22

IV. A REFORMULATED CONTENT IRREALISM The question arises whether there is some other, more salutary, way of formulating an irrealism about content and truth. Well, one set of views we may simply set aside: no version of an error conception of content can be made to yield anything satisfactory. So long as there is no hope of confining a skepticism about content purely to the psychological domain, an error thesis about content will yield a contradiction. But it might seem that the non-factualist conception could be modified to yield something more promising. A standard non-factualist conception, applied to content discourse consists, as we saw, of the following pair of claims: (7) The predicate “true” does not refer to a property and (6) “S has truth condition p” is not truth-conditional. The difficulty was that (6) presupposes that (7) is false, so they cannot both be true. Given, however, that the difficulties for this position appear to stem solely from the joint assertion of (6) and (7), it is natural to wonder whether an acceptable content irrealism might not be formulable with the aid of only one of these propositions. After all, the non-factualist position currently under consideration was generated by a fairly mechanical application of standard nonfactualist recipes. It remains conceivable, therefore, that there exists some non-standard way of expressing an irrealism about content properties, one that will not fall prey to the difficulties uncovered above. One unpromising strategy would be to give up (7) in favor of (6). The idea here is that an irrealism about content could be secured by asserting-with (6)-that truth condition-attributing sentences are non-factual (non-truthconditional), without having to say-with (7)-that truth is not robust, that the predicate “true” does not name a genuine property. But the position is unstable, for familiar reasons. The trouble is that since truth condition-attributing sentences are declarative, denying of them-with (6)-that they are truthconditional presupposes that truth is robust, that “true” does name a genuine property. But if “true” does name a genuine property, how could sentences which attribute such properties be-as (6) claims-non-factual, not capable of genuine truth and falsehood? The opposite strategy of giving up (6) but retaining (7) seems much more promising. For a deflationism about truth-the thesis expressed by (7)-would appear to be an impeccably content irrealist position. And although now it will be impossible for the content irrealist to say, given the loss of (6), that truth condition-attributing sentences are non-truth-conditional, it should also be clear that there is no longer any need to say that. He can happily admit that such sentences are factstating and even that they are occasionally true. For with the denial that “true” refers to a robust semantical property, the admission is innocuous: it is perfectly consistent with his irrealism about the central semantic concepts. The proposal that recommends itself would appear to be, then, that content irrealism be reformulated so as to consist solely in (7)-in the denial that the predicate “true” expresses a property. The remainder of this paper will be devoted to an examination of this proposal. I shall first question whether it can accommodate the sorts of motivation that have traditionally fueled content irrealism. Then I shall question whether it manages to evade the sorts of problem that beset the earlier formulations. V. DEFLATIONARY CONCEPTIONS OF TRUTH AND CONTENT IRREALISM Most proponents of content irrealism came to that view by way of the conviction that neuroscience, or something else similarly physically basic, will ultimately provide the true story about the etiology of human behavior and cognitive activity, and that ordinary content-based psychology will not, for one reason or another, reduce to that story. In short, irrealists about content tend to be realists about physics, and, indeed, the former because the latter. But there is a serious difficulty seeing how this original motivation for content irrealism can be conserved, on the adjusted understanding of what the view consists in. For how is a realist/irrealist contrast between physics and semantics to be formulated, if content irrealism is expressed as a deflationism about truth? The arguments against error theories of content bar the irrealist from saying that contentascribing sentences differ from the sentences of physics in that the former, but not the latter, are systematically false. And the adjusted formulation of content irrealism bars him from saying that the difference consists in their differential capacities for stating facts. For if an irrealism about content is simply a deflationism about truth, then, as the discussion of Part I showed, such a view will entail that all declarative sentences, regardless of subject matter, must be treated on a par: there can be no interesting distinction between sentences that are genuinely in the business of stating facts and those that aren’t. But how now to express the conviction that inspired the whole program in the first place: namely, that there is something specifically suspect about content? If content irrealism is formulated as a deflationism about truth, no latitude is left for expressing invidious distinctions between content discourse and physics. In particular, the suggestion championed by Churchland, that content irrealism might lead to the