, Volume 40, Issue 2, pp 357–363

Missed It By That Much: Austin on Norms of Truth


    • Department of PhilosophyWichita State University

DOI: 10.1007/s11406-011-9336-3

Cite this article as:
Hershfield, J. Philosophia (2012) 40: 357. doi:10.1007/s11406-011-9336-3


A principal challenge for a deflationary theory is to explain the value of truth: why we aim for true beliefs, abhor dishonesty, and so on. The problem arises because deflationism sees truth as a mere logical property and the truth predicate as serving primarily as a device of generalization. Paul Horwich, attempts to show how deflationism can account for the value of truth. Drawing on the work of J. L. Austin, I argue that his account, which focuses on belief, cannot adequately accommodate the complex role that truth plays in the norms governing assertion and similar speech acts.


TruthDeflationary theoryAustin, J. L.Horwich, PaulDummett, MichaelSpeech acts

The relation between a rough, loose, inaccurate, imprecise, &c. statement and the facts is a most important matter for elucidation.

J. L. Austin

Michael Dummett argues in “Truth” that “we cannot in general suppose that we give a proper account of a concept by describing those circumstances in which we do, and those in which we do not, make use of the relevant word…we must also give an account of the point of the concept, explain what we use the word for” (1959, p. 3; emphasis in original). This is supposed to point up the inadequacy of what he calls the “redundancy theory of truth,” which sees nothing more to the content of truth than is captured in instances of the disquotational schema “It is true that p if and only if p.” Such a view, now more commonly known as the deflationary theory of truth, does not imply that the truth predicate is logically superfluous, however. A disquotational truth predicate still serves what Williams (2002) calls an “expressive function.” It confers an ability to say in finite terms what would otherwise require conjunctions or disjunctions of infinite length to state.1 Nevertheless, the utility of truth is not exhausted by its expressive function, a fact, Dummett contends, that goes unacknowledged by the deflationary theory.2 He draws on an analogy with games. It is essential to the very concept of a game (or, more cautiously, an important class of games) that the participants try to win. Thus, e.g., someone who learned all the rules of chess, knew all the possible moves, understood what checkmate and check meant, but who failed to see that chess matches are competitive, with players attempting to beat their opponents, could not be said to have fully grasped the concept of chess.3 Just so with a person whose understanding of the truth predicate was limited to the instances of the disquotational schema. Such an individual would know the truth conditions of all the sentences of the object language for which the truth predicate is defined, but his knowledge of truth would be incomplete. What he would be lacking, Dummett suggests, is the idea of truth being a goal of assertion. The deflationary theory of truth fails to capture the normative dimension of the concept of truth, i.e., the fact that truth marks an important form of success for assertoric speech acts.

Paul Horwich has argued that it is indeed possible to explain truth’s value to us, why we aim for truth, without abandoning deflationism. The key is to see that nothing more is required of the truth predicate than its use as a device of disquotation. Horwich traces the normative dimension of truth to the contribution that true belief plays in the explanation of success: Action-guiding beliefs of the form “If you do x, you will get y” tend, in the presence of a desire for y, to lead agents to do x, ceteris paribus. Thus if doing x will indeed result in y--if the action-guiding belief is true--the agent gets what she wants. Furthermore, we enhance the likelihood of our action-guiding beliefs being true by inferring them from true beliefs in accordance with truth-preserving rules of inference, and since practically any belief could serve as evidence for our action-guiding beliefs, we arrive at the result that it is desirable for our beliefs to be true, i.e., that we believe that p only if p.4 It is obvious, however, that truth is only figuring here in its generalizing role, since it is really the instances of the above schema that are valuable. What we want, in other words, is to believe that drinking a glass of water will quench our thirst only if drinking a glass of water will quench our thirst, that getting vaccinated will immunize us against H1N1 flu only if getting vaccinated will immunize us against H1N1 flu, and so on. Moreover, since these action-guiding beliefs are more likely to have the desired feature provided the beliefs they are derived from do, we also want to believe, e.g., that feelings of thirst are normally caused by dehydration only if feelings of thirst are normally caused by dehydration, and that the vaccination is safe only if the vaccination is safe, and so on. The truth predicate, because of its disquotational feature, allows us to capture this potentially infinite set of normative facts in finite terms.

