In this paper we describe patterns of spatial co-ordination that, we propose, are a distinctive characteristic of multi-person face-to-face interactions. The data come from a task in which participants describe some simple, non-spatial, computer code in an instructor / learner scenario. Three participants take part; 2 instructors and 1 learner. Using excerpts from these interactions we show that participants make frequent use of combinations of head angle, gesture and participants' positions to 'triangulate' their contributions. We propose that these examples show how face-to-face interaction is distinguished, in part, by the potential it offers for using physical space to create shared 'interactional maps' that provide a structured resource for tracking conversational states.