Innovation systems can be defined in a variety of ways: they can be national, regional, sectoral, or technological. They all involve the creation, diffusion, and use of knowledge. Systems consist of components, relationships among these, and their characteristics or attributes. The focus of this paper is on the analytical and methodological issues arising from various system concepts. There are three issues that stand out as problematic. First, what is the appropriate level of analysis for the purpose at hand?

This report is concerned with the ‘who?’ ‘what?’ and ‘how?’ of pro-poor extension. It builds on the analytical framework proposed in the Inception Report of the same study (Christoplos, Farrington and Kidd, 2001), taking it forward by fleshing out the analysis with empirical information gathered from several countries during the course of the study (from primary data in Bolivia, Colombia, Nicaragua, Uganda and Vietnam, and from secondary sources in a range of other countries, including India), and drawing conclusions on the scope for action by governments and donors in a range of contexts.

The paper explores the implications of rural livelihood diversity for agricultural innovation policies. It summarises literature on the nature of rural poverty, with particular emphasis on the relative roles of farm and non-farm income. It also reviews the various roles, direct and indirect, that agricultural innovation can play in rural poverty reduction. Finally, it uses an agricultural knowledge and information systems (AKIS) perspective to argue for a differentiated approach to targeting agricultural innovations, based on an analysis of rural assets.