Abstract

In my paper ‘A Logic for Evidential Support’ (this Journal, 17 (1966), 21 ff.) the argument on page 25 is illustrated by wrong and misleading examples.1 The argument proceeds by considering statements logically equivalent to a universal hypothesis U1 that are formed by generalising analogously not about the individual elements of U1's domain of discourse, but about pairs, trios, or n-membered classes of these elements, where the domain of U1 has at least n elements. But the generalisations must be understood distributively : not collectively, as the examples in the original version suggested. Such an equivalent for ‘All ravens are black’ will not be ‘All pairs of ravens are pairs of black things’, but ‘All pairs are pairs of things each of which is a raven only if black’. Similarly, instead of ‘If the pair a, b is a pair of ravens it is a pair of black things’ on page 26,1 should have written ‘The pair a, b is a pair of things each of which is a raven only if black’. If these alterations are made, the argument is unobjectionable, and the qualification about truth-functional conditionals, on page 25, is no longer required. (A corresponding correction should be made in the argument of ‘What has Confirmation to do with Probabilities ?’, Mind, 75 (1966), 474.)