On the Proposed Exhaustion of Truth

John Collins


In the paper’s first part, I present an umbrella thesis - the exhaustion thesis - which captures the core component of the various deflationary positions on truth: the claim that the content of the truth predicate is exhausted by the content of that to which the predicate applies. I argue that this thesis is only trivially supported by the common thought that truth withstands substantive analysis, for predicates in general do not succumb to such analysis. I then consider two positive articulations of the exhaustion thesis  and argue that neither begins to show that the thesis is explanatorily adequate.


1: Introduction

Deflationism is perhaps the prevailing conception of truth within contemporary philosophy. The chief reason for this ascendancy, I think, is that deflationary theories present themselves to be neutral between all disputes in epistemology and metaphysics. This offers deflationism a straightforward dialectical advantage over the more traditional theories that seek to explicate being true via notions which essentially embroil the concept in substantive disputes to do with the nature of mind, knowledge and the world. In crude terms, deflationism allows one to say something simple about truth without concomitantly deciding upon the character of nigh-on every other concept of philosophical concern. To a great degree, then, the accrued value of deflationism is due to a favourable contrast with more traditional theories. While I think that such invidious asymmetries are true enough, they furnish little reason to favour deflationism in itself. My argument will be that an anti-substantive account of the property of truth may, indeed should, be common ground for all. What else is required from deflationism as proprietarily understood is a thesis about the content of truth - the exhaustion thesis. But insofar as this thesis has been articulated and defended it is beset by insuperable conceptual and empirical problems, or so I shall contend.


2: The Exhaustion Thesis

Deflationism is a familiarly mixed bag; nonetheless, its core may be captured by what I shall call the exhaustion thesis:


(ET) The content (/sense/cognitive significance) of is true is exhausted by the content of

         that to which it applies, or a specification thereof.[1]


ET does not tell us what true means, what subjects grasp when they are competent with the concept of truth, for ET gives no suggestion what content amounts to; its message, rather, is that truth brings with it no more content than must already be grasped to entertain that which is being evaluated as true. This thought is, for sure, somewhat impressionistic, although I offer it not as a precise analysis but more as the identification of a schematic relationship which may be variously realised. What is neatly captured though, I think, is the characteristic flatness of deflationary theories: the content of P is true is not explicated by mentioning  P or any of its constituent words as subjects or relata of a further predicate; rather, the content of the predication is explicated by the use of P itself: there is no underlying predication which may articulate that P is true other than the predication contained in P itself.[2] In this sense, ET should be understood as a proposed  adequacy condition on accounts of truth. The condition has both a negative and positive aspect. If an account of truth proposes that the concept has to do, say, with the employment of ideal rationality in epistemically ideal conditions, then ET rules against the account, for surely the content of that to which we apply the truth predicate is not typically  concerned with ideal rationality under any conditions. The point holds generally: tendentious notions from epistemology or metaphysics appear to be expressed by just those judgments which are transparently about such issues. On the other hand, various contrary accounts are compatible with ET and the extent to which they are is the extent to which I shall understand them to be deflationist.

      Roughly, Ramsey (1927), Ayer (1936), Strawson (1950), and Williams (1976) hold that the truth predicate plays a purely pragmatic or grammatical role that does not involve the expression of a property. On such an account, ET is vacuously satisfied, or perhaps may be offered as an explanation of the semantic inertness of truth. Alternatively, Horwich (1990), Field (1994), and Soames (1999) suggest that is true does have an independent content, expresses a property, but that such a content is fully specified via the specification of the content to which the predicate applies (as we shall see, this idea can be variously realised). Similarly, prosentential theories of truth (e.g., Grover (1992) and Brandom (1994)) analyse truth in terms of an anaphoric role relaised by predications of truth, such that the predications inherit the content of that to which the predication is made. Again, ET is transparently satisfied.[3] One may think of inter-deflationary disputes and differences as disagreements over how to cleave properly to ET.       

       ET can be usefully understood as commending a degenerate version of what is perceived to be the traditional analytical endeavour. This view of conceptual analysis has it that a concept C is explicated (in some sense) by the articulation of C in terms of a uniform set of more primitive concepts which apply just when the target concept applies; that is, analyses seek instantiations of the schema TM:


(TM) ("x)[F(x) ↔ …x…].


The predicate on the left flank expresses our target concept C, and the blank on the right flank is to be filled by a complex predicate that express the content of the target predicate. Among the restrictions on the filling is that it should be antecedently understood and not include ‘F’ itself as a proper part. As regards truth, the following are paradigm cases:


(CT) ("x)[TRUE(x) ↔ ($p)(x represents p and p obtains)]

(PT) ("x)[TRUE(x) ↔ x is a member of a maximally coherent belief set]

(JT) ("x)[TRUE(x) ↔ x would be rationally agreed upon at the end of inquiry]


It is moot whether such proposals are best thought of as strict instances of the TM paradigm or as loser elucidations (Wright (1999)). Still, the crucial feature is retained that truth is a non-primitive concept which may be unpacked into metaphysical and/or epistemological notions which uniformly hold for all truths.

        Now ET amounts to a deflation of the advertised analyses in the sense that the blank of TM is not to be filled; rather, we need to find a way of letting that claimed to be true to stand alone on the right flank of the equivalence. We may thus think of ET as commending a null instance of TM vis-à-vis truth, not in the sense that nothing at all can go on the right flank, but that there is no predicate which uniformly occurs there to capture the cognitive import of is true across all potential subjects. This is a direct consequence of the ET injunction that we don’t require any explicatory concepts other than those expressed in the sentences (/propositions) claimed to be true. This approach is cognate with what I early termed the flatness of deflationary accounts: a sentence’s truth is not explained by a further predication to the sentence or its parts, but by the articulation of its content - the predication of is true is explained by flattening it onto that claimed to be true.  Thus, any (partial) uniformity we might find in accounting for what it would take for a range of sentences to be true will simply be a reflection of the co-occurrence of concepts expressed in the sentences. With this in mind, we do not need to have a commitment to any particular theory of content to see that the analyses above contravene ET, for if the concepts expressed in the subjects of the truth predicate  are sufficient to capture the predicate’s content, then uniform notions of rationality or correspondence are explanatorily otiose precisely because they are not common features of all concepts.[4] But to deny any such unpacking is just to relinquish TM on its proprietary understanding, for no particular predicate, whether complex or simple, is called upon to fill the blank in every instance. The nomenclature, ‘deflationism’, signals that any such putative filling will speciously inflate truth. Consider Horwich (1990, p.6):


Unlike most other properties, being true is insusceptible to conceptual or scientific analysis. No wonder that its ‘underlying nature’ has so stubbornly resisted philosophical elaboration; for there simply is no such thing.


Similar thoughts are expressed by Quine (1960, 1970/86), Leeds (1978), Grover, et al. (1975),  Soames (1999), and Field (1994), among many others. As indicated above, there are many subtle and fundamental distinctions between these thinkers. Nevertheless, all may be understood to reject the uniform TM model on the grounds of ET. This common theme is typically expressed by the thought that an adequate account of truth will consist of a specification (in some form, implicit or explicit) of each instance of the schema DT:


(DT) TRUE(P) iff P,


where ‘TRUE(P)’ picks out sentences which variously attribute truth to a clause P:


(1)a. It is true that snow is white iff snow is white

     b.‘Grass is green’ is true iff grass is green

c. The proposition (claim/belief/conjecture, etc.) that the sky is blue is true iff the sky

    is blue.


      Thus, the content of P is true is given by P itself, not P in relation to any other elements; there is no general, uniform analysis that answers the question What is truth?.[5] It does not follow, though, that there is no property of truth, that ET can only be vacuously satisfied.

