A number of breeding institutions developed a project to assess importance of participatory plant breeding approaches for rainfed rice improvement in eastern India. The results of the first two years of participatory varietal selection are reported here. The objective was to evaluate the respective effects of participation of farmers in varietal evaluation and decentralization of varietal testing from breeding stations to farmers' fields on varietal ranking. Fields representing various hydrological situations were chosen in two to three villages at four rainfed lowland sites and one upland site. Sets of 15 to 25 varieties were tested both in farmers' fields and on-station in 1997 and 1998 and ranked by both farmers and breeders. The effect of participation was judged by comparing the rankings attributed by farmers and breeders to a given set of material in a given trial. The effect of decentralization was determined through comparisons between individual breeders' rankings across trials. Farmers' rankings were not randomly allocated, but agreement within the farmers' group was not always very strong. Except at one site, concordance among breeders' ranking was high, but, because of the limited number of breeders involved, it was seldom significant. In about two-thirds of the trials, there was a good agreement between farmers' and breeders' mean rankings. The consensus was particularly strong when severe constraints induced contrasting behavior in the genotypes. The decentralization effect appeared to be moderate, but variations due to a breeder effect were recorded. The part of genotype by environment interactions for grain yield due to location within one site and year was evaluated through various methods, showing more effect of G x E interactions at some sites than at others. Crossover interactions inducing changes in ranks represented a limited part of the yearly G x E interactions at all sites. Both farmer participation and decentralization of varietal testing i