This book addresses a question fundamental to any discussion of grammatical theory and grammatical variation: to what extent can principles of grammar be explained through language use? John A. Hawkins argues that there is a profound correspondence between performance data and the fixed conventions of grammars. Preferences and patterns found in the one, he shows, are reflected in constraints and variation patterns in the other. The theoretical consequences of the proposed 'performance-grammar correspondence hypothesis' are far-reaching -- for current grammatical formalisms, for the innateness hypothesis, and for psycholinguistic models of performance and learning. Drawing on empirical generalizations and insights from language typology, generative grammar, psycholinguistics, and historical linguistics, Professor Hawkins demonstrates that the assumption that grammars are immune to performance is false.