Analyticity and Nondescriptionality[*]
Barbara Abbott
Michigan State University
One of the widely accepted and quite influential conclusions of modern Anglo-American philosophy is that there is no sharp distinction between analytic truths and statements that are true only [by] virtue of the facts; what had been called analytic truths in earlier work, it is alleged, are simply expressions of deeply held belief. This conclusion seems quite erroneous. There is no fact about the world that I could discover that would convince me that you persuaded John to go to college even though he never intended or decided to go to college; nor is there any fact of experience even relevant to the judgment that you failed to persuade him if he never intended or decided to go to college. The relation between [persuade] and [intend] or [decide][1] is one of conceptual structure, independent of experience--though experience is necessary to determine which labels a particular language uses for the concepts that enter into such relations. The philosophical debate over these matters has been misleading because it has focused on very simple examples, examples involving words that lack the relational structure of such terms as chase and persuade. Thus there is much debate over whether the statement "cats are animals" is a truth of meaning or of fact (if we discovered that what we call cats are really robots controlled by Martians, would the sentence "Cats are animals" now be considered false, or would we conclude that what we have called cats are not really cats?). In such cases a decision is not easy to reach, but in others it seems quite straightforward.
--Chomsky 1988, 33-34 (bolding added).
It may not be clear what Chomsky means by 'relational structure' in the bolded sentence of the quote above. Chase and persuade differ from cat in being verbs and in expressing relations, and these facts might be relevant to a semantic difference of the type Chomsky is claiming. However the sentence immediately preceding speaks of a 'relation between persuadeand intendor decide' which is 'one of conceptual structure', strongly suggesting that the phrase 'relational structure' in the bolded sentence refers to semantic structure, a structure of semantic components or meaning postulates or the like. In the first case we might postulate a semantic component intend, of which the meaning of the verb intend is exhaustively composed. We could then analyze the meaning of decide as come to intend (where come to represents a change-of-state component), and persuade (in one of its senses) as cause to come to intend. In the second case, where appeal is made to meaning postulates, our constraints on appropriate models for English would include postulates similar to those in (1) (ignoring details concerning times of eventualities, etc.).

(1)a. [] " xyz [persuade (x, y, z) -> decide (y, z)]

    b. [] " xy [decide (x, y) -> intend (x, y)]

On the other hand for the word cat the issue is exactly whether or not it can be adequately decomposed into components, one of which is animal, or whether the meaning postulate in (2) belongs to the semantics of English.

(2)     [] " x [cat (x) -> animal (x)]

So we have an evident contrast between persuade, which seems clearly to enjoy semantic relations with other words of English, whether via semantic components or meaning postulates, and cat, for which the case is not so clear.

The first two thirds of Quine's classic paper 'Two Dogmas of Empiricism' explores various explicational cul-de-sacs (meaning, synonymy, definition, 'semantical rules') en route to his conclusion that analyticity is not a notion which can be clarified to his satisfaction. The examples most often chosen by Quine to illustrate this futile path use the word bachelor, e.g. No bachelor is married (p. 23), All and only bachelors are unmarried men (p. 29). Although not a relational term or a verb, bachelor has often been the prototype of words with a comfortably tidy analysis into semantic components (human, adult, male, not married) or for which uncontroversial meaning postulates, as in (3), can easily be devised.

(3)     [] " x [bachelor (x) -> ¬married (x)]

Given that 'Two Dogmas of Empiricism' is universally acknowledged as the most forceful statement opposing the analytic/synthetic distinction, the suggestion of Chomsky's bolded in the opening quotation would seem to be simply wrong. However there is more to the matter than this.

The main issue is whether there are analytic sentences or not. In his classic paper Quine pursued three lines of argument to the effect that there are not. The first line was that alluded to above, in which Quine tracked successive analyses of the notion of analyticity, only to observe that each time the analysis crucially included notions like meaning, synonymy, or necessary truth, which were themselves equally problematic from his point of view. This is not a demonstrative argument against the existence of analytic sentences, but only an argument that the concept of analyticity is as unclear as the other concepts. One who found the other concepts relatively clear would not be persuaded by this line.

For a moment we will skip from Quine's first argument to his third, which requires a bit of background. Reductionism, part of the then widely received verificationist theory of meaning, was the other dogma of empiricism which Quine was determined to defeat in 'Two Dogmas'. Reductionism is 'the belief that each meaningful statement is equivalent to some logical construct upon terms which refer to immediate experience' (Quine 1953, 20). In the last two sections of his paper Quine brought the two dogmas together and asserted that they 'are, indeed, at root identical' (p. 41). The idea is that both dogmas require that the truth of statements be divisible into a linguistic portion (the meaning, which states the experiences necessary to confirm or disconfirm the statement) and a factual portion (the confirmatory experiences). But this analysis of a single statement in isolation from all others was inconsistent with Quine's holistic view of knowledge and belief -- his confirmation holism; and assertion of the latter constituted Quine's third line of argument against analyticity. '[O]ur statements about the external world face the tribunal of sense experience not individually but only as a corporate body' (Quine 1953, 41).

This argument against analyticity, unlike the first one, is intended to be demonstrative. However, also unlike the first one, this argument depends on a particular view of what meaning is. Here we see that Quine is not opposed to the essence of verificationism, the idea that meaningfulness consists in being subject to confirmation or disconfirmation on the basis of experience, but only its application to language sentence by sentence. Indeed, a broadly verificationist view of meaning is necessary to making this argument at all, since only then would we get the necessary conflict between confirmation holism and sentence meaning. One opposed to verificationism in all its forms would not be convinced by this argument.[2]

In laying out the first and third lines of argument against analyticity Quine moves from the point of view of a myopic bookkeeper to that of God. Wedged between the two, and visible only for a brief moment, we see Quine as an ordinary human being.

