Response to three cycles of recurrent selection for yield in a soybean Glycine max (L.) Merrill population was evaluated. The source population was derived from three generations of intermating among forty high-yielding lines and cultivars of maturity groups 0 to IV. The parents of the cycle 1 population were identified by testing 300 F(,5)-derived lines of the cycle 0 population, and selecting the 10 early, mid-season, and late lines with the highest yield. The 30 parents were recombined in diallel and 300 F(,4)-derived lines of the cycle 1 population were tested for yield. The 10 highest-yielding lines of early, mid-season, and late maturity were chosen as the parents of cycle 2. The 10 parents of each maturity were mated in a separate diallel, 100 F(,4)-derived lines from each maturity were tested for yield, and 10 lines with the highest yield of each maturity were chosen as parents of cycle 3. Each cycle of selection required 2 years;Response to selection was evaluated by testing the parents of cycles 0, 1, 2, and 3 as three maturity sets, early, midseason, and late, at six Iowa environments in 1979 and 1980. Response to selection also was evaluated by testing composites of parents from each cycle and composites of unselected lines from each cycle;There was a significant increase in yield based on the mean of parents of each cycle of 120 (+OR-) 10 kg/ha per cycle for the early and 24 (+OR-) 9 kg/ha per cycle for the late sets, but no significant change in yield for the midseason (-14 (+OR-) 8 kg/ha per cycle) set. Response to selection evaluated using composites of parents within each maturity set indicated similar results to those obtained from testing individual parents. Improvement in yield was observed only for the late set when selection was evaluated using composites of 100 unselected lines within each maturity set. No improvement in yield was observed when evaluation was based on composites of parents across maturity sets, or composites of 300 unselected lines across maturity sets;There was no consistent change in maturity, lodging, plant height, seed size, protein percentage, or oil percentage caused by selection for yield.