Water deficits result in large yield reductions on the one-half of the world's rice land that is rainfed. The most severe water deficits tend to occur during the reproductive stage. Field studies were conducted over a three-year period to develop a method to screen rices for reproductive-phase drought stress resistance. Staggered planting date was effective in synchronizing the flowering of test entries during the stress treatment period. Grain yield, relative grain yield, and spikelet fertility were significantly and negatively related to the number of days beyond the beginning of the stress period that a cultivar flowered. A drought index based on the deviation of entry performance from regression of the plant character on flowering date (expressed as days from application of the drought stress treatment) was used to adjust cultivar performance. Rankings were similar for the index values of the grain yield, relative grain yield, and spikelet fertility. Percent spikelet fertility was highly correlated with grain yield in the event of reproductive-stage water stress (r2= 0.64**, 0.84**and0.66** during three years of tests). Spikelet fertility was judged the most practical character by which to score cultivar performance. A mass screening method is proposed that involves staggered planting, interruption of irrigation during the flowering period, visual scoring for spikelet fertility, and regression analysis to adjust for differences in flowering date.