SORITES ISSN 1135-1349

Issue #03. November 1995. Pp. 48-63.

A Note on Truth, Deflationism and Irrealism

Copyright © by SORITES and Pierluigi Miraglia

A Note on Truth, Deflationism and Irrealism

Pierluigi Miraglia<16>Foot note 2_1

Section 1.--.

It seems reasonable to require of a naturalistic account of a given region of scientific or ordinary discourse that it construe the reference of expressions central to that region in terms of naturalistically acceptable entities (ultimately, physical objects or states). In many areas of discourse, however, this project stumbles upon notorious difficulties, not all attributable to what appear to be contingent gaps in the current state of scientific knowledge. It is appealing in such cases to regard the utterances and predicates of the given area of discourse as playing a different, non-descriptive role -- that of evincing the speaker's stance, for example, or expressing an attitude. The thought is that this move offsets the need to specify physicalist denotata of the predicates in question, for the «point» of uttering judgments in the given area of discourse would not be, in effect, to denote anything. Such is the thrust of an irrealist approach to an area of discourse. Of course, irrealists immediately face a further question: if the predicates under examination are non-descriptive and non-denoting, can they still play a legitimate role in our conceptual ecosystem, or is it simply a mistake to go on using them? A choice must then be made: should we preserve the discourse in question or should we rather «quine» it, possibly consigning it to extinction? The conservationist option requires, at a minimum, that we explain how discourse involving predicates that fail to denote may still support some standards of cognitive legitimacy, justificatory procedures and so on. This strategy characterizes the conservationist variety of irrealism known as non-factualism. The alternative strategy -- to acknowledge the mistake intrinsic to non-denoting discourse -- is distinctive of an error-theoretical approach.

An important issue in irrealism is: how far can one go? Paul Boghossian's «The Status of Content» (1990a; SOC henceforth) lays out a powerful argument against irrealist conceptions of content: he believes, in a nutshell, that no irrealist about a certain area of discourse (different from content) can be irrealist about content. Let irrealism about an area of discourse F be the doctrine according to which there are no real properties corresponding to (or denoted by) the predicates of F. Irrealism about content consequently holds that such characteristic semantic predicates as «has truth-conditions p» or «means that p» (which apply to sentences or utterances) do not denote real properties. SOC aims at showing that irrealism about content is an «unstable», intrinsically incoherent doctrine. I shall counter this claim by showing that Boghossian's arguments are based on a distorted view of the commitments attendant upon irrealist views. But our scope will of necessity be broader. The interest of the SOC argument resides in its forcing us to confront us some deep and far-reaching issues in metaphysics and semantics. It involves a detailed discussion of a deflationary conception of truth, which is alleged to have dire consequences for irrealist views. Now, that issues surrounding truth and semantics are central to irrealist projects in general hardly needs emphasizing:<17>Foot note 2_2 the point urged by Boghossian is therefore a crucial one and, if sustained, would have significant consequences for the very possibility of irrealism. I hold out hope that, by clarifying the relations between irrealist (and specifically non-factualist) projects and philosophical views about truth, the present discussion may also serve the broader purpose of deepening our understanding of irrealism.

Now for a preview of things to come: Section 2 presents a reconstruction of the central argument of SOC, stated around p. 175; Section 3 investigates aspects of the deflationary conception of truth which seem to me essential for the purpose at hand; IV lays out the main objection to the SOC argument. Much of what Boghossian says about content, irrealism and related matters will be granted without argument. I take no issue with his contention that irrealist views about content apply as much to mental content as to linguistic meaning, so I shall assume in what follows that we may restrict our attention to the linguistic case. Similarly, I assume that «the essential core of the ordinary notion of content does consist simply in the idea of a truth condition» (1990a, p. 173). Thus, such content-ascribing locutions as «S has truth condition p» (where S is the name of a sentence), «S is true» and the like will be the focus. The viability of irrealism about more comprehensive notions of content (if any exist) will not be discussed.

Furthermore, I briefly hinted at the fact that irrealist doctrines about any subject matter F come in two varieties: error theories -- according to which the predicates of F purport to be genuinely referential but are in fact systematically empty -- and non-factualist theories. But my discussion will concern only the latter; nothing substantial is said in this paper about the arguments in SOC that deal with the alleged difficulties of error theories. My concern is with the objections against non-factualist theories of content. In the remainder of this paper, «non-factualism» and «irrealism» are used interchangeably.

