16
Truth, Knowability, and Neutrality Author(s): Tim Kenyon Source: Noûs, Vol. 33, No. 1 (Mar., 1999), pp. 103-117 Published by: Wiley Stable URL: http://www.jstor.org/stable/2672017 . Accessed: 14/06/2014 06:04 Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at . http://www.jstor.org/page/info/about/policies/terms.jsp . JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship. For more information about JSTOR, please contact [email protected]. . Wiley is collaborating with JSTOR to digitize, preserve and extend access to Noûs. http://www.jstor.org This content downloaded from 188.72.126.41 on Sat, 14 Jun 2014 06:04:07 AM All use subject to JSTOR Terms and Conditions

Truth, Knowability, and Neutrality

Embed Size (px)

Citation preview

Truth, Knowability, and NeutralityAuthor(s): Tim KenyonSource: Noûs, Vol. 33, No. 1 (Mar., 1999), pp. 103-117Published by: WileyStable URL: http://www.jstor.org/stable/2672017 .

Accessed: 14/06/2014 06:04

Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at .http://www.jstor.org/page/info/about/policies/terms.jsp

.JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range ofcontent in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new formsof scholarship. For more information about JSTOR, please contact [email protected].

.

Wiley is collaborating with JSTOR to digitize, preserve and extend access to Noûs.

http://www.jstor.org

This content downloaded from 188.72.126.41 on Sat, 14 Jun 2014 06:04:07 AMAll use subject to JSTOR Terms and Conditions

NOUS 33:1 (1999) 103-117

Truth, Knowability, and Neutrality

TIM KENYON

University of Aberdeen

1. Introduction

Whatever the ambiguities of the terms 'realism' and 'anti-realism', and they are many, there is an obvious traffic between that debate and the analysis of the concept of truth. For the interesting cases of disagreement between realists and anti-realists are those involving regions of discourse in good standing-that is, where there is a strong feeling that statements within that fragment of the lan- guage can express truths quite literally. A metaphysical realist may exploit this intuition by taking such statements at face value, so that, for example, to say truly that a picture is humorous is correctly to ascribe the property of hilarity to the picture, in the same way that one might predicate a particular mass of the picture as well. On the univocal understanding of truth as this classical notion, with connotations of metaphysical "depth", anti-realism towards a discourse is forced to express itself either as the denial of the truth-aptitude of the relevant state- ments, as in emotivism, or as the assertion of their unexceptioned falsity, as in error theory.

In Truth and Objectivity, Crispin Wright responds by pressing the view that 'truth need not be the exclusive property of realism' (1992, p. 12). Wright pro- poses 'to outline a framework for the expression and development of anti-realist intuition...' (1992, p. 12), a framework intended to be a venue for the pursuit of discourse-specific debates between realists and anti-realists. To this end, he in- troduces the core notion of superassertibility (SA) as a minimalist truth predicate 'neutral on the preferability of a broadly realist or anti-realist view' (1992, p. 33). This proposal is an attempt to tweak the truth predicate into something stronger than the notion of warranted assertibility (WA) Wright uses to undermine defla- tionism towards truth; SA is constructed to capture the stability and absoluteness of truth, in such a manner that it satisfies certain intuitions ('platitudes') about the relation between language and the world, and behaves as we expect truth to be-

t 1999 Blackwell Publishers Inc., 350 Main Street, Malden, MA 02148, USA, and 108 Cowley Road, Oxford OX4 1JF, UK.

103

This content downloaded from 188.72.126.41 on Sat, 14 Jun 2014 06:04:07 AMAll use subject to JSTOR Terms and Conditions

104 NOUS

have with respect to the logical operators. It is intended to be the weakest notion capable of going in for T in the Disquotational Schema, while yet conforming to the Epistemic Constraint.

(DS) 'P' is T iff P. (EC) P iff 'P'is knowable.

This proposal is akin in spirit to Michael Dummett's (1991) argument that truth is 'central' to a meaning-theory in a sense compatible with its being a verification-based notion. And as with Dummett's view, there is no shortage of prima facie objections to Wright's epistemically constrained truth predicate. One objection to Wright is made by Terence Horgan (1995), who raises the logical possibility of brains in vats as a challenge to SA. This challenge, I argue, is flawed. In fact, it is flawed for reasons that relate to other prominent objections to SA, made by Stewart Shapiro and William Taschek (1996), and by Neil Tennant (1995). Showing how these criticisms fly wide of the mark is one purpose of the following discussion, and is accomplished mostly by clarifying the structure of Wright' s argument. Another aim, however, is to articulate a concern of my own, which arises from the need to modify the definition of SA. I argue that the mod- ification points up a tension implicit in Wright's truth predicate, having to do with the notion of warrant or evidence to which he appeals.

2. The inflationary argument

To begin, here is a brief summary of Wright's argument against deflationism, given in the first chapter of Truth and Objectivity. Notice that if every proposition has a significant negation, it follows from DS that

(i) 'not-P' is T iff not-P.