       The problem with such nihilism, first pointed out by Tarski (1944, 1956), although noted by Ramsey (1927), is that the predicate ‘is true’ takes subjects in which that (sentence, proposition, etc.) claimed to be true is not recoverable by the mere deletion of lexical or punctuational material. The standard examples of such subjects are quantifier and quasi-quantifier NPs:


(2)a. Everything/Something Bill said is true

b.  What Bill said is true

c.  The sentence Bill uttered is true

d.  Bill’s claim is true.


‘True’ ineliminably occurs in such constructions, it cannot be parsed away on the model of (1). This feature of ineliminability is most often highlighted by the failure of a potential regimentation of such constructions into first-order quantification theory: a quantifier would be asked to bind two variables ranging over truth bearers (sayings, utterances, etc.), but such variables are objectual, and so in need of predicates. The elision of ‘is true’ would issue in ungrammatical nonsense; e.g.,


(3) ("x)[Bill said xx].[6]


        So much, of course, is not a demonstration that there is a property of truth, still less is to it say anything about what the property might amount to. It does, though, undercut any inference from the eliminability of the truth predicate in constructions such as (1)a-c to truth’s not being a property. But if there is a property of truth, then why does it, as revealed in (1)a-c, appear not to enter ineliminably into the determination of the truth of truth predications?

       The resolution of this quandary, due, in essence, to Quine (1960, 1970/86) and endorsed by Leeds (1978), Soames (1984, 1999), Horwich (1990), and Field (1994),  is to give truth no more credit than that which is due from its provision of a predicate for the stranded variable in constructions such as (3). That is, truth earns its conceptual keep, as it were, by allowing us to generalise, or be otherwise indefinite or circumlocutory about that we wish to assert. If we were always only concerned with presented individual sentences, then, by DT, truth would be a superfluous indirection. Sometimes, of course, the service is indispensable, when, say, our truth claim covers an infinity of sentences/propositions.

       So, rather than being anomalous, (2)a-d highlight the need for truth, notwithstanding its insubstantiality. Still, truth is merely ‘useful’.[7] This is the message of DT. Given any sentence as value of the bound variable argument of the truth predicate, we have an instance of the left-hand side of DT, which may then be substituted for its disquotation or, more generally, the clause to which truth is attributed. We may model this logic by a simple argument form, the a priori acceptability of whose instances putatively reveals that the truth predicate, even where not merely an indirection, has a content still caught by the content it applies to, and so remains in line with ET. Thus, where (i) is the form of a truth predication to a quantifier NP (e.g., ‘Everything Bill says is true’):


            (i) ("x)[F(x) → TRUE(x)]                      (Assumption)   

           (ii) F(P)                                  (Assumption)

           (iii) F(P) → TRUE(P)                           (UI (i))

           (iii) TRUE(P)                                  (MPP (ii), (iii))

           (iv)  P                                      (DT)

The argument goes through whatever ‘P’ we chose; effectively, from (i), the restriction, F, on the quantifier subjects of is true acts as a disquotational licence for the assertion of that claimed to be true. This idea is quite explicit in Field (1986) and Horwich (1990), where, in terms of content, the truth predicate serves as an abbreviation of the DT instances of the sentences (/propositions) covered by the quantifier’s restriction. Thus, (i) says no more than (AT):


(AT) If F(S1), then S1; and if F(S2), then S2; and if F(S3), then S3


The truth predicate acts as a proxy for a substitutional quantifier; that is, truth bearers can be values of objectual variables rather than substitutends for sentential place holders, but the effect of the truth predicate is to reduce the generalisation about the sentences/propositions covered to the compendious assertion of the infinite conjunctions/disjunctions of the truth bears; and so we commit ourselves to no more than that which the sentences covered say. Following Quine (1960; 1970/86), let us refer to this means as semantic ascent: truth allows us to sally back and forth between language talk and world talk without extending the content of truth beyond the contents expressed in the ‘language’.

      Now,  semantic ascent might provide an explanation for why we should have a deflated predicate, but it hardly demonstrates that the truth predicate is deflated. It is perfectly coherent to think of truth as a concept that is not adequately captured by DT’s instances but which also enables semantic ascent. There is some confusion about this.

      Horwich, as quoted above, bases the claim that truth has no ‘underlying nature’ on the fact that, “unlike most other properties”, it fails to succumb to TM in the sense that it has no uniform analysis. Clearly, if most properties do have essences which my be picked out on the right flank of instances of TM, then the recalcitrance of truth - the ‘fact’ that the truth of ‘Snow is white’ is flat with the whiteness of snow, while the truth of ‘Grass is green’ is flat with the greenness of grass, etc. - would alone perhaps license the deflationary reading. But, I should say, it is not unreasonable to think that properties in general fail to succumb to TM. To show this in detail is beyond my present scope. I am only currently interested in arguing that it is more reasonable to view the failure of TM vis-à-vis truth as a symptom of the general inadequacy of TM, rather than as a revelation about truth in particular, still less as a direct argument for deflationism. Let us begin by taking Horwich at his word in 1990.

      Horwich neglects to mention any of those other properties apart from truth that have given way under analysis, but he explicitly intends the claim to cover both colloquial and scientific or formal notions. But does Horwich seriously think that we have analyses of chair, beauty, to play, to paint, to be, bold, funny, on, from, and so on indefinitely? And what of the history of the philosophical analysis of the notions of knowledge, cause, freedom, good, probable, meaning, consciousness, self, etc.? On a simple inductive measure, it appears that our failure to find an underlying nature to truth tells us more about analysis than it does about truth. We can, of course, explicate the notions which have traditionally occupied philosophers in various ways, some of which might commend broad agreement, but the endeavor to yoke such explications under TM has been singularly unsuccessful.[8] One does not require any particular explanation of this failure  - that, say, there is no analytic/synthetic distinction (Quine (1953)) - to recognise the failure in itself (but see below). What of ‘scientific analysis’?

         Presumably, Horwich has in mind the identification of natural kinds. Now it might be that we understand our colloquial natural kind terms to denote essences, to be rigid (Kripke (1980); Putnam (1975)), but it would be illegitimate to move from this semantic thesis to the thought that science attempts to discover such essences, still more so that it has been successful in doing so (Putnam (1990); Chomsky (2000)). Let us agree that the very reason some natural kind term K features in a scientific theory is that it is projectible: through it we can frame counterfactual supporting predictive generalisations about the phenomena at issue. In this sense, K is a primitive predicate as far as the containing theory is concerned. We do not employ K simply as a convenient proxy for some more complex predicate G. To do so would be to settle for fundamentally superficial generalisations. On very rare occasions we can ‘reduce’ one theory to another, but this typically involves either modifications to both theories (e.g., the explanation of chemical elements at the atomic level) or does not issue in predicates for predicates but manageable disjunctions, which themselves do not feature in the proprietary generalisations. (e.g., heat-in-gases or heat-in-solids…). If we do find underlying properties, then, ipso facto, we have found better theories of the phenomena at issue - it is far from clear that we have analysed anything at all.