Section 2.--

According to Boghossian, non-factualism about a predicate P belonging to a given area of discourse F is characterized by adherence to the following two claims:

(i) the predicate P does not denote a property;

(ii) a declarative sentence containing P (such as the atomic sentence «x is P») is not truth-conditional, i.e. does not have a truth-condition.

It is a belief in (ii) that sets non-factualism apart from error theories about P, for those are committed to the view that statements containing P are to be evaluated as if they had a «genuine» truth value, although their truth-conditions are never satisfied (i.e. they are always false). Non-factualism about content, then, will be characterized by the following two theses about the content-ascribing predicate «has truth conditions p» (given the modest construal of content, having truth conditions p amounts to having content p):

(1) The predicate «has truth condition p» does not denote a property;

(2) «S has truth condition p» (where S is an appropriate name for a given sentence) is not truth-conditional.

The central argument against non-factualism about content unfolds in two stages, separately developing the conflicting implications of (1) and (2). More precisely, the trouble with non-factualism is that (1) presupposes a deflationary view of truth conditions, while only a non-deflationary, correspondence conception can justify acceptance of (2). We shall consider deflationism more closely in the next section. Let us now examine Boghossian's argument in some detail.<18>Foot note 2_3

(A) If «true» does not refer to a property, then any declarative sentence is (trivially) truth-conditional.

That «true» does not refer to a property is, for Boghossian, the central tenet of deflationism. The important point is that

there is no more to a sentence's being truth-conditional -- genuinely apt for (deflationary) truth or falsity -- than its being a significant sentence possessing the appropriate syntactic potentialities (SOC, p. 164).

What Boghossian seems to have in mind is a disquotational or «homophonic» construal of (a sentence's having) truth conditions. On a disquotational reading the truth condition for any given declarative sentence is that expressed by the sentence itself. In effect, nothing more is needed in order for S to be able to express a truth condition than just that S be assertible at all. S must be merely a candidate for assertion in order to be truth-conditional. From a deflationist standpoint the qualifying requisite for such candidacy must be, according to Boghossian, extremely weak: just being «meaningful» (a syntactic property) and «declarative» should suffice. Hence:

Any meaningful declarative sentence would be (at a minimum) a candidate for assertion... Any such sentence would count, therefore, as truth-conditional in a deflationary sense (SOC, p. 165).

(B) By contraposition of (A), (2) implies that «true» does refer to a property. Were the non-factualist to adopt a deflationist reading of locutions such as «expresses a truth condition», she would have no room to deny that any meaningful declarative sentence is truth-conditional -- but that is precisely what non-factualists deny.

The idea is to show that, while (2) rules out deflationism for the reasons just given, (1) entails it. The case is condensed in the following passage:

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). (1990a, p. 175)

This can be reconstructed as follows:

(C) The predicate «true» denotes a property if and only if «a sentence's possessing a truth value» is a «thoroughly factual matter».

Strictly speaking, the argument requires only the left-to-right direction. Prima facie, though, the right-to-left direction fails on a disquotational notion of truth: I shall argue in the next section that a disquotationalist should maintain that a sentence's truth value can be a perfectly factual matter without conceding that «true» denotes a property. On the other hand, if the deflationist envisioned by Boghossian adheres to a view of truth as some kind of pragmatic «virtue» of statements, (C) might well be in order: for this kind of deflationist, the truth of an assertion is indeed a thoroughly non-factual matter. One senses here the need to take a closer look at deflationism. At any rate, since the equivalence is not necessary to the SOC argument, these are for the time being definitional quibbles.

(D) (1), the claim that the predicate «has truth condition p» does not denote a property, entails that a sentence's possessing a truth value is not a thoroughly factual matter.

(E) Therefore (by the left-to-right direction of (C)), (1) entails that «true» does not refer to a property, i.e. a deflationary conception of truth. By (B), we obtain a contradiction. Hence, irrealism (i.e. non factualism) about content is inconsistent.

Section 3.--

If standard usage is to offer any guidance in semantics, it appears that the predicate «true» (or «is true») as applied to sentences in the indicative mood satisfies equivalences of this form:

(T) «Snow is white» is true if and only if snow is white.

