingContenť,
forthcoming in Zfeiiriirig iri IIin~i:E S S N ) , ~the 11 01.6 n f . 7 ~ 1 ~ ) .O N Fn~ior,Oxford, Baiil Blacl\\ell, I argue
that this theor! is subject to the same difliculties ac confront standard optimalit! ~ersions.
5 ( I For theories of this firm see: Da\id Papineau. Reillit), iiri'i Oxford. BaiilRepreserit~itii~r~.
Black\\ell, 198;; J. Fodor: 'Pr!chosemantics'. 11s. AIIT. 1984. I sh! awa! from sa!ing ~rhetherR.
llillikan. Liitpiii:e.. Tl~ii~ifihtNIW/01Jii.r B!o/ofiricii Ciiti:,.ol-ie.s. Lambridge, \IIT Press. 1987. preients a
theor! of' thii fi~rm.
5 1
K..p. 32.
" g a i n i t Drctslie see Fodor. Psj,ihnsei?zizntris, loc. cit.: against teleological theories see m! Essn),s
on Ilriirirrlg IN^ Belt</;and Fodor, '..\ Theor! of Contenť. Part I, lac, cir.
53
Thii amounts to sa!ing that such theories cannot meet the extensional requirement; so I shall
not elen coniider nhether the! mcct the intensional one.
538 Paul A. Boghossian
24.It will be worthwhile to lay the problem out with some care. Consider
Neil and a particular expression, say, 'horse', in Neiľs mental repertoire.
And suppose that Neil is disposed to token that expression 'in the belief
mode' both in respect of horses and in respect of deceptively horsey
looking cows on dark nights. Let it be clear, furthermore, that 'horse' for
Neil means horse, and that on those occasions when he applies 'horse' to
cows, this amounts to his mistaking a cow for a horse. Now, the thought
behind an optimality version of a dispositional theory is that there is a set
of naturalistically specifiable conditions under which Neil cannot make
mistakes in the identification of presented items.54 Under those conditions,
then, he would believe that there is a horse in front of him only if
there is one. But that in turn implies that, under those conditions, 'horse'
will get tokened (in the belief mode) only in respect of the property it
expresses. So, to figure out what any expression means: look at the
properties Neil is disposed to apply the expression to, when conditions are
in this sense optimal. The end result is a dispositional reconstruction of
meaning facts: for Neil to mean horse by 'horse' is for Neil to be disposed
to call only horses 'horse', when conditions are optimal. Clearly, two
conditions must be satisfied: (i) the specified conditions must really be
such as to preclude the possibility of error-otherwise, it will be false that
under those conditions 'horse' will get applied only to what it means; (ii)
the conditions must be specified purely naturalistically, without the use of
any semantic or intentional materials-otherwise, the theory will have assumed
the very properties it was supposed to provide a reconstruction of.
What I propose to argue is that it is impossible to satisfy both of these
conditions simultaneously.
Optimal dispositions and objective contents
25. The dispositionalist is after a non-semantically, non-intentionally
specifiable set of conditions 0 , which will be such as to yield true, a priori
optimality equations of the form:
(8) For any subject S and concept R: O+(S judges Rx+Rx).
Could there be such a set of conditions?
Notice, to begin with, that where R is the concept of an objective
property, we ought not to expect optimality equations for R, even if 0
were not required to meet the rather stiff constraints imposed by a
reductive dispositionalism-namely, specification in non-semantic and
non-intentional terms. For, intuitively, the very idea of a wholly objective
property (or object or relation) is the idea of a property (object, relation)
5 4 This restriction to perceptually fixed beliefs stems partly from a desire to simplify exposition and
partly from a desire to consider such theories at their strongest.
The Rule-Following Conszderations j39
whose nature is independent of any given person's abilities or judgements:
for such a property, in other words, there is no necessary function from a
given person's abilities and judgements to truths about that property.55
The contrast is with a class of contents for which there does exist a range
of circumstances such that, appropriate subjects are necessarily authoritative
about those contents under those circumstances. Philosophers disagree,
of course, about what contents fall where, but it is typical to think of
judgements about shape as wholl! objective and of judgements about pain
as representing an extreme example of the contrasting class. Let us call this
a distinction between accessible versus inaccesszble contents.56
We are now in a position to see, however, that a dispositional theor! of
meaning, by virtue of being committed to the existence of optimality
equations for every concept, is committed thereby to treating every
concept as if it were accessible. It is thus committed to obliterating the
distinction between accessible and inaccessible contents.
