AbstractSix pearl millet (Pennisetum glaucum[L.] R. Br.) composites were subjected to three or more cycles of selection in multilocational yield trials and downy mildew disease nurseries in India. The base and selected populations were tested (a) over four years at three locations in India (11° to 29°N), (b) under terminal‐drought and optimal moisture conditions for two years, and (c) under induced downy mildew infections to determine the impacts of selection on grain yield, agronomic traits, and resistance to downy mildew (Sclerospora graminicolaSacc. Schroet.). Mean grain yield increases for four composites undergoing three to six cycles of selection ranged between 23 to 94 kg ha‐1cycle‐1(0.9 to 4.9 % cycle‐1) which, averaged over composites, amounted to 3.3 % cycle‐1. Yield gains were generally expressed at all test locations and under both terminal‐drought and optimal moisture environments. The gains in grain yield were associated with increases of both biomass and harvest index without extending the growth duration except in the earliest composite. Susceptibility to downy mildew remained below 10 % in all selected populations. Thus, the effectiveness of recurrent selection for increasing yield and yield stability of pearl millet is clearly