The propensity of citizens to engage in political discussion is classically considered to be a basic component of democratic political systems (Lane, 1961; Almond & Verba, 1963). However, the renewal of academic interest in political discussion over the last two decades has been stimulated by research focused both on political participation and deliberative democracy. In the current debate over the consequences of everyday political discussion on citizens and democracy, the most common way to conduct research is to base it on declarations and find out patterns of causality between, on the one hand, how much citizens say they talk politics and with whom, and on the other hand, respondents' characteristics (Mutz, 2004; Searing et alii, 2008; Huckfeldt and Mendez, 2008). Other researchers try to find out more precisely what these discussions are really made up of, what they are like, how citizens manage to discuss politics with others. There are different methods to approach this question: asking people to tell us about it (Conover, Searing and Crewe, 2002), observing people discussing politics (Gamson, 1992, Cramer Walsh, 2004), or both (Eliasoph, 1998). Our contribution belongs to the second category: we conducted focus groups in order to observe how "ordinary" citizens discuss politics. The design of this project differs from others notably because we set up the discussion in such a way that participants knew they were expected to discuss politics, but we then let them free to do it in their own way (or not). We will present and explain the research design in the first section of the paper. Our analysis of how participants did talk politics emphasises two processes: a process that which recalls deliberation, as debated in the framework of deliberative democracy, because it is a cooperative process of opinion building; another process that we call "conflictualisation" (section 2). We will explain how these two processes involve "raw material" of different natures, respectively opinions and cleavages (section ...