Similarly,

(ii) not-('P' is T) iff not-P,

and hence

(iii) 'not-P' is T iff not-('P' is T). 1

Wright's deflationist holds that DS exhausts the concept of truth. So the truth predicate is merely a device by means of which we endorse assertions; to assert that a proposition is true is to assert the proposition itself, though perhaps with emphasis. Wright observes that truth coincides with WA with respect to such assertion endorsement, even when we understand WA in the narrow sense of 'assertible under warrants held in our current state of information' (henceforth,

This content downloaded from 188.72.126.41 on Sat, 14 Jun 2014 06:04:07 AMAll use subject to JSTOR Terms and Conditions

TRUTH, KNOWABILITY, AND NEUTRALITY 105

WA-in-i). The conditions that license the assertion of a proposition P just are the conditions licensing the assertion that P is true, so WA and truth coincide in their norms of endorsement-to aim at satisfying one is to aim at satisfying the other. This may be what motivates the view of Wright's deflationist, that truth is liter- ally nothing i.e., to predicate 'true' of a sentence is to predicate nothing over and above one's willingness to assert it. The crucial datum, then, is that when T is understood as WA-in-i, (iii) fails left-to-right (1992, p. 20). (Whether (i) and (ii) fail as well depends on whether warrant is understood as defeasible or abso- lute; we will return to this question in the final section.) WA-in-i does not com- mute with negation, in other words, while truth does, and so the two diverge extensionally, at least for any discourse admitting of neutral states of information. Since the commutativity of truth with negation is required in order to generate all intuitive instances of DS, the thesis that DS exhausts the concept of truth entails that truth is a concept imposing constraints more restrictive than those imposed by WA-in-i.2 So truth might have a minimal content, but it cannot have no content at all. The moral for Wright is that deflationism is inadequate.3

This inflationary argument serves to launch Wright' s discussion of epistemi- cally constrained truth. The proposal for massaging WA into a plausible truth predicate is given in the definition of SA:

A statement is superassertible... if and only if it is, or can be, warranted and some warrant for it would survive arbitrarily close scrutiny of its pedigree and arbitrarily extensive increments to or other forms of improvement of our information (1992, p. 48).

This candidate for an epistemically constrained truth predicate, and its distinct- ness from WA-in-i, are the central issues in the discussions that follow.

3. Horgan (1995)

The proposal of SA as a satisfier of both DS and EC prompts Horgan to raise a brain-in-a-vat response to such epistemic conceptions of truth. We are asked to consider Smith, who was kidnapped by aliens, had his brain placed in a vat, and was hooked up to a device that continues to provide his brain with information consistent with the details of his mundane life.

We seem to have here a case of a cognizer many of whose beliefs are both superas- sertible and yet false. Superassertibility, after all, is a matter of "durably meeting... standards of warranted assertion"...; and this durability is guaranteed by the quality of the brain-computer interface and the fact that this interface will persist for the rest of the brain's life. So much the worse, it appears, for Wright's epistemically reductive construal of truth as superassertibility. Moreover, this kind of brain-in-a-vat scenario presumably could be invoked to undermine any epistemically reductive notion of truth (1995, pp. 133-134).

This content downloaded from 188.72.126.41 on Sat, 14 Jun 2014 06:04:07 AMAll use subject to JSTOR Terms and Conditions

106 NOUS

There are two preliminary points to make about this thought experiment. One is that it implicitly rejects Hilary Putnam's (1981) argument to the effect that brains in vats would not instantiate false beliefs-an argument springing not from Putnam's own 'epistemically reductive construal of truth' but from his ex- ternalist view of content.4 Setting this aside, there remains the surprising fact that Horgan fails to acknowledge Wright's explicit consideration of how SA would perform against precisely this thought experiment. Wright anticipates the possi- bility that a 'suitably chosen proposition ... (perhaps that we are not brains-in-a- vat) may be undetectably false. Since [the Disquotational Schema under SA] is hostage to counterexample, and so not a priori true, superassertibility has no case to be a truth predicate' (1992, p. 51). Having provisionally floated the same case against SA as that made by Horgan, Wright attempts to puncture the problem.

The envatted brains thought experiment is taken by Wright to express the charge that SA fails to substitute for H in the schema:

H: It is true that it is rI that P iff it is true that P.

The 'illusion of failure' with respect to the envatted brain case is then put down to the plurality of truth predicates operating in the instance of the schema that results when SA goes in for H. The thought experiment assumes that the first and last instances of 'true' are distinct from SA. But, Wright argues, if we instead take the first instance of 'true' to mean SA, then it is no longer clear that substituting SA for H is problematic.

Wright goes on to link this point to pluralism with respect to truth; he claims that the plurality of truth predicates tolerated in the fl-schema is one that his minimalism can accept, and even encourage (1992, pp. 52, 21n.ff). But the con- nection between Horgan' s problem and pluralism is not especially illuminating, in my view. A better response is to use the fl-schema to show that SA is safe from at least one sort of objection: that of internal incoherence. Even should one find the idea of multiple truth predicates problematic (for whatever reason), one must recognize the stability of a position that takes every use of 'true' to mean SA; the project of showing some internal difficulty with such a position is not promising in the least. If such a position is secure from internal incoherence, the remaining threat to SA is the charge of general counter-intuitivity, like that levelled in Hor- gan's vat comments. Now the field is clear to consider whether the thought ex- periment succeeds in its aim.

