Protein meal and oil are the two commodities produced from soybean [Glycine max (L.) Merr.] that give the crop its value. Increasing seed concentrations of either or both may add value. Objectives of this study were to investigate the effectiveness of recurrent half-sib selection for increased seed oil, to evaluate the effect of tester oil content on selection response, and to investigate testcross heterosis and inbreeding depression for seed oil content. A recurrent half-sib selection system was devised for soybean and selection for increased oil content was conducted in a population for seven and three cycles using a high and a low-oil tester, respectively. The base population was a high-oil composite with gray pubescence (tt) that was segregating for nuclear genetic [ms.sub.t] male sterility. In summer, the base population was planted in single plant hills and bordered with the tester ([Ms.sub.1][Ms.sub.1]TT) in a random mating block in North Carolina. About 100 to 200 random male-sterile plants with hybrid seeds were harvested. Half-sib families derived from each male-sterile plant were then grown in Puerto Rico in winter. At maturity, seeds from tawny plants (tester hybrid) were used to identify half-sib families with high-oil content. Corresponding gray plant hybrids from sib matings within the population were hulked to start the next cycle of selection. Random progenies from the base populations and selected progenies from each cycle of selection were evaluated in a replicated field experiment at three locations in North Carolina. Cycle x tester hybrids and cycle x cycle sib hybrids were also included in the tests. The results showed that oil content was significantly increased at a rate of 1.1 [+ or -] 0.2 g [kg.sup.-1] [cycle.sup.-1] in the high-oil tester populations but not in the low-oil tester populations. The realized heritability estimate for the high-oil tester population was 0.12 [+ or -] 0.03. Evidence of heterosis indicated that some dominance effects on oil content existed. Dominance effects may affect the evaluation accuracy of the genotypes being tested. A high-oil tester and high-oil populations may have many common alleles resulting in less dominance and more additive effects in their hybrids. Consequently, a high-oil tester can lead to better evaluation and selection precision, compared with a low-oil tester which could mask additive effects and reduce selection precision.