Description: In Geneva, there is no formal education for becoming a theatre technician. Property managers working at Le Grand Théâtre de Genève (The Geneva opera house) have first followed a basic technician education (carpentry, cabinet-making, locksmithing, painting, ...) and then non-formal education in the workplace. Their job is to construct objects (props) for the producer and set designer (with innovative forms or dimensions, new tools, new materials, ...). The orders placed by the producer and set designer are often allusive and each producer or designer has its own way of placing an order. Most of the job of the property managers consists in elaborating on these allusive orders, making them operational and realizable, and adapting this part of the job to each dimension of the order (form, level of requirement, ...), and this under time pressure. So the property managers must become self-directed learners. They must also find a way of stabilizing activities which are at each occurrence very different and new. This self-directed learning develops through the participation in a community of practice, characterized by mutual engagement, joint enterprise and shared repertoire (Wenger 1998), which allows the experiences to stabilize and obliges this stabilization to occur. The community of practice in this context is both an individual and collective resource for the property managers, allowing them to cope with problems, and the orders placed. Employed on temporary limited contracts, newcomers must succeed in making a highly complex object : allusive yet demanding, they must invent new tools, explore new materials, ... The newcomer must show that he can anticipate and take calculated risks. In such a context, how do newcomers succeed in belonging to the community of practice, in becoming theater technicians, and specially property managers ? According to Lave & Wenger, learning occurs through legitimate peripheral participation (Lave & Wenger 1991), but what is legitimate peripheral participation in such a context ? Methodology: Participants: 5 experienced property managers, 1 newcomer (30 years old). All technicians have volunteered to participate in the study. Ethnographic and participant observations (1 day a week over a period of one year), fieldnotes, audio and video recordings (interaction between set designer and property managers' chief, daylong recordings), interviews and self-confrontations with props' photos. The analysis of data (fieldnotes, recordings, interviews), based on a linguistic approach to interview verbalization, is validated with the team of property managers. Conclusions: Our paper shows that the way of learning of newcomers in the context of the property managers workplace does question some of the characteristics mentioned by Wenger (1998, p. 100) for legitimate peripheral participation : lessened intensity (here the intensity is high), lessened cost of error (it's in our case a test, if the newcomer is not successful, no more contracts will be established), close supervision (the newcomer must show he's independent and can manage the task alone), or lessened production pressures (except time, here the production pressure is similar or even higher than for experimented ones).