Detailed analysis of face to face computer mediated interactions implies a study of a corpus composed of the interaction traces produced by the collaborative tool and video and audio recordings. Because of their separate nature, it is frequently difficult to perform this joint analysis. In order to facilitate such analyses we propose a model for describing corpora composed of sources from different media; we also describe how certain types of analyses can be modelled and reified for future use. We show how a tool based on these models facilitates a number of analysis methodologies by simplifying situations where entities are multiple: several researchers (inter-coder reliability), several sources for the corpus (multi-modal interactions) or several analyses (for re-use of prior analyses or for the confrontation and comparison of two different analyses. We believe that capitalizing on previous analyses and affording parallel analysis of video and computer-mediated traces will open new avenues for researchers and teachers.