According to Quine, abandoning these dogmas will pave the way for an appropriately pragmatic and "enlightened" epistemology.

The overall question: How can we make sense of a distinct class of statements that are true (or false) by dint of meaning?

The direction this is going: If we can’t make satisfactory sense of analytically true sentences, then the very notion of meaning itself must be called into question.

Preliminary attempts to understand analyticity:

A. Hume: Analytically true statements are those whose negations are self-contradictory.

B. Kant: Analytic truths are those whose "predicates contain their subjects."

Quine: -unacceptably metaphorical- What does "containment" amount to in this context?

C. Leibniz: Analytic statements are those that are true "in all possible worlds."

D. Analytic truths are those which are true ‘by definition"

A. The general Idea: Analytic truths are those which are converted into logical truths when we substitute synonymous terms for one another.

Comments:

(1) This idea presupposes that we can make antecedent sense of the notion of a logical truth, which would seem to require that we are already able to grasp the meanings of logical terms. Quine does not dwell upon this point (but perhaps he should have, for I’d suggest that it would allow him to see his way through some of the later dilemmas he constructs).

(2) This approach also requires us to have an antecedent grasp of the notion of synonymy - a sense of when two terms or predicates mean the same thing. It is the understanding of this notion to which Quine now turns.

B. Understanding Synonymy: Quine now asks, "What then is the meaning of ‘meaning’?" When can we say that two terms or predicates have the same or different meanings?

(i) A non-starter (psychologism): Two terms have the same meaning when they are associated with the same idea or mental item (also known as "the myth of the museum").

(ii) Another dead-end: Synonymous terms are those that have the same reference oo extension.

Frege-type counterexamples: Cases in which two co-referring terms (e.g. The first Postmaster General, the inventor of the bifocals or the morning star and the evening star) or co-extensive predicates (e.g. "has a heart", "has a kidney") clearly have different meanings.

Same thing about names (Superman, Clark Kent): Different names for the same think don’t seem synonymous or "cognitive value" (Frege); It seems then as if there must be something more to the meaning of a name than mere reference. (We’ll talk more about this when we discuss Kripke).

(iii) Perhaps synonymous terms are those which are intersubstitutable salva veritate (saving truth-value):

Here Quine asks, in what contexts does this principle apply?

- Surely it doesn’t extend to cases in which we are talking about the words involved:

"’Bachelor’ begins with a ‘B.’"/ "An unmarred man begins with a ‘B’."

-But it can’t be limited to just "extensional" context, for once again, that would seem to imply that merely co-referring terms are synonymous.

Unlike merely co-referential or co-extensive terms, synonymous terms must then be those that preserve truth value in certain non-extensional contexts, such as modal contexts. In other words, one might be tempted to draw the distinction between the genuinely synonymous and merely co-extensive or co-referential in terms of necessity or possibility (i.e., while all bachelors are of necessity unmarried man, it is somehow possible for a creature with a heart not to be a creature with a kidney). But once again, that threaten to rest our account of synonymy upon some notion of necessity or possibility, which itself would seem to rest upon some notion of analyticity. Once more, we seem to have made little genuine progress in our attempt to understand analyticity via synonymy.

A. Here’s another try at unpacking the notion of analyticity: First suppose that all sentences carry with them their own conditions of verification (or confirmation). Then we might understand as analytically true any statement that could be confirmed by any experience whatsoever (or possibly, couldn’t be disconfirmed by any experience).

Synonymous terms might then also be understood as those with equivalent contributions to the verification conditions of the sentences of which they form parts.

B. The trouble Quine finds with this idea is that individual sentences simply don’t have intrinsic confirmation conditions, independent of the verification conditions we understand other sentences to have. As Quine puts it, sentences face "the tribunal of experience," not one-by-one, but rather as a corporate whole. The minimal unit of meaning is thus not the individual sentence (though it could be a minimal unit of truth), but rather a complete battery or constellation of sentences (perhaps one might follow Quine and call it a "theory"-- though I wouldn’t).

Comments:

1. This is mostly just a picture of a position, not really an argument for it.

If the notion of synonymy is ultimately suspect, then what of the idea that "red" (in English) is synonymous with "rot" (in German), or that "Schnee ist weiss." (in German) means that "Snow is white." (in English)?

Quine concludes (in his book, Word and Object,1960) that his failure to find a satisfactory account of synonymy and analytic truth renders suspect the very idea that there are determinate translation schemes between languages. In principle, for any language, one could come up with incompatible, yet empirically equivalent, translation schemes for that language.

-Example of "Gavagai"

- This is the thesis that so many have found so exciting and have thoroughly abused.

Final comment: Note that Quine’s account here mostly takes on the form of an assertion rather than an argument. He hasn’t shown that there couldn’t be any satisfactory account of analytycity so much as show that he hasn’t found such an account. Consider how Sellars might respond to this conclusion (think about the section in EPM on the logic of ‘means’ and the functions of linguistic items).