The natural response from a friend of EC is to charge that the envatted brain fails as a counterexample because, contrary to Horgan's basic premise, Smith's beliefs are not superassertible. Horgan seems to miss that SA is not relativized to the performance limitations or situatedness of any single cognizer, nor to a set of actual agents; like truth it is a property of utterances, not of utterers. To see this, provisionally assume a distinction between truth and SA. We may note that if the statements expressing the aliens' beliefs about Smith' s predicament are true, then those statements are also SA. But this licenses us to infer, on Horgan 's reasoning,

This content downloaded from 188.72.126.41 on Sat, 14 Jun 2014 06:04:07 AMAll use subject to JSTOR Terms and Conditions

TRUTH, KNOWABILITY, AND NEUTRALITY 107

that Smith's beliefs are both SA and not-SA. Proof: The former holds by hypoth- esis from Horgan while the latter follows from two considerations: that the aliens speak the truth, and that they will deny the superassertibility of Smith's beliefs. Thus we have a reductio of the assumption that truth and SA diverge.

Horgan's only recourse, again, would be to interpret SA as relativized to in- dividuals or groups. So the SA predicates in the metalinguistic components of

(iv) 'Smith's beliefs are true,' uttered by Smith, is SA

and

(v) 'Smith's beliefs are true,' uttered by aliens, is not-SA

would be crucially distinct, in that each would be relativized to the epistemic state of the utterer, and contradiction would be avoided through ambiguity. This inter- pretation, though, is explicitly at odds with Wright's definition of SA as 'abso- lute' (1992, pp. 47-49). In short: the interpretation of SAby which Horgan generates his apparent counterexample appears unfaithful to Wright's definition, and thus fails to engage the minimalist position.5

This response does not depend for its efficacy upon the presence of the aliens in the thought experiment. Even were there no aliens, and the whole human race comprised brains in vats, nevertheless it is clear that placing one such brain into a body like the ones we believe ourselves to have, and showing that person the other brains floating in vats, would enable such an agent to re-evaluate and reject her earlier belief that she had not been a brain in a vat. Our recognition of this possibility demonstrates that, were we brains in vats, our beliefs to the opposite effect would not survive 'arbitrarily extensive... improvements of our information'.

Horgan anticipates this sort of reply to his thought experiment, but contends that it 'gives the game away' (1995, p. 134). The appeal to 'improved epistemic vantage points' on which rests SA's claim to be absolute can only succeed if we suppose the superiority of some vantage points to be a superiority at gener- ating true beliefs. So SA apparently conceals a reliance upon a metaphysically deep notion of truth, and thus cannot serve to underwrite an alternative (mini- malist or otherwise) to truth so conceived. To deal with the envatted brain, SA requires that some viewpoints on the world are more apt to allow us to see how things really are. The appeal to such viewpoints commits SA to a realist notion of warrant evidence is there to be had, independent of observation thus corrupt- ing its claim to be an epistemically constrained truth predicate.6

This line of reasoning cannot succeed, however, without showing Horgan's thought experiment to be entirely superfluous. For the availability of such rea- soning would licensefrom the outset the following dilemma: SA either is or is not coextensive with our ordinary notion of truth; if it is, then it cannot be epistemi- cally constrained; if not, then it cannot be a truth predicate. But there is no such dilemma a priori. Horgan's response only follows on the assumption that there

This content downloaded from 188.72.126.41 on Sat, 14 Jun 2014 06:04:07 AMAll use subject to JSTOR Terms and Conditions

108 NOUS

are cogent applications of our truth predicate to statements unknowable in prin- ciple (if he is to engage Wright's position), such that a predicate whose applica- bility mirrors that of 'true' thereby defies EC. Unsurprisingly, this assumption fits poorly with the notion of an epistemically constrained truth predicate, but the power of the observation is dubious. SA' s potential to perform as a truth predicate is, after all, precisely what is at issue. To assume a coherent evidence-transcendent element in every proper notion of truth is to beg the question straightforwardly against this, and any other, epistemically constrained truth predicate.

These considerations do not remove Wright' s obligation to advance an inter- pretation of knowability that denies the superassertibility of Smith's thoughts, and yet which is principled in a manner distinguishable from truth as deployed by the metaphysical realist. The feature of Wright's proposal intended to fulfil this obligation is his realist characterization of states of information. In the definition of SA, an existential quantification over warrants lurks in the appeal to a partic- ular sort of waffant. Wright subsequently makes this explicit, claiming that

...the range of the states-of-information quantifier in the characterisation of superas- sertibility does not comprise all merely possible states of information ... nor is it restricted to actually occurring such states. Rather it comprises an intermediate set: the actually accessible states of information which this world, constituted as it is, would generate in a suitably receptive, investigating subject (1996, p. 922, n. 15, italics in original).

