Abstract:
Three crosses of soybean involving United States and other germplasm were examined in the F3 and F4 generations at two locations, one near the coast, the other on the Darling Downs. Line and line x location interaction variances within crosses for most traits were almost all significant, except for protein, oil, and protein plus oil yield where line variances were non-significant when data were combined over locations. Line variances were considerably smaller for all traits when estimated across, rather than within, locations. For protein, oil, and protein plus oil percentage, the line x location interaction variance component generally was smaller than the line variance component in two crosses, but larger in a third even though the latter cross had the largest line variance for these traits. The line x location interactions may have been related in some cases to lower protein and oil percentages of late-maturing lines in that location where cold conditions prevailed late in the season. Heritability percentages computed from F4 variance components ranged from 49 to 89% and were largest for the chemical pexentage traits within locations. Standard unit heritability percentages estimated from F3-F4 generations were lower than those computed via the variance component method, and approached zero for oil percentage, protein plus oil percentage, and the chemical yield traits in some crosses if the generations were grown in different locations. Genotypic correlations between protein percentage and seed yield ranged from - 0.50 to 0.16 across crosses and were opposite in sign to those between oil percentage and seed yield. Genotypic and phenotypic correlations between protein and oil percentage were highly negative in two crosses (-0.62 to -0.96) but much less negative in the third cross (-0.24 to -0.35). Predicted genetic advance (percentage of population mean) ranged from 5.9 to 26.1 % for the chemical yield traits and from 1.5 to 6.3 % for protein or oil percentage. Indirect selection for chemical yield through seed yield was 90-109 % as efficient as direct selection, and a selection index that combined seed yield and either protein, oil, or protein plus oil percentage generally was no more efficient than direct selection for the corresponding chemical yield trait. Indirect selection for chemical yield through chemical percentage was much less efficient than direct selection for chemical yield.