The electronic corpus of naturally occurring texts in machine-readable format has become a central concept in the computer-assisted analysis of language. However, research has so far tended to focus either on computational stylistic analyses of distinct but relatively small corpora. This paper suggests ways of bringing the two kinds of corpora together in translation, using the large corpora as ‘controls’ against which to evaluate specific examples from texts on which the bilingual lexicographer, translation critic or translator is working, and serving as an aid to the selection of a translation equivalent. The examples in this paper focus on the author’s own experience as a lexicographer, as a researcher on translation and as a practising translator.