Peirce describes himself as a 'contrite fallibilist' (see 1.14, c. 1897)1;
he stresses that any of our beliefs may be mistaken, that our beliefs can never be absolutely certain, perfectly precise, or completely universal, and he urges that we always be prepared to revise our beliefs, to 'dump the whole cartload', as he picturesquely puts it, 'the moment experience is against them' (1.55, c. 1896). When it comes, however, to the question of whether we are fallible, not only with respect to our ordinary, empirical beliefs, but also with respect to our mathematical beliefs, Peirce's confident anti-dogmatism seems to falter. Peirce believes that the truths of mathematics are necessary. And he seems to suspect that the necessity of mathematical truths somehow precludes the possibility of our being mistaken in our mathematical beliefs; for when (e.g. 7.108, 1892, 1.248, 1902) he claims that fallibilism does extend even to mathematics he is tempted to compromise his commitment to the necessity of mathematical truths, and to hint that mathematical inference is, after all, only probable, and when, elsewhere (e.g. 1.149, c. 1897) he stresses the necessary character of mathematical truths, he also hints that we are fallible only with respect to our factual beliefs. 2 What is puzzling about this, to my mind, is why Peirce should think that there is a difficulty about reconciling fallibilism with the necessity of mathematical truths. Why should one n o t allow that we may hold false mathematical beliefs, and at the same time grant that mathema- tical truths are necessarily t r u e ? - o f course, if mathematical truths are necessarily true, then, if we hold false mathematical beliefs, those beliefs are necessarily false; but there is still no obvious difficulty in that. (If I add up a long column of figures incorrectly, then, if the correct answer is necessarily true, my answer is necessarily false. But why should it be supposed that t h a t precludes my adding up the column incorrectly?)
So my object, in the present paper, is, first, to understand why
Peirce should have suspected that there is a difficulty about reconcil- ing fallibilism and necessity, and, secondly, to discover whether the difficulty he suspects is genuine. Both issues, I shall suggest, turn upon the question of exactly what fallibilism amounts to. I shall argue that there is a thesis which is rather easily mistaken for fallibilism, which is incompatible with the necessity of mathematical (or any other) truths; but that a correct formulation of fallibilism shows that it can be reconciled with necessity. I shall then suggest that Peirce's uncertainty about whether fallibilism can be extended to mathematics can be explained by his failure to specify fallibilism very carefully. I shall conclude with some remarks about affinities between my arguments and those that support Lakatos' fallibilist philosophy of mathematics, and some comments on the significance of the pos- sibility of reconciling fallibilism and necessity for Quine's philosophy of logic. First, though, I'll set the stage by sketching Peirce's arguments, on the one hand, for the necessity of mathematical truths, and, on the other, for fallibilism.
2. N E C E S S I T Y
In early papers Peirce offers an account of the status of mathematical
truths which rather strikingly anticipates some central theses of logicism. There are, he claims (3.20 ff, 1867) certain propositions, propositions which may be taken as definitions of mathematical objects, from which the truths of mathematics follow 'syllogistically'; so the truths of mathematics can be reduced to logic and definitions, and require no reference to experience or intuition. In a paper of 1881 (3.252 ff) Peirce works out this idea in more detail, urging, notably, that in the derivation of mathematical truths traditional, Aristotelian logic will need to be supplemented by the logic of relatives, (Usually, the beginning of modern logicism is set at 1879, with the publication of Frege's Begriffsschrift.) Subsequently, however, Peirce came to think that these ideas were quite misguided (indeed, to say-4.333, c. 1905-that the 1867 paper