In French-speaking Switzerland, a Competence Center for Didactics was created in 2017. It received federal funding for supporting (a) staff and successors qualification in the partner teacher education schools and universities, and (b) constitution of research teams, national and international networks in didactics. It is in this context that our research team in didactics of physical education (PE) was constituted. It is composed of nine researchers and/or teacher trainers with various research trainings, skills and experiences, and one PhD student. Within the framework of a didactic engineering for development and training (Perrin-Glorian, 2011), the research team (the restricted group) first elaborated “scholastic forms of practice” (SFP – Mascret, 2011) in middle-distance race. Voluntary primary and secondary teachers (the extended group) experimented these SFP with their students within a research process led by the restricted group. Engaging all researchers and/or teacher trainers in the co-analysis of data produced within the experimentation was supposed to allow the less experimented and/or trained members of the restricted group to acquire didactical research skills, according to the remit of the Competence Center for Didactics. The SFP were designed to serve as a starting point and an assessment tool of a teaching sequence. Teachers were expected to provide students with additional learning tasks and experiences that target the learning issues related to the SFP. For each teacher, one member of the restricted group recorded four lessons. After two lessons, he/she confronted the teacher with the video recording of his/her actions. Then, through a priori analysis of the learning tasks (Amade-Escot, 2005) and analysis of teacher and students’ joint action (Ligozat, 2011), he/she recorded in synopsis if the learning issues were (a) absent, (b) potential, (c) addressed or (d) worked. When comparing their respective data and analyses during collective meetings of the restricted group, the members observed that they had conducted differently the self-confrontation interviews and categorized differently the data. Of course, this could be seen as a problem in terms of inter-analyst-reliability, but it also revealed differences between members of the restricted group in terms of didactical and methodological knowledge. We will illustrate with concrete examples how they acted as co-constructors of meaning and how, as an outcome of this collaborative process, they were brought to simplify their co-constructed analysis tool. With conceptual transposition (Amade-Escot et al., 2014), we will discuss these findings regarding the practical epistemology (Sensevy, 2006) of researchers and teacher trainers.