SORITES, ISSN 1135-1349

Issue #13 -- October 2001. Pp. 72-79 Identity, Analyticity and Epistemic Conservatism

Copyright © by SORITES and Paul Tomassi

Identity, Analyticity and Epistemic Conservatism

Paul Tomassi


Introduction

Rigorous, i.e. non-metaphorical, formulations of the analytic/synthetic distinction are thin on the ground. Moreover, many will sympathise with Quine's explanation of that fact.Foot note 5_1 Famously, Quine distinguishes two purported types of analytic statement:

Those of the first class, which may be called logically true, are typified by:

(1) No unmarried man is married.

...But there is also a second class of analytic statements, typified by:

(2) No bachelor is married.

The characteristic of such a statement is that it can be turned into a logical truth by putting synonyms for synonyms; thus (2) can be turned into (1) by putting `unmarried man' for its synonym `bachelor'.Foot note 5_2

It follows, in Quine's view, that analyticity is not properly defined in the absence of an adequate definition of cognitive synonymy. The attempt to construct such a definition invariably brings (vicious rather than virtuous) circularity in its wake. The correct explanation of that fact, Quine argues, is that the analytic/synthetic distinction is merely (one) empiricist dogma. Quine's critique of a second empiricist dogma (reductionism) reinforces his earlier conclusion: there is no such distinction to be drawn. Consequently, no statement, laws of logic included, is ultimately immune to revision in the light of experience.Foot note 5_3 Here I suggest that even if much of Quine's reasoning is cogent his conclusion is not warranted. Complete repudiation of the analytic/synthetic distinction is not the only conclusion which can validly be drawn. While I accept that there is a genuine problem of demarcation here I hope to show that the particular demarcation considered by Quine is the result of epistemic aggrandisement. The account of analyticity proposed may also cast new light on questions of warrant as regards essentialist accounts of identity-statements; these consequences are considered in the final section of the paper.

I

With no intended sleight of hand, the account of analyticity considered by Quine might be crystallised as a distinction between two kinds of identity-statement. The first of these is generally taken to exemplify the law of identity, i.e. the thought that everything is identical with itself:

Type I a = a

In `Two Dogmas of Empiricism', Quine's attitude to statements of this form is ambiguous. Laws of logic are revisable in the light of experience. More specifically, it is suggested that this fate might already have befallen the law of excluded middle.Foot note 5_4

While it is now clear that the latter claim is erroneous (at least for the reasons Quine gives), by the former claim, Quine would appear committed to the potential revocability of the law of identity on empirical grounds. Current physical theory may offer no basis for any such revision but science is progressive and, in the longer run, who knows. On the other hand, Quine's case against the first dogma is not made modulo logical truths but rather on the basis of a second type of purportedly analytic statement:

Type II a = bwhere and only where (a, b) is a synonym-pair.

As Quine puts it:

... the major difficulty lies not in the first class of analytic statements, the logical truths, but rather in the second class, which depends on the notion of synonymy.Foot note 5_5

Given the recalcitrance of cognitive synonymy as regards (non-circular) definition, Quine is right to claim that no analytic/synthetic distinction has yet been cogently drawn. He may, however, be wrong to conclude that there is no such distinction. Quine's basic concern is how scientifically-informed belief-sets warp in experience's light. Given that belief-sets warp under guiding principles which are pragmatic, conservative and simplicity-seeking, it is difficult to see what could possibly motivate rejection of the law of identity, and more difficult still, to imagine how any belief-set could actually be formed (or adjusted) in ways which did not keep faith with the law of identity.

The matter will not be settled by consensus but, I take it, few if any will want to dispute the analyticity of statements of type I. I further assume that the reason underpinning that reaction is constituted by the form of the sentence itself. Type I statements are epistemically reassuring in the sense that we generally grasp the truth of any such sentence as soon as we grasp the meaning of the symbols which compose it. The source of that reassurance consists in our understanding of the identity sign and our recognition that the same term fills both gaps in the identity relation. Indeed, in this we have a very clear definition of type I identities, i.e. identity-statements in which one and the same term plugs both gaps in the relation. Thus, it is very natural to assert that every such sentence is true in virtue of the meaning of the symbols which compose it, i.e. at first blush, at least, every sentence of that form seems guaranteed to be true. Moreover, it would appear that the content of any such statement can be paired down very considerably while retaining (intuitively, at least) self-evidence. Even here, however, there are limits. For example, the poverty of current flubjub-theory notwithstanding, one might be tempted to accept the truth of the following statement:

Flubjubs are flubjubs.