Of course, this objection will not impress anyone reluctant to
countenance wholly objective, inaccessible contents in the first place. I
turn, therefore, to arguing against the dispositional theory on neutral
ground: for any concept, subjective or objective, it is impossible to satisfy
dispositionalism's basic requirement: the specification of a set of conditions
0, zn non-semantzc, and non-zntentzonal terms, such that, under 0,
subjects are immune from error about judgements involving that
concept.
Optimal dispositions and belief holism
26. The basic difficulty derives from the holistic character of the processes
which fix belief. The point is that, under normal circumstances, belief
fixation is typically mediated by background theory-what contents a
thinker is prepared to judge will depend upon what other contents he is
prepared to judge. And this dependence is, again typically, arbitrarily
robust: just about any stimulus can cause just about any belief, given a
suitably mediating set of background assumptions. Thus, Neil may come
to believe Lo, a magpie, as a result of seeing a currawong, because of his
further belief that that is just what magpies look like; or because of his
belief that the only birds in the immediate vicinity are magpies; or because
of his belief that whatever the Pope says goes and his belief that the Pope
5 5 See, for example, Tyler Burge, 'Cartesian Error and the Objectivity of Perception', in Sub~ect,
Thought and Context, ed. J . hlcDowell and P. Pettit, Oxford, Clarendon Press, 1986, p. 125,for a
similar formulation of the concept of objectke an property.
56 It is important to appreciate that this is an epistetnological distinction, not a constitutive one. It
does not follow from the fact that a content is accessible, that it is therefore constituted by our best
judgements about it. (I take it no one is tempted to conclude from the fact that we are authoritative
about our pains, that pains are constituted bl- the judgements we make about them. We shall ha\-e
occasion to discuss constitutke claims of this sort later on in the paper.
j40 Paul A. Boghossian
says that this presented currawong is a magpie. And so on. The thought
that something is a magpie can get triggered by a currawong in any of an
indefinite number of ways, corresponding to the potentially indefinite
number of background beliefs which could mediate the transition. Now,
how does all this bear on the prospects for a dispositional theory of
meaning?
A dispositional theorist has to specify, without use of semantic or
intentional materials, a situation in which a thinker will be disposed to
think, Lo, a magpie only in respect of magpies. But the observation that
beliefs are fixed holistically implies that a thinker will be disposed to think
Lo, a magpie in respect of an indefinite number of non-magpies, provided
only that the appropriate background beliefs are present. Specifying an
optimality condition for 'magpie', therefore, will involve, at a minimum,
specifying a situation characterized by the absence of all the beliefs which
could potentially mediate the transition from non-magpies to magpie
beliefs. Since, however, there looks to be a potential infinity of such
mediating background clusters of belief, a non-semantically, non-intentionally
specified optimality situation is a non-semantically, non-intentionally
specified situation in which it is guaranteed that none of this potential
infinity of background clusters of belief is present. But how is such a
situation to be specified? What is needed is precisely \\-hat a dispositional
theory was supposed to provide: namely, a set of naturalistic necessary and
sufficient conditions for being a belief with a certain content. But, of
course, if we had that we would already have a reductive theory of
meaning-we would not need a dispositional theory! Which is to say that,
if there is to be any sort of reductive story about meaning at all, it cannot
take the form of a dispositional theory.
VI 

ANTI-REDUCTIONIST CONCEPTIONS O F 

MEANING 

A n argument from queerness?
27. If these considerations are correct, there would appear to be plenty of
reason to doubt the reducibility of content properties to naturalistic
properties. But Kripke's sceptic does not merely draw an anti-reductionist
conclusion; he concludes, far more radically, that there simply could not be
any content properties. Suppose we grant the anti-reductionism; what
justifies the content scepticism? Not, of course, the anti-reductionism by
itself. At a minimum one of two further things is needed. Either an
independent argument to the effect that only naturalistic properties are real.
Or, failing that, a frontal assault on the irreducible property in question,
showing that it is, in Mackie's phrase, somehow inherently 'queer'.