It is routinely presumed, after Tarski, that a theory of truth for a given language L should be able to derive all equivalences of this form, one for each declarative sentence of L. In other words, it is an adequacy constraint on the definition of a truth predicate for L that the schema resulting by writing a sentential variable in place of the sentence «snow is white» in (T) above be validated by such a predicate. Tarski's Convention T is of course such a schema. This being the case, it might be tempting to conceive of Convention T as telling us in effect all we need to know (and all we can expect to know) about the semantic role of the predicate «true»: this conception I call deflationism about truth.<19>Foot note 2_4

I take it that someone embracing a deflationary conception of truth must at least be committed to the following thesis:

(DEFL) for any (declarative) sentence S, the assertoric content of «S is true» is the same as the content of S; i.e., to say that S is true is to say no more and no less than what is expressed by asserting S.<20>Foot note 2_5

This thesis puts the deflationist at variance with what is usually labeled as the «correspondence theory» of truth, according to which to say of a sentence S that it is true amounts to saying that there is a special relation of «correspondence» between S and some parcel of ontology -- a state of affairs, a fact, a combination of objects, etc. This conception then opens a gap between what I have called the assertoric content of «S is true» and the content of S itself, for on the face of it the latter contains no reference to «correspondence». One might retort perhaps that to say of S that it corresponds in the appropriate way with the facts is just to say that S (putting aside the fact that such a claim runs counter to the intuitive judgement of most speakers), i.e. that correspondence is itself a «disappearing» or redundant property. This might have been Ramsey's view when he said:

We can, if we like, say that [the proposition aRb] is true if there exists a corresponding fact that a has R to b, but this is essentially not an analysis but a periphrasis, for «The fact that a has R to b exists» is no different from «a has R to b». (1927, p. 39)

But it seems fair to say that a correspondence theory couched up along these lines is no longer a conception of truth as correspondence: it is just the old sheep in wolf's clothing, that is, the deflationary conception repackaged.

My critique of the SOC argument depends on a number of more or less elementary facts about deflationism, which I now proceed to explain. A vindication of the deflationary conception of truth, worthwhile as it is, is clearly beyond the scope of this paper. Yet certain basic aspects of the deflationary conception are made very short shrift of in Boghossian's SOC, as well as in several of the critical papers it generated. By taking such aspects into account, we shall have made some progress toward a clearer understanding of the issue.

Firstly: The most straightforward embodiment of the deflationary conception is the so-called disquotationalist theory of truth; I take disquotational truth to be the paradigmatic example of a deflationary notion of truth, and I shall use disquotationalism to show that irrealism about content is ultimately immune to Boghossian's objections. This is in contrast with most other critics of Boghossian's views, who have generally conceded his point about disquotationalism even while disputing his assessment of deflationism. According to the disquotationalist theory, «S is true» is equivalent to a (typically infinite) conjunction of the form

if S is «snow is white», then snow is white; if S is «the cat is on the mat», then the cat is on the mat; ...

(where, as usual, S stands for any declarative sentence). Thus, disquotationalism about truth involves in effect little more than strict adherence to DEFL.<21>Foot note 2_6 By contrast, another view of the notion of truth, also classified by many as deflationary, offers a quite different outlook. According to a pragmatist conception of truth, truth is a special compliment paid to a select class of sentences -- perhaps those that we are prepared to «defend against all comers», or those that we deem «explanatorily indispensable», and so on.<22>Foot note 2_7 Such a conception is clearly incompatible with the correspondence theory. But is it a deflationary conception? If I say of «snow is white» that it deserves to be treated by me with the respect I reserve for beliefs that are indispensable to my world picture (which is roughly what I would imply, according to the pragmatist, by asserting that «snow is white» is true), I do not seem to be saying just that snow is white. On the other hand, a careful pragmatist could reply that the compliment paid to the sentences in the select class does not add to the content of these sentences, so the contents of the two relevant assertion -- that S and that S is true -- remains the same. At any rate, I mentioned a pragmatist conception of the truth predicate to evince a certain contrast that might emerge in relation to my next point.

