|This project aims at describing doctor-patient interaction when breaking bad news. Consultation with medical physician is organised in several phases (Byrne and Long 1976, Ten Have 1989, Cosnier 1993, Traverso, 2001). The different phases the authors are distinguishing are almost always similar. Even though the phases are presented in an organised and linear way, the interlocutors may navigate from one phase to another, in particular if different problems are evocated along the interaction (Ten Have 1989). The different phases have been well described in the literature (Heritage and Maynard 2005, Ten Have 2001) but little has been said about the transition from one phase to another: when does it appear? How is it done? By whom? This study aims at exploring this particular moment in the interaction looking at the different modalities that may be involved. We intend to do so using the analysis of interaction in a multimodal perspective. Previous studies have shown that the behaviour of the participants is evolving through the interaction (Guardiola et al. 2012). For example, during medical encounter, the doctor is usually the main speaker but appears to give the floor to the patient in the second half of the interaction when the latter thus become the main speaker (Vergely et al. 2011). Therefore, we suppose that the transitions might be produced differently through the interaction. The corpus analysed was created by and belongs to the Institut Paoli-Calmette (Marseille, France) and is an authentic corpus of training sessions for doctors involved in role plays with an actor playing the role of a patient. It is composed by 6 audio-video recorded interactions involving one doctor in training and one actor/patient. Each dialogue is about 15 minutes long. The participants are involved in a consultation to break bad news. In our case, the actor actually plays the role of the patient’s relative. We conduct both quantitative and qualitative analyses. Transitions are described in terms of the phases involved, speech production, co-speech gestures production (McNeill, 1992), gaze, head movements and posture (Tan et al. 2010) for each participant. Linguistic phenomena such as feedback (Bunt 2012, Bavelas 2002) or Repair (Schegloff et al. 1977, Uhmann 2001) involved in transitions are described as well.