Most linguistic and interactional studies on public and media debates focus on the way talk-in- interaction is locally managed by the moderator. For example, they analyze the extent to which the questions asked are both sequentially and categorically relevant. The present paper aims at enriching these studies by discussing the profitability of a multimodal and longitudinal approach to membership categorization practices. The approach is multimodal in the sense that it does not focus exclusively on verbal features. The approach is also longitudinal by considering the extent to which a debate is pre- configured. To this end, particular attention is paid to the posters promoting the encounters and to the way the participants are positioned in physical space.