The Rule-Following Considerations j41
The single greatest weakness in Kripke's sceptical argument is that he
fails to bring off either requirement. He does not even try to defend a
reductionist principle about the intentional; and his brief attempt at a
'queerness' argument is half-hearted and unconvincing:
Perhaps we may try to recoup, by arguing that meaning addition by 'plus' is a
state even more sui generis than we have argued before. Perhaps it is simply a
primitive state, not to be assimilated to sensations or headaches or any 'qualitative'
states, nor to be assimilated to dispositions, but a state of a unique kind of its own.
Such a move may in a sense be irrefutable, and if it is taken in an appropriate
way Wittgenstein may even accept it. But it seems desperate: it leaves the nature
of this postulated primitive state-the primitive state of 'meaning addition by
"plusn'-completely mysterious. It is not supposed to be an introspectible state,
yet we supposedly are aware of it with some fair degree of certainty whenever it
occurs. For how else can each of us be confident that he does, at present, mean
addition by 'plus'? Even more important is the logical difficulty implicit in
Wittgenstein's sceptical argument. I think that Wittgenstein argues, not merely as
we have said hitherto, that introspection shows that the alleged 'qualitative' state
of understanding is a chimera, but also that it is logically impossible (or at least
that there is a considerable logical difficulty) for there to be a state of 'meaning
addition by "plus"' at all.
Such a state would have to be a finite object, contained in our finite minds. It
does not consist in my explicitly thinking of each case of the addition table. . . .
Can we conceive of a finite state which could not be interpreted in a quus-like way?
How could that be?57
There are several problems with this passage. In the first place, it
misconstrues the appropriate anti-reductionist suggestion. I take it that it
really is not plausible that there are 'primitive states' of meaning public
language expressions in certain ways, one state per expression. The process
by which the inscriptions and vocables of a public language acquire
meaning is a manifestly complex process-involving an enormous array of
appropriate propositional attitudes-the outlines of which may arguably
be found in the writings of Paul Grice and othemS8 A plausible antireductionism
about meaning would not wish to deny that there is an
interesting story to be told about the relation between linguistic content and
mental content; what it maintains, rather, is that there is no interesting
reduction of mental content p~operties to physical/finctional properties.
According to anti-reductionism, in other words, at some appropriate level
mental content properties must simply be taken for granted, without
prospect of identification with properties otherwise described.
Does Kripke manage to create a difficulty for this suggestion? The
passage contains a couple of considerations that may be so construed.
The first charge is that we would have no idea how to explain our ability
5 7
K., pp. 51-2
j8 See the papers cited under n. 3 a b o ~ e
542 Paul A. Boghossian
to know our thoughts, if we endorsed a non-reductionist conception of
their content. Now, no one who has contemplated the problem of selfknowledge
can fail to be impressed by its diffi~ulty.~'But I think that we
would be forgiven if, before we allowed this to drive us to a dubiously
coherent irrealism about content, we required something on the order of a
proof that no satisfactory epistemology was ultimately to be had.
Kripke, however, provides no such proof. He merely notes that the nonphenomenal
character of contentful states precludes an introspective
account of their epistemology. And this is problematic for two reasons.
First, because there may be non-introspective accounts of self-knowledge.60
And second, because it does not obviously follow from the fact
that a mental state lacks an individuative phenomenology, that it is not
intr~s~ectible.~'
Kripke's second objection to the anti-reductionist suggestion is that it is
utterly mysterious how there could be a finite state, realized in a finite
mind, that nevertheless contains information about the correct applicability
of a sign in literally no end of distinct situations. But, again, this
amounts merely to insisting that we find the idea of a contentful state
problematic, without adducing any independent reason why we should.
We know that mental states with general contents are states with infinitary
normative characters; it is precisely with that observation that the entire
discussion began. What Kripke needs, if he is to pull off an argument from
queerness, is some substantive argument, distinct from his anti-reductionist
considerations, why we should not countenance such states. But this he
does not provide.
None of this should be understood as suggesting that an antireductionism
about content is unproblematic, for it is far from it. There
are, for example, familiar, and serious, difficulties reconciling an antireductionism
about content properties with a satisfying conception of their
causal efficacy.62 But in the context of Kripke's dialectic, the antireductionist
suggestion emerges as a stable response to the sceptical
conclusion, one that is seemingly untouched by all the considerations
adduced in the latter's favour.
5 9 For discussion of some of the difficulties see my 'Content and Self-Knowledge', Philosophical
Topics, Spring 1989.
