On the semantic status of mathematical truth (Q1061112)

Are mathematical truths analytic or synthetic? - The question has been discussed by philosophers of mathematics since Kant at least. \textit{E. Snapper} [in the article reviewed above] argues that any mathematical theorem worthy of the name is synthetic. The authors have reservations concerning his argument as well as the question. Three dubious presuppositions are mentioned: that the notion of analytic truth is sufficiently well defined; that the truths of mathematics are exclusively analytic or exclusively synthetic; and that the set of mathematical statements is well defined. The latter difficulty is skirted by considering only a very restricted version of the question: What is the semantic status of the truths of elementary arithmetic (EA)? As Gödel has shown, EA is not axiomatizable. It follows that mathematical truth cannot be identified with mathematical theoremhood. The restricted problem concerns only those truths of EA that are not logically true: are they analytic? The authors indicate some flaws in Snapper's argument. One difficulty with it is that it applies only to axiomatizable theories, unlike EA. Further, on the authors' interpretation of him, he claims that the analytic theorems are those that reduce to logical truths when defined vocabulary is eliminated in favor of undefined vocabulary. The problem is that definability is a relative concept. One theory's primitive term is another's defined term, and vice versa. As a result, reducibility to logical truth by elimination of defined vocabulary is also relative to theory, and so is analyticity, on Snapper's criterion. In criticizing Snapper, the authors do not wish to maintain that some mathematical truths worthy of the name are analytic. Rather, they have reservations concerning the intelligibility of the question. - The question of the semantic status of mathematical truths has an epistemological point. How do we come to know the truths of mathematics? This question is embarrassing for empiricists. It would have a relatively straightforward answer if all mathematical truths were analytic: we come to know the truths of mathematics through an understanding of language. The philosophical import of Snapper's thesis is that the empiricist does not have this easy way out. But this threat to empiricism can be posed by defending a much weaker thesis: that some mathematical truths are synthetic. And this weaker thesis seems easy to defend. For, given any acceptable account of the distinction analytic/synthetic, some mathematical truths, the existential ones, are synthetic. One could not come to know that something exists merely through an understanding of language.