Given the logical form of the statement and the validity of existential generalisation, one might then infer:

Flubjubs exist.

Given that there are no flubjubs, such reasoning is invalid. Hence, we may require extraneous reassurance in some form that the name in question is not vacuous; in other words, that the concept in question has an extension. Given such minimal extraneous information, type I statements are robustly epistemically reassuring, i.e. Type I statements as defined which satisfy the reference-condition wear their analyticity on their faces; as we might put it, such statements exhibit orthographic identity.

Whether or not Quine ultimately shares with his opponents recognition of the epistemically reassuring character of type I statements, it is clear that, in Quine's view, the real trouble kicks in with the attempt to extend analyticity to type II statements. The key assumption here is just that there is a class of such sentences. However, if there is any such class, no member of that class exhibits orthographic identity. Type II identity-statements do not wear their analyticity on their faces. Indeed, the logical form of a type II statement is generally definitive of syntheticity rather than analyticity. Further, if the claim is that a certain class of such statements has this form but is nonetheless properly understood as analytic rather than synthetic, the key question is exactly which class? In other words, which substitution-instances are properly included in the analytic `box' and which excluded?

The answer seems obvious. Type II statements of the form a = b are analytic where and only where (a, b) is a synonym-pair. But now Quine's points as regards the recalcitrance of cognitive synonymy all become pressing. For singular terms in the relation, for example, it is now not enough simply to know that both terms refer. Equally, for general terms, it is not enough to know that the relevant extensions are non-empty or even that such extensions are identical. Thus, there is not only a challenge to clearly pin down type II analytic truths but also a fundamental epistemic contrast between the two purported forms of analytic truth. The challenge here is not adequately met by pointing to one or two statements of the relevant form which, most would agree, assert true identities. Rather, what is required is clear demarcation of the relevant class.

Type II identity-statements strongly contrast with those of type I. The former, unlike the latter, clearly are not orthographic identities. We noted above that recognition of the truth of a type I statement requires only that we are aware that the statement has type I form and that the first term in the relation refers. In contrast, knowing that a statement is of type II form and that the first term in that relation refers is not a state of information adequate to establishing the truth of any type II identity. Moreover, if the first term in a type I identity refers it is certain that the second term also refers. Again, however, the referential success of the first term in a type II identity offers no guarantee of reference as regards the second term. Therefore, there are clear epistemic differences between the two kinds of identity-statement. Given the lack of a clear demarcation of the purported class of type II analytic truths together with the fact that there is no reassuring orthographic identity in the case of any type II statement, the question naturally arises: why should we accept that there is any such class? Thus Quine's boundary problem may indeed be one of aggrandisement, i.e. the notion of analyticity has been stretched beyond the genuinely self-evident. Traditional attempts to draw the distinction fail with Quine's attempt precisely because these strategies try to draw the line between the analytic and the synthetic in the wrong place.

II

Given the foregoing, we may rationally accept Quine's critique of type II `analytic' statements while rejecting his conclusion, i.e. we may choose to warp our belief-set in the light of our experience rather differently. Given the Duhem-Quine thesis, our hand is not forced to any particular local fix and thus the move may even be Quinean in spirit. To be quite clear, the alternate conclusion is a position which confines analyticity to orthographic identity; given the background to this debate, a position aptly entitled epistemic conservatism, i.e. conservatism modulo analyticity. In Quinean spirit still, we ought to keep an eye to the knock-on effects of such an adjustment to, and in, the bigger picture. What does philosophical cost-benefit analysis reveal here?