This is to explicate two features of SA so central to the project that a failure to appreciate their significance is probably the easiest way to misunderstand exactly what Wright is committed to. One is stability: suitably indexed for time and place, true statements remain true. The other is absoluteness: truth does not admit of degrees. Wright defines SA as having both properties. In Horgan's defence, however, his misunderstanding is invited by Wright's definition. For if the exis- tence of states of information explains the stability and absoluteness of SA, then the distinction is lost between a statement that is warranted and one that can be warranted. This contrast survives only if the 'is, or can be' distinction hinges on there being some distinct notion of defeasible warrant in the first disjunct. And if such a notion of warrant is crucial to the definition, it will indeed appear that any false statement uttered under warrants possessed in this sense, and never over- thrown in the actual run of events (or even nearby possible runs of events), would still count as superassertible.

As it stands, Wright's definition misleads by drawing an apparent contrast where none is relevant. A first pass at a solution is to amend the definition of SA, most simply by deleting 'can be', and construing 'is' tenselessly. This modifica- tion renders the information-realism of Wright's view significantly more trans- parent. As I will argue in my concluding section, however, making clear the spirit of Wright's proposal in this manner involves giving up much of the letter of his definition.

This content downloaded from 188.72.126.41 on Sat, 14 Jun 2014 06:04:07 AMAll use subject to JSTOR Terms and Conditions

TRUTH, KNOWABILITY, AND NEUTRALITY 109

4. Shapiro and Taschek (1996)

Horgan misses that stability and absoluteness crucially characterize SA. By con- trast, and by design, Wright's notion of WA-in-i exhibits neither property. Stew- art Shapiro and William Taschek (1996) do not recognize the importance of this contrast when mounting their objection to the inflationary argument that sets the stage for SA. Here is the problem as they see it.7

Wright's argument proceeds from the claim that there are, in at least some discourses, propositions for which certain states of information are neutral with respect to the warrant they offer. In such states of information, a lack of warrant to assert the relevant proposition does not license the assertion of that proposi- tion' s negation. WA-in-i, as we have seen, does not commute with negation, and thus cannot support DS. So long as DS is taken to be characteristic of truth, only a predicate diverging extensionally from WA-in-i could be a truth predicate.

So far so good. But Shapiro and Taschek point out that on a proof conditional semantics, like that proposed by Heyting, to negate a proposition just is to show the absurdity of the assumption that there is a proof for it. And if we follow Dummett in adapting this semantics to natural language, the likeliest notion of proof available to us is WA. For such a semantics, the condition under which P is not assertible is exactly the condition under which not-P is assertible: namely, when there is no warrant for P. The assumption that neutral states of information abound collapses. Shapiro and Taschek conclude:

Either Wright is wrong that nothing deserves the title of a truth predicate-and thus, that nothing deserves the title of a minimal truth predicate-if it does not allow for the potential divergence of truth and warranted assertibility, or he is mistaken in suppos- ing that his notion of minimal truth provides a neutral point of departure for under- standing and assessing various realisms and antirealisms. For it would seem to rule out a Dummett-style semantic antirealism from the start (1996, p. 80).

If Wright has made a mistake, it is not the one claimed by Shapiro and Taschek. The mistake is purely tactical, and consists in his having made insufficiently clear the point of his discussion of assertoric norms. The intention, recall, is not to select some particular semantics and show where it goes wrong. It is first to consider how WA-in-i (a suitably minimal candidate for a truth predicate) suc- cessfully governs assertion, inasmuch as to aim at the satisfaction of 'WA-in-i' is to aim at the satisfaction of 'true' however maximally conceived. Wright then argues that WA-in-i fails to support our central intuitions about truth. The concept of truth, then, has at least a greater content than mere assertion-governance, and thus we are driven to be at least minimalists, and not deflationists, towards truth. This point can be lost in the dialectic of Wright's discussion, since rather than flag his assertibility predicate as I have done, he simply explains that he intends 'WA' to be understood as 'WA-in-i' (p. 20). Of course Shapiro and Taschek are correct that WA may serve as the natural language analogue of provability only if it provides the same crucial features that proofs do in the mathematical case: sta-

This content downloaded from 188.72.126.41 on Sat, 14 Jun 2014 06:04:07 AMAll use subject to JSTOR Terms and Conditions

110 NOUS

bility and absoluteness. It is central to Wright's case that WA-in-i does not pro- vide these, so to offer Dummettian semantics as a counterexample to the inflationary argument is to miss the point.

Shapiro and Taschek do consider something like the account just given, and, as already mentioned, argue against its usefulness.

On this interpretation, Wright's argument correctly shows that true is not to be iden- tified with proved. Against this interpretation, however, note that both provability and warranted assertibility are modal notions, while 'proved' is not modal. Moreover, the extension of 'proved' changes from moment to moment, as long as competent math- ematicians are at work, while warranted assertibility should be at least relatively stable over time. It is for this reason, perhaps, that as far as we know no one has tried to identify 'true' and 'proved' in any mathematical context. So if Wright's argument is rescued in this way, it attacks a straw man when it comes to mathematics (1996, p. 81).