60 See, for example, Tyler Burge, 'Individualism and Self-Knowledge', Journal of Philosophy,
November 1988, and D. Davidson, 'Knowing One's Own Minď, Proceedings of the APA, January
1987.
61 It is interesting to note, incidentally, that one of the more striking examples of the introspective
discernment of a non-qualitative mental feature is provided by, of all things, an experiential
phenomenon. I have in mind the phenomenon, much discussed by Wittgenstein himself, of seeing-as.
We see the duck-rabbit now as a duck, now as a rabbit; we see the Necker cube now with one face
forward, now with another. And we know immediately precisely how we are seeing these objects as,
when we see them now in one way, now in the other. But this change of 'aspecť, although manifestly
introspectible, is nevertheless not a change in something qualitative, for the qualitative character of the
visual experience remains the same even as the aspect changes.
62 See below.
The Rule-Following Consideratzons 543
McDowell on privacy and community
28. If we endorse a non-reductionist conception of meaning, does that
mean that the rule-following considerations disturb nothing in our
ordinary conception of that notion? A number of writers who have found
an anti-reductionist suggestion attractive have certainly not thought so;
they have discerned in those considerations important lessons for the
correct understanding of the possibility of meaning, while rejecting
substantive reductive answers to the constitutive question: in virtue of
what do expressions possess meaning?
John McDowell, for example, has written that:
By Wittgenstein's lights, it is a mistake to think we can dig down to a level at
which we no longer ha\-e application for normative notions (like 'following
according to the rule').63
We have to resist the temptation, according to McDowelľs Wittgenstein,
to form a picture of 'bedrock'-'of how things are at the deepest level at
which we may sensibly contemplate the place of [meaning] in the
worlď-which does not already employ the idea of the correct (or
incorrect) use of an expression.
Oddly, however, McDowell does not take this to commit him to a
quietism about meaning, a position from which no substantive results about
the conditions for the possibility of meaning can be gleaned. On the
contrary, he claims that it is the discernible moral of the rule-following
considerations that correctness, and hence meaning, can exist only in the
context of a communalp~actice,thus precluding the possibility of a private
language. He writes:
Wittgenstein fiarns us not to trq to dig below 'bedrock'. But it is difficult, in
reading him, to avoid acquiring a sense of what, as it were, lies down there: a web
of facts about behavior and 'inner' episodes, describable without using the notion
of meaning. One is likely to be struck by the sheer contingency of the
resemblances between indijiduals on which, in this vision, the possibility of
meaning seems to depend .. ..64
And:
It is true that a certain disorderliness below 'bedrock' would undermine the
applicability of the notion of rule-following. So the underlying contingencies bear
an intimate relation to the notion of rule-following . . ..65
This is, of course, McDowelľs characterization of the familiar Wittgensteinian
claim that a certain measure of agreement in communal responses
is a precondition for meaning. But how is such a thesis to be motivated?
How, in light of the rejection of substantive answers to the constitutive
6 3 'Wittgenstein on Following a Rule', loc. cit., p. 341.
64 Ibid., p. 348. 6 5 Ibid., p. 349.
544 Paul A. Boghossian
question, is it to be argued for? The claim that communal practice is
necessary for meaning is a surprising claim; mere reflection on the concept
of meaning does not reveal it. And what, short of a substantive constitutive
account, could conceivably ground it?
Consider the contrast with the communitarian view considered above.
That view engages the constitutive question, offers a substantive answer to
it, and generates, thereby, a straightforward argument for the necessity of
a communal practice: since correctness is said to consist in conformity with
one's fellows, correctness, and with it meaning, are possible only where
there are others with whom one may conform. But McDowell, rightly in
my view. rejects the suggestion that correct application might be analysed
in terms of communal dispositions. Indeed, as I have already noted, he
rejects the very demand for a substantive account of correctness: norms are
part of the 'bedrock', beneath which we must not dig. But if we are simply
to be allowed to take the idea of correctness for granted, unreduced and
without any prospect of reconstruction in terms of, say, actual and
counterfactual truths about communal use, how is the necessity of an
'orderly communaľ practice to be defended? From what does the demand
for orderliness flow? And from what the demand for community?
