||Keywords: Agricultural R&D, field experimentation, crossbreeding, dairy, feeding technologies, mixed farming, farming systems research, modelling.Smallholder mixed crop livestock systems continue to be a dominant agricultural production system in many developing countries, includingIndia. Dairy farming is part and parcel of many such systems, and it is often seen as an important livelihood option to increase household income and to therefore contribute to poverty alleviation in rural areas. As a result, substantial efforts in agricultural R&D have been directed towards design of new technologies for smallholder dairy farming. Variable success in technology transfer has clearly shown that adoption is context-specific, related to the physical and socio-economic environment, access of farmers to resources, access to information and personal attitudes. A series of concepts and methods were developed to incorporate these considerations, and to replace narrow technology-driven approaches by broader ones such as Farming Systems Research (FSR).This thesis describes and analyses experiences of BAIF, an Indian NGO, with the use of FSR methodology in livestock development programmes inGujarat,India. The objectives were to identify criteria and methodologies for selection of appropriate livestock technologies for farm level, and to identify differences in the methods of selection of appropriate technology. Section 1 describes the variation in livestock production systems inIndiain general and in Gujarat-state. Livestock comprises defined and undefined breeds of cattle and buffalo. Total livestock population increased annually by over 1% in the last four decades, with buffalo and goat populations increasing faster than cattle. This section also gives background to the BAIF organization and to FSR methodologies. Section 2 more specifically describes theGujaratresearch area with agro-ecological zone-wise information on animal breeds, herd composition, feed resources, crops, and trends in seasonal availability of feed as derived from transects, Participatory Rural Appraisals, and mapping. Constraint analysis and modelling indicated limited genetic potential of the local breeds and shortage of feed resources, both quantitatively and qualitatively, as major constraints for livestock development. Crossbreeding for breed improvement and use of (improved) local feed resources were identified as suitable technologies to alleviate these constraints.Ex-post performance monitoring of some BAIF crossbreeding programmes show that crossbred cattle fitted well in the smallholder mixed farming systems of both tribal and non-tribal farmers in all three selected agro-ecological zones (Section 3). Milk production of crossbreds was substantially higher, as was livestock gross margin and household income. Although quality of the roughages is a major limiting factor, farmers owning crossbreds tried to adjust to the needs of the cows by feeding concentrates. There was no difference in workload and labour division between households with and without crossbreds. Crossbreeding thus proved a techno-economically and socially viable livelihood option for both mixed and landless farming systems inGujarat.Various modelling approaches were used in Section 4 to explore,ex ante,the suitability of feeding technologies such as urea supplementation, use of local and commercial concentrates, urea-treated straw with concentrates, and leuceana tree leaves for crop-livestock systems inGujarat,India. Major conclusions were that (i) concentrate feeding is beneficial to farmers with market access and crossbred cows, (ii) crossbreeding interventions are more remunerative for landless and tribal farmers than for non-tribal farmers; feeding interventions are more effective for crossbreds than for local cows, (iii) maximum farm income is achieved at medium milk yields per animal; higher milk yields require use of better feeds, which renders the straws of the grains useless for feeding; at farm level, the (economically) optimum cropping pattern would then shift from grain crops to cotton.This section continues with a narrative on BAIF's experiences with field testing of technologies at animal, at herd, at farm and watershed level, including a shift to crop research when dictated by local needs.Over a period of roughly 30 years, three phases in on-field testing can be distinguished, i.e., starting with a period of predominantly top-down approaches, moving to a phase with emphasis on participatory identification and testing of technologies, and then into a phase with work at community and watershed level. A few cases are discussed for each phase, illustrating the processes, methods and types of technologies involved, and drawing lessons on field experimentation for livestock and rural development in general. The studies brought out, among others, that adoption of technologies is facilitated when these use local (feed) resources, that are readily available, requires only small changes in farm practices, are relatively simple to implement, and yield tangible results in the short term.Section 5 analyses the dynamics in methods and approaches of BAIF's work on livestock development, as it grew from Gandhian roots into a large development organization. It emphasises the dynamics in approaches between top-down, objectivist and reductionist approaches on one hand and bottom-up, constructivist, holistic and self-organized approaches on the other hand. These experiences are set against similar developments on the (inter)national scene and in industrialized countries, along with factors that influence the changes, suggesting that agricultural R&D behaves as a complex adaptive system with its own dynamics and associated paradigm shifts. It also discusses a number of cross-cutting issues such as the notion of real versus perceived problems, hierarchy and grid, phases in development and aspects of holism versus reductionism, also reflected in notions of goal and process orientation.Concluding, the thesis considers development as a continuous process, of which the goals change over time-and-space. This reflects a paradigm issue, and if development is indeed a dynamic process it implies that choice of methodology and technology) should go along with changes occurring in that process. Some guidelines regarding the usefulness of approaches and technologies are given. But agricultural R&D is ultimately considered to be a complex adaptive system, also inGujarat, and development organizations such as BAIF have to, therefore, show dynamic behaviour.