This passage combines the summary of one of Wright's key points with the un- intentional recapitulation of another. It then adds to this the mistaken idea that Wright' s discussion of assertoric norms is intended as an attack on absolutely all interpretations of WA. Of course, that is not the point; SA itself is intended to be an attenuated version of WA. And Wright's argument does more slightly, but importantly, more than show that true is not to be identified with proved: it presses the idea that the two are not to be identified despite coinciding in positive normative force. Thus, Wright thinks, deflationism inflates into minimalism.

That the inflationary argument is marred by unclarity is suggested by the fact that Shapiro and Taschek offer the instability of WA-in-i as if it weighed against the view put forward in Truth and Objectivity. Perhaps, among other factors, Wright too readily assumes some familiarity with his previous writings on SA. On the other hand, he makes specific reference to these (1992, p. 48). Turning to the referenced passage, we find an exposition of the very point Shapiro and Taschek advance as problematic.

If we reflect that 'provable', unlike 'proved', is also intuitively a timeless property, we are pointed towards an obvious moral. The proposal has been to interpret 'T' in terms of current warranted assertibility. But what would suffice for a trouble-free construal of the recursions for the connectives is a notion of assertibility which stands to that notion as provability stands to 'is proved' (Wright 1993, p. 411).

Wright might have made this point more clearly, and at greater length, or his critics might have followed up the references. But he didn't; and they didn't. The result is an unnecessary confusion about what Wright intends. This confusion is carried to the extreme in Neil Tennant's (1995) objection to the anti-deflationism argument.

This content downloaded from 188.72.126.41 on Sat, 14 Jun 2014 06:04:07 AMAll use subject to JSTOR Terms and Conditions

TRUTH, KNOWABILITY, AND NEUTRALITY 111

5. Tennant (1995)8

Consider again the synoptic version of the argument against deflationism. Ten- nant, anticipating Shapiro and Taschek, charges that Wright cannot resist the inference from not-('P' is WA) to 'not-P' is WA. The objection is as follows. By taking T to be WA, Wright commits himself to saying that if we assume

(vi) not-('P' is WA)

we are assuming (vi) to be, not true, but WA (p. 103). And by our understanding of WA, we must not only have reason not to assert

(vii) 'P' is WA,

but must have reason to assert

(viii) 'not-P' is WA.

Tennant concludes that no problem arises in the inference from (vi) to (viii), so that Wright's principal objection to deflationism fails.

The confusion begins with the supposition that a disproof of (vii) constitutes a proof of (viii). One might observe that as a principle of practical reasoning, we often take a lack of evidence for a proposition to provide grounds for asserting its negation. But clearly the power of this intuition varies with our sense of how informed we are. All Wright needs is the supposition that in some cases we are so relatively uninformed that a lack of warrant for a proposition cannot be taken as warrant for rejecting that proposition. Clearly, many examples can be constructed to bear out Wright's concern (1992, p. 20) with the behaviour of WA-in-i on states of information neutral with respect to a given proposition.

Tennant acknowledges such cases, but contends that they do not speak to the 'real issue' of WA. Such cases support only the 'weak' claim

(ix) not-('P' is WA in state of information J),

and not the 'much stronger' claim made by (vi) (1995 p. 103). But what, and whose, is this stronger reading of WA, and how does it compel Wright to admit that evidence against (vii) constitutes evidence in favour of (viii)? Tennant's weak strong distinction demonstrates that he too has simply misunderstood what Wright means by 'WA'. In the first instance, Tennant's error is the same as that made by Shapiro and Taschek: he holds that Wright 'cannot demur at [the step from 'not-p' to "'not-p' is WA"] without begging an extremely important ques- tion against the anti-realist' (p. 101). But, as we have seen, Wright can and ought to demur at such an inference, since he takes WA in the minimal sense already discussed.

This content downloaded from 188.72.126.41 on Sat, 14 Jun 2014 06:04:07 AMAll use subject to JSTOR Terms and Conditions

112 NOUS

Indeed, the error (the Ohio State Error, it seems) is all the more surprising here, for Tennant effectively reformulates the notion of SA in the course of explaining how WA-pace Wright, he thinks-ought to be understood. On the proper un- derstanding of an epistemically constrained truth predicate, Tennant writes,

...to deny that P is to deny that there is a warrant for the assertion of P. The latter denial's warrant, in state of information I, must consist in a reductio of any assumed walTant, in any state of iniformation consistently extending I, for the assertion of P. Note that we have to cover any state of information consistently extending I (p. 104, Tennant' s italics and exclamation point)!

This emphasis on the exact details of his proposal stands in contrast with Ten- nant' s somewhat inexact awareness of SA, a core notion of Truth and Objectivity. Recall the definition of SA:

A statement is superassertible...if and only if it is, or can be, warranted and some warrant for it would survive arbitrarily close scrutiny of its pedigree and arbitrarily extensive increments to or other forms of improvement of our information (1992, p. 48).

Clearly Tennant's version of WA closely resembles SA. He has misconstrued Wright's attempt to motivate SA, to such a degree that he offers a notion plainly similar to SA as if it were a response to Wright. The negative claim fails to engage, while the positive claim reduplicates Wright's proposal; the only lasting effect is a muddying of the waters.