        These points, I think, should be trivially acceptable. That they may not be is perhaps due to the feeling that if being a K is not to be pleonastic,  then K must have some constitution, as if there is some underlying property to, say, beauty or swimming, even if our science will not stretch to it. Well, science explains what it can, but it does not proceed from our colloquial concepts and categories, it proceeds from whatever concepts and categories - extant or created - capture generalisations that explain and predict the data at hand and any novel data as it arises. It does not begin with our intuitive ontological schema (or any other for that matter) and attempt to discover the ‘underlying reality’. The point, then, is not so much that we should not hold our breath for a scientific theory of, say, swimming in general, but that there is no reason whatsoever to think that being a swimmer is a natural kind. Theories, we may say, mold their concepts in the search for finer explanations; any correspondence with our intuitive concepts about the same phenomenon will be accidental.[9]

      Notwithstanding these essentially empirical complaints, the TM model suffers from a  general conceptual problem to which Fodor (1975, 1981, 1998a) has long been drawing our attention. Not all concepts can have analyses, for that would lead to endless analysis in that any concept featured in an analysans would be the analysand for some further analysans. There must, therefore, be primitive concepts. But to be primitive just means not to be defined, not to possess an analysis. The TM format, therefore, cannot apply to every concept. Which ones are exempt? Well, wherever the line is drawn between the defined and the defining appears to be arbitrary in at least two ways. First, it is not clear that there is any principle which determines that possession of a given concept entails possession of another. One may, of course, appeal to standard containment taxonomies: e.g., dogs ` mammals ` animals ` natural objects (as opposed to artifacts). This might take us a certain distance, but clearly not all the way. Examples from biology are unduly convincing because there is a natural organisation into species, phyla, etc., although the nature of such containment is highly contentious and patently does not reflect our intuitive concepts. However, for most concepts there is not such a natural hierarchy. Moreover, it is unlikely that possession of one concept involves possessing the containing concepts. If we use lexical acquisition as a guide, then, ontogenically, the most basic concepts appear to be ‘middle terms’: dogs, say, as opposed to poodles or mammals. Accordingly, paradigm theories of concepts, for example, would reverse the kind of containment indicated above: our notion of an animal, say, is based upon the features of certain common and easily discriminable cases or paradigms; perhaps we thus arrive at animal from dog! I do not want to suggest that such accounts are correct (see n.9), only that independent of a (to some degree) confirmed theory of human conceptuality, it is difficult to make sense of the notion that a range of concepts are constitutively basic.

        Second, unless there is some principle in play which dictates when analysis should terminate, then to stop just here appears unjustified: Why should one have begun the procedure at all or not stopped at any earlier stage? With regard to truth in particular, why should representation or ideal rationality or facts or coherence or… cabbage be thought to be more primitive than truth or a suitable place to stop? The concepts seem to be as much in need of analysis as truth itself. But such remarks hold, I think, for all purported analyses. Consider knowledge and its traditional justified true belief gloss. If we leave to one side the familiar counterexamples (beginning with Gettier (1963)) where justification and truth come apart, it is still unclear why we should find the constituents of the analysans any less opaque (or transparent) than the target notion. The traditional empiricist answer to this quandary has been to base all  concepts on phenomenal features - the Given. This at least has the virtue of positing a base which is not analysable further (albeit by definition). Yet if the philosophy of the previous century taught us anything, it is surely that such phenomenalism is false. It would not be stretching the point, I think, to say that the failure of phenomenalism signals the failure of the TM model in general, for the problem with it is not so much that being a grandmother, say, is not definable in phenomenal terms, although it is not, but that there is no reason whatsoever for holding that all we can think about unpacks into a shared primitive base: the real problem is: What does being a grandmother have to do with being a quark or being a carburetor?

         Horwich (1999, pp.239-40, n.2) has latterly revised his 1990 judgment in apparent response to objections to his implied commitment to TM; his present claim is merely that “there is nothing to be said - not even very roughly speaking - about what [truth] consists in”. So, even on the assumption that the TM format is forlorn, truth is still peculiar in that, again unlike most other properties, presumably, we cannot even sketch, gloss or elucidate what truth is. In essence, the thought seems to be that whereas most concepts at least present themselves as being analysable - they possess some rough and ready gloss that might potentially be made precise - truth fails to meet even this necessary condition for analysis.

      Horwich’s earlier remarks asked us to notice that truth is unlike “most other properties” so that we might be drawn to deflationism as a good (the best?) explanation for why truth is so peculiar. The historical failure to analyse truth, let us agree, does militate for deflationism, at least against the traditional analyses. But the weaker present case against a substantive account of truth appears to be wholly unsupported. Familiarly, associated with truth are a set of intuitions or platitudes, such as its being the goal of inquiry, its being glossable as correspondence to the facts (the correspondence intuition), its relations with the concepts of proof, knowledge, justification, reliability, etc., its role in deduction (truth is that which validity preserves), its role in semantic theorising (e.g., the relation between sub-sentential satisfaction conditions and truth being taken to be explanatory of compositionality), and so on. Prima facie, these associations are not to be denied; the question for everyone is how they are to be accommodated into one’s general account. The  deflationist is free to argue, as Horwich (1990, 1998), Field (1994), and Soames (1999) do, that all these associations can be explained without departing from DT, or be otherwise diagnosed, but that is precisely to admit that, pre-theoretically, we can say quite a bit - roughly speaking - about truth.

      Clearly, then, it is not that deflationism is supported by the independently established ‘fact’ that we cannot say anything about truth, as deflationism is supported by the failure of TM, albeit trivially; rather, the ‘fact’ is being offered as a supposed happy consequence of deflationism, under which we are enjoined to diagnose the apparent associations truth supports as essentially separate notions which have been fallaciously yoked to truth under inflationist dogma. Yet we are not given any reason to welcome this consequence, still less a reason to think that deflationism is to any degree confirmed. Of course, it is perfectly legitimate for the deflationist to seek to explain as much as he can with the minimum of conceptual resources. I am not here disputing deflationism as a methodology, although, as I shall shortly argue, I think the core exhaustion thesis faces insuperable problems. My present complaint, rather, is that, independent of what else deflationism might have going for it, once we are free of the pull of TM in general, there is no reason to view DT as licensed by default or even so much as commended because truth is not amenable to analysis. There is a growing awareness of the false dichotomy - TM or deflationism - that lies behind Howich’s reasoning. For example, in opposition to deflationism, Wiggins (1980), Wright (1992), and Davidson (1990, 1997), in their different ways, have sought to integrate the kind of associations pointed to into a richer conception of truth while also explicitly shunning TM. Similarly, those of a pragmatist bent seek to stand apart from deflationism vs. the TM format (e.g., Misak (1998)).[10] But none of these parties are perforce drawn to think that there is an essence to truth, quite the opposite in fact. The message appears to be that of  Karl Krauss: ‘If I’m asked to choose the lesser of two evils, I’ll choose neither’.

            I am not, therefore, suggesting that truth is indeed substantive or robust, nor even that the deflationists are incorrect to think that they can help themselves to a property as thin as truth is on their model. Boghossian (1990) has insisted that deflationism’s property of truth is, in some sense, degenerate, language dependent, non-factual. On this matter, at least, I am with the deflationists[11], but for general reasons: for the vast majority of conceivable properties, if we want to provide an analysis or give necessary and sufficient conditions for their realisation (in short, something more robust than an explicatory gloss), then we can do no better than disquotationally read them off corresponding predicates. We do of course depart from such disquotational purity for various explanatory ends. If one is not sure of the difference between being melancholic and being sad, then it is of no help to be told that ‘is melancholic’ applies to x just if x is melancholic. Likewise, we often stipulate in, for example, legal or official contexts, where we require some way of discriminating cases (e.g., we shall say that the tenant is the undersigned who…’). Similarly, for the purposes of discussion or the avoidance of vagueness we often agree on some paraphrase of a vague or contentious concept (e.g., being tall is being over 6ft; being a realist is being someone who accepts classical logic over the area of discourse in question). Indeed, it might be that we never concern ourselves with disquotational specifications of properties precisely because they are trivial, but, and this is the point, it is only such specifications we understand to be unhedged, non-stipulative, indisputable, unrevisable, independent of our particular and variable capacities of discrimination. I can, then, see no distinction between the deflationist attitude to the property of being true vis-à-vis analysis and the one which is appropriate in general. My complaint is not that truth is more robust than deflationism claims, but that the non-robustness of truth is quite trivial; the status of truth vis-à-vis TM is in no way peculiar. If some property P does have a significant constitution, then it will not be unearthed by conceptual analysis of ‘is P’.