Before considering that question, it is worth forestalling one obvious objection. Thus far, the only restriction imposed upon orthographic identities over and above logical form is the reference-condition, i.e. the term in that relation must be non-vacuous. Those who sympathise with Kripkean accounts of the semantics of names and natural kind terms in terms of rigid designation, for example, may rest easy given that condition alone.Foot note 5_6 While I cannot fully make the case out here, given the work of Gareth Evans and others, it seems likely that some descriptive content (however minimal, e.g. sortal) must attach to names.Foot note 5_7 However, the move to a (weakly) descriptivist, anti-Kripkean, position does nothing to undermine the key epistemic differences between the two types of identity-statement highlighted here. Assuming that there is some such content, that content will attach to the terms in any identity-statement. As regards orthographic identities, we can be equally certain ex hypothesi that exactly the same content attaches to each term in every such case. By the same assumption, there is no guarantee whatsoever that the same descriptive content will attach to the terms of any non-orthographic identity. Thus, a descriptivist stance preserves key epistemic differences between the two types of identity-statement.Foot note 5_8

To sum up: the account of analyticity proposed here confines the analytic to orthographic identity-statements satisfying the reference condition. Clearly, every such identity is necessarily true in a highly intuitive sense. Moreover, that intuition can be made more precise. In every possible world in which the first term in an orthographic identity-statement refers that statement is guaranteed true. Therefore, there is no world in which any such orthographic identity is false. Given the familiar interpretations of the notions of possibility and negation, it follows that true orthographic identities are true necessarily, i.e. no such orthographic identity is false in any possible world. Further, while it is clear that the epistemic character of type I identities is underwritten merely by the form of the statement and satisfaction of the reference-condition, it is equally clear that the same basis is inadequate as regards establishing the necessity of any type II identity-statement. Here, merely grasping the form of the statement and knowing that the first term refers is plainly insufficient to establish the truth, let alone the necessity, of any type II identity-statement. There remain, therefore, important epistemic differences between the two kinds of identity-statement precisely as regards warrant for analyticity and, by the same token, necessity.

III

The account of analyticity proposed here not only allows a clear (if modest) analytic/synthetic distinction to be drawn but also provides a useful framework within which questions of warrant as regards essentialist accounts of identity-statements can usefully be considered. As is very well known, certain authors (chief among them Saul Kripke and Hilary Putnam) have argued for an account of a certain class of type II identity-statements which, if true, are true necessarily.Foot note 5_9 Famously, the terms in such identity-statements are the prime candidates for rigid designation, i.e. names and natural kind terms. Of course, such identity-statements are not canvassed as simply analytic in the sense proposed here. Rather, in the light of identities of this type, we are urged to distinguish epistemic necessity from metaphysical necessity. Identities of the relevant kind are metaphysical necessities known a posteriori.Foot note 5_10 It follows that there is, indeed, a class of properly type II identities which, if true, are necessarily true. Given the foregoing discussion, knowledge of the logical form of any such statement and the fact that the first term in any such statement refers is, just as such, insufficient to warrant recognition of the necessity of that statement. In these terms, the key question is exactly how that state of information is supplemented to constitute warrant? To that end, I separate cases. Consider, for example, Kripke's claim to the necessity of the statement:

The lectern is not made of ice.

Were the truth of the statement established by fiat, there could be no sense in which our knowledge (if indeed we have such knowledge) is the result of empirical discovery. That this is not Kripke's position is quite clear:

What we know is that first, lecterns usually are not made of ice, they are usually made of wood. This looks like wood. It does not feel cold and it probably would if it were made of ice. Therefore, I conclude, probably this is not made of ice. Here my entire judgement is a posteriori.Foot note 5_11

Here, the key point is that to establish the necessity of the statement in question is to establish that it is not possible that the statement be false, i.e. that there is no possible world in which the statement is false. But what in the passage quoted assures us of that fact? The statement is highly probable. But any degree of probability less than 1 is plainly inadequate to Kripke's task. The statement's being .99 probable and yet being false is perfectly consistent. It follows that there is a possible world in which the negation of the statement holds; whence, then, its necessity? Kripke outlines an argument-form which, under substitution, is intended to establish the necessity of the statement as conclusion:

