A cross-classification (North Carolina Design II) mating design was used to produce half-sib and full-sib families in three hybrid maize (Zea mays L.) populations: BSSSCO x BSCB1C0, BSSS(R)C9 x BSCB1(R)C9, and BS13(S)C3 x BSCB1(R)C9. Families were evaluated In a replicatlons-wlthin-sets, randomized Incomplete block design with two replications at five locations. Data were collected for grain yield, grain moisture, root and stalk lodging, dropped ears, plant and ear height, days from planting to anthesis and silking, and number of tassel branches. The objective was to estimate genetic variance components, heritabillties, genetic correlations, and genetic gains for the three hybrid populations. The additive genetic variance component was the most Important component of genetic variability for all traits. Except for grain yield, additive by environment Interpopulation variance estimates were smaller than their corresponding additive variances. For grain yield, the additive variance components increased with selection; however, the original population cross showed greater additive by environment interaction variance than the improved population crosses. The estimates of additive genetic variance Increased for plant and ear height and decreased for root and stalk lodging with reciprocal recurrent selection for grain yield. Except for grain yield, all traits exhibited a decrease In estimates of the dominance variance component after recurrent selection. The BSSS additive genetic contribution to the total interpopulation additive variance became more important with selection except for days to anthesis, days to silking, and number of tassel branches. Heritabiltty estimates for grain yield increased with recurrent selection. Herltability estimates of the original and selected cross populations were similar for the other traits. Grain yield had positive genctyplc correlations with grain moisture, plant height, and ear height and negative correlations with stalk lodging. Correlations between traits were not affected by selection for yield. The results suggest that reciprocal recurrent selection was an effective method for improving the population cross for yield without decreasing genetic variability.From the Department o( Agronomy, Iowa State University, Ames, IA 50011. This Is a contribution of the Department of Agronomy and journal paper no. J-16352 of the Iowa Agricultural and Home Economics Experiment Station, Ames, project 3082, and part of a dissertation submitted by F. J. B. In partial fulfillment of the requirements for a PhD degree.