Believing in language

We propose that the generalizations of linguistic theory serve to ascribe beliefs to humans. Ordinary speakers would explicitly (and sincerely) deny having these rather esoteric beliefs about language-e.g., the belief that an anaphor must be bound in its governing category. Such ascriptions can also seem problematic in light of certain theoretical considerations having to do with concept possession, revisability, and so on. Nonetheless, we argue that ordinary speakers believe the propositions expressed by certain sentences of linguistic theory, and that linguistics can therefore teach us something about belief as well as language. Rather than insisting that ordinary speakers lack the linguistic beliefs in question, philosophers should try to show how these empirically motivated belief ascriptions can be correct. We argue that Stalnaker's (1984) ''pragmatic'' account-according to which beliefs are dispositions, and propositions are sets of possible worlds-does just this. Moreover, our construal of explanation in linguistics motivates (and helps provide) responses to two difficulties for the pragmatic account of belief: the phenomenon of opacity, and the so-called problem of deduction.