Why the Distinction between Analytic and Synthetic Statements?

Abstract

The distinction has occasioned a long controversy between Carnap and W.V. Quine. The latter distinguishes two sorts of analytic statements: the logical truths, characterized by their remaining true under all reinterpretations of the descriptive terms; and the statements, which reduce to logical truths with the help of definitions or by substitution of synonyms for synonyms. In “Two Dogmas of Empiricism”, he directs his criticism mainly against the latter arguing that the explications so far provided move in a circle, since, in the end, the concept of synonymy is explained in terms of analyticity. He reaches the conclusion that the only form of synonymy not exhibiting this deficiency is that introduced by explicit conventions. But according to him, this procedure is of no avail, because the philosophically interesting content of the general term “analytic in L” (with variable L) cannot be grasped with its help. Carnap’s suggestion of defining a new basic concept “analytic in L0” does not solve the problem since we could just as well designate this with a neutral word, say “K”, thus avoiding the illusion of an explanation. The attempt to define synonymy with the help of semantic rules fails in a similar way: it accomplishes nothing more than to place certain statements under the rubric “Semantic rules” — a notion which is no clearer than the explicandum. Carnap replies that his definition of “analytically true” is given with regard to artificial or formalized languages for which we can determine syntactic and semantic rules in such a way that the concept of truth, in Tarski’s sense, can be applied to them. Analytic statements other than logical truths will thus be distinguished by means of so-called meaning postulates which serve to make the truth-value of such statements independent of empirical matters. I share Carnap’s astonishment that Quine finds the term “semantic rule” unintelligible while formulating no complaint against similar notions such as that of a syntactically well-formed formula, or that of an axiom etc. It appears that he cannot be satisfied with rules, because he expects a theory which would give us an empirical, i.e., a behavioristic, criterion of analyticity for natural languages. His chief dissatisfaction seems to come from the fact that no clear explicandum has been put forward — in contrast, e.g., with the notion of truth, where we have a sufficiently clear pre-theoretical concept. Thus, according to him, there is nothing that could be clarified by an explication.