       If, therefore, deflationism’s specific claim that the instances of DT constitute, in some sense, an adequate account of truth is to be supported, more is required than the mere renunciation of analysis. Otherwise put, that nothing of interest can be said about the constitution of being true does not mean that an adequate account of the content of ‘is true’ is similarly trivial.


3: On the Proposed Meaning of Truth

If ET is to be confirmed, then it is not enough merely to appeal to the failure of the TM model, for that should be common ground. What is required is a positive account of the content of truth that does not go beyond the instances of DT. There are perhaps just two broad approaches to this desideratum that are currently being pursued: the equivalence and dispositional views.

          The equivalence view is simply that instances of DT express an equivalence of meaning; that is, P is true means the same as/is synonymous with/is cognitively equivalent to P. This view is consonant with the nihilism of Ramsey, Ayer, Strawson, and Williams; for clearly, if ‘is true’ is empty, fails to express a property, then its predication to a sentence adds at best a pragmatic dimension, rather than a change of content. Perhaps Frege (1952, 1976) also held an equivalence view, although it would be inaccurate to regard him as a nihilist; rather, his view appears to be that truth is an inherent aspect of content, so that an attribution of truth is simply making explicit what is already present in the thought. The prosentential theory is committed to the  equivalence view in that it claims that the predication of truth across all subjects forms a sentential inheritor which anaphorically inherits its content from an antecedent sentence. Thus, P is true  inherits its whole content from P; What Bill said is true inherits its whole content from whatever Bill said, etc.; the difference between the members of the pairs is at best pragmatic. Field (1994) has also commended an equivalence view, where P and P is true play the same cognitive role in speakers’ idiolects: defined over the sentences one understands, ‘is true’ is semantically inert (I shall presently look more closer at Field). Quine (1960, 1970/86) also appears to hold an equivalence view, but since his views on content are orthogonal to the typical deflationist position, he is awkward to place. He certainly would resist any claims of synonymy.

      It bears emphasis that, while the present view is one of equivalence, it is not of the traditional TM stripe, where a predicate is offered as an explicit definiens. An open sentence, however, is available in terms of which we can generalise DT:


(ST) ("x)[TRUE(x) ↔ (Sp)(x = ‘p’ & p)],


where ‘p’ is a substitutional variable ranging over the sentences of a given language or idiolect. This, however, is a mere schematic abbreviation of the infinite set of instances of DT; it offers no conceptual  enlightenment, still less does it answer to TM on its proprietary understanding.

     The dispositional view does not seek to establish a relation of synonymy between truth predications  and their disquotations; even so, DT is understood to fix the content of truth. The position I have in mind is that of Horwich (1990). His claim is that “a[n] [English] person’s understanding of the truth predicate, ‘is true’ - his knowledge of its meaning - consists in his disposition to accept, without evidence” every instance of (propositional) DT (op cit., p.36). The idea is flush with Horwich’s (1998) general account of meaning, under which the meaning of a word W is constituted by a community’s disposition to accept a set G of sentences featuring W, where this disposition is understood to explain all uses of W outside of G. In the present case, the set is the instances of DT, which, per deflationism, explain all other uses of ‘is true’. Soames (1999) has a similar view, based on the analytical and a priori status of the propositional version of DT (i.e., (1)c.); to possess the concept of truth is to a priori accept the instances of DT for those sentences one understands. Soames is not committed to the dispositional analysis of Horwich; indeed, refreshingly, Soames (op cit., p.231) avers that deflationism is perhaps too vague a label to admit precise analysis, and that the equivalence relation of DT is itself obscure. Still, like Horwich, Soames rejects the equivalence view and contends that a mere recognition of the primitive status of instances of DT constitutes  possession of the concept of truth, although one should hope for a more detailed  analysis.

       Whatever general virtues these two views may possess as regards elegance, simplicity, etc., they are untenable.

        Let us grant that instances of DT do have a peculiar status; it is unclear though, just what such a status is. Consider the sentential version of DT, an instance of which is


(4) ‘Kleptomaniacs are mendacious’ is true iff kleptomaniacs are mendacious


Now if, according to the equivalence view, the two flanks of the equivalence have the same meaning, then (4) must, in some sense, be a priori or necessarily true, it cannot be a mere contingent truth we could learn or forget. But that ‘mendacious’ has to do with mendacity is wholly contingent. Nor does (4) acquire an a priori status if we index the predicate to English, for an English speaker would not ipso facto know that the quoted sentence is English. Likewise, even if the predicate is indexed to ‘my language’, the sentence fails to be a priori true, for one may not have a clue what the embedded sentence means. The problem with (4) is simply that it cannot be a priori for a speaker unless she knows what ‘Kleptomaniacs are mendacious’ means, but (4) in no way reflects this fact. Otherwise put, one can perfectly understand, and indeed accept, the left-hand side without understanding the right-hand side; the two flanks are thus hardly synonymous. In general, sentential instances of DT will be a priori only for those sentences one understands. This feature is now widely recognised (Field (1986, 1994), Wright (1992), Gupta (1993), Soames (1999), Collins (forthcoming)). In response to this problem, Field (1994) proposes pure disquotationism:


(PD) (i) The content of the truth predicate, as understood by speaker S, is restricted to the

              sentences S understands.

         (ii) For each speaker S, P is true is cognitively equivalent to P.[12]


Heuristically, (4) may now be couched as (5):


(5) ‘Kleptomaniacs are mendacious’ is true iff it is true-as-I-understand-it


From PD(i), if S understands the left flank of (5), then S understands the quoted sentence; by PD(ii), the truth predication, as S understands it, means the same as the quoted sentence; and so the right flank of (5) amounts to ‘Kleptomaniacs are mendacious’.

       An alternative response to the problem is to reject sentential DT in favour of a propositional version:


(6) The proposition that kleptomaniacs are mendacious is true iff kleptomaniacs are mendacious.


The thought here is that one cannot understand the left flank unless one understands its subordinate clause.[13] Thus, (6) is inherently a priori, one does not need to relativise truth explicitly to what one understands.

        There is, however, a large fly in the ointment (Collins (2002)). Field’s proposal amounts to relativising truth to idiolects. This feature alone has been read as having the absurd consequence that one’s own concept of truth waxes and wanes with the width of one’s linguistic competence (e.g., Gupta (1993) and David (1994)). But the problem I have in mind also applies to the non-idiolectic propositional version: one can clearly predicate truth to sentences/propositions one cannot formulate. So, simply as a datum about our use of ‘is true’, we apply the concept outside of our idiolects. How, then, might our grasp of truth be exhausted by the contents we can otherwise entertain when the concept applies beyond such contents? This problem appears insuperable for Field’s proposal, and the propositional ‘solution’ goes no way to resolving it: it appears to be simply a datum about truth that its predication does not entail that one understands its subject, so any account that restricts truth so that such an entailment holds appears to have missed a central feature of the concept. I shall not dwell on this quandary for deflationism, although I do think it is inescapable.[14] Instead, I want to pursue a broader complaint that is suggested by this problem and question the very idea that an equivalence is what we should be after in explaining truth.

     Let us grant with Field that, if a speaker S does understand P, then S would hold an attitude A to P is true just if he held A to P. To concede this is to do no more than acknowledge that the left-right, right-left entailments of DT instances would not typically be denied by competent speakers. Clearly, we have not advanced beyond data, for in no way are we enlightened as to why we find this pattern of judgment. Specifically, an account of the meaning of true should enter into an explanation of why we find such a relation between P and P is true.

