Agricultural research - from recommendation domains to arenas for interaction

Agricultural research designs tend to be bounded by agroecological conditions, farming systems and other dimensions assumed to be homogeneous for the population of interest (that is, a recommendation domain or population for whom a technology or practice is expected to be relevant). Scaling is then a question of 'rolling out' results across the domain. But what if technology adoption and institutional context explain the variance in the output of smallholders, and agricultural development is also a question of institutional innovation? What if a domain is seen as a system of interest among actors who have a stake in the system and as an arena for concerted action and institutional innovation? This paper reports on six years of action research that attempts to answer these questions. It compares experimental interventions and subsequent systemic changes within each of nine agroenterprise domains. The experience suggests that the research approach used can explain variance in smallholder output that, in present-day West Africa, is not explained by technology adoption.