A maize-breeding programme was targeted at resource-poor farmers of eastern India using a client-oriented approach (often called participatory plant breeding). Farmers tested a range of varieties in a participatory varietal selection programme but none proved to be very popular. Information from these trials was used in a breeding programme to develop a broadly-based population from three yellow- and three white-grained maize varieties that were either identified in the trials or had traits liked by farmers. The population was subjected to several cycles of random mating. In the C3, C4 and C6 cycles, subpopulations were extracted and subjected to mass selection over several generations in environments that represented well the target population of environments of resource-poor farmers' fields in eastern India for traits identified by farmers. To identify better the required traits, in some generations the selection was carried out by farmers on a research farm under fertility levels that approximated farmers' practice. The improvement of the subpopulations resulted in several varieties that performed well in research station and on-farm trials. One of them, BVM 2, was released in Jharkhand state, India. In multilocational research station trials, it yielded more than the control variety BM 1 but silked earlier. In the less favourable environments of on-farm trials, its yield superiority, in percentage terms, was higher. Farmers perceived BVM 2 to have better grain quality and stover yield than the local varieties. BVM 2 was specifically bred to meet the needs of the clients (resource-poor farmers with no access to irrigation) and has earlier maturity combined with higher grain yield. The returns were higher from this highly client-oriented approach, than by classical breeding, mainly because uptake was faster as a result of research and extension being done in tandem.