Truth is more than merely instrumentally valuable, however; it is considered valuable for its own sake as well. While Horwich (2006) concedes that truth’s intrinsic value “may well be epistemologically and explanatorily fundamental” (p. 351), he suggests that it has its genesis in truth’s practical benefits. And in this he sees a similar pattern at work with many moral norms: their undeniable utility leads to an elevated status as things to be valued for their own sakes. Thus for example, he speculates that it is because we derive enormous practical benefits from living with people who “are considerate of others and abhor the infliction of pain” (ibid., p. 352) that we come to view these traits as valuable in and of themselves.5 Nevertheless, to say that truth is valued for its own sake is potentially misleading. What has intrinsic value are the instances of the schema S believes that p only if p; truth comes in, as it did in the instrumental case, only as a way of generalizing on this potentially infinitely large set.

Dummett’s original challenge to deflationism was put in terms of the value of true assertions, so the question is whether Horwich’s analysis can be extended to cover these as well. He makes some backhanded progress towards this goal by linking the wrongness of lying to the disutility of false belief. Since acting on false beliefs increases the likelihood of failure, lying, the purpose of which is to mislead a hearer, can be glossed as an attempt to harm someone.6 What is important to note in this explanation is that it is the instances of the pattern of believing that p only if not p that have disvalue; consistent with the deflationist strategy, the truth predicate figures here only as a device of generalization.7 Of course, lying is not the only way in which one can deliberately fail to state truths, and in some such cases, at least, a speaker cannot be said to have done anything wrong. While the norm “Assert that p only if p” (or only if one has good reason to believe that p) may indeed operate in a variety of conversational settings, it does not have unrestricted application. There are perfectly commonplace circumstances where the prevailing truth norms are more forgiving. In such cases, it is acceptable for a speaker deliberately to say something false provided the statement falls within an acceptable range of inaccuracy. Thus we can sometimes exaggerate, speak imprecisely, describe loosely or roughly, understate or overstate, approximate, and so on, without those speech acts being in any way defective. The secretary of the committee can record that the meeting began at 3:00 (when in fact it was 3:03), the teacher can describe France as hexagonal, the hungry co-worker can pointedly remark that he hasn’t eaten in three days, and in each case the speaker can be said to have provided information sufficient for the purposes at hand. Examples of this kind raise difficulties for deflationism precisely because they reveal the truth predicate to be more than just the simple device of generalization that deflationists propose, with the implication that truth is not, as Horwich (1998a) calls it, a mere “logical property.” Truth turns out to have a substantive nature amenable to philosophical analysis.

It was J. L. Austin who delineated the importance of these norms for the theory of truth. Truth and falsity, he argues, measure “how the words stand in respect of satisfactoriness to the facts, events, situations, &c., to which they refer” (1975, p. 149). But in that they are hardly unique, belonging instead to a “galaxy of words…connected with the notion of fitting and measuring” (1961, p. 161). That loose descriptions, exaggerations, approximations, and so on, sometimes “work” reveals the existence of “various degrees and dimensions of success in making statements” (1950, p. 130; emphasis in original), where what is being evaluated is the nature of the “fit” between an assertion and the facts. The important point is that these are context-sensitive norms concerning forms of language-world correspondence, with truth serving an inelminable regulative role. Thus an approximation cannot be too rough nor a description too loose, lest it fail to be informative, and exaggeration and understatement only work, or work best, when what is said falls within a certain range relative to the facts. Indeed, the very notions of looseness, roughness, approximation, and so on, capture ways of deviating from a standard of language-world correspondence that is couched in terms of truth.

