Reconciling corpus data and elicitation data in FLT

Gilquin, Gaëtanelle [UCL] Shortall, Terry []

Because they tap into basically different things, corpora and elicitation tests may at times provide diverging results. In this paper, we will show that, far from being opposed to each other, corpus data and elicitation data should be seen as complementary and that their reconciliation can actually prove useful in a field such as Foreign Language Teaching (FLT). Two case studies will illustrate this. In the first one, we will use data coming from a learner corpus and from fill-in and evaluation exercises to investigate learners’ knowledge of make-collocations. We will see that only the combination of corpus and elicitation data can give the full picture, i.e. performance and competence, and that an analysis relying on just one source of language data therefore runs the risk of being biased. In the second case study, we continue to investigate the differences between competence and performance through a comparison of the existential structure in elicited data from native speakers and in a spoken corpus. The divergences between the prototype effects displayed by the elicited data and the frequency effects found in the corpus data are discussed in the light of Foreign Language Teaching, and it is argued that prototypes should be taught first and that structures should be revisited in a cyclical fashion as proficiency increases, with all extensions of a structure being eventually introduced to students.