Cassava (Manihot esculenta Crantz) is an important food security crop in the semi-arid areas of Eastern Kenya. It provides food for more days in a calendar year than any other crop grown. Kenya has relied on varieties bred in other countries and because of this, local breeding methodologies and expertise are lacking. Access to appropriate varieties and adequate planting materials are major limiting factors to cassava production. Farmers grow late bulking landraces that take up to 18 mo to harvest. Efforts to introduce early bulking genotypes from IITA failed because of poor end-use quality. Local cassava breeding is necessary to alleviate the production constraints. Before a local breeding program can be established, farmers' preferences and production constraints must be identified and methodology appropriate to the Kenyan environment must be developed. The aims of this study were to identify farmer production constraints and preferences, to develop methods appropriate for cassava breeding in the semi-arid areas of Kenya, develop a population segregating for bulking period to estimate genetic variances that would explain the gene effects controlling yield components, and through participatory selection identify varieties that combine early bulking and preferred end-user traits. PRA tools, focus groups and individual interviews were used to identify production constraints and farmer preferences for cassava varieties. The PRA found that farmers grow 13 landraces in the area and 11 production constraints were identified and prioritised. The four most limiting in the order of importance were drought, lack of planting material, pests and diseases. Crosses between cassava varieties often do not produce much seed and the seed produced does not germinate well. Germination studies were done with open pollinated seeds to identify conditions favourable for seed germination in Kenya. The highest germination of the seeds was at 36°C. The control seeds had a higher germination percent (77%) compared to the seeds which were pre-heated at 36°C (57%). Crosses were made between selected IITA and local Kenyan genotypes following the NC 11 mating design to develop new genotypes which combine early bulking along with other farmer/end-user preferred characteristics. The hybrid progenies were evaluated in a seedling trial and clone genotypes advanced to a clonal trial and performance trial. The clonal trial was destroyed by red spider mites and cassava green mites, and only the tolerant 225 genotypes were planted in a performance trial that was harvested at 6, 7 and 8 mo after planting. The SCA effects were estimated to be 57% to 75% for most of the traits, except root number, which was mainly controlled by GCA effects (55%). Participatory selection of genotypes that combined early bulking and end-user qualities at the 7 and 8 mo after planting was done by farmers. Thirty genotypes that combined early bulking and end-user qualities were identified and ranked according to their performance in both agronomic and end-use traits using a selection index. A number of selected genotypes yielded more than three times the yield of the best parents, showing strong progress in breeding. Combining the farmers' preference aggregate score and the selection index based on the agronomic data, assisted in the final identification of the best genotypes developed in the breeding process. These results clearly demonstrated that it is possible to breed early bulking varieties with good end-use quality in the semiarid areas. / Thesis (Ph.D.)-University of KwaZulu-Natal, Pietermaritzburg, 2006.