McDowelľs paper contains no helpful answers.66
Wright on the judgement-dependence of meaning
29. Crispin Wright has written about the anti-reductionist conception that:
[tlhis somewhat flat-footed response to Kripke's Sceptic may seem to provide a
good example of 'loss of problems.' . . . In fact, though, and on the contrary, I
think the real problem posed by the Sceptical Argument is acute, and is one of
Wittgenstein's fundamental concerns. But the problem is not that of answering the
Argument. The problem is that of seeing how and why the correct answer gij-en
can be correct.67
Wrighťs intriguing suggestion is that there are important constitutive
results to be gleaned from the epistemological question we shelved some
pages back: namely, how, if content properties are simply to be taken for
6 6 Though see his remarks-which I am afraid I do not understand-on a 'linguistic community
[that] is conceived as bound together, not by a match in mere externals (facts accessible to just anyone),
but by a meeting of minds'. McDowelľs problems here echo, I think, Wittgenstein's own. The main
difficulty confronting a would-be interpreter of Wittgenstein is how to reconcile his rejection of
substantive constitutive accounts-especially of meaning, see Zettel #16: 'The mistake is to say that
there is anything that meaning something consists'-with the obvious constitutive and transcendental
pretensions of the rule-following considerations. It is fashionable to soft-pedal the rejection of
constitutive questions, representing it as displaying a mere 'distrusť on Wittgenstein's part. But this
ignores the fact that the rejection of analyses and necessary and sufficient conditons is tied to extremely
important first-order theses about meaning, including, most centrally, the family-resemblance view of
concepts.
6' 'Wittgenstein's Rule-Following consideration and the Central Project of Theoretical Linguistics',
loc. cit., p. 237.
The Rule-Following Considerations 545
granted, without prospect of reconstruction either in experiential or
dispositional terms, can they be known? As we saw, Kripke attempted to
use this question to embarrass his anti-reductionist opponent. Wright,
however, has a more constructive project in mind. Pressing the epistemological
question will reveal, so he claims, that facts about content are
essentially 'judgement-dependenť.
What does it mean for a class of facts to be judgement-dependent?
Wrighťs explanation is framed in terms of a failure to pass the 'order-ofdetermination
tesť:
The order-of-determination test concerns the relation between best judgements-judgements
made in what are, with respect to their particular subject
matter, cognitively ideal conditions of both judger and circumstance-and truth.
. . . Truth, for judgements which pass the test, is a standard constituted
independently of any considerations concerning cognitive pedigree. For judgements
which fail the test, by contrast, there is no distance between being true and
being best; truth, for such judgements, is constitutively what we judge to be true
when we operate under cognitively ideal condition^.^^
We may explain the contrast Wright has in mind here by recurring to the
idea of an accessible content (see above). An accessible content is one
about which subjects are necessarily authoritative under cognitively
optimal circumstances. Now, a question may be raised about the correct
explanation for this authority: is it that, under those optimal circumstances,
subjects are exceptionally well-equipped to track the relevant,
independently constituted facts? or is it, rather, that judgements under
those circumstances simply constitute the facts in question? A fact is
judgement-independent if the former, judgement-dependent if the latter.
The contrast, then, is between facts which are constituted independently
of our judgements, however optimal, and facts which are constituted
precisely by the judgements we would form under cognitively ideal
circumstances. And the claim is that facts about content have to be
construed on the latter model. Pace Kripke, the target of the rulefollowing
considerations is not the reality of content facts, but, rather, a
judgement-independent (or Platonist, if you think these come to the same
thing) conception of their constitution. Best judgements constitutively
determine the truth-value of sentences ascribing content to mental states;
they do not track independently constituted states of affairs which confer
truth or falsity upon them.
Wright argues for this 'judgement-dependenť conception of content by
attacking the epistemologies available on the alternative model. Drawing
extensively on Wittgenstein's actual text, Wright reconstructs an interesting
set of considerations against both int~ospectiveand inferential conceptions
of self-knowledge, thus, presumably, exhausting the epistemologies
6 8 Ibid., p. 246
546 Paul A. Boghossian
available to his opponent. So long as facts about our mental states are
construed as independent of, and, hence, as tracked by our self-regarding
judgements, we can have no satisfactory explanation of our ability to know
them. On the assumption, then, that Kripke's unstable content irrealism is
to be avoided at all costs, that leaves the judgement-dependent conception
as the only contender. So goes Wrighťs argument.
Wrighťs discussion raises a number of interesting and difficult questions.
