This paper identifies key characteristics of participatory research and development (R&D): it is client-driven, requires decentralised technology development, devolves to farmers the major responsibility for adaptive testing, and requires institutions and individuals to become accountable for the relevance and quality of technology on offer. The paper then reviews ways by which institutions can respond to these characteristics. For creating clientdriven agenda options commonly include: representation by clients on their governing boards, joint research planning and the establishment of researchextension liaison units. However, more effective than this "representation" strategy might be to place a significant proportion of the available research resources directly under client control: client groups would then contract as they deem appropriate. Decentralising technology development requires scientists to shift away from a "pipeline" model which defines a limited number of products towards the development of menus of options, and prototypes, which are then adapted to "niche" conditions by others. Such localised testing, requiring a community-based adaptive research capacity, can be achieved through working with groups of fanners (rather than individuals) and with producer organisations. The devolution of trial testing is discussed in reference to experimental methods, statistical validity and cost. Institutionalising accountability sharing is probably the most challenging issue. While scientists' rewards might be tied to the success or failure of technologies, clients' contracts with research or third party evaluations probably serve as more binding options. Three issues are signalled for future attention. First, clarification is needed of the respective roles of fanners and scientists in prototype screening: e.g. what features should scientists be screening for and at what stage?; how. early in the process can farmers be involved and to what degree can they control decisions on trial design and measurement? Second, decentralised technology development requires corresponding reorientation in service provision (e.g. credit, extension and seed multiplication). Third, steps need to be taken to safeguard equity, both between the more and less vocal groups of fanners,and between the requirements of present and future generations (the latter referring particularly to environmental concerns). Finally, participatory R&D alone is insufficient to deliver innovations relevant to diverse client groups: explicit procedures are required to define which clients are to participate, whose agenda are to drive the process, and what organisational innovations are needed to move agricultural R&D in these directions.