The school, established in a different community of itself and different in itself, is in confluence with multiple systems and it cannot be alienated, needing to find structures of interaction and figures that facilitate this interaction. Mediation in education tends to facilitate interactions by solving and preventing problems and conflicts. We understand mediation in a very wide sense, but in the first place it constitutes a preventive process. "The real mediator is someone who, like a 'prophet', possesses a sense of anticipation, by creating passerelles that could restore communication and by solving misunderstandings and conflicts that may emerge" ( Six, J- F, 1995: 97). Continuing the research presented last year in EERA, in Crete (Caetano e Freire, 2004), we now present a new multi- case study, focused in class assemblies as a device of conflict mediation and in formal mediation devices inside school, implemented to prevent and to deal with children and adolescents at risk. Problems of the indiscipline, violence, dropout, poverty, multicultural conflicts, are some of the issues that school has to deal with. Here, the mediation process intends to establish relationships between several instances inside the school (teachers, tutors, students…) and between the school and the community (families, support institutions, local authorities…). Besides the goal to contribute to the knowledge of the Portuguese reality in the field of mediation in education, we pretend to conceptualise it. This is, to conceptualise the field and the concept of mediation, where groups are organised to deal with their problems and conflicts and formal devices are created to deal with several problems in a wide-school perspective. We also develop the model of analysis early presented, applying and modifying it with new transformations. The questions raised are now deepened, with new data, and restricted to mediation conflict. One of the privileged research strategies (in the study of class assemblies) is the organisation of research partnerships with professional in the field, through the development of action-research projects. In this way we'll sustain a more direct relation between research, theory and action. Documental analyses (of class diaries, class assembly reports), the video and naturalistic observation, questionnaires to the pupils are some of the main research methods and techniques. The other privileged strategy is the ethnographic approach. Here we continue using interviews and naturalistic observations. We also develop the model of analysis early presented, applying and modifying it with new transformations. The questions raised are now deepened, with new data. The studies are ongoing, so we have not yet their results. We'll present them in September, at Dublin.