CONTEXTThe “Living Lab” (LL) model of innovation practices is increasingly applied and strongly supported by innovation policies in the agricultural sector. But while the main principles of LLs imply specific approaches to collectively implementing open innovation processes, few studies have developed an understanding of the different ways of doing so, in relation to the experimental practices on which they rely. While the specificities of agricultural LLs are beginning to be described, little is known about how they more particularly reconfigure experimentation in agricultural innovation systems.OBJECTIVEThe purpose of this paper is to investigate the diversity of experimentation practices in agricultural LLs. It looks at how these practices correspond to different contextualized appropriations of an open innovation model, as translated into key principles for LL implementation.METHODSFirst, we produced an analytical framework that combines the principles of the LL model as described in the LL literature with generic dimensions of experimentation as analyzed in Science and Technology Studies. Second, we used this framework to analyze the diversity of experimentation practices in eighteen cases of LLs in the agricultural and food sectors. Cases were analyzed via two approaches: 1- three small and recent ongoing LLs located in France were studied during a one-year immersive observation; and 2- a collection of 15 other cases (identified in scientific and technical literature, each case corresponding to experiments that are forms of appropriation of the LL model of innovation, sometimes in several similar LLs) were analyzed, mostly based on scientific papers (describing the LLs or their outcomes), technical documents and available information such as videos online, with additional visits and discussions with LL actors for a subset of cases.RESULTS AND CONCLUSIONSOur results enable us to characterize three ideal-types of experimentation practices within agricultural LLs: 1- experimentation as a game of creativity in a predefined space; 2- experimentation as a progressive contextual adaptation for innovation adoption; and 3- experimentation as a catalyst for long-term local collective action. Each ideal-type corresponds to a particular combination of ways of (i) controlling what happens within the experimentation space, (ii) defining a perimeter for experimentation, (iii) integrating plural knowledge in problematization, and (iv) integrating various actors' evaluations of experiments. Beyond a clear distinction among ideal-types, appropriations of the LL model reveal different ways of anchoring the experimental process in an explicit temporality and situation, and of considering the construction of the LL itself as experimental and open-ended.SIGNIFICANCEOur analytical framework opens a new way to distinguish between various experimentation practices claiming to develop open innovation processes in agriculture. The three ideal-types that we have identified show the need to pay attention to the relation between collective experimentation practices and the LL expected outcomes.Graphical abstract