IAP1R grain sorghum Sorghum bicolor (L.) Moench random-mating population was constituted in 1973 using 80% adapted U.S. germplasm and 20% converted exotic sorghums. One hundred twenty half-sib and 120 C(,1) families chosen randomly from the C3 were evaluated in Experiment I at Ames (central Iowa), and a further sample of 120 S(,1) families was evaluated in Experiment II at Ames and Castana (western Iowa);Population means and genotype ranges for grain yield, components of grain yield, and other agronomic traits indicated that mass selection with enforced outcrossing had released latent variability through recombination and breakage of linkage blocks without deleterious effects on population means. Means and ranges were large enough to indicate that IAP1R should be a useful breeding population;Estimates of inbreeding depression at 100% homozygosity were significant for grain yield traits (-13.7%), seed number traits (-8.2%), and 100-seed weight (-4.9%). All other traits displayed additive gene action. Progeny mean estimates of heritability for S(,1) families were much lower for several traits in Experiment II than in Experiment I. The lower values resulted from large genotype-environment interactions and environments less favorable for expression of genetic variance in Experiment II. Heritability of grain yield/unit area was 0.74 in Experiment II vs. 0.85 in Experiment I. Estimates for the two experiments were similar for seed number traits (0.79 vs 0.77) and 100-seed weight (0.78 vs 0.82). Estimates of heritability on an individual plant basis for grain yield/unit area were lower in Experiment II (0.06 vs 0.13), but the estimates varied little for seeds/panicle (0.23 vs 0.20) and 100-seed weight (0.43 vs 0.41);Genetic correlations with grain yield/unit area were highest for seeds/panicle (0.53-0.65). The coefficients were low for panicles/plant (0.07-0.27) and 100-seed weight (0.32 to -0.16). Phenotypic correlations of grain yield/unit area were positive with height (0.41) and maturity (0.29);Gridded mass selection was estimated to be the most effective procedure for improving seed size. S(,1) testing seemed preferable for improvement of grain yield, seeds/panicle, and panicles/plant. Selection for yield components was estimated to be less effective than direct selection for grain yield.