A corpus-based study of idioms in academic speech

A mastery of idioms is often equated with native speaker fluency (Fernando, 1996; Schmitt, 2000; Wray, 2000), but it is difficult for language teachers and material writers to make principled decisions about which idioms should be taught, given the vast inventory of idioms in a native speaker's repertoire. This article addresses the advantages and limitations of a corpus-based approach to researching and teaching idioms in a specific genre by drawing on a specialized corpus of 1.7 million words of academic discourse, the Michigan Corpus of Academic Spoken English. We argue that evidence from such a corpus can be quite informative for language teachers when the primary target language domain matches that of the corpus. In terms of pedagogical applications, we demonstrate the use of corpus data to construct teaching materials aimed not only at helping students learn unfamiliar idioms but also at raising their awareness of the speech contexts idioms occur in and the discourse functions they perform.