One diagnostic comment recommends itself. The problem seems to be Ten- nant' s certainty that the logic of anti-realist negation is overlooked or abused in the inflationary argument. Wright claims that negation is not the issue, since the only negation rules he needs are shared by classical and intuitionistic logics (1992, p. 32), and here Shapiro and Taschek agree with Wright (1996, p. 78). Tennant, too, quotes this passage from Wright, but says of it that '[t]he very form of words beguiles the reader into not looking for the main fault elsewhere' (1995, p. 100). This conviction that there is a main fault elsewhere appears to have led to the reduplicative proposal.

6. Information and neutrality

Returning to a theme raised by Horgan, it is worth noting that taking a realist view of states of information need not constitute an admission of failure by the anti- realist, since the relevant sort of realism at issue for this tradition is that very strong variety identified by Michael Dummett.

The very minimum that realism can be held to involve is that statements in the given class relate to some reality that exists independently of our knowledge of it, in such

This content downloaded from 188.72.126.41 on Sat, 14 Jun 2014 06:04:07 AMAll use subject to JSTOR Terms and Conditions

TRUTH, KNOWABILITY, AND NEUTRALITY 113

a way that that reality renders each statement in the class determinately true or false, again independently of whether we know, or are even able to discover, its truth-value (1993, p. 230, my italics).

To be an anti-realist, in the sense relevant to this debate, is just to reject this very strong doctrine of the principled transcendence of particular truths. By this cri- terion, minimalism in the form of information-realist SA is well-situated to deny any commitment to Dummettian realism. Plausibly, it is essential to something' s being an informational state that it be accessible.9 Augmented with the thought that such accessibility is indexed, at least in principle, to our cognitive capacities, this produces a realism about states of information too weak to support the pos- sibility of truths in principle beyond our ability to grasp; Dummettian realism towards states of information, so construed, is self-contradictory.

Of course, the idea that the world consists of states of information will inev- itably draw questions about what it is for a state of information to obtain inde- pendent of anyone's actually accessing it. This is an excellent question, but, it must be noted, one that arises with equal force for any realism whatever towards objects, facts, or substance. There need be no special problem for the ontology of information. We have here just a metaphysical corollary of EC, in which the knowability of all truths is replaced with the comprehensibility of all states of affairs. Or rather, all states of affairs relative to a discourse, since Wright refrains from committing himself to the generality of this view. Wright's claim is only that SA is a truth predicate in any discourse where such an informational realism is an accurate characterization of the relevant states of affairs. If it can be made out that some discourse is not so characterizable, then SA cannot be a truth pred- icate in that region (1992, pp. 52-3). Explicating SA, says Wright, demands no more than this; it would be to press a distinct global anti-realism to claim that every region of discourse is governed by minimal truth.

Grant the tenability of this distinction. Should we then think of SA as non- partisan, rather than an anti-realist truth predicate? How, in other words, are we to understand the claim that Wright's minimalism is 'neutral on the preferability of a broadly realist or anti-realist view'? Neutrality is a rather delicate issue in this context; the sense in which Wright's proposal displays it is not altogether clear.

It is tempting to think of neutrality as agreeability to all sides. On this under- standing, disputants could first agree on the truth predicate operating over a dis- course, and then pursue their dispute over the metaphysical import of the discourse itself. SA does not seem apt to fulfil such a role. It is clear that such a dispute implies some antecedent disagreement, at least implicitly, over the strength of the modal component of SA. That is, there could not be such a dispute save by there being a disagreement over which of the possible ways of transcending our im- mediate warrants is the weakest that accounts for the formal and intuitive aspects of the discourse. So there is some danger that when the realist and anti-realist agree on SA as a neutral truth predicate, their agreement is made in ignorance.

This content downloaded from 188.72.126.41 on Sat, 14 Jun 2014 06:04:07 AMAll use subject to JSTOR Terms and Conditions

114 NOUS

They are constantly punning on SA. Each mistakenly supposes the other to mean the same thing, when in fact one has in mind a notion that satisfies EC, while the other takes the way things stand with states of information to depend upon facts that can obtain independently of our capacity to know.

This distinction is bound up with a Euthyphronic contrast-the contrast be- tween regarding SA as a predicate that completely recovers the familiar and cru- cial features of truth, or regarding it as a predicate whose applicability depends wholly upon the antecedent applicability of a distinct truth predicate. Wright considers this contrast in some depth, and I will not follow the thread here. But there is little to be made of the thought that SA can be neutral in the formal sense that disputants agree on its formulation, while recognizing that their opponents hold different interpretations of the metaphysical weightiness of the predicate. This just looks too similar to disagreement over whether SA really governs the disputed discourse. Each disputant might agree that SA satisfies the platitude 'that to be true is to correspond to the facts', and yet may-must-disagree about what that means. So the project of showing that 'truth need not be the exclusive property of realism' could not, upon reflection, have consisted in showing how realists might accept EC within a disputed discourse; that is plainly not in the cards.

