Top-down and bottom-up approaches to corpora in language teaching

Both native-speaker and learner corpora are exploited in language teaching. The activities associated with these types of corpora typically proceed in opposing directions: 'downwards' from a supposed model of native-speaker performance, helping the learner to discern and assimilate the lexico-grammatical patterns that lie behind this performance; and 'upwards' from the learners' collective interlanguage productions, guiding them towards a closer approximation with native-speaker proficiency. There is, ideally, a convergence between these two movements, by which language learners will become better able to perceive discrepancies between their own patterns of use and those of native speakers, starting from what they already know about the target language, and from what they themselves are trying to use it for. The purpose of this paper is to suggest ways in which this convergence can be encouraged, by constructing activities based on both native-speaker and learner corpus data.