The point is, secondly, that a deflationist about truth may be an irrealist with respect to (the predicates of) a given area of discourse, but she need not be one. Deflationism is perfectly compatible with realism. The disquotationalist holds that to assert that «snow is white» is true is the same as to assert that snow is white. But must she be an irrealist about the color of snow? Trivially, no. I conjecture that the reason why this fact may on occasion appear less than completely trivial is that it tends to be obscured when the deflationist view one has in mind is a pragmatist one. Suppose someone asserts that it is true that there are electrons in my kitchen. If I say, along with a certain kind of pragmatist, that the speaker is thereby expressing his conviction that the sentence asserted is explanatorily indispensable, I may easily slip into making the existence of electrons in my kitchen a matter of explanatory expediency -- and this would seem to be close to an irrealist conception of electrons. Thus, if this kind of pragmatist conception of truth has intuitive appeal, it often is on the score of a more general anti-realist project which recommends it. Whatever the matter with pragmatism, however, deflationism in general is independent from irrealism (or realism). In the same case, a disquotationalist would have no need to finesse her stance with respect to the existence of electrons in my kitchen; she simply asserts the identity of such a fact with the fact that «there are electrons in my kitchen» is true -- assuming, of course, that there is such a fact. In other words, the disquotationalist need not import any sui generis conception of the truthmakers of such a sentence. Thus, her account of truth is, to a large extent, metaphysically flexible. How flexible, of course, is precisely the question addressed by Boghossian's paper: an essential part of the argument there is that deflationism is incompatible with mixed accounts (irrealism about some areas of discourse but not others), because drawing the boundary line between factual and non-factual areas requires a non-deflationary conception of truth.

Thirdly, and lastly: The deflationary conception faces no serious or special difficulty in explaining the concept of truth-conditions of a sentence. To put things in terms of Tarski's theory of truth: the truth-conditions of a sentence in the first sense are simply given by the respective T-sentence, a biconditional having the same form as (T) above. The truth-conditions of the sentence named on the lefthand side of the biconditional are just the righthand side of the biconditional. So, the truth-conditions of «snow is white», as well as «la neve è bianca», are that snow is white. That, of course, is precisely what the deflationist (disquotationalist) theory will predict. And yet there is a sense that some important element is missing from deflationary truth-conditions, something that «real» truth-conditions should have. I shall proceed here with a bit more caution.

A brief reflection on Tarski's contribution may help here. Tarski notoriously claims, in the opening paragraphs of his (1935), to be concerned exclusively with the «classical» notion of truth, i.e. (in his words) truth as correspondence. But the condition for the «material adequacy» of a definition of truth he stipulates (i.e., Convention T) merely prescribes that the theory entail all T-sentences;<23>Foot note 2_8 truth as defined by the disquotationalist would satisfy Tarski's criterion. This should be no surprise, since what the disquotationalist does is, in effect, to define truth with the help of a homophonic translation of the object language into the metalanguage, and that is pretty much what Tarski did.<24>Foot note 2_9 Still, one might object at this point that the analysis of truth conditions in the Tarskian approach is not exhausted by mere consideration of the T-sentences; it also matters how the T-sentences are derived from the theory, and the base axioms used in such derivation will be (in compliance with compositional requirements) statements about the denotation of terms and predicates. Can the deflationary conception accommodate such denotation axioms? Isn»t denotation «intrinsically» a correspondence notion? Well, the theory of denotation that can be extracted from Tarski's work is something like the following (in the case of, e.g., proper names; other cases are defined accordingly):

To say that the name N denotes a given object a is the same as to stipulate that either a is France and N is «France», or ... or a is Germany and N is «Germany» (Field (1972), p. 365).

In the same passage Field continues:

This is Tarski's account of denotations for English proper names...[Such theories of denotation as the above] satisfy criteria of adequacy exactly analogous to the criteria of adequacy that Tarski accepted for theories of truth.

Now, a theory of denotation couched in these terms is a theory that will be perfectly congenial to a deflationist: we could call such a theory a disquotational or homophonic theory of denotation. We may raise all sorts of grievances against such a theory. For example, we may wonder, as Field does, «what a real explication of denotation in nonsemantic terms would be like» (ibid.), in which case we should look at a different theory to supply that. Nevertheless, the point is clear: to the extent that talk of truth conditions is cast in terms of a Tarskian truth theory (i.e. as talk about the righthand side of the T-sentences), truth-conditions pose no special threat to the deflationist. But now a problem arises.