       I am not here insisting on any form of reductive account, my claim is only that meaning should stand in an explanatory relation to our linguistic judgments; the nature of meaning is left open by this demand. What is ruled out is the thought that an adequate account will not go beyond a specification of the kind of sentences we are find intuitively acceptable. This demand, I think is quite natural and applies generally. Consider:


(7)a. If Joe expects to shave himself, then Joe expects to be shaved.

b.  If Harry wonders who Joe expects to shave himself, then Joe doesn’t expect to be shaved.

c.  If Bill was persuaded by Mary to leave, then Bill left (not Mary).

d.  If Bill ate, then Bill ate something.


Every competent speaker of English knows a priori that the sentences of (7) are true, but that is the beginning of the investigation. We want to know why speakers make such judgments. We need theories of anaphora, empty categories (to act, inter alia, as subjects for infinitive verbs), and semantic selection features (e.g., that eat always selects for a patient, i.e., the thing eaten. Cf., sleep), as well as, of course, an account of the conceptual features of lexical entries: the factive character of persuade, etc. The case of ‘true’ is no different. The equivalence view merely presents data as explanation. Field (1994), however, appears to have a rejoinder to this complaint.

      The equivalence view does not say what either P or P is true mean, it simply tells us that they mean the same thing. Field (1994), though, suggests that this claim is compatible with various accounts of meaning, such as verificationist and inferential role theories.[15] In itself, however, this suggestion is question begging, for it assumes that any adequate account of meaning would establish that DT instances express a cognitive equivalence and that this would tell us all we want from a theory of truth. As readily conceded above, DT instances are not to be denied, and if we do understand the quoted sentence, then an acceptance of one flank should rationally lead to the acceptance of the other. But it is precisely this kind of pattern of judgment, inter alia, we want explained, and it is far from clear that there are independent reasons for thinking that any investigation will turn out as Field expects. For one thing, Field’s account is restricted to particular speaker’s idiolects - the sentences they understand. But, as we just saw, surely any theory of meaning will assign some interpretation to P is true, where a given speaker does not understand P. For any such speaker and theory, therefore, cognitive equivalence will not be upheld. It is far from clear, then, that any theory of meaning could immediately support Field’s PD. Even if we put this problem to one side (see n.14), while the instances of DT are not to be denied,  we cannot treat cognitive equivalence as an independent datum; that such equivalence is supported by a theory of meaning might reasonably be taken to militate against the theory. For example, correspondence theorists such as Davidson (1969), Devitt (1984), Musgrave (1989) and Alston (1996) accept the DT instances, but deny that they express a cognitive equivalence: the instances are intuitively acceptable just because we understand the language in question, but truth is not explicated by the acceptability of the instances in themselves but by language-world relations which DT instances reflect.  Davidson (1984, 1990, 1997), Wiggins (1980), and others likewise accept the instances, but suggest that they have an interpretive role, rather than being mere a priori trivialities. For example, according to Davidson’s interpretive conception of the job of a truth theory, heterophonic DT instances - e.g., ‘Der Schnee ist weiss’ is true iff snow is white - are equally characteristic of truth as their homophonic siblings, although such instances are certainly not disquotational and only support cognitive equivalence if one is an English-German bilingual.  Even deflationists, such as Horwich (1990, 1999) and Soames (1999), deny a cognitive equivalence relation (see below). In short, while the acceptability of DT instances might be a datum, their proper interpretation certainly is not. The options are wide open: an adequate theory of meaning might or might not confirm a cognitive equivalence, or the instances of DT might not have a settled status; perhaps they can be read in a number of ways, as, indeed, appears to be the case. This would not be surprising, for, presumably, it is one of the jobs of a theory of meaning to fix the notion of sameness of meaning (or its cognates) in such a way as to explain our judgments; it is not the job of the theory to tell us what we really mean, whatever that might mean. It bears emphasis that Soames, as alluded to above, essentially shares this view that it is obscure just what the relation is between P is true and P.

        Finally, it is of no help to appeal to semantic ascent as constituting, in some sense,  the meaning of true. The thought I have in mind is often expressed by the claim that true is a formal device for semantic ascent or compendious assertion, perhaps to be accounted for in terms of attachment and detachment rules, as if it were a logical constant. Some such rules might account for the meaning of true, but the lesson above recapitulates: it is far from obvious if such rules would constitute the sameness of meaning between P and P is true. Semantic ascent certainly does not establish any such equivalence. Indeed, the very ‘usefulness’ claimed for the truth predicate appears to involve P is true expressing a different thought from P.

      On Field’s (1994) proposal, the thought that Everything Gödel said is true reduces to the DT instances of what Gödel said, for ‘is true’ offers us a mere abbreviation by which we can assert an indefinitely large set of sentences picked out by the subject of the predicate, in this case: ‘everything Gödel said’. Thus:


(GT) If Gödel said ‘S1’, then S1; and if Gödel said ‘S2’, then S2;…


But if the generalisation is just an abbreviation of GT, and we can always ‘restore the effect of objective reference’, then we appear to understand everything Gödel did (and much more): each consequent of each conditional in GT. But does the thought that everything Gödel said is true allow us thoughts about the continuum hypothesis? Obviously not: the semantic competence drawn upon to understand the lexical items and syntactic form of the generalisation does not involve competence with any set theoretical hypotheses. This simple point might be missed by a use-mention confusion. If we accept that everything Gödel said is true, then we should certainly assent to each conjunct of GT, and, if we know that Gödel said ‘The continuum hypothesis is consistent with ZFC’, then we should assent to ‘The continuum hypothesis is consistent with ZFC’, but so much does not establish that we understand what the sentence means, and  so does not establish that it is cognitively equivalent to its truth predication. In this respect, GT and the argument form DISQ are highly misleading. We can entertain the generalisation and, with supplied premises, assent to each consequent of each conditional conjunct without having competence with each of the entailed conjuncts of GT. The very ‘usefulness’ of truth, therefore, appears to involve us in sentences or propositions we do not understand, ones which are not, for us, cognitively equivalent to their truth predications. Of course, Field believes that he can account for truth predications to sentences one does not understand, but whether he can or not, the general thought that the meaning of truth is constituted by its usefulness now depends on the adequacy of the account for what were perceived to be marginal cases. As it happens, Field account of these now central cases faces serious problems (see n.14).  Does the dispositional view fare better?

          The dispositional view does not claim that P is cognitively equivalent to P is true; rather, the thesis, as expressed by Horwich (1990, 1998, 1999), is that a speaker’s a priori disposition to assent to (propositional) instances of DT is what constitutes a speaker’s possession of the concept of truth. It does so because such a disposition is “explanatorily basic”, i.e., the disposition explains all other uses of ‘true’. Soames (1999) does not accept this appeal to dispositions, but does similarly claim that it is one’s primitive acceptance of (propositional) DT instances that constitutes one’s grasp of truth.

       There is little to recommend this view of truth. Firstly, the claim that speakers are disposed to accept a priori instances of DT is an unsupported empirical claim. It might be true, but it is far from obvious. Consider: would the mathematically ignorant be as ready to accept a DT instance whose embedded sentence tokens concerned topology as one concerned with the colour of snow? One might want to say that they should accept the instances equally, but that would be due to some antecedent thought about what the instances mean; that is, the putative dispositions would reflect an independent semantic competence rather than be constitutive of it.[16]

       Secondly, the claim that acceptance of instances of DT explains all other uses of ‘true’ appears to be simply false. How does DT relate to ‘true’ as used attributively of ‘friend’ or of ‘volume’ in measurement theory?; or when ‘true’ heads a phrase such as ‘true of Bill’? These are common uses of the adjective, but  they clearly cannot be accommodated by the DT schema. No doubt Horwich would appeal to the standard bracketing which isolates predicative uses of ‘true’. Some such bracketing at some stage of investigation will no doubt have to be made as the various semantic aspects of the word are delineated. Note, though, that Horwich is supposed to be telling a story about the word ‘true’, not an exclusive story about the VP ‘is true’ when it takes subjects with clausal complements. Are we to suppose that ‘true’ divides into a range of homonymous words ‘true1, ‘true2’, etc., each with different semantic properties? Surely not; but if not, then one’s accepting DT is not “explanatorily basic”.

