under the less provocative guise of “conceptual truth.” The terms “analytic” and “conceptual” will henceforth be used interchangeably.
Quine’s arguments are generally found much less compelling than they once appeared. Although he may succeed in showing that “analytic” is caught in a circle with other semantic terms, such as “synonymous,” he does not adequately motivate his jump from that point to the conclusion that the terms in the circle all lack scientific respectability, as opposed to the contrary conclusion that they all have it. Given any science, someone may insist that it define its terms, and the terms used to define them, and so on until it is driven round in a circle. By itself, that hardly demonstrates the illegitimacy of the science. Every discipline must use undefined terms somewhere or other. “Two Dogmas of Empiricism” does not explain why we should regard the undefined terms of semantics as worse off than the undefined terms of other disciplines, except by dogmatic charges of unclarity. After all, semantics is now a thriving branch of empirical linguistics. It is not to be trashed without very good reason.2
Some terms may be so unclear by ordinary working standards that no circle of definitions will render them scientifically useful. But semantic terms are not like that. By ordinary working standards, the word “synonymous” is quite clear enough to be useful. Although it is not perfectly precise – surely it has borderline cases – its degree of vagueness seems no worse than that of undefined terms in many other sciences. When clarification is needed in some specific respect, it can be achieved by stipulation or otherwise, as elsewhere in science. Indeed, few contemporary philosophers feel special qualms in using the term “synonymous.” Thus any objection they have to “analytic” can hardly be based on Quine”s arguments, since his only objection to defining “analytic” in terms of “synonymous” is to the use of “synonymous” (1951: 24, 35).
The feeling remains that “analytic,” unlike “synonymous,” carries obsolescent philosophical baggage. For “analytic,” unlike “synonymous,” was once a central term in philosophical theorizing, notably in the work of logical positivists, such as Carnap, and of postwar linguistic philosophers, such as Strawson. The reason why it cannot recover that position lies not in Quine’s critique, which no longer seems compelling, but rather in Kripke”s widely accepted clarification of the differences between analyticity, apriority and necessity. Kripke did not deny that there is a boundary between the analytic and the synthetic; he merely distinguished it from other boundaries, such as the epistemological boundary between the a priori and the a posteriori and the metaphysical boundary between the necessary and the contingent (Kripke 1980: 39). He stipulated that “analytic” entails both “a priori” and “necessary.” Since he argued that neither of “a priori” and “necessary” entails the other, he was committed to denying that either of them entails “analytic” (by the transitivity of entailment).3 Thus “analytic” does neither the purely epistemological work of “a priori” nor the purely metaphysical work of “necessary.” Its current role inevitably looks less central than the one it occupied when “a priori” and “necessary” were treated as pretty much interchangeable and “analytic” was taken to do the work of both. But that does not yet imply that no work remains for it to do.
If we try to sort sentences as “analytic” or “synthetic” in the manner of chicken-sexers, we can usually achieve a rough consensus. Of course borderline cases will occur, but so they do for virtually every distinction worth making: perfect precision is an unreasonable demand. The issue is what theoretical significance, if any, attaches to the rough boundary thus drawn. Even if “analytic” is defined in terms of “synonymous” and other expressions under better control than “analytic,” we should not assume without checking that it has any of the consequences sometimes associated with it. In particular, we should not assume that analytic truths are insubstantial in any further sense.
Nothing in this book challenges the legitimacy of familiar semantic terms such as “synonymous.” They will be used without apology, and they permit various senses of “analytic” to be defined. But none of them makes sense of the idea that analytic truths are less substantial than synthetic ones, or that core philosophical truths are less substantial than the truths of most other disciplines. There is something robust about “Two Dogmas of Empiricism”: insights remain even when its skepticism towards meaning is stripped away.
On some conceptions, analytic sentences are true simply in virtue of their meaning, and analytic thoughts simply in virtue of their constituent concepts. They impose no constraint on the world, not even on that part of it which consists of words and concepts. That is why it is unnecessary to get up out of one’s armchair to investigate whether such a constraint is met. Analytic truths are less substantial than synthetic ones because the latter do impose constraints on the world, which it may or may not meet. This is another way of putting the idea that analytic truths are true in virtue of meaning alone while synthetic truths are true in virtue of a combination of meaning and fact, for if analytic truths did impose constraints on the world, they would be true partly in virtue of the fact that the world met those constraints, and so not true in virtue of meaning alone. Call such conceptions of analyticity metaphysical. Other conceptions dispense with the idea of truth in virtue of meaning, and treat analyticity as a privileged status in respect of knowledge or justification which a sentence or thought has in virtue of the conditions for understanding its constituent words or possessing its constituent concepts. Although the privileged truths impose constraints on the world, the task of checking that they are met is somehow less substantial than for other truths, for those who understand the relevant words or possess the relevant concepts. Call such conceptions of analyticity epistemological.4
This chapter examines a variety of attempts to develop a metaphysical account of analyticity. Some depend on misconceptions about meaning or truth. Others yield intelligible notions of analyticity, by watering down the traditional account to a point where it loses most of its usually supposed implications. They provide no reason to regard analytic truths as in any way insubstantial.5 Even if core philosophical truths are analytic in such a sense, that does not explain how we can know or justifiably believe them.6 At best it reduces the problem to the epistemology of another class of truths, such as necessary truths or logical truths. The next chapter will examine attempts to develop an epistemological account of analyticity, also with negative results. The overall upshot is that philosophical truths are analytic at most in senses too weak to be of much explanatory value or to justify conceiving contemporary philosophy in terms of a linguistic or conceptual turn.
The conclusion is not best put by calling purportedly analytic truths “substantial,” because in this context the term “substantial” is hopelessly vague. Rather, appeals in epistemology to a metaphysical conception of analyticity tend to rely on a picture of analytic truths as imposing no genuine constraint on the world, in order to explain the supposed fact that knowing them poses no serious cognitive challenge. If that account could be made good, it would provide a useful sense for “insubstantial,” which would refer to the pictured property, epistemological not in its nature but in its explanatory power. Substantial truths would be the ones that lacked this property. But the account cannot be made good. The metaphysical picture cannot be filled in so as to have the required explanatory power in epistemology. Thus “substantial” and “insubstantial” are not provided with useful senses. The negation of a picture is not itself a picture. That is a problem for appeals to metaphysical analyticity, not for the present critique.
2
The distinction between analytic truth and synthetic truth does not distinguish different senses of “true”: analytic and synthetic truths are true in the very same sense of “true.” That should be obvious. Nevertheless, it is hard to reconcile with what many logical positivists, Wittgensteinians and others have said about analytic truths. For they have described them as stipulations, implicit definitions (partial or complete), disguised rules of grammar and the like. On such a conception, enunciating an analytic truth is not stating a fact but something more like fixing or recalling
