Whilst supporting Schegloff's (1992:215) call for context's ‘demonstrable relevance to participants’ as well as his notion of procedural consequentiality, this paper on context in communicative interaction, puts forward cases where the relevance for participants is not always demonstrable within the observation under study, but in previous observations. This problematizes considering context in distal and proximate terms and considering the analytic tasks as one of subsuming the former within the latter. The proposed solution is the notion of a messo context, whereby the analyst is no longer restricted by the confines of one observation, but nevertheless obliged to show relevance to participants.