       Horwich (1990, p.36) issues his claim as a “stipulation”, but I am at a loss to know what this means other than its being code for ‘Regardless of whether people are in fact disposed to accept a priori instances of DT, and regardless of whether such instances have any bearing on most uses of ‘true’, to so accept instances of DT constitutes what it is to possess the concept of truth; if one is not so disposed than one does not possess the concept’. Fine; one is free to define new meanings.

        If a dispositional story could be coherently told for a range of words, one might feel more sympathetic to Horwich’s general approach vis-à-vis ‘true’. Horwich’s (1998) most persuasive examples are of colour words, where one is supposedly disposed to accept ‘That’s red’, say, when confronted with red surfaces in normal light, etc., and connectives, where the disposition is  to accept every instance of, say, A and B, when one accepts A and B, and vice versa. There are problems with the former case because it is highly tendentious, even with colour words, whether a recognitional ability can constitute possession of a concept, for it is moot whether recognitional abilities compose. One might be able to recognise red things and recognise apples, but it does not necessarily follow that one could recognise red apples, yet it seems that possession of the constituent concepts (plus syntax) is necessary and sufficient to grasp the complex concept (see, e.g., Fodor (1998b, Chps.4, 5)). Horwich’s (1998, Chp.7) response to this problem is to suggest that what constitutes a complex meaning need not be the same kind of property that constitutes the meanings of its parts. Thus, while the meanings of ‘red’ and ‘apple’ might be constituted by recognitional abilities, the meaning of ‘red apple’ need not be; so, the account we give of individual words is not constrained by their composition into phrases and sentences. Even at this schematic level, however, a profound problem looms. This remedy to the initial problem essentially divorces a subject’s competence with words from her competence with the complexes in which the words occur. But if there are two different kinds of property at issue, then how do they relate such that grasp of the primitive properties is necessary and sufficient for grasp of the complex? That is the apparent datum we want explained, and multiplying properties, far from answering it, threatens to make it unanswerable (see Higginbotham (1999) and Collins (2001)).

     A further problem is intimated by the ‘etc.’ in my statement of Horwich’s account. The conditions under which one would be correct to say ‘That’s red’ are open ended, for no actual situation is so accommodating that mistakes are precluded, but ‘is red’ simply does not mean ‘looks red to me now’. The worry is a traditional one and is acute for Horwich. The very point of Horwich’s account is to explain meaning in terms of our use of words, but if the prevailing situation is never sufficiently accommodating, then the “explanatory basic uses” are never realised. Of course, no-one other than philosophers are to be found asserting  instances of DT, but we can easily specify the conditions under which a ‘normal’ speaker would be correct to do likewise (modulo the worries raised above), but we simply cannot do this with colour words, at least not in a way which would remotely give us the meaning of the words. The root of Horwich’s problem, here made vivid, is that meaning should explain why we use words the way we do, not the converse.  Horwich’s notion of disposition is ungrounded, i.e., it is not an expression of an underlying semantic competence. I shall return to this point shortly.

        Horwich’s other chief example is logical connectives. These terms do offer the best case for a dispositional account of meaning, but do so for the same reason that they offer the best case for a conceptual/inferential role theory: connectives serve the purpose of coordinating clauses - their function is their meaning. It is thus possible, if we put pragmatic issues to one side, to specify  the tautological schemata whose instances one should accept. But connectives are peculiar precisely in this regard; a theory which works best for such words is perhaps not one we should expect to work well in general. They are exceptions which prove the rule.

       Horwich offers no other examples, although in (1990) he suggests that a disposition to accept Peano arithmetic constitutes one’s possession of the concept of natural number. I am not sure if this is a serious proposal. The concept of a natural number was not invented by Peano (or Dedekind); also, does Horwich’s proposal mean that if one accepted Robinson’s Q, say, as opposed to Peano arithmetic, then one would lack the concept of natural number? The reason this suggestion is absurd is that we can decide between axiomatisations and definitions, not just about natural number, but throughout mathematics. We might construe the choice as deciding between alternative concepts or as choosing the right codification of the one concept, but we clearly do not think the decision determines whether we possess a concept tout court. Of course, words which have associated axioms or inferential rules are a tiny minority; for the vast majority of words nothing whatsoever springs to mind. For closed class words, such as the determiners, articles, prepositions, pronouns, reflexives, affixes, etc., the very idea of there being respective sets of sentences, which our acceptance of constitutes what we mean by the words, strikes me as simply incoherent.

    . The above problems are really symptoms of the fact that dispositions are not explanatory; specifically, a speaker/hearer’s dispositions are at best data as to what the subject understands. Imagine an alien speaker who is disposed to accept a priori DT instances. We would, on the basis of this disposition alone, perhaps conjecture that he shares our concept of truth. But imagine that, on further investigation, we found that he dissented from, say, the left-right entailment of (propositional) instances of DT; there would be numerous ways to proceed: maybe the alien has a different concept of the biconditional from us, or maybe he has a different concept of proposition, or maybe he has a different concept of truth and our initial conjecture was mistaken. Horwich’s advice to find the “explanatory basic” disposition makes no sense here, for it is wholly unclear whether dispositions stand in any explanatory relations to one another. What appears to be required is a notion of the alien’s underlying semantic competence,  for none of the options adumbrated are a priori preferable to any other, and without a structure of which the dispositions are expressive, it makes as much sense for the dispositions to stand alone as it does for them to be integrated in any of numerous ways. The point, of course, applies to ourselves, where we antecedently know what we mean.

        Consider the sentences of (7). The best explanations we have of why we find them intuitively obvious involve quite abstract principles, some of which the human child appears to possess innately. One is free to deny this and attempt another explanation, but a dispositional one is not available precisely because it is our typical  judgments we want explained. Consider the following grammatical chestnut.

     Let us say that anyone who is disposed to assent to Bill is eager to please is similarly disposed to Bill is eager to please someone, but not It is eager to please Bill; while anyone who is disposed to assent to Bill is easy to please is similarly disposed to It is easy to please Bill, but not the deviant Bill is easy to please someone. Now what assent dispositions for sets of sentences respectively featuring easy and eager will explain this difference? The question cannot be eluded by a bracketing of syntax from semantics; that would be illicit, for Horwich’s ‘explanatory basic’ dispositions are supposed to explain all meaning-relevant differences between easy and eager, i.e., every feature which enters into the determination of our normal pattern of usage. Well, the pattern is hardly self-explanatory; the dispositions as they stand appear to reflect an ‘underlying’ feature of the respective adjectives that enters into the interpretation of every construction in which they occur. Thus, no given number of respective constructions for the adjectives will be apt to be explanatory basic. Nor, patently, will a schematic approach, as offered for connectives, work: it is unclear whether jointly accepting the members of the respective paraphrase pairs is necessary for an understanding of the respective adjectives, but it is not sufficient, for if it were, then, absurdly, understanding easy, say, would reduce to knowing that its host sentences may be paraphrased one way as opposed to another.  This becomes clear if we take a closer look at the constructions.

      Roughly, easy (on the present reading) does not take external NP arguments, as the acceptability of the pleonastic subject it demonstrates; thus the whole complement clause - to please - is the single argument of easy. In simple terms, the sentence means that it is Bill’s being pleased that is easy to bring about; Bill himself is not easy, hence the nonsensical Bill is easy to please someone. On the other hand, eager does take external NP arguments - it is Bill who is eager - which, in the example at hand, may thus control the empty subject of the infinitive complement to please. Again, this explains why It is eager to please Bill is clearly deviant if the pronoun is read pleonastically (the sentence is perfectly acceptable with it referring to a pet dog, say), and so why it is not a paraphrase of Bill is eager to please. In short, the difference is to do with the argument structure of the adjectives, but such structure is general in the sense of being independent of what any particular construction is ‘about’. One’s disposition (conditional or schematic or otherwise), therefore, to assent to this or that set of sentences is of no moment: any set would do as well as another, for any set which is not gibberish will reflect the argument structure. This conclusion, of course, holds across the board for any lexical item that selects arguments.  