This does not mean that there is no important neutrality involved in Wright's project, however. The neutrality is just the goal of the project, and not a feature of the specific proposals. Once superassertibility is on the field, a dispute over the metaphysical weightiness of a discourse may be pursued on specific issues like, e.g., what Wright terms the 'width of cosmological role' of the discourse's central postulata. The realist is denied the cheap response, that we know from the nature of truth that the anti-realist position is forlorn. SA is not metaphysically neutral, in other words, but its availability makes truth, broadly conceived, a metaphys- ically neutral issue. So, notwithstanding his apparent claim to do so, Wright ought not be taken to advance a brand of minimalism 'neutral on the preferability of a broadly realist or anti-realist view'. Rather, the relevant neutrality just derives from the claim that truth is not 'the exclusive property of realism'.

7. Conclusion: warrant as belief and information

I wish to conclude with some brief remarks on the nature of evidence. Specifi- cally, it is important to be more careful with the notion of warrant than we have so far been. On the most common understanding of the term, it denotes a defea- sible commodity; one's warrants for asserting 'P' are just those of one's beliefs which seem to bear out that P. Such beliefs may themselves turn out to be false; indeed, this defeasibility is partly constitutive of empirical or inductive reason- ing. It is on the notion of defeasible warrant that we can recognize a difference between beliefs that are true and beliefs that are warranted.

In contrast, there is the sort of warrant discussed in the previous section: the sort that may be assimilated to information like that constituting nodes in Kripke

This content downloaded from 188.72.126.41 on Sat, 14 Jun 2014 06:04:07 AMAll use subject to JSTOR Terms and Conditions

TRUTH, KNOWABILITI AND NEUTRALITY 115

models or Beth trees. A more familiar example is available if we equate warrant with evidence, and consider how a police detective might say, perfectly correctly, "The evidence is out there we just have to find it." Warrant in this sense is not defeasible, but rather is absolute, denoting information that obtains indepen- dently of its being accessed. The principled accessibility of such information, on which EC depends, may be characterized as resulting from the model's having a certain structure; that is, not containing branches of an order-type that would elude an arbitrarily extensive finite search of the tree. On this understanding, SA is the truth predicate for a discourse just in case that discourse has a model with such a structure. The identification of warrant with information in the absolute sense seems most reasonable in relation to mathematics, if anywhere, since it is there that 'warrant for not-P' is open to construal as 'proof that P leads to an absurdity', so that not-P is forced, or true in every extension of that state of information. But it carries an implication of indefeasibility that is out of place in our everyday understanding of evidence, the understanding on which it is 'un- remarkable', to use Wright's term (1992, p. 69), for warranted statements to fail to be superassertible. Which notion do we read into SA? His quantification over states of information suggests that Wright employs an absolute conception of warrant, as does his entire anti-deflationism strategy. After all, why else empha- size that biconditional (iii) fails left-to-right on neutral states of information, when clearly (i), (ii) and (iii) would all fail if actually defeasible warrants figured in the predicate 'WA-in-i'?

These observations together suggest a fairly radical modification, or attenua- tion, of the definition of SA.

A statement is superassertible... if and only if it is, or can be, warranted and some warrant for it would survive arbitrarily close scrutiny of its pedigree and arbitrarily extensive increments to or other forms of improvement of our information (1992, p. 48).

Given the need to delete Wright's 'can be' clause and construe 'is' tenselessly, we ought to wonder at the labour subsequently performed by the portion of the def- inition following the word 'warranted'. For on the absolute understanding of warrant, there is little sense to be made of warrant surviving improvements to one's informational state. That is just what it is for a statement to be warranted in a state of information; one says nothing additional in stating that, moreover, the relevant statement continues to be warranted. The idea of evidentiary survival is clearly linked to the defeasible sense of warrant: it is the idea that some of the beliefs or judgements that motivated an assertion will remain rationally tenable through arbitrarily extensive investigation. Once we excise the equivocation be- tween these two senses, defeasible warrant is no longer relevant to SA. To think otherwise is to invite Horgan's objection again.

We are left with the view that a statement is SA if and only if it is warranted, where 'is' is timeless and 'warrant' means absolute warrant. If this eliminates

This content downloaded from 188.72.126.41 on Sat, 14 Jun 2014 06:04:07 AMAll use subject to JSTOR Terms and Conditions

116 NOUS

some needless complexity from Wright's initial definition, it must be regarded as an advance. But it also seems to involve a certain idealization of the notion of assertibility, so that the relation between SA and the sort of warrants people generally take themselves to have in making assertions is not pellucid. Conse- quently, it is not clear whether SA is best conceived as an assertibility predicate at all. 10

Notes

1. The move from (i) to (ii) and (iii) is not trivial. In particular, (ii) and (iii) involve truth- functional transformations using the notions of negation and logical equivalence in the metalanguage. Plausibly, (i) holds independent of the truth-functional connectives of the metalanguage.

2. A discussion anticipating and encapsulating these points is found in Dummett (1991, p. 64). 3. Wright's inflationary argument serves as the springboard to his quite independent consider-

ation of how the distinction between truth and WA illuminates what is required of any plausible epistemically constrained truth predicate. However, see Anil Gupta (1993a, 1993b), for a distinct case against deflationism employing no analogue of Wright's premise that deflationism is committed to the equivalence of the truth and WA. If Gupta is correct, then there are anti-deflationist arguments to be had in any case-though he prefers to characterize his target as minirnalism, and would likely see Wright as espousing the view he rejects.

