Since 1996, we have used Role Playing Games (RPG) with our Multi-Agent System (MAS) platform, CORMAS (Bousquet et al. 1998), in different experiments about local land use management, water management, negotiations between foresters and breeders, and preservation of wild genetic resources by local peasants. Each experiment tested a different sort of coupling between RPG and CORMAS, from a RPG simply explaining the MAS model to a RPG strongly evolving in the different steps of the modeling itself. Each experiment tested also different objectives for this linking, from the improvement of the modeler's knowledge about the issue to the support of an endogenous debate within the local community. For a better knowledge of the use of these games, we varied in each experiment (i) the links between the games and the reality; (ii) the nature of knowledge put at players' disposal (players' knowledge or external expert's knowledge, technical or social knowledge,...); (iii) the adaptability and the upgrade ability of rules; (iv) and of course the supports of the game (board game; links with MAS; type of rules, players and organizer,...). Finally, the difficult question of validation was started on by several ways, from computer and statistical validation as well as the different possibilities of players' validation. In this paper, we have begun a formalized appraisal of these six experiments, allowing us stressing the advantages and disadvantages of each methodological step as regards the different possibly purposes of the participatory modeling.