I do not know whether the statement 'Everything green is extended' is analytic. Now does my indecision over this example really betray an incomplete understanding, an incomplete grasp of the "meanings", of 'green' and 'extended'? I think not. The trouble is not with 'green' or 'extended', but with 'analytic'. (Quine 1953, 32) (This is the second argument.) Grice & Strawson, in their 1956 defense of analyticity, take Quine to be indecisive not specifically concerning the analyticity of the statement in question, but more generally about its truth, wondering possibly whether a point of green light should be taken to be extended (Grice & Strawson 1956, 90). But this seems wrong. Note that Quine says the trouble is not with 'green' or 'extended', which it would be if Grice & Strawson were correct in their construal. Instead, I think Quine assumes that the statement is true and is genuinely unclear about its analyticity.

I believe that it is significant that when Quine slips into the position of the ordinary human being who is genuinely puzzled about meaning, a position which could perhaps have the most force with the most readers, he switches examples. Green is a word that seems to be more like cat than it is like bachelor or persuade in the relevant respects. For one thing, it is an observational term, whose denotation would typically be learned by ostension rather than explanation. So it seems that Chomsky may be right, that attention to the difficult cases may have carried more weight in the argument against analyticity than it should have.

If we could give a good account of Quine's green example perhaps we could (hypothetically of course) convince him that there really are analytic sentences.[3] That is too much to attempt for this paper, but a few things can be mentioned briefly. One is that color words were included among the natural kind terms argued by Kripke (1972, 1980) to be nondescriptional -- that is to have their reference secured without the mediation of something like a Fregean sense.[4] Such words -- green, cat, gold -- denote things of a kind that exist in nature but whose real essence has historically been unknown to speakers of languages. It happens to be typical of such terms that some properties may hold necessarily of their referents, without nevertheless constituting a part of their meaning. Thus Cats are animals was argued by Kripke to state a metaphysically necessary truth, but not an analytic one.[5] One who thought that the notions of necessity and analyticity were roughly coextensive, such as the author of Quine 1953 appears to be,[6] might well be puzzled over such examples.

Everything green is extended strikes my intuitions as being like Cats are animals in stating a necessary synthetic truth. On the other hand it is difficult to put through the same kinds of thought experiments that Kripke used for Cats are animals; it is easier to imagine discovering that cats are actually mechanical creatures than discovering that the green things which we thought were extended actually are not. Whatever the upshot of these cases (which I have argued elsewhere are relatively rare (Abbott 1989)[7]), Chomsky's point remains. Hard cases make bad law.


[*] I am very much indebted to Gene Cline, Rich Hall, Larry Hauser, Myles McNally, Paul Rusnock, and Carol Slater for reading an earlier draft and providing me with many helpful comments which have substantially improved this paper.

[1] Perhaps I should explain these square brackets. This passage which I am quoting comes from lectures originally presented to a Spanish-speaking audience. I have replaced Chomsky's Spanish language examples with the corresponding English words. The square brackets above in the third line of the quote are because of a misprint in the original.

[2] In getting clearer on Quine's argument here I am indebted to Chapter 2 of Fodor & Lepore 1992. Boghossian 1997 has also been a useful source on Quine and analyticity.

[3] This achievement is hypothetical for several reasons, one of which is the fact that Quine has already admitted the existence of analytic sentences (see e.g. Quine 1974, 78-80).

[4] The term 'nondescriptional' in this sense was introduced in Salmon 1981.

[5] This example (Cats are animals) had appeared earlier in Putnam 1962 but Putnam's views on it are slightly different from Kripke's. See the discussion in Kripke 1972, 1980, 122-6.

[6] But cf. Quine's careful statements concerning the relation between analyticity and necessity on pp. 29-30 of 'Two dogmas'. I am grateful to Rich Hall for pointing these out to me.

[7] There is widespread disagreement over the extent of nondescriptionality in the lexicon of a typical language. This issue bears a close relation to the question of the existence of analytic sentences, and one which needs to be looked into.



Abbott, Barbara. 1989. Nondescriptionality and natural kind terms. Linguistics and Philosophy 12:3, 269-92.

Boghossian, Paul Artin. 1997. Analyticity. In Bob Hale and Crispin Wright, eds., A companion to the philosophy of language, Oxford: Blackwell, 331-68.

Chomsky, Noam. 1988. Language and problems of knowledge: The Managua lectures. Cambridge, MA: MIT Press.

Fodor, Jerry & Ernest Lepore. 1992. Holism: A shopper's guide. Oxford: Blackwell.

Grice, H. Paul & P.F. Strawson. 1956. In defense of a dogma. The Philosophical Review 65:2, 141-58. Reference is to the reprint in Jay F. Rosenberg & Charles Travis, eds., 1971, Readings in the philosophy of language, Englewood Cliffs, NJ: Prentice-Hall, 81-94.

Kripke, Saul A. 1972, 1980. Naming and necessity. Cambridge, MA: Harvard University Press.

Putnam, Hilary. 1962. It ain't necessarily so. Journal of Philosophy 59, 658-71.

Quine, W.V. 1953. Two dogmas of empiricism. In From a logical point of view, Cambridge, MA: Harvard University Press, 20-46.

Quine, W.V. 1974. The roots of reference. La Salle, IL: Open Court.

Salmon, Nathan U. 1981. Reference and essence. Princeton: Princeton University Press.