ABSTRACT
This chapter indicates some of the ways in which the study of conversational data requires modification of the linguistic paradigm itself, and thus makes a significant contribution to linguistic theory. It looks at the related puzzles of native like selection and nativelike fluency: the ability of native speakers to convey meanings by expressions that are not only grammatical but also natural and idiomatic and the ability of native speakers to produce fluent stretches of spontaneous connected discourse. It argues that the linguists view of native speaker knowledge must be expanded to include memorized sequences and lexicalized sentence stems in addition to the usual apparatus of linguistic description. Such a modification of linguistic theory will require rejection of the principle of parsimony in the evaluation of grammars, but will more accurately reflect native speakers real linguistic knowledge.