P P

P;

Therefore

P

The conclusion -- `P' -- is that it is necessary that the table not be made of iceFoot note 5_12

To be clear, the validity of the argument-form outlined is not in question here. However, to draw the conclusion soundly requires establishing the truth of all the premises. Thus, the key question is the nature of the warrant for P itself. Certainly, given P, and given that if P then necessarily P, necessarily P. But, again, the truth of P is by no means conclusively established by Kripke's argument. In fairness to the position in question, and lest anything significant should hang merely on one particular example, consider Putnam's candidate identity:

Water is H2O

Putnam is clearly aware that establishing the necessity of the identity-statement requires ruling out the possible non-obtaining of this truth (if it is one) in some possible world:

... we can perfectly well imagine having experiences that would convince us (and that would make it rational to believe that) water isn't H2O. In that sense, it is conceivable that water isn't H2O. It is conceivable but it isn't possible! Conceivability is no proof of possibility.Foot note 5_13

Again, were the truth of this identity-statement established by fiat, there could be no sense in which such knowledge is the result of empirical discovery. But Putnam is equally clear that the truth of the identity-statement in question is an empirical matter: `Once we have discovered that water (in the actual world) is H2O, nothing counts as a possible world in which water isn't H2O.'Foot note 5_14 Given that the discovery in question is empirical in character it remains to ask what it is that ultimately assures us of the truth of the identity-statement in question to the extent that we may confidently assert the necessity of that statement?

Undoubtedly, that water is H2O is currently an uncontroversial commonplace of a mature science. However, it does not follow that the truth of that statement is thereby guaranteed. Scientific statements are, generally, defeasible statements. Therefore, it remains possible that any such statement will turn out to be false. In the bigger picture, Larry Laudan's `confutations', may show that the appropriate general meta-induction is pessimistic rather than optimistic.Foot note 5_15 Certainly, chemistry is no exception to the rule. The history of science does not bear out the thought that the percentage composition of familiar compounds within established chemical theories is immune to revision. Indeed, just that point was effectively exploited by Thomas Kuhn precisely in order to establish the revocability not only of scientific theory but of scientific data.Foot note 5_16 Certain such changes could force us to revise our view of the claim that water is H2O and any such revision would undermine the plausibility of Putnam's candidate type II identity.

I do not claim here that water is likely to turn out not to be H2O. Rather, my point is that any appeal to empirical knowledge as regards bridging the justification-gap between type I and type II necessities faces the apparently recalcitrant problem of the principled defeasibility of scientific claims. If the Kripke-Putnam position is underpinned by the thought that there are points at which scientific hypotheses pass from being hypothetical to being written in metaphysical stone then we must ask at precisely which points that transition is accomplished? In other words, we should recognise here a surrogate for Quine's original challenge to demarcate that class of type II identities which are properly understood to be necessary rather than contingent. Moreover, in this arena, any positive answer to the question would also tell us just when and where science could stop.

Two further important points are worth emphasis. We may very well agree that if a true type II identity is to be had then there is a valid inference to the necessity of that identity-statement. But the question remains: what assures us of the truth of the antecedent and thereby licenses modus ponens? Most sentences of this form are synthetic and license no such inference. At root, the problem is epistemic: under exactly which kinds of circumstance are we warranted in allowing that we have grasped the truth of any such identity-statement? Further, to the claim that the critical discussion of Kripke and Putnam in this section simply conflates epistemic necessity with metaphysical necessity, the rejoinder is that the candidate type II identity-statements considered here are supposed by both authors to motivate just that bifurcation of modalities. Unless we can recognise that those statements do enjoy the special status claimed on their behalf the motivation to draw the modal distinction would appear to be lost.Foot note 5_17

Ultimately, the soundness of the particular arguments presented here is less important than the fact that kinds of doubt can be raised as regards warrant for type II identity-statements to which type I identity-statements are immune. Given the fundamental epistemic differences between the two, type II identity-statements will always require special pleading re necessity. I have argued here that, in just this respect, the Kripke-Putnam case remains not proven.



Paul Tomassi
<p.tomassi@abdn.ac.uk>