Public sector research and extension agencies are frequently unable (because of cost constraints) or unwilling to serve the needs of small farmers living in complex, diverse and risk-prone environments. This paper describes one attempt to increase these farmers' demand-pull on public sector and non-governmental technology-providing agencies by harnessing their capacity for local experimentation and combining it with some of the techniques of more formal agricultural research. The CIAT IPRA programme has provided resources, technical support and management training to locally elected committees (CIALs), composed of 'experimenting' farmers, which manage and conduct research on behalf of the community as a whole. The community is responsible not only for the election of the committee, but also for setting the agenda for research and helping to evaluate the results. The risks of conducting research are reduced by the existence of a fund — which is 'owned' by the community as a whole — on which the committees can draw should the experimentation result in financial losses. So far 55 of these committees have been established in Colombia since the programme began in 1990. A number of these have evolved into significant local seed production enterprises. The committees as a group have now formed their own corporation. This enhances the scope for information exchange and puts them on an independent financial footing. Other committees have been established in Bolivia, Ecuador, Honduras and Peru. The CIAT programme has monitored the progress of the Colombian CIALs and developed a range of training materials which can be used by others interested in establishing committees. This paper looks at the results of the programme, focusing in particular on the support needs of CIALs and the relative costs and quality of CIAL research. It also notes some of the distributional and institutional issues which are expected to be important to the future adoption of the CIAL approach to adaptive agricultural research.