4. Are the two related? Probably: Truth as a deep relation connecting belief contents with the world is perhaps unnecessary on a view holding that the world is itself constitutive of belief contents.

5. This may require some unpacking. It is perfectly knowable that Smith's expressions of his beliefs are false; therefore, they fail to satisfy Wright's definition of superassertible statements, for such expressions would not 'survive arbitrarily close scrutiny' under 'arbitrarily extensive incre- ments' to Smith's state of information. The aliens themselves surely apprehend that Smith's beliefs fail to be superassertible, even as they monitor Smith's non-superassertible second-order belief that his beliefs are superassertible. And the accessibility in principle of this awareness is not confined to the aliens. Put Smith's brain back in his body, show him the vat with a few other brains floating in it, and explain what was done to him; he will (or could, in a strong sense of 'could') revise his previous beliefs in light of his new informational state. That the evidence undermining his beliefs is not known to Smith, as Horgan tells the story, is irrelevant. Smith might survive for fifty years in the vat, holding beliefs that he would repudiate were he appropriately situated. But this is no more remarkable than the fact that I, holding many false beliefs, might be flattened by a meteor tomorrow. In either case, it would be wrong to characterize those false beliefs as superassertible, and thus counterexemplary of Wright's account, merely on the grounds that neither Smith nor I would personally get a chance to revise them. I need not immediately have evidence for a statement in order for it to be superassertible; SA preserves this feature of truth.

6. To expand a bit. The can-be claise is what crucially enables SA to allow for warrants not currently possessed but accessible nevertheless. And this confers upon SA the power to go proxy for truth in situations where a statement turns out to be true, though first uttered, e.g., as a guess, or in one's sleep. Now, if we take the modal component of the clause to mean current or practical possi- bility, then of course Horgan's thought experiment shows an extensional divergence between SA and truth. But if we take the operator to mean principled possibility, allowing SA to 'see' evidence beyond the practical accessibility of Smith, we seem to be assuming realism-the way things actually are, independent of the beliefs we may justifiedly hold-as the governing principle in question. We may round out this dilemma by noting that Wright's rejection of so-called Peircian truth (which he asso- ciates with Putnam) is based on the obscurity of the notion of 'epistemically ideal circumstances'; but, in this regard, the superiority of the notion of evidence 'in principle accessible' is not immediately obvious (1992, pp. 45-47). As I argue below, there are mitigating considerations; Horgan has not shown that a problematic realism must intrude.

This content downloaded from 188.72.126.41 on Sat, 14 Jun 2014 06:04:07 AMAll use subject to JSTOR Terms and Conditions

TRUTH, KNOWABILITY, AND NEUTRALITY 117

7. Actually, they advance two problems. I review only the first, since the second has a good deal in common with a criticism raised by Timothy Williamson (1996), to which Wright (1996) has re- sponded directly.

8. The following section was jointly written with Jason Holt. 9. See below, however, for the suggestion that Wright need not rely even on this rather plausible

claim. Accessibility may be seen as a property, not of information generally, but of particular states of information as characterized in a semantic model.

10. Various sections of this paper have benefited from the useful and generous comments of William Demopoulos, David DeVidi, Jason Holt, Paul Markwick, Crispin Wright, and a referee from Nous. Any remaining errors are my own. This work was supported by Social Sciences and Humanities Research Council of Canada Fellowship 752-95-1150.

References

Dummett, Michael. (1993) The Seas of Language. Oxford: Oxford University Press.

. (1991) The Logical Basis of Metaphysics. Cambridge, MA.: Harvard University Press. Gupta, Anil. (1993a) 'A Critique of Deflationism'. Philosophical Topics, Vol. 21, No. 2.

. (1993b) 'Minimalism'. Philosophical Perspectives, Vol.7. Horgan, Terence. (1995) 'Wright's Truth and Objectivity'. Nofls, Vol. 29, No. 1. Cambridge MA:

Blackwell.. Putnam, Hilary. (1981) Reason, Truth, and History. Cambridge: Cambridge University Press.

Shapiro, Stewart, and Taschek, William. (1996) 'Intuitionism, Pluralism, and Cognitive Command' The Journal of Philosophy, Vol. 93, No. 2.

Tennant, Neil. (1995) 'On Negation, Truth, and Warranted Assertibility' Analysis, Vol. 55, No. 2. Williamson, Timothy. (1996) 'Unreflective Realism'. Philosophy and Phenornenological Research.

Vol. 56, No. 4. Wright, Crispin. (1996) 'Reply to Critics' Philosophy and Phenomenological Research. Vol. 56,

No. 4. (1992) Truth and Objectivity. Cambridge MA: Harvard University Press. (1993) Realism, Meaning and Truth. 2nd ed. Cambridge MA: Blackwell.

This content downloaded from 188.72.126.41 on Sat, 14 Jun 2014 06:04:07 AMAll use subject to JSTOR Terms and Conditions