Austin’s discussion of these matters is framed by his larger goal of constructing a theoretical framework for the study of language use, and in the process overturning long-standing philosophical traditions. One such is the sharp division of utterances into constatives and performatives. The former, which included, inter alia, statements, assertions, and descriptions, were thought to say something about the world, whereas the latter were simply a matter of doing something. Austin held that this distinction fails on both counts: making a statement and describing a scene are as much performances as are determining someone’s guilt by saying “I find the defendant guilty as charged” or estimating the time by saying “I estimate that the meeting will take the entire hour.” Conversely, evaluations are possible for many performatives where “there is an obvious slide towards truth or falsity” (1975, p. 141): did the defendant really commit the crime, did the meeting take less than an hour, and so on. Perhaps it was to counter what he felt were deeply entrenched positions that he tended to raise the rhetorical heat beyond what was necessary. Thus he argues that in the sorts of cases we have been considering, where standards of accuracy are somewhat relaxed, it would be inappropriate to evaluate them for truth and falsity, saying of the characterization of France as hexagonal, e.g., that “it is a rough description; it is not a true or false one” (1975, p. 143). Even more remarkable is what he has to say about those situations where ascriptions of truth and falsity are thought to be appropriate. In How to do Things With Words he gives two different characterizations of truth that appear, on the surface at least, to be inconsistent with one another. On the one hand, truth is seen as something to which most assertions cannot hope to aspire, since it is glossed as “the ideal of what would be right to say in all circumstances, for any purpose, to any audience, &c.,” something he claims occurs “perhaps with mathematical formulas in physics books” (1975, p. 146). And yet on an earlier page truth is held out as an all-purpose form of approbation, as standing for a “general dimension of being a right or proper thing to say as opposed to a wrong thing, in these circumstances, to this audience, for these purposes and with these intentions” (ibid., p. 145).8

Here I think we must concur with Warnock’s (1989) observation that these characterizations are not terribly compelling. A statement can be true but not the “right or proper thing to say” in a given context, let alone in all speech situations, because it is off-topic, boring, offensive, impertinent or obvious, to name just a few possibilities. Moreover, while it may be inapposite, it is not incoherent to evaluate exaggerations, approximations, etc., for truth and falsity.9 None of this, however, need detract from Austin’s principal insights that there are a variety of expressions for measuring the fit between language and the world, that truth plays a regulative role in many of these evaluations, and that it can do so precisely because it is itself a species of language-world fit (the limiting case in which an utterance corresponds to the facts).

Austin ends his article “Unfair to Facts” by describing an imaginary practice of signaling targets with differently colored flags, where there are preset conventions determining which colored flags are to be used with which targets. This flag-target correspondence makes it possible to define a form of success and failure in signaling. There are, he suggests, important parallels between the practice of signaling targets and that of making statements, and these afford a way of countering Strawson’s contention that because facts are so closely tied conceptually to statements, they have no ontological standing of their own. Targets are no less strongly linked to signals, yet they clearly exist independently of signals. Moreover, just as signals can be said to be correct because they signal targets but not conversely, so the truth of statements turns on their stating facts and not the other way around. The analogy between stating facts and hitting targets is even closer than Austin’s flag-signaling conventions would indicate. The problem with his example is that there appear to be only two possibilities: correctly signaling a target with the appropriate flag or failing to do so. But there are many other games involving targets where the criteria of success are more forgiving, and it is possible to score points even when a throw, shot, etc. is not perfectly on target. In curling, e.g., a stone that lands in the “house” (the area covered by the concentric rings) but not on the “button” (the smallest circle at the center) can count towards the score provided no opposing stones lies closer to the button. Austin’s insight is to recognize a similar pattern in the manner in which assertions and similar speech acts are evaluated. A statement falling short of truth (or containing only part of the truth) can still count as successful provided it meets the standards of accuracy demanded by the context--“what may score full marks in a general knowledge test,” Austin avers, “may in other circumstances get a gamma” (1950, p. 130). With assertoric speech acts, as with horseshoes, close sometimes counts, and this has implications for the analysis of truth.

The upshot of the discussion so far is that Dummett’s characterization of the aim of assertion as that of uttering truths is oversimplified, and that a subtler analysis will reveal a more complex picture of truth’s regulative function. This is important, because the ability of deflationism to account for norms of truth depends on limiting the utility of truth to its generalizing function. Now, it might appear as if the same problems beset Horwich’s deflationary treatment of the truth norms governing belief, and consequently that his proposed solution for belief can easily be extended to cover assertoric speech acts. As we noted earlier, Horwich (2006) credits true belief as having non-instrumental value, as being valuable for its own sake. He encapsulates this assessment in the following principle:

VT It is desirable to believe what is true and only what is true (2006, p. 347).

