Professional experience, as well as a great deal of published research (e.g. Gass & Varonis, 1991; Varonis & Gass, 1985a), suggests that even successful users of English as a second language unwittingly give rise to communication problems when encoding and decoding certain features of language, both linguistic and pragmatic. Among the latter is the issue of referring (for example, by pronouns or lexical substitution) to people, places and objects not in the immediate context. This paper, based on a wider study, outlines a procedure in which teachers were asked, via stimulated recall sessions, to reflect on previously-recorded pair work interactions and to discuss occasions where problematic communication occurred, or did not occur. Close analysis of transcript data reveals that potentially significant misleading stimuli may not eventuate in any failure of communication, while apparently trivial slips could cause considerable misunderstanding. It is suggested that teacher-researchers might usefully adapt a procedure such as that used in the study jar their own professional development, for use in class, and as the basis jar action research projects.