A discussion of the role of linguistic theory in explaining language acquisition proposes that theory draws too narrow a picture of language to adequately account for the developmental phenomena of acquisition. While recognizing the importance of descriptive linguistic research, a new approach cautions against embracing description to the exclusion of information suggesting what processes may be at work in language learning. Six theoretically derived theses on acquisition are presented and discussed: (1) the concept of language changes as new data come to light; (2) the crucial question for acquisition is not how the child learns which expressions are syntactically grammatical, but rather which interpretations he excludes; (3) the child cannot receive significant negative evidence; (4) input information is contradictory; (5) subset violations show where principles apply; and (6) change in grammar over time occurs because an item is either ambiguous or parametrically contingent. (MSE)