Our study explores the similarities and differences between a discourse completion task (DCT) and corpus data and discusses potential implications for using the two in a pedagogic context. The DCT has traditionally been used as an instrument for the study of interlanguage pragmatics while both spoken and written corpora have mainly been used to facilitate language description. Corpora also have a place in the language-teaching context as they can provide the learner with patterns of language use in social interactions some of which are not open to intuition. By contrasting native speakers' expressions of gratitude elicited by a DCT with those found in a five million word corpus of spoken English, we examine the advantages and disadvantages of both data sets with regard to the language-teaching context. The results suggest that a combined use of both instruments might aid the teaching of formulaic sequences in the classroom. (Contains 3 tables.)