Estimated Gains from Selection for Protein and Yield in the F 4 Generation of Six Soybean Populations 1

Seventy‐eight F 2 ‐derived F 4 lines from each of six soybean [ Glycine max (L.) Merr.] populations were evaluated to determine which population had the best potential for improving yield, protein, and yield and protein in combination. The six populations, one high protein ✕ high protein, four high protein ✕ high yield, and one high yield ✕ high yield, were derived from diallel crosses among two adapted high‐protein (P 1 and P 2 ) and two adapted high‐yield parents (Y 1 and Y 2 ). P 1 P 2 contained more lines high in percent protein and in protein per hectare, contained more lines that combined high protein with high yield, showed the greatest expected advances for yield and protein per hectare, was exceeded only by P 1 Y 2 for predicted progress in percent protein, and was the only population in which expected advance exceeded parental means for percent protein and protein per hectare. Its highest‐yielding lines, however, were below the highest lines in other populations, particularly P 1 Y 1 and P 2 Y 2 . Genotypic and phenotypic correlation coefficients were computed among yield, percent protein, and percent oil. Highly significant negative correlations were observed for protein and oil in all six populations. Correlations between yield and percent oil were positive, but significant in only two populations. Significant coefficients between yield and percent protein were negative in P 1 P 2 and P 2 Y 2 and positive in Y 1 Y 2 .