elimination of content discourse, can no longer be coherently motivated, for the basis on which an invidious distinction between our talk of content and our talk about any other subject matter was to be constructed, no longer exists. Content irrealism itself guarantees that. The difficulties encountered here-in the attempt to preserve a special place for physics while avoiding the pitfalls of the sorts of standard irrealist construals of content lately discussed-are nicely illustrated by Stephen Schiffer’s recent book Remnants ofMeaning.33 Schiffer there argues for an irrealism about content characterized by the following pair of theses: Ontological Physicalism, which holds that there are no extra-linguistic, irreducibly psychological entities of any ontological category; and Sentential Dualism, which holds that there are true, but irreducible, belief-ascribing sentences. The attractions of the view are unmistakable: it promises to satisfy simultaneously the conviction that there are no non-physical properties, and the conviction that psychological descriptions play an indispensable role in our selfconception, and all this without relying on an implausible reductionism. 14 But can it be brought off? One way to appreciate its difficulties is as follows. Ontological Physicalism is the view that there are no extra-linguistic psychological properties. But let us ask this: are there any extra-linguistic physical properties according to Schiffer? If he says “yes,” then Sentential Dualism-the view that there are true belief-ascribing sentences-will have been exposed as a sham. For if there are extra-linguistic physical properties for the sentences of physics to answer to, but no extra-linguistic psychological properties for the sentences of psychology to answer to, then it isn’t true, in the strict and literal sense, that there are true sentences of psychology, and the overall view is indistinguishable from the sorts of standard irrealism about content that we recently found so problematic. On the other hand, if Schiffer denies the existence of properties altogether, then the view can hardly be described as a combination of Ontological Physicalism and Sentential Dualism: it would be more appropriate to describe it as a combination of Ontological Nihilism and Sentential Dualism. Or, if we prefer, and on the assumption that we can now provide a deflationary construal of “ontology,” we may describe it as a combination of Ontological Dualism and Sentential Dualism. Either way, we would have been unable to preserve a special and privileged place for physics to occupy relative to the rest of discourse.35 An irrealism about content can only be a deflationism about truth. And this is a significantly different view from what we had been led to expect. But is it at least stable in this guise? The suggestion that it isn’t is likely to meet with some resistance. Deflationary conceptions of truth, although of relatively recent provenance, have had many distinguished proponents, including many of the Vienna positivists, some of the American pragmatists, Ayer, Quine, Rorty, and others. And although one may be inclined to believe that these philosophers are wrong about truth, it seems hard to believe that their view is not even a coherent option. Be that as it may, it is actually implicit in the foregoing discussion that a deflationism about truth is an inherently unstable position. A deflationary conception of truth is the view that there really is no such property, that talk about truth and truth conditions must be understood in some way other than as talk about genuine, language-independent properties that sentences or thoughts may enjoy. It is typically expressed, as Ayer and others have expressed it, like this: (7) The predicate “true” does not refer to a property. But there is a serious problem seeing how any such view could itself be true. The point to bear in mind, from the discussion in Part I, is that the denial that a given predicate refers to, or expresses, a property, only makes sense on a robust construal of predicate reference; on a deflationary construal, there is, simply, no space for denying, of a significant, predicative expression, that it expresses a property. But, then, if this is correct, the denial-expressed in (7)-that the truth predicate refers to a property, must itself be understood as framed in terms of a robust notion of reference. Otherwise, it would amount to the false claim that a significant, predicative expression-in this case “true”-fails of deflationary reference. But this result, in conjunction with the platitude connecting reference and truth noted above-namely, that “x is P” is true if and only if the object denoted by “x” has the property expressed by “P”-implies that (7) presupposes that truth is robust. So the denial that truth is robust attempted in (7) can succeed only if it fails. It is natural to wonder whether the difficulty can be got around like this: first, deny that predicate reference is a robust notion; and then define a deflationary notion of truth in terms of this deflationary notion of reference. Notice that in so doing there is no longer any need to use the offending sentence (7); for granted the deflationism about reference, the concession that “true” does refer to a property would be innocuous: it would be perfectly consistent with a deflationism about the central semantic notions. The trouble is that an offending sentence is bound to crop up somewhere. For