        Such simple reflection on fairly elementary syntactic features of words brings into relief the fundamental shallowness of the idea that meaning may be explained by use. 

        A final worry for Horwich is that while he rejects the cognitive equivalence view, he neglects to explain what kind of relation the ‘iff’ of DT is. If the instances of DT are to be a priori or analytic, then the relation clearly cannot be the material biconditional, for that carries no more information than sameness of truth value. The relation must be, in some sense, definitive or meaning based. This is what worries Soames (1999, p.231): there are no ideas about the relation which are free of deep problems. Horwich, however, does not consider the relation at all; it is enough if we simply accept the instances. But such indifference threatens the very rationality of semantic competence. One accepts a DT instance because, in part, of how one understands the relation. If one thought that the relation was not some form of equivalence, then, presumably, one’s conditional assent dispositions based on DT instances would be quite different. But the mere specification of the instances is insensitive to the relation, it is as if the instances were single words, or perhaps grunts. Horwich would presumably appeal to another disposition to accept some set of sentences featuring ‘iff’ that constitutes a speaker’s understanding of the equivalence relation, but the same problem recapitulates: What dispositions would differentiate between a material and a meaning-constituting relation? This question appears unanswerable, because the dispositions are supposed to be explanatorily prior to semantic categories and relations. A range of further compositional problems also arise, touched on above,  concerning how two sets of dispositions relate to one another to account for a complex expression. I have looked at these problems elsewhere (Collins (2001)). Suffice it to say that, as it stands, there is little reason to think that dispositions enter into an explanation of the meaning of ‘true’, or any other word, for that matter. This is hardly surprising: dispositions constitute nether the theoretical nor evidential base for any serious proposal in linguistics.


4: Concluding Remarks

I have not attempted to essay a positive story about truth. I should say that we need to await a genuinely explanatory semantic theory or, perhaps concomitantly, a working account of the conceptual systems that underlie our thought, before any worthwhile story about truth can be told. Deflationism is popular at the moment; it is thought by many to be the end of the story, or perhaps at least the beginning. This popularity is principally due to the problems with the substantive alternatives, but these problems just do not translate into reasons for deflationism. When we look at the positive claims concerning the content of truth we find insuperable problems of both a conceptual and empirical nature. Such, at least, has been my argument.[17]



[1] For the purposes of this paper, I assume that there is no issue about the extension of the truth predicate, which may be specified disquotationally. This is not an innocent assumption, not least because of the semantic paradoxes; furthermore, it suggests nothing as to how other forms of ‘true’ are to be accommodated, e.g., where the adjective modifies a singular noun (see below).

[2] Such flatness is the chief characteristic of what Field (1986, 1994) calls pure disquotationism. 

[3] Cf., Grover (1992, p.180): “a token of a sentence like… ‘”Snow is white” is true’, has as its propositional content - in a given context - the content of its antecedent [viz., ‘Snow is white’].

[4] One way of understanding this thought is that ET says that the content of ‘S’ also accounts for the truth of ‘S’. Thus, if ‘S’ means that p, then ‘S’ is true just if p.

[5] Deflationism, when expressed directly in terms of DT as opposed to ET, creates certain taxonomic problems. For instance, Tarski (1956) does furnish an answer to What is truth? in terms of satisfaction. This has led to a debate over whether Tarski’s definitional format is deflationary or not (see Etchemendy (1988), Davidson (1990), Horwich (1990), Field (1986, 1994), and Schantz (2000)). Once our focus is on ET, I think it is clear that Tarski should count as a deflationist, even though there is much more to his format than Convention T. Similar remarks hold also for Kripke’s (1975) fixed point model. 

[6] The ineliminability of ‘is true’ is not dependent on first-order readings of (2)a-d. The same result obviously holds for the treatment of quantifiers as cardinality functions (so-called, generalised quantifiers) which take an n-tuple of sets as argument and give a natural number as value as a Boolean function of the cardinality of the sets. On this understanding, the elision of ‘is true’ would result in a  set missing from the argument of the quantifier, viz., the set of truths in the domain over which the quantifier is defined.

[7] It is not obvious what is meant by the thought that we have a truth predicate because it’s useful. The ‘explanation’ is hardly genetically adequate. If it is intended as a synchronic functional explanation, then usefulness does not translate into a theory of the predicate’s content (cf., Chomsky (1975) on the communicative ‘purpose’ of language). 

[8] Fodor (1998a) attacks paradigm theories (as accounts of concept/property constitution), along with network theories, theory theories, et al. just as if they were species of TM (also see Leslie (2000)). Fodor might be right to reject all extra-denotational accounts of concepts, but, for my purposes, what is interesting about the proliferation of these alternative accounts of concepts is that they are motivated precisely by the failure to advance with TM.

[9] For an instructive elaboration of this thought concerning our concept of light, see Churchland (1996); for similar thoughts concerning biological concepts, see Hull (2000).

[10] An anonymous referee drew my attention to Misak’s interesting paper, which criticises deflationism in much the same spirit as my previous points. It is worth noting, though, that the current complaint is not premised on any particular alternative to deflationism, such as a form of pragmatism. The point is a general one: we are not in an either-or situation, so a rejection of one limb - TM - does not commend the other.

[11] For deflationary responses to Boghossian, see, e.g., Kraut (1993) and Soames (1999, chp.8).

[12] Field lets his notion of cognitive equivalence remain somewhat impressionistic; he does not, however, associate it with synonymy for familiar Quinean reasons. So to have something definite in mind, the relation may be expressed as (CE):

(CE) ("S)("C)[P is cognitively equivalent to P* for speaker S ↔

                         (i) S holds P under cognitive attitude C iff S holds P* under C;

                        (ii) there is no Q such that if S were to hold Q under C, then S would hold

                              P under C but not P*, or P* under C but not P].

The second clause says that the equivalence in the first clause is non-defeasible: there are no cognitive grounds which would lead S to adopt differing attitudes towards P and P*.  

[13] Recanati (2000) dubs this feature of ‘that-clause’ reports (inter alia) iconicity.

[14] See Collins (forthcoming). Field (2001) suggests that his (1994) response to the problem is inadequate, but that it can be easily rectified by the thought that one’s understanding of a truth predication to a sentence one does not understand is equally indeterminate as one’s understanding of the sentence itself. This thought, however, appears to be simply a stipulation. Patently, there is some difference between English monolingual A and English-German bilingual B in their understanding of ‘”Der Schnee ist weiss” is true’, but it just does not follow that A’s understanding of  the truth predication is of the same status as his understanding of ‘Der Schnee ist weiss’, and Field gives no reason to think that it is.

[15] In principle, any non-truth conditional account of meaning is compatible with deflationism. But such an account, of course, constitutes perhaps the dominant tradition. Horwich’s (1998) semantic deflationism and Field’s (1994) methodological deflationism correctly see that a deflationism about truth is half the job - truth must also be removed as the core component of a theory of meaning.

[16] Horwich, in line with his (1998) commitments, might make appeal to the dispositions of the community of English speakers. This would recapitulate the problem. Even if we were able to talk sensibly about such a community, most of its members would surely be leery of assenting to any claim about topology.

[17] My thanks go to an anonymous referee for constructive criticism and a number of suggestions from which this paper has greatly benefited.



