ABSTRACT
Present-day theories about the nature of a priori knowledge fall for the most part under one of two heads. On the one side, in accordance with the tradition of logical empiricism, it is still held by some that such knowledge is essentially linguistic, and that necessary truths are 'analytically' true, by definition of the terms involved. On the other side, that general proposal has been widely rejected, largely on the ground that it is impossible to justify a sufficiently tight conception of the meaning of a term to support a firm formal distinction between analytic and synthetic propositions or sentences. The most that can be said in favour of the distinction, it is commonly held, or in explanation of the propensity of philosophers to draw it, is that there are some sentences that we are (and would be in almost all circumstances) particularly loathe to give up as we trim and modify our system of beliefs under the pressure to adopt that system which constitutes the most
coherent and useful response to sensory input to date: useful, that is to say, for the purpose of prediction.
