AbstractGenetic variance, heritability, and expected response from selection arc useful in devising alternative methods and criteria of: selection. The objectives of this study were to estimate these for seed yield and its components from 200 F2: populations involving 80 cultivars and lines of mostly small‐seeded dry bush bean (Phaseolus vulgaris L.) of habits growth I, II, and III of Middle‐American origin. All cultivars and lines were crossed in eight sets of ten parents each in a Design II mating system. The F2 populations, without parents, were evaluated in the field in a replicates‐in‐sets design at two locations in Colombia in 1983.Estimates of additive genetic variance were significant for yield, pods/m2, seeds/pod, and seed weight. Interaction with environments was also significant. Values for nonadditive genetic variance were not significant for either yield or yield components. The estimates of narrow sense heritability, based on the F2 population mean and unbiased by genotype x environment interaction, were 0.21 ± 0.13 for yield. 20 ± 0.13 for pods/m2, 0.57 ± 0.13 for seeds/pod, and 0.74 ± 0.15 for seed weight.The expected direct response from selection of the top 20 % of F2 populations for yield per se would result in a 4.30 % increase in yield with a correlated response of 0.21 % in seed weight. In contrast, the expected gain from direct selection for seed weight would result in a 11.76 % increase in seed weight with a, correlated gain of 0.28 % for yield. Direct selection for pods/m2 would decrease yield, seeds/pod and seed weight, while direct selection for seeds/pod would reduce pods/m2 and seed weight but increase seed yield by 0.37 %.Data on yield from replicated trials in the early segregating generations could be utilized for identification and selection of promising crosses and families or lines with crosses for dry bean yield improvement.