Talk of truth-conditions is often perceived as ambiguous: on the one hand, we seem to have in mind the righthand side of Tarskian biconditionals -- that snow is white; on the other hand, many seem to expect and demand something more «robust», or at least as robust as real snow. Yet attempts to clarify the perceived ambiguity and to strengthen the expression of truth-conditions in one way or another tend to be vacuous.<25>Foot note 2_10 What could the robust truth-conditions of «snow is white» be like? Perhaps the fact that snow is white? Certainly we could (and should) say that the truth-conditions of «snow is white» are the fact that snow is white, but this would in no way mark a difference between us and a deflationist: the described fact is precisely that expressed by the righthand side of the relevant biconditional. Could we then protest, with M. Devitt, that on the deflationary picture of truth-conditions truth or falsity «do not apply to sentences partly in virtue of contingent properties of sentences determined by facts about language» (1990, 254)? This will not take us very far: it may well be in virtue of a causal connection (a primeval baptism, or whatnot) between the word «snow» and snow that the truth-conditions of «snow is white» are what they are, but the baptism itself is not the truth-conditions of the sentence, although it explains why the truth-conditions are in fact that snow is white. But, more generally, it is not as though talk of baptism, causal connections or other «contingent properties» will scare off the deflationist: it is not part of the deflationary conception that sentences acquire their truth-conditions by magic or by fiat. Well, if all this is right it will be difficult to embarrass the deflationist by producing something (the «robust» truth-conditions) which she cannot express or refer to. But we still have the opposite problem to deal with: she may trivialize the notion by being too prodigal, by expressing too many truth-conditions. In fact, this is the charge levelled by Boghossian: that a deflationist will not be in a position to deny that any sentence is truth-conditional.

Section 4.--

Thus, the serious problem turns out to be this: a deflationary conception of truth precludes an understanding of what it is for a sentence to be truth-conditional, because a deflationist cannot make sense of a sentence's failing to be such. For a deflationist must classify any sentence as truth-conditional (see steps (A) and (B) in our reconstruction of SOC). But we will se that, whatever the merit of this claim, the problem for the deflationist is much overblown. My response to the allegation will be that, to the extent that the deflationist has trouble with the notion of truth-conditionality in question, so does everybody else; and to the extent that a correspondence theorist can make sense of the claim that some sentences are not truth-conditional, so can the deflationist. A consequence of this thesis will be that claims about what sentences are not truth-conditional have little, if anything, to do with one's conception of truth. This is a mildly surprising conclusion, because it has been customary in irrealist quarters to label non-factual discourse as «non-truth-conditional». Yet I submit that what matters is not the label, but the features of sentences and utterances that the label is meant to designate -- and I hold that these features have in fact little to do with a particular conception of truth. So, non-factualists can be accused of loose talk, but not -- if I am right -- of incoherence.

What do people mean when they declare a sentence or utterance to be truth-conditional? Or, perhaps better: what can they mean? The safe, if trivial, answer is of course: that the sentence in question has truth conditions. By which in turn it is meant, without straying too far from triviality, that the sentence expresses a proposition that at least sometimes is true, and thus occupies the right side of the biconditional in the relevant T-sentence.

So far, so good for the deflationist: we have not exceeded the resources available under the deflationary conception. As noted in Section 3, this notion of truth conditions is perfectly compatible with deflationism. But the notion of truth conditions so easily wrung out of Tarskian biconditionals is also quite weak, or so it seems. It applies to any sentence for which the theory of truth is to yield a corresponding T-sentence. Here lies the problem, or so the criticism goes. For we may want to deny that a sentence is truth-conditional even though that sentence «has truth conditions» in the weak sense palatable to the deflationist. It must then be the case that such denials, if we are to make sense of them, involve a different notion of truth conditions and being truth-conditional. Thus Boghossian:

declarative sentences cannot fail to possess truth conditions except against the background of a robust [i.e. correspondence] notion of truth (1990, p. 166).

I agree that there is a non-trivial notion of being truth-conditional at play here, but I deny that a correspondence theorist is any better equipped to capture it than a deflationist. The non-trivial notion that we may want to capture is that of a sentence or a class of sentences being fact-stating, or part of fact-stating discourse. We may suspect (correctly, I believe) that the utterances of certain areas of discourse perform a different function than that of asserting that a given state of affairs obtains; such utterances would not be fact-stating. As we saw previously, to hold such a view of the sentences belonging to a given region of discourse is precisely to be an irrealist (and a non-factualist) about that region. Now, it seems to me essential that we understand what is going on when someone takes a non-factualist stance with respect to a certain area of discourse, that is, what is being asserted by declaring that some sentences are not truth-conditional.