Is it really true that Wittgenstein's discussion destroys all 'cognitive
accomplishmenť theories of self-knowledge? Supposing it does, does this
inevitably drive us to a judgement-dependent conception of content? Are
there not other conceptions that would equally accommodate the rejection
of a tracking epistemology? Unfortunately, none of these questions can be
adequately addressed within the confines of the present essay. Here I have
to settle for raising a question about whether a judgement-dependent
conception of content could ever be the cogent moral of any argument.
30. The suggestion is that we must not construe facts concerning mental
content as genuine objects of cognition, and that this is to be accomplished
by regarding them as constituted by truths concerning our best judgements
about mental content. Well, what does this amount to? For
illustrative purposes, Wright offers the case of colour. What would have to
be true, if facts about colour are to judgement-dependent? We would need,
first and foremost, to secure the accessibility of colour facts, and so a
biconditional of the following form:
if C: S would judge x to be bluettx is blue.
But not just any biconditional of this form will serve to secure the
accessibility of colour. For example, unless restrictions are placed on the
permissible specifications of C, every property will turn out to be
accessible; just let C be: conditions under which S is infallible about
colour. So, it must be further required that C be specified in substantial
terms, avoiding a 'whatever-it-takes' formulation.
Now, what it would take to ground not merely the accessibility of colour
facts, but their judgement-dependence? What is needed, as Wright points
out, is that
the question whether the C-conditions, so substantially specified, are satisfied in a
particular case is logically independent of any truths concerning the details of the
extension of colour concepts.69
This seems right. For unless the specification of the C-conditions, or,
indeed, of anything else on the left-hand-side, is precluded from presupposing
facts about the colours of objects, it will remain entirely open
whether subjects' judgements, formed under the relevant C-conditions,
6 9 Ihid., pp. 247-8,
The Rule-Following Considerations 547
really did determine facts about colour. For satisfaction of the conditions
described on the left-hand-side would always presuppose some antecedently
fixed constitution of colour facts, thus undermining the claim that it
is precisely truths about best judgement that fix those facts.
No doubt, other requirements are in order as But it is, I trust,
already clear that there is a serious difficulty seeing how facts about mental
content could conceivably satisfj- the stated requirements on judgementdependence.
For it is inconceivable, given what judgement-dependence
amounts to, that the biconditionals in the case of mental content should
satisfy the requirement that their left-hand-sides be free of any assumptions
about mental content. For, at a minimum, the content of the
judgements said to fix the facts about mental content have to be presupposed.
And that means that any such biconditional will always presuppose
a constitution of mental content quite independent of constitution by best
judgement.
In a way, an intuitive difficulty should have been clear from the start. A
'judgement-dependenť conception of a given fact is, by definition, a
conception of that fact according to which it is constituted by our
judgements. The idea is clearly appropriate in connection with facts about
the chzc or the fashionable; familiar, though less clearly appropriate, in
connection with facts about colour or sound; and, it would appear,
impossible as a conception of facts about mental content. For it cannot in
general be true that facts about content are constituted by our judgements
about content: facts about content, constituted independently of the
judgements, are presupposed by the model itself.
Conclusion: robust realism-problems and prospects
3I . Let robust realism designate the view that judgements about meaning
are factual, irreducible, and judgement-independent. Then the moral of
this paper-if it has one-is that the major alternatives to robust realism
are beset by very serious difficulties.
Irrealism-the view, advocated by Kripke's Wittgenstein, that judgements
about meaning are non-factual-appears not even to be a coherent
option. (An error-theoretic variant, as promoted, for example, by Paul
Churchland, seems no better.)
Reductionist versions of realism appear to be false. The proposal that
judgements about meaning concern communal dispositions is unsatisfactory
not merely because, implausibly, it precludes the possibility of
communal error, but because it appears bound to misconstrue the meaning
of every expression in the language. The rather more promising (and
rather more popular) proposal, that judgements about meaning concern a
'O For a very illuminating discussion of the conditions that would have to be met, see ibid.,
PP,246-54 

548 Paul A. Boghossian
certain sort of idealized disposition, also appears to confront serious
difficulties: it is hard to see how the idealizations are to be specified in a
non-question-begging way.
And, finally, ajudgement-dependent conception of meaning seems not to
be a stable option, because the very idea of constitution by best judgement
appears to presuppose a judgement-independent conception of meaning.
