Recurrent selection is a cyclical breeding technique that has become popular in comprehensive maize improvement strategies. The magnitude of response in recurrent selection depends upon the modes of gene action and their frequencies in effective population sizes. We studied four effective population sizes with the objectives: (1) to estimate response to selection and inbreeding effects, (2) to partition the response to selection into that due to additive and dominance genetic components, and (3) to examine changes in the correlated traits. Four effective population sizes 5, 10, 20, and 30 were developed by S1-progeny selection in the BS11 maize population using a 20% selection intensity designated as 5-S1, 10-S1 , 20-S1, and 30-S1 programs, respectively. After five cycles of selection in each effective size, population per se, populations per se selfed, populations top-crossed to B97 and B73, C0 x C1 of populations, diallel cross among C0, C3, and C5 of populations including five checks were tested in three Iowa locations in 1997 and 1998. There was positive and significant response for grain yield in 30-S1, 20-S 1, and 10-S1. The response in 5-S1 was significantly unfavorable due to genetic drift and inbreeding. The highest response for grain yield was 0.11 Mg ha-1 year-1 obtained in 30-S1, which was double that in 20-S1 and 10-S1. The S1 selection increased yield in 10-S 1, 20-S1, and 30-S1 due to favorable additive effects. In BS11C0 population both additive and dominance genetic effects were responsible for grain yield; however, the additive effect was greater than the dominance effect. Larger effective sizes (20-S1, 30-S 1) had loci with more favorable additive effects and showed better response to selection than the smaller effective sizes (5-S1, 10-S 1). The inbreeding depression effects were significant in all the populations as the result of genetic drift. Grain moisture, root lodging, stalk lodging, and all other correlated agronomic traits were improved in the desired direction. Direct and indirect responses were more favorable in the larger effective population sizes than in the smaller. The results suggest that effective size of 30 is preferable for recurrent selection programs planned for at least ten or more years.