Computer-mediated-communication (CMC) tools allowing learners to be in contact with native speakers of their target language in other locations are becoming increasingly flexible, often combining different modes of communication in a single web- and internet-based environment. The literature on telecollaborative exchanges reveals, however, that online intercultural communication between language learners "often fails to achieve the intended pedagogical goals" (O'Dowd & Ritter, 2006:624) and warns that "exposure and awareness of difference seem to reinforce, rather than bridge, feelings of difference" (Kern, 2000:256). Yet, research into the reasons for lack of success in CMC-based partnership-learning has, so far, only been carried out on a relatively small scale (see, for example, Thorne, 2005, Ware, 2005, O'Dowd & Ritter, 2006). In autumn 2005, students of French at Carnegie Mellon University (CMU), USA and adult learners of French at the Open University (OU), UK were joined by native French speakers studying for an MA in distance education at the Universite de Franche Comte (UFC), France in a pilot Tridem project in which all participants worked on the completion of a series of collaborative tasks. The Tridem partners met over several weeks in an internet-mediated, audio-graphic conferencing environment. The project output, a shared reflection in French and English on cultural similarities and differences, took the form of several collaborative blogs. The paper draws on data from pre- and post-questionnaires, from the work published by the learners in the blogs and from post-treatment, semi-structured interviews with volunteer participants. Beyond considering some of the known factors influencing success and failure in CMC-based collaborations such as discrepancies in target language competence among learners, this article also explores affective issues and difficulties arising from varying levels of multimodal communicative competence. The insights gained are mapped against O'Dowd and Ritter's (2006) "inventory of pitfalls" in telecollaboration. The result is a tentative framework which allows those involved in setting up and running telecollaborative exchanges to gauge both degree and nature of some of the risks they are likely to encounter.