21

II. THE STATUS OF CONTENT DISCOURSE Irrealist construals of content-based psychology have been formulated in both error-theoretic and non-factualist versions; and the error alternative, at least, in both eliminative and instrumentalist guises. Thus, Paul Churchland has defended the thesis that our common sense psychological framework is a false and radically misleading conception of the causes of human behavior and cognitive activity. On this view, folk psychology is not just an incomplete representation of our inner states; it is an outright misrepresentation of our internal states and activities. Consequently, we cannot expect a truly adequate neuroscientific account of our lives to provide theoretical categories that match up nicely with the categories of our common sense framework. Accordingly, we must expect that the older framework will simply be eliminated, rather than reduced, by a matured neuroscience. 19 And Daniel Dennett and Kripke’s Wittgenstein have endorsed a conception of psychological discourse that seems best understood along non-factualist lines. According to Dennett, the “attribution” of beliefs and desires to something consists in nothing over and above the adoption of a certain sort of (predictive) “stance” toward it; it ought not to be understood as an attempt to describe properties of that thing, an attempt that would misfire if the relevant properties were not to obtain. A system counts as having beliefs and desires if it is an intentional system; and it counts as an intentional system if it can be successfully predicted from the intentional stance. We do quite successfully treat these [chess-playing] computers as intentional systems, and we do this quite independently of any considerations about what substance they are composed of, their origin, their position or lack of position in the community of moral agents, their consciousness or self-consciousness. … The decision to adopt the [intentional] strategy is pragmatic, and is not intrinsically right or wrong.20 And Kripke has interpreted Wittgenstein as holding that statements involving the notion of meaning or content have no truth conditions, but only conditions of warranted or justified use. All that is needed to legitimize assertions that someone means something is that there be roughly specified circumstances under which they are legitimately assertable, and that the game of asserting them has a role in our lives. No supposition that `facts correspond’ to those assertions is needed.21 It is important to emphasize that these conceptions of content-based psychological discourse have been presented as conforming fully to the paradigm examples of irrealism about other domains. The motivations for psychological irrealisms conform to the traditional model: the worry is that nothing in the world answers to our talk of belief and desire. And the irrealist reactions have been formulated in traditional ways: as recommending either that we view content attributions as false, or that we view them as not fact stating. In Parts III-V, I shall turn to a discussion of the cogency of such conceptions. But first I want to clear up a couple of questions about how precisely they are to be understood. The first question comes up as follows. Irrealists about content tend to restrict their thesis to our ordinary talk of the psychological; missing, usually, is any suggestion that we should also regard it as the correct conception of the idioms we employ in characterizing linguistic behavior.22 Indeed, not only is such a suggestion typically missing, it is occasionally explicitly denied. (This comes up particularly sharply in some of Churchland’s writings, where he goes so far as to call for the creation of a new theory of meaning for natural languages, one that would be consistent with the falsity of ordinary psychological discourse.)23 The suggestion that we treat the notions of linguistic and mental meaning differentially is surprising, however, given the very close affinity between them. Could an irrealism about mental content really be made to cohabit with a realism about linguistic meaning? One argument for the conclusion that it cannot has been given by Lynne Rudder Baker. She has argued that the claim that public-language expressions derive their content from the content properties of mental states is platitudinous and non-optional: [L]anguage can be meaningful only if it is possible that someone mean something. This is a platitude, not a theory. It is clearly incumbent upon anyone who wants to deny the platitude to show that there can be meaningful language even if no one has meant anything, even if no one has ever intended anything.24 Unfortunately, the argument doesn’t convince. Of course, we may grant Baker that there are platitudes connecting our talk of linguistic meaning with our talk of contentful mental states; so that it is, let us suppose, always permissible to move from “S means something in O’s mouth,” to “0 intends something by the use of S.” But such platitudes, by themselves, can do nothing to secure Baker’s claim; for they cannot, by themselves, constrain how they are to be understood. In particular, they cannot ensure