When non-factualists and irrealists in general claim that sentences belonging to a given area of discourse are not truth-conditional, they seem to mean, first, that no fact (state of affairs, etc.) answers to the sentences of that area. This strikes me as a characteristically metaphysical thesis, i.e. one that has to do with ontology, with objects and combination of objects and their presence or absence in the world we inhabit. There is, however, a further step that one needs to have taken before being in a position to declare a sentence non-truth-conditional. For the simple absence of a fact to make a sentence true might just be taken to indicate that that sentence is false. Now, false sentences are seemingly truth-conditional. So the non-factualist claim about the non-truth-conditionality of the area of discourse in question must contain a further element: the non-factualist must mean that the communicative function of sentences in that area of discourse is radically different from that of sentences that are to be evaluated for truth or falsity. This is more of a semantic claim,<26>Foot note 2_11 having to do with the proper way to interpret certain sentences or utterances: it is a claim, for example, about people do when they utter judgments of a certain type. Notice that what is proprietary to non-factualists is the second (semantic) claim, for the first (metaphysical) claim they share with the error theorists. The error theorists (about an area of discourse F) also hold that there are no things answering to the predicates of F, but from this realization they conclude, roughly, that everything said in F is false. Non-factualists want to escape precisely that conclusion. So they need to make both points, if non-factualism is to be at all viable.

A question then naturally arises as to the precise connection between the ontological claim and the semantic claim. The question could be stated this way: how does the thesis about what facts or objects there are translate into a view about the proper applicability of truth and falsity to an area of discourse? Or if you prefer: does the ontological non-factualist thesis entail the semantic non-factualist thesis? If it did, then the realization that the world does not contain facts about F, say, would be sufficient to conclude that F-discourse is non-factual, and therefore not truth-conditional. Consequently: If the correspondence theorist had a better grip on the ontology (i.e. the world) than the deflationist, as those dissatisfied by deflationism seem to suggest, then he (the correspondence theorist) could at least make clear sense of a F-sentence's not being truth-conditional: the things or properties asserted by F to exist are not there. But if the ontological thesis does not imply the semantic thesis, then even if the correspondence theorist has a better grip on the ontology, he has no significant advantage over a non-correspondence theorist -- simply because the ontology in his grip is by itself inconclusive as to the semantic status (truth-conditionality) of F-sentences. I argue that, at least prima facie, there is no direct implication between the ontological claim and the semantic claim.

One may well protest the vagueness of such expressions as «having a better grip on ontology», by which I have characterized the correspondence theorist. I am sympathetic: I am not sure what it means. But the burden of clarification falls on the correspondence theorist. The strategy is this: the fans of correspondence complain that deflationary truth really does not establish some needed connection discourse-ontology. Let us grant that, and let us grant that a correspondence theorist will have privileged, unfettered access to such connection; even though we have really no idea of what connection we are talking about here, this is the kind of things one hears from the fans of correspondence, and this we concede: we are stacking the deck in favor of correspondence. Then let us show that even under such assumption the correspondence theory will offer no significant advantage over the competing deflationary notion.

I ask the reader to imagine a correspondence theorist as singularly well-equipped to discover the real properties and objects to which predicates and terms refer in the given language. He is an ontological detector: if there is a property answering to a certain predicate, this theorist will see that there is, and viceversa.<27>Foot note 2_12 The question is whether this peculiar ability provides him with a criterion to discriminate factual (or «truth-conditional») from non-factual discourse. Now let us imagine this theorist confronted with the realization that there is no property of wrongness and therefore no fact about, say, the wrongness of telling lies. In accordance with his view of truth, he would presumably regard such a fact as the truth conditions of «lying is wrong». But he cannot take the lack of fact about the wrongness of lying as a direct indication that a sentence like «lying is wrong» is not truth-conditional. Why not say, more economically (at least from a semantic point of view), that the sentence is plainly false? Perhaps our correspondence theorist might reply that the situation is peculiar in that the lack of a fact making «lying is wrong» true does not seem to translate into a fact making «lying is not wrong» true, as we might expect for sentences and areas of discourse of the ordinary sort. But this is no good reason to relegate sentences about the wrongness of lying in a class by themselves, since the same peculiar situation obtains for very many sentences about, e.g., the present king of France (or perhaps the future king of England).<28>Foot note 2_13 But those sentences surely seem as truth-conditional as any. Nor would it matter very much to establish that the truth conditions of (i.e. the fact that) «lying is wrong» can never be realized (assuming this could be established), since that is also the plight of a lot of necessarily false sentences (such as «water is not H2O»). But impossible sentences also are truth-conditional. Anyway, we have supposed that there is no object in the world around us with the property of wrongness. But there is also no object with the property of being a winged horse, nor could there be. By assumption, the correspondence theorist is aware of this situation; but then how will he be able to determine that utterances concerning winged horses are to be counted as false, while utterances about the wrongness of lying are to be put in the non-factual limbo? What fact about the world could justify this disparity of treatment?

