Most formal research in support of agricultural development has focused on the alleviation of farm-level productivity constraints, with problem diagnosis often occurring through a single disciplinary lens. There is a strong push within national and international arenas to move toward broader units of analysis and intervention, including the landscape, catchment and watershed. However, there is a current imbalance in the strong momentum behind this shift and the paucity of methodological guidelines for operationalising these new approaches within research and development (R&D) circles. This paper outlines an approach for grounding watershed management in local incentives for improved natural resource management (NRM) beyond the farm level, addressing component-specific contributions to landscape degradation, and bringing formal research contributions to bear on a demand-driven NRM agenda. Following a description of a methodology used to diagnose problems at landscape or watershed level, a case study from the highlands of central Ethiopia is presented to illustrate the application of the approach within agroforestry. The case study provides a concrete example of how to move from participatory problem diagnosis to a modified R&D agenda at the landscape level.