This paper describes how plant breeders and farmers worked together to produce improved varieties of maize for the low-resource farmers of the Panchmahals district of Gujarat, India. Initially, farmers tested a range of maize varieties in a participatory varietal selection (PVS) programme. However, none of these proved to be very popular with farmers, although farmers who had more fertile fields adopted the variety Shweta from Uttar Pradesh. Hence, in 1994 a participatory plant breeding (PPB)programme was begun to generate new, more appropriate varieties. Yellow- and white-endospermed maize varieties were crossed that had been either adopted to some extent following PVS or had attributes, such as very early maturity,that farmers had said were desirable. In subsequent generations, the population was improved by mass selection for traits identified by farmers. In some generations,farmers did this in populations which were grown by breeders on land rented from a farmer. Soil fertility management was lower than that normally used on the research-station. The breeding programme produced several varieties that have performed well in research-station and on-farm trials. One of them, GDRM-187, has been officially released as GM-6 for cultivation in hill areas of Gujarat state,India. It yielded 18% more than the local control in research-station trials, while being seven days earlier to silk. In farmers' fields, where average yields were lower, the yield advantage was 28% and farmers perceived GDRM-187 to have better grain quality than local landraces. PPB produced a variety that was earlier to mature than any of those produced by conventional maize breeding, and took fewer years to do so. The returns from PPB,compared to conventional breeding, are higher because it is cheaper and benefits to farmers are realised earlier.