what, after all, is a deflationary conception of reference? Presumably, it is the view that (8) The expression “refers to a property” does not itself refer to a property. And here, it would seem, the same problem simply recurs. For, again, if the notion of “refers to a property” that is used in (8) is a deflationary notion, then (8) amounts to the false claim that a significant, predicative expression-in this case, “refers to a property”-fails of deflationary reference. It must, therefore, be understood as expressing a robust conception of reference. On the terms of this understanding, however, what it says is false, too. So yet again the attempt to state a deflationism about the central semantic notions results in incoherence. VI. CONCLUDING REMARKS By way of closing, I would like to recapitulate the principal conclusions and to indicate, however briefly, where we require more understanding than we currently possess. The history of philosophy has furnished us with two ways of thinking about a region of discourse, once we have become convinced that nothing in the world answers to its characteristic predicates-an error theory and a non-factualist theory. Both these models are constructed with the aid of certain theses about the concepts of truth, truth conditions and reference. Thus, error theories about a given region presuppose that the target sentences possess truth conditions; and nonfactualist theories presuppose robust conceptions of truth and reference. Recently, the suggestion that one or another of these models might apply also to ordinary, content-based psychology has become increasingly influential. The suggestion, however, rightly understood, is tantamount to the proposal that we construe semantic discourse as a whole along irrealist lines. And this proposal is of dubious coherence; for any irrealist conception presupposes certain claims about truth and truth conditions, which an irrealism precisely about truth entails the denial of. Perhaps the right way to express an irrealism about content is through deflationism? But this won’t help, for a deflationism about truth and reference is just a version of a non-factualist thesis about semantic concepts, and is subject to the same sorts of difficulty. If these considerations are correct, the upshot is that we have not been shown how to make sense of the question: do any properties answer to our talk about truth, truth conditions and reference? On the face of it, this is a surprising result; for it seems that we should be able to wonder whether anything answers to our semantic discourse, much as we have profitably wondered whether anything answers to our evaluative discourse. But if the argument of the present paper is correct, the whole enterprise of asking such questions is itself based on a realist understanding of semantic discourse; the suggestion, then, that irrealism might turn out to be the correct model for semantic discourse itself is a suggestion that we cannot coherently entertain. What recourse might a content irrealist have in the face of these arguments? Two main possibilities suggest themselves. He may seek to deny that it follows inexorably from a deflationary conception of reference, that every significant predicative expression expresses a property. Or, he may attempt to deny that the nature of his worries about content discourse is best expressed in terms of our standard irrealist models. The first option strikes me as extremely unpromising. I simply cannot see how, if we refuse to think of truth and reference as substantive properties, we can motivate stronger-than-minimal requirements on eligibility for truth and reference. It seems to me-as it has seemed to most deflationists-almost definitional of a deflationary conception, that eligibility should in this way be trivial.36 What of the second suggestion? The contrast between realism and irrealism has traditionally served as the contrast of choice between cognitively reputable and cognitively disreputable discourse. It ought to be noted, however, that some philosophers have recently begun to explore the possibility that a cognitively interesting contrast may be drawn in non-truth-theoretic terms.37 It is too early to say whether this project will succeed, or, for that matter, whether it would ultimately be of use in the present context, if it did. But, in light of the difficulties confronting a truth-theoretic formulation of content irrealism, it seems an avenue well worth exploring. I am inclined to believe, however, that the correct moral of the considerations on offer here is just what it appears to be: that we really cannot make sense of the suggestion that our thoughts and utterances do not possess robust truth conditions. Much as Descartes’s cogito argument may be understood to have shown that I cannot make sense of the suggestion that I do not exist, by showing that the claim that I do exist is a presupposition of the most refined attempt to deny that I do; so the present argument should be understood as showing that we cannot make sense of the claim that our thoughts and utterances do not possess robust truth conditions by showing that the claim that they do possess robust truth conditions is a presupposition of the most refined attempt to deny that they do. In either case, whether the argument ultimately succeeds depends upon whether an alternative formulation of the disputed thesis can be found, one which does not carry the self-defeating presupposition.38