It is sometimes said that an anti-reductionist conception is too facile a
response to the problem about meaning. It is hard not to sympathize with
this sentiment. But if the considerations canvassed against the alternatives
are correct, and if it is true that the 'rule-following' considerations leave an
anti-reductionist conception untouched, it is hard, ultimately, also to agree
with it. Meaning properties appear to be neither eliminable, nor reducible.
Perhaps it is time that we learned to live with that fact.
I do not pretend that this will be easy. Robust realism harbours some
unanswered questions, the solutions to which appear not to be trivial.
There are three main difficulties. First: what sort of room is left for
theorizing about meaning, if reductionist programs are eschewed? Second:
how are we to reconcile an anti-reductionism about meaning properties
with a satisfying conception of their causal or explanatory efficacy?And,
finally: how are we to explain our (first-person) knowledge of them?
I cannot, of course, hope to address any of these questions adequately
here. A few brief remarks will have to suffice.
T o begin with the last question first, I cannot see that an antireductionist
conception of content has a special problem about selfknowledge.
As far as I am concerned, no one has a satisfactory explanation
of our ability to know our own But I do not see that the antireductionist
need feel any special embarrassment about this. If anything, it
seems to me, the prospects are better for him than for his opponent. A
reductionist would have it that meanings are fixed by certain kinds of
dispositional fact, the sort of fact that could hardly be known observationally.
It would appear to follow that the reductionist is committed, if he is
to have a substantial epistemology of self-knowledge, to an inferential
conception-a conception that may be, as I have argued elsewhere, worse
than implausible.72 The anti-reductionist labours under no comparable
burden.
As for the charge that there would be nothing left for a theory of
meaning to be, if reductionism is eschewed, it seems to me simply false.
Let me here mention just a few of the questions that survive the rejection
of reductionist programmes. For one thing, as I have stressed, a nonreductionism
about meaning is best understood as a thesis about mental
meaning, not about linguistic meaning. So anti-reductionism, as I understand
it, is not only consistent with, but positively invites, a theory about
7 2 A ." See my 'Content and Self-Knowledge', loc. cit. galn see my ibid
The Rule-Following Considerations 549
the relation between thought and language. How do public language
symbols come to acquire meaning and what role does thought play in that
process? Secondly, anti-reductionism in my sense is consistent with wanting
a general account of the principles by which we interpret other people. The
important work of Quine, Davidson, Lewis, Grandy, and others on the
theory of radical interpretation neither needs, nor is best understood in
terms of, reductionist aspirations. Its proper goal is the articulation of the
principles we evidently successfully employ in interpreting the speech and
minds of others. And, finally, an anti-reductionism about mental content is
perfectly consistent both with substantive theories of the nature of the
propositional attitudes-that is, of what makes a given mental state a belie&
as opposed to a wish or a desire; and with the claim that the grasping of
certain mental contents depends on the grasping of others, and so with
theories of the compositional structure of mental content.
There is hardly any fear, then, that we shall run out of things to do, if
we forego reductionist programmes in the theory of mental content.
Finally, though, there is the question of mental causation: how are we to
reconcile an anti-reductionism about content properties with a satisfying
conception of their causal efficacy? It is a view long associated with
Wittgenstein himself, of course, that propositional attitude explanations
are not causal explanations. But, whether or not the view was Wittgenstein's,
it has justifiably few adherents today. As Davidson showed, if
propositional attitude explanations are to rationalize behaviour at all, then
:' they must do so by causing it.73 But propositional attitudes rationalize
partly by virtue of their content-it is partly because Neiľs belief is that
there is wine in his glass, that he reaches for it; so, propositional attitude
explanations commit us to holding that content properties have a genuine
causal role in the explanation of intentional action. But, now, how is an
anti-reductionist about content properties to accord them a genuine causal
role without committing himself, implausibly, to the essential incompleteness
of physics?
This is, I believe, the single greatest difficulty for an anti-reductionist
conception of content. It may be that it will eventually prove its undoing.
But the subject is relatively unexplored, and much interesting work
remains to be done.74
Department of Philosophy P A U L A. B O G H O S S I A N
University of Michigan
Ann Arbor, MI, USA 48109
7 3 See 'Actions, Reasons and Causes', in his Essajis on Actzons and Eoents, Oxford, Oxford
University Press, 1980.
74 For some recent papers see E. LePore and B. Loelqer, 'Mind Matters',Journal of PAtlosphy, 1987,
and Jerry Fodor, 'Making Mind Matter More', Ph~losophicalTopics, Spring 1989.