While the inferential dimension of ethos has been studied extensively, its relationship with multimodality, i.e. the fact that linguistic devices used in verbal interaction combine with other semiotic resources such as gestures or shifts in gaze direction, remains largely unknown. Stepping from a language-oriented approach to argumentation, the paper describes a theoretical framework for the multimodal analysis of ethos in argumentative talk-in-interaction. An example taken from a video-recorded corpus of French public debates is provided.