Two main approaches to design and testing exist in F.S.R.; a "Top down" approach and a “Bottom-up" approach. In F.S.R. in Southeast Asia, design and testing have been dominated by “Top-down” approaches in which the design of technical, improvements to be tested is assumed from a preliminary site description and mainly concerns the development of new cropping patterns. Testing methods are based on analysis of variance used to compare yield, gross return, ect., of new cropping patterns or superimposed technical components with those of the existing cropping patterns and techniques performed by farmers. Despite its undoubted practicability this approach raises several objections from a methodological point of view. In "Bottom-up” approaches the design is late and is Preceded by an agronomic “on-field" survey. This survey uses the existing variability of technical itineraries and of situations created by farmers as an experimental design. It aims at an identification and an understanding of the factors that affect yield elaboration of crops under farmers' conditions, and leads to the construction of a qualitative model used for the design phase. Technical improvements are designed in order to reach optimum states of the crop defined by the model and to fit to farming systems functioning characteristics. (farmers’ decision making, labour availability... ) During the testing, an analysis of the differences between the results and the predictions from the model is used to improve the design of techniques and the model. Some features of the agronomic “on-field” survey are illustrated with data from a F.S.R. project carried out in the Songkhla Lake Basin (Southern Thailand). The notion of variability and its usefulness for experimental design is developed, The principles of analysis of such an experimental design are described and guidelines for the construction of a qualitative model presented and discussed.