Corpus analysis has not yet been greatly exploited in the field of translation studies and, more generally, corpus-based studies of text imagery are rare. No comparative study of the structure of similes, metaphors and analogies in French and English has been done to date. It therefore seems relevant, from a translational perspective, to undertake such a corpus-based study, a means of establishing optimally objective models based on actual usage. This paper reports on the preliminary results of research which aims to highlight the lexical, syntactic and referential characteristics of images used in popularized scientific texts, in order to identify possible differences in the creative use of French and English, with a view to eventually proposing certain orientations for idiomatic translation of imagery in such texts. (Isabelle Collombat)