The principal theme of this thesis stems from the failure of farming systems research and extension (FSR/E) methods to deliver agroforestry interventions that are widely adopted by farmers. The possibility of using a knowledge-based systems approach to augment conventional FSR/E methods in considering the improvement of complex agroforestry practices was, therefore, evaluated by sequential adoption of ICRAF's Diagnosis and Design Methodology and a novel knowledge-based systems approach. This was done on a case study basis in the Kandy district of Sri Lanka, where the traditional forest gardens are an important feature of land use. It has been previously suggested that analysis of an explicit record of farmers knowledge might be used to identify deficiencies constraining productivity and sustainability of agroforestry practices and hence develop appropriate research and extension strategies, this thesis records the first attempt to do this. There were difficulties in applying conventional FSR/E methods to farming systems involving complex agroforestry practices, specifically because the classification at the level of the farming system did not result in identification of sufficiently homogeneous groups of agroforestry practices for the purposes of designing research and extension. The extent to which a knowledge-based systems approach generated new information which changed what was known about farming practice from that available from a conventional diagnosis was explored and the knowledge-based systems methodology further developed and tested by: " comparing alternative methods of testing representativeness of knowledge collected from a purposive sample of key informants against the distribution of knowledge in the whole farming community, " exploring the validity of items of indigenous knowledge in an objective sense through experimentation to determine their scientific validity, and exploring the possibility of generating and introducing new knowledge to fill gaps identified through analysis of farmers' existing knowledge.
Th" e findings of the research have, plications for the role of knowledge-based methods in agroforestry research and extension which are discussed in the light of contemporary developments in approaches to research and development in agriculture and forestry more generally.