This paper addresses issues of institutionalising participatory approaches to research. Experience in Botswana has led to careful consideration of the roles of different types of trial and the distinction between experimental and nonexperimental variables in each. Examples from Botswana distinguish between "high leverage" interventions which address critical constraints in the farming system but whose implementation may require major changes in farmer practice, and "low" or "non-leverage" interventions which have restricted impact but are easier to implement. Requests to test the latter arise from on-station research, whereas a principal objective of on-farm research is to seek solutions to the former. The paper also describes institutional changes designed to enhance the accountability and credibility of research, and to enhance the speed and accuracy of technology recommendations which are essential to the institutionalisation of participatory approaches.