This paper concerns the use of computational methods for collection and analysis of language data as a basis for research in pragmatics. The aim is to describe advantages and problems associated with the use of corpora, and to provide both a bird’s-eye view of the field as well as an in-depth account of a limited set of topics of particular salience. A main claim to be argued here, with reference to several papers, is that, as a method for empirical research, corpus-based pragmatic analysis is essentially inductive and interpretative and can never be performed without the reliance on the analyst’s intuition.