In this paper we propose a theoretical reassessment of the discourse-analytic tools that are used in classroom research and of the data that are addressed. To that end we present in-depth analyses of 1) an IRE structure in a first English (EFL) lesson in a Dutch secondary school; 2) episodes from Dutch as a second language classes in a school for refugee children; 3) two instructional units that were used as reflection data in a university teacher training course. All of these exhibit complex footing changes (Goffman, 1981) and instances of co-speaking, muttering, whispering, overhearing etc. that are largely unexplored in classroom research. Yet these phenomena co-construct the interactional matrix that organizes participation and learning in multiparty settings. To shed light on what happens in the structural niches of turns and at the fuzzy edges of speaking-slots-in-progress we need a more dynamic and complex notion of context. Speaker/hearer roles may be embedded in or parasitic on each other along a continuum of official and informal or collusive stances. In our proposals to model emerging discourse complexity, subtle changes in register, bodily posture, tone of voice, rhythm and pitch are an integral part of what constitutes "the data".