VT essentially combines two different ideas: the desirability of having true beliefs about anything whatsoever, and the desirability of avoiding false beliefs. Both of these, he concedes, can be challenged. Many things are simply not worth bothering about, lacking practical or theoretical import. In other cases, the truth might be so horrific or disturbing that one would be better off not knowing. And sometimes a belief is so essential to one’s psychological or emotional well-being, or to one’s ability to act, that there is positive value in having the belief even if it is false, as when a soldier on a dangerous mission believes that he is likely to escape death or injury. Horwich argues that these kinds of examples can be accommodated without modifying VT. What they reveal, he claims, is the unsurprising fact that true belief is not the only thing we value: we also care about, inter alia, self-preservation, our own happiness, behaving honorably, and pursuing interesting and worthwhile subjects. It is possible for these goals to clash with our aim of believing all and only truths. Far from showing the need to dilute VT, Horwich argues, the aforementioned examples are illustrative of how the pursuit of truth can be trumped by these other desiderata. The crucial point to keep in mind is that on such occasions “the sacrificed values continue to matter,” it is just that “they are outweighed by more important considerations” (ibid., p. 348).

The idea of extending this solution to cover assertoric speech acts might seem promising. Horwich could agree with Dummett that a basic norm of truthfulness governs assertion, but add that this sometimes can be overridden by other imperatives, such as, inter alia, the need to be polite, kind, concise, or sensitive to the expectations of one’s audience. Unfortunately for the deflationist, the analogy breaks down at this very point, since the aforementioned pragmatic factors do not so much as trump a norm of truthfulness as interact with it, and thereby alter it. While pinpoint accuracy may not be to the point in such conversational settings, it is also not the case that just anything goes. Truth is still playing a regulative role, since the acceptability of an utterance will turn in part on the relationship between what it says and the facts. Some departures from the literal truth--certain forms of inaccuracy or incompleteness--will pass muster in a given conversational context and others won’t. The net effect of these pragmatic factors is that the scope of truth as a normative concept is broadened, so that in certain situations being approximately true or partially true is enough to confer success on an assertion; literal truth is not always necessary. In other words, if saying something true is tantamount to hitting a bull’s eye, to recur to Austin’s comparison with games involving targets, then there are many ways of coming close to doing this, and, in the right conversational settings, these near misses and partial hits count for something.

It would be tempting at this point for a deflationist to dismiss these considerations as nothing more than by-products of a metaphysically extravagant theory of truth, one that mistakes weak analogies for deep theoretical insights. In this case temptation is best resisted; the problem cannot be made to go away so easily. The crucial issue is coming to understand the nature of the norms governing assertoric speech acts. Austin’s insight is to recognize the extent to which these are shaped by pragmatic and conventional factors, the net result of which are context-dependent norms that sometimes confer success on utterances falling short of literal truth. Truth is still operative here, though in an attenuated capacity: for such assertoric speech acts the aim is not “the whole truth and nothing but the truth,” but rather “lying an acceptable distance from, or containing a portion of, the truth.” The strategy Horwich employs in the case of belief, viz., to explain away seeming exceptions to VT as those in which the value of true belief is trumped by other values, won’t work here. It is not that the value of true assertion is being overridden so much as it is being diluted. The irony is that truth matters even in norms that otherwise sanction falsehood. It remains to be seen whether the deflationist has the requisite tools in his toolbox to accommodate these curious linguistic norms.


See, e.g., Dummett 1959, p. 4; Quine 1970, pp. 10–13; Horwich 1998a, pp. 2–5.


This is not quite correct. Strawson (1950) suggests uses of the truth predicate, e.g., as a device for signaling agreement, that are consistent with a deflationary treatment of the semantics of truth. This poses no threat to the Dummettian criticism discussed in the text.


In matches involving a series of games, players who are leading in wins may sometimes adopt a conservative strategy of playing not to lose. Nevertheless, this is still part of an overall strategy of winning the match.


Obviously, this account is too crude as it stands. See Horwich (1998a), pp. 44–46, for an outline of a more refined variant of the explanation of success.


In Horwich 2005a he makes the same point using the virtues of courage and perseverance (pp. 116–118). See also Horwich 1998b, chap. 8; Horwich 2002.


See Horwich 2006, pp. 347–348.


This needn’t imply that lying is always wrong, since the norm against lying could sometimes be overridden by other norms; see below.


In “Truth” he seems to combine both ideas in saying that truth is either a “bare minimum or an illusory ideal” (1950, p. 130).


Warnock (1989) argues that Austin’s mistake is treating the illocutionary act as the proper locus of truth and falsity, whereas it is really the locutionary act which determines what was said in the utterance; cf. pp. 138–139.


Copyright information

© Springer Science+Business Media B.V. 2011