A recurrent selection program in winter wheat (Triticum aestivum L. em Thell) was initiated by intermating 12 two, three, and fourway hybrids involving 16 cultlvars and lines to determine gain from selection for early heading, to develop genotypes which are earlier heading than currently available types, and to demonstrate the feasibility of recurrent selection in wheat improvement. The hybrid population produced by intermating these 12 hybrids was the nonselected base population. Three cycles of selection and intermating of S0 plants were carried out in the greenhouse. Hybrid plants from each of the three cycles, S1 plants or the selfed progeny from S0 plants selected to intermate in each cycle, plants of the base population, and the 16 parents were planted in the field at two locations in fall of 1979.The largest gain from selection was realized in the first selection cycle. Additional progress was made in the following two cycles, but at slower rates. Mean days to heading after 30 April for the base population were 14.2 and 23.4 at Sullivan and Lafayette, Indiana, respectively. Mean days to heading for hybrid populations from cycles 1, 2, and 3 at the 10% selection level were, respectively, 12.0, 11.7, and 11.2 at Sullivan; and 22.2, 21.6, and 21.6 at Lafayette.Realized heritability estimates from cycle 1 were 0.92 and 0.49 for the selected S0 populations, and 0.31 and 0.36 for the selfed progenies at Sullivan and Lafayette, respectively. The heritabilitles of cycles 2 and 3 were lower, 0.24 to 0.32, and were similar for the S0 and S1 generations. These data suggest that in the first cycle, selection acted largely on major genes which showed some nonadditive gene action, while selection in cycles 2 and 3 acted on minor genes governed by additive and/or additive × additive effects.Recurrent selection for early heading in winter wheat was effective in shifting the population means in the desired direction.