The moral is clear enough, I think: the mere lack of a suitable fact or «truthmaker» for a sentence is not sufficient evidence for us to determine that the sentence is not truth-conditional in the sense required by non-factualist claims. But notice that we have worked from within the constraints of a correspondence conception of truth and truth-conditions. Thus we come to this realization: the special liaison between facts and sentences, to which the correspondence theorist claims privileged access, cannot be the criterion by which we single out truth-conditional or fact-stating discourse. That criterion must advert to issues of use, interpretation, conventions, etc. -- to the function of an area of discourse in communication. But these issues do not require for their clarification a correspondence conception of truth. For that matter, they do not impose a deflationary conception either: the point is that the crux of this matter relatively independent from any particular conception of truth.<29>Foot note 2_14

Now one might still reply that, while the ontological component of the non-factualist thesis (the claim that there are no facts of a certain type) is not sufficient to ground the semantic component, yet it is necessary: and the correspondence theorist has the resources to make at least sense of the ontological thesis, while the deflationist lacks even those. This is in effect a version of the misconception we discussed in III, according to which a deflationary conception issues into generalized anti-realism, so that a grasp of facts in the world (of ontology) is quite out of its reach. But we have shown this prejudice to be mistaken. The deflationist (i.e., the disquotationalist about truth) can very well make sense of the absence or presence of certain facts. She can admit of some and reject others, so that her conception of facts is certainly non-trivial. Given this, she has as much of a right to the ontological thesis as the correspondence theorist does. Hence she can make as much sense of a non-factualist claim that some sentences are not truth-conditional as the correspondence theorist can, for the ontological thesis that there are no facts answering to those sentences is the only component of such a claim that can ever presuppose a correspondence conception.

The upshot is this, I believe: There is a certain ambiguity in the concept of truth-conditionality, as it applies to sentences or utterances (and perhaps in the very concept of truth conditions). There is a somewhat weak notion of being truth-conditional, which is associated with Tarskian biconditionals. Tarskian biconditionals are deflationary «in spirit», so to speak, but Boghossian is right in claiming that, on this notion, almost any sentence would be truth-conditional. So this cannot be what people mean when they deny that sentences of a certain region of discourse are not truth-conditional. So, either non-factualism makes no sense as it is, or there also is a non-trivial reading of claims of truth-conditionality. But the non-trivial notion of truth-conditionality appealed to by non-factualist claims turns out, on closer inspection, to portend (i) a straightforward ontological thesis and (ii) a thesis about the role of certain sentences in communicative discourse. I have argued in this section that (i) does not imply (ii), and that, since (i) does not imply (ii), the correspondence theorist is in no better shape than the deflationist to account for the non-trivial notion of truth-conditionality. In effect, we might conclude that either non-factualism makes no sense at all (a stronger claim, which even Boghossian seems unwilling to underwrite), or the distinction between factual and non-factual discourse has less to do with a conception of truth than is often presumed.

Let us show now that Boghossian's argument is neutralized. To see this, let TC stand for the predicate «is truth-conditional». Thus, the expression TC(S) «says» that the sentence named by S is truth-conditional. Let TC(Stcp) be the sentence TC(«S has truth conditions p»). According to step (A) in the SOC argument, deflationism entails TC(S) for any declarative S; so in particular TC(Stcp). But the non-factualist holds (2), which is the negation of TC(Stcp). Therefore, (2) implies that deflationism is false. This is the SOC argument in a nutshell. The considerations developed in this and the previous section, however, show that there are two distinct predicates (or senses) of truth-conditionality available to the non-factualist: there is the «trivial» version borrowed from Tarski and the non-trivial one involving the «ontological thesis».

