A number of disparate approaches to language, ranging from cognitive linguistics to stochastic implementations of optimality theory, have challenged the classical distinction between knowledge of language and use of language. Supporters of such approaches point to the functional motivation of grammatical structure, language users' sensitivity to the frequency of occurrence of grammatical elements, and the great disparity between sentences that grammars generate and speakers' actual utterances. In this article I defend the classical position, and provide evidence from a number of sources that speakers mentally represent full grammatical structure, however fragmentary their utterances might be. The article also questions the relevance of most corpus-based frequency and probability studies to models of individual grammatical competence. I propose a scenario for the origins and evolution of language that helps to explain why grammar and usage are as distinct as they are.