Alston, W. (1996). A Realist Conception of Truth. Ithaca: Cornell University Press.

Ayer, A. (1936). Language, Truth, and Logic. London: Penguin.

Boghossian, P. (1990). ‘The Status of Content.’ Philosophical Review, 99: 157-184.

Brandon, R. (1994). Making it Explicit: Reasoning, Representing, and Discursive

     Commitment. Cambridge, MA: Harvard University Press.

Chomsky, N. (1975). Reflections on Language. London: Fontana.

Chomsky, N. (2000). New Horizons in the Study of Language and Mind. Cambridge:

      Cambridge University Press.

Churchland, P. (1996). ‘The Rediscovery of Light.’ Journal of Philosophy, 93: 211-228.

Collins, J. (2001). ‘Truth Conditions without Interpretation.’ Sorites, 13: 52-72

Collins, J. (2002). ‘Truth or Meaning? A Question of Priority.’ Philosophy and

      Phenomenological Research.

David, M. (1994). Correspondence and Disquotation: An Essay on the Nature of Truth.

      Oxford: Oxford University Press.

Davidson, D. (1984). Inquires into Truth and Interpretation. Oxford: Clarendon Press.

Davidson, D. (1990). 'The Structure and Content of Truth.' Journal of Philosophy, 87:


Davidson, D. (1996). ‘The Folly of Trying to Define Truth.’ Journal of Philosophy, 94:


Devitt, M. (1984). Realism and Truth. Oxford: Basil Blackwell.

Etchemendy, J. (1988). ‘Tarski on Truth and Logical Consequence.’ Journal of Symbolic

      Logic, 53: 51-79.

Field, H. (1986). 'The Deflationary Conception of Truth.' In G. MacDonald and C. Wright

    (eds.), Fact, Science, and Morality (pp. 55-117). Oxford: Oxford University Press.

Field, H. (1994). 'Deflationist Views of Meaning and Content.' Mind, 103: 249-258.

Field, H. (2001). ‘Postscript to Deflationist Views of Meaning and Content.’ In Truth and

     the Absence of Fact (pp. 141-156). Oxford: Oxford University Press.

Fodor, J. (1975). The Language of Thought. Cambridge, MA: Harvard University Press.

Fodor, J. (1981). ‘The Present Status of the Innateness Controversy.’ In Representations:

       Philosophical Essays on the Foundations of Cognitive Science (pp. 257-316).

       Cambridge, MA: MIT Press.

Fodor, J. (1998a). Concepts: Where Cognitive Science Went Wrong. Oxford: Oxford

      University Press.

Fodor, J. (1998b). In Critical Condition: Polemical Essays on Cognitive Science and the

     Philosophy of Mind. Cambridge, MA: MIT Press.

Frege, G. (1952). ‘On Sense and Meaning.’ In P. Geach and M. Black (eds.), Translations

     from the Philosophical Writings of Gottlob Frege (pp.56-78). Oxford: Basil


Frege, G. (1976). ‘The Thought: A Logical Inquiry.’ In P. Strawson (ed.), Philosophical

     Logic (17-38). Oxford: Oxford University Press.

Gettier, (1963). ‘Is Justified True Belief Knowledge?’ Analysis, 23:121-123.

Grover, D. (1992). A Prosentential Theory of Truth. Princeton: Princeton University


Grover, D. et al. (1975). 'A Prosentential Theory of Truth.' Philosophical Studies, 27:

     73-124. Reprinted in Grover (1992): pp. 70-120.

Gupta, A. (1993). ‘A Critique of Deflationism.’ Philosophical Topics, 21: 57-81.

Higginbotham, J. (1999). ‘A Perspective on Truth and Meaning.’ In L. Hahn (ed.), The

     Philosophy of Donald Davidson (pp. 671-686). Peru, Ill.: Open Court.

Horwich, P. (1990). Truth. Oxford: Blackwell.

Horwich, P. (1998). Meaning. Oxford: Blackwell.

Horwich, P. (1999). 'The Minimalist Conception of Truth.' In S. Blackburn and K.

      Simmons (eds.), Truth (pp. 239-63). Oxford: Oxford University Press.

Hull, D. (2000). Science and Selection: Essays on Biological Evolution and the

       Philosophy of Science. Cambridge: Cambridge University Press.

Kripke, S. (1975). ‘Outline of a Theory of Truth.’ Journal of Philosophy, 72: 690-716.

Kripke, S. (1980). Naming and Necessity. Oxford: Basil Blackwell.

Kraut, R. (1993). ‘Robust Deflationism.’ Philosophical Review,103: 247-263.

Leeds, S. (1978). ‘ Theories of Reference and Truth.’ Erkenntnis, 13: 111-127.

Leslie, A. (2000). ‘How to Acquire a Representational Theory of Mind.’ In D. Sperber

      (ed.), Metarepresentations: A Multidisciplinary Perspective. Oxford: Oxford

      University Press.

Misak, C. (1998). ‘Deflating Truth: Pragmatism vs. Minimalism.’ The Monist, 81: 407-


Musgrave, A. (1989). ‘NOA’s Ark - Fine for Realism.’ Philosophical Quarterly, 39: 383-


Quine, W. V. O. (1953). From a Logical Point of View. Cambridge, MA: Harvard

      University Press.

Quine, W. V. O. (1960). Word and Object. Cambridge, MA: MIT Press.

Quine, W. V. O. (1970/86). Philosophy of Logic (Second Edition). Cambridge, MA:

     Harvard University Press.

Quine, W. V. O. (1987). Quiddities: An Intermittently Philosophical Dictionary.

     Cambridge, MA: Harvard University Press.

Recanati, F. (2000). Oratio Obliqua, Oratio Recta: An Essay on Metarepresentation.

      Cambridge, MA: MIT Press.

Putnam, H. (1975). ‘The Meaning of ‘Meaning’.’ In Mind, Language and Reality:

      Philosophical Papers, Volume 2 (pp. 215-271). Cambridge: Cambridge University


Putnam, H. (1990). ‘Is Water Necessarily H20?’ In Realism with a Human Face (pp. 54-

     79). Cambridge, MA: Harvard University Press.

Ramsey, F. (1927). 'Facts and Propositions.' Proceedings of the Aristotelian Society,

     supp. vol. 7: 152-170.

Schantz, R. (2000). ‘Truth, Correspondence, and Reference: How Deflationists Diverge

    from Tarski.’ In A. Chapuis and A. Gupta (eds.), Circularity, Definition, and Truth

    (pp.317-332). New Delhi: Indian Council of Philosophical Research.

Soames, S. (1999). Understanding Truth. Oxford: Oxford University Press.

Strawson, P. F. (1950). 'Truth.' Proceedings of the Aristotelian Society, supp. vol. 24:


Tarski, A. (1956). 'The Concept of Truth in Formalized Languages.' In A. Tarski (1956),

    Logic, Semantics, Metamathematics (pp. 152-277). Oxford: Oxford University Press.

Tarski, A. (1944). 'The Semantic Conception of Truth.' In L. Linsky (ed.) (1952),

     Semantics and the Philosophy of Language (pp. 13-47). Chicago: University of

     Illinois Press.

Wiggins, D. (1980). ‘What Would be a Substantial Theory of Truth?’ In Z. van Straaten

     (ed.), Philosophical Subjects: Essays Presented to P. F. Strawson (pp. 189-221).

     Oxford: Oxford University Press.

Williams, C. J. F. (1976). What is Truth? Cambridge: Cambridge University Press.

Wright, C. (1992). Truth and Objectivity. Cambridge, MA: Harvard University Press.

Wright, C. (1999). ‘Truth: A Traditional Debate Reviewed.’ In S. Blackburn and K.

     Simmons (eds.), Truth (pp.203-238). Oxford: Oxford University Press.