Let TC and TC* represent, respectively, the trivial and the non-trivial predicate, and return to the argument. On Boghossian's own account (2) involves a non-trivial notion of being truth-conditional: in other words, (2) is not the negation of TC(Stcp), but rather of TC*(Stcp). The negation of TC*(Stcp), however, does not contradict TC(Stcp), short of equivocation.<30>Foot note 2_15 On the other hand, deflationism does not entail the non-trivial thesis, i.e. TC*(«S has truth conditions p»), for the reasons explained. So on the non-trivial interpretation step (A) is not a valid inference. Therefore we have two possible cases: Either (i) step (A) is valid, but its consequence is TC(S) and successively TC(Stcp); in which case there is no contradiction with (2), which denies TC*(Stcp). Or (ii) we take the conclusion of step (A) to be TC*(Stcp), thereby producing a contradiction with (2); in which case, though, step (A) is invalid, for deflationism does not entail that any sentence is truth-conditional in the non-trivial, ontologically significant sense.

The significance of these results may perhaps be better appreciated by a brief comparison with some of the critical literature generated by SOC. M. Devitt and G. Rey have generally conceded one of Boghossian's crucial points; they agree with him that «the idea that a sentence lacks truth conditions presupposes that the sense of truth in question is robust, not deflationary».<31>Foot note 2_16 But I have shown in this section that «the idea that a sentence lacks truth conditions» either presupposes little about truth, or else is perfectly intelligible on a deflationary notion. On the other hand, Robert Kraut (1993) attacks the thesis that for a deflationist any sentence is trivially truth-conditional, i.e. step (A) in my reconstruction (notice that Devitt and Rey concede this, too). Thus his rejoinder to Boghossian's attack is parallel to that formulated in this paper. In order to provide a counterexample, however, Kraut adopts a pragmatist, not a disquotationalist, conception of truth -- truth, on this view, is a special «compliment» paid to select utterances. Kraut is effective in characterizing a pragmatist conception immune to Boghossian's charge, but he admits that Boghossian's point is valid against disquotationalism and other «promiscuous» varieties of deflationism. If I am correct, however, either Boghossian's point about disquotationalism is incorrect, or else it does not mean what Boghossian takes it to mean. At any rate, I disagree with Kraut on the issue of promiscuity: the «non-promiscuity» that is necessary to formulate non-trivial irrealist and non-factualist theses about given areas of discourse has little to do, if I am correct, with truth itself.


S. Blackburn 1984: Spreading the Word (Oxford: Clarendon Press).

P. Boghossian 1990a: «The Status of Content», The Philosophical Review XCIX, 159-184.

P. Boghossian 1990b: «The Status of Content Revisited», Pacific Philosophical Quarterly 71, 264-278.

M. Devitt 1990: «Transcendentalism about Content», Pacific Philosophical Quarterly 71, 247-263.

M. Devitt and G. Rey 1991: «Transcending Transcendentalism: a Response to Boghossian», Pacific Philosophical Quarterly 72, 87-100.

H. Field 1972: «Tarski's Theory of Truth», The Journal of Philosophy LXIX, 347-375.

H. Field 1986: «The Deflationary Conception of Truth», in Fact, Science and Morality, ed. by McDonald and Wright (Oxford: Blackwell). 55-117.

D. Grover, J.L. Camp, N. Belnap 1975: «A Prosentential Theory of Truth», Philosophical Studies 27, 73-125.

F. Jackson, G. Oppy, M. Smith 1994: «Minimalism and Truth Aptness», Mind 103, 287-302.

R. Kraut 1993: «Robust Deflationism», The Philosophical Review CII.

F.P. Ramsey 1927: «Facts and Propositions», in Philosophical Papers, ed. by D.H. Mellor (Cambridge: Cambridge University Press, 1990).

A. Tarski 1935: «The Concept of Truth in Formalized Languages». English translation by J.H. Woodger, in Logic, Semantic, Metamathematics, ed. by J. Corcoran (Indianapolis: Hackett, 1983).

N. Tennant (unpublished), «On Two Transcendental Arguments about Truth and Truth-Conditionality».

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

Pierluigi Miraglia

Department of Philosophy

The Ohio State University

350 University Hall

Columbus, OH 43210-1365, USA