that the sentence “0 intends something by the use of S” has to be understood realistically, as describing a genuine mental fact from which the meaning of S is derived, rather than being understood as a mere notational variant for the sentence-“S means something in O’s mouth”-to which it is, by assumption, platitudinously connected. And so, for all that the platitudes show, it remains wide open that the meaning of S is determined by non-mental factors, by facts about “use,” for example. None of this, of course, amounts to an endorsement of non-mental theories of linguistic meaning. It amounts simply to the insistence that the question of the truth of such theories is not settled by the availability of platitudes connecting linguistic and mental concepts. The real difficulty with the suggestion that one may sustain differential attitudes toward mental and linguistic content stems from the fact that the best arguments for the claim that nothing mental possesses content would count as equally good arguments for the claim that nothing linguistic does. For these arguments have nothing much to do with the items being mental and everything to do with their being contentful: they are considerations, of a wholly general character, against the existence of items individuated by content. If successful, then, they should tend to undermine the idea of linguistic content just as much as they threaten its mental counterpart.25 The considerations in question are classifiable into four kinds: arguments from the indeterminacy of content, arguments from the holistic character of content, arguments from the irreducibility of content and arguments from the “queerness” of content. It is a famous claim of Quine’s that for any mental state or linguistic expression, a pair of content ascriptions can always be devised which would be such that, although they could not both be true, no rational considerations could decide between them. He took this to show that ascriptions of meaning were not a genuinely factual matter.26 Considerations of the second kind, due to Stephen Stich, attempt to show that attributions of contentful states are governed by holistic and contextsensitive criteria, and that this feature militates against the existence of content properties.27 A third kind of content-skeptical argument, common to many philosophers, proceeds from the presumed failure of naturalistic reductions of content: content properties are not genuine properties because they are not reducible to the only properties that are. 2″ And, finally, arguments from “queerness”-advocated recently by Kripke’s Wittgenstein-claim that the content properties envisaged by common sense could not be real, because no real property could have the sorts of feature that common sense considers constitutive of content.29 I have no interest at the moment in the soundness of these arguments. My only concern is to point out that, if effective at all, they should be as effective against linguistic content as they are against mental content. This is evident from the fact that the arguments construct their skeptical case by exploiting features of content properties, but without exploiting any facts about the putative bearers of those properties. Thus, they would apply to anything said to possess content, whether it was mental or not. I have been arguing that one ought not to be an irrealist about mental content attributions, unless one is prepared to be an irrealist about all content attributions. But what notion of content is in question, exactly? It is a question of considerable contemporary controversy whether the ordinary notion of content may be understood to consist simply in the idea of a truth condition, or whether it has to be conceived as consisting in something more fine-grained. Fregean opacity phenomena pull, of course, in the latter, more ambitious, direction; but it remains unclear whether those phenomena are decisive or whether they can be handled in a way that conserves the more modest, truthconditional construal.30 What is not controversial, however, is that the essential core of the ordinary notion of content does consist simply in the idea of a truth condition; so that even if, for whatever reason, we had to forego something more fine-grained, that would still count as a significant commitment to the ordinary notion. And, for all we now know, truth conditions may be all the ordinary notion calls for in any event. What this suggests is that a skepticism about content, if it is to be interesting, must be directed primarily at the idea of a truth condition and not at any more ambitious construal of the ordinary notion. Otherwise, it will be easy to deflect the skepticism by settling for the modest construal, a course of action that may be forced upon us in any case by considerations internal to the theory of content. Fortunately for contemporary content skeptics, it seems clear that the standard arguments for content skepticism (reviewed above) do not exploit the complexities that opacity phenomena induce: they would apply even on a modest, truth-conditional construal of content.31 For the remainder of this paper, therefore, I will assume that contents just