This article deals with the use of video-recordings of spontaneous interactions in workplaces as resources for training courses. It focuses on a multilingual psychotherapy consultation with a refugee in France, in which an interpreter was present. Drawing on multimodal interaction analysis, this contribution shows the coordinated and collaborative dimension of meaning construction in a therapist-initiated question-answer sequence and the way it is dealt with by her co-participants. The detailed analysis of the interaction highlights the challenges of translating the therapist's talk when it embeds medical as well as cultural assumptions regarding the patient. A discussion of the implications of this type of analysis for interpreter-mediated consultations and of the interest of an interdisciplinary approach to data analysis concludes the article.