are truth conditions. III. IRREALIST CONCEPTIONS OF CONTENT A summary may be useful at this stage. An examination of the standard recipes for constructing irrealist conceptions revealed that non factualist theories presuppose robust conceptions of truth and reference; and that error theories presuppose that the target sentences possess truth conditions (on either a robust or a deflationary construal of truth). I have also argued that one ought not to be an irrealist about psychological content without being an irrealist about all content attributions; and that the relevant notion of content may be assumed to consist simply in the idea of a truth condition. We are finally in a position to assess the cogency of irrealist construals of content. Consider first an error conception. As the preceding discussion has argued, this amounts finally to the claim that (4) All sentences of the form “S has truth condition p” are false, where S is to be understood as ranging over sentences in the language of thought, or neural structures, as well as over public-language sentences. But, now, (4) would seem to have the immediate consequence that no sentence has a truth condition. For whatever one’s conception of “true”-whether robust or deflationary-a sentence of the form “S has truth condition p” will be true if and only if S really does have truth condition p; this is, of course, nothing but a reflection of the truth predicate’s disquotational properties, properties it possesses on any conception of truth. And so, since “S has truth condition p” is true if and only if S has truth condition p, then, since all sentences of that form are held to be false, for no S and for no p does S have truth condition p. Now, however, a problem would seem immediate. For (4) implies, that is, that no sentence whatever has a truth condition. But what (4) says is that all truth condition-attributing sentences are false. And these sentences cannot be false unless they have truth conditions to begin with. Hence, (4) implies both that truth condition-attributing sentences have truth conditions and that they don’t have them. This is a contradiction. What about a non-factualist conception of content? Applying the standard recipe for constructing such conceptions-namely, (1) and (2)-to this case, we see that a non-factualism about content comes to the view that content predicates do not express properties and (hence) that content-attributing sentences are not genuinely truth-conditional. That is, the view consists in the pair of claims: (5) The predicate “has truth condition p” does not refer to a property, and (6) “S has truth condition p” is not truth-conditional. Notice, however, that (5) entails (7) “true” does not refer to a property. For the truth value of a sentence is fully determined by its truth condition and the relevant worldly facts. There is no way, then, that a sentence’s possessing a truth value could be a thoroughly factual matter (“true” does express a property) if there is non-factuality in one of its determinants (“has truth condition p” does not express a property). A non-factualism about content amounts, therefore, to (6) and (7).32 But now here too a contradiction seems apparent. For we saw in Part I that the idea of a significant declarative sentence failing to possess truth conditions is an idea that presupposes that “true” does refer to a property: it presupposes a robust, as opposed to deflationary, conception of truth. It follows, therefore, that a non-factualism about content is seen to consist in a pair of claims, one of which presupposes the negation of the other. For (6) is the denial that a declarative sentence possesses truth conditions, which presupposes that truth is robust; whereas (7) is the denial that truth is robust. Now, this seems an extremely curious result, doesn’t it?-no irrealist conception of content, modelled on standard formulations of irrealist theses about other subject matters, yields a coherent view. How can this be? Irrealist conceptions of other domains-of ethics, for example, or of mathematics-may not be particularly appealing or plausible; but they’re not incoherent. Why should matters stand differently with content discourse? The source of the asymmetry is not hard to find. It derives from the fact that error and nonfactualist theories about any subject matter presuppose certain claims about truth and truth conditions, which an error or non-factualist conception directed precisely at truth ends up denying. Not surprisingly, the ensuing result is unstable. Thus, an error thesis about any subject matter presupposes, by its very nature, that the target sentences are truth-conditional. But an error thesis directed precisely at our talk of truth conditions themselves entails the denial of that presupposition. Thus, also, a non-factualism about any subject matter presupposes a robust conception of truth and reference. But a non-factualism directed precisely at truth entails the denial of that presupposition. The conclusion, it seems to me, is inescapable: if there is a genuine issue about the status of content discourse, it cannot be formulated in accordance with our standard irrealist models.