SELECTION AMONG S(1) FAMILIES VS SELFED HALF-SIB OR FULL-SIB FAMILIES IN AUTOGAMOUS CROPS

In autogamous crops, half- and full-sib families are routinely generated in various systems of intermating. Usually, these families have not been used as units of selection, due to low expected gains relative to other selection units (e.g., S1 families). The use of selfed half- (SHS) or full-sib (SFS) families, and differences in phenotypic variance between these and S1 families, might reduce the difference in expected gains between the selection methods. Our primary objective was to predict the expected gain from selection on each family unit, based on a genetic model with additive and homozygous dominance effects. A second objective was to empirically estimate phenotypic variances among SHS, SFS, and S, families derived from random-mating populations of wheat (Triticum aestivum L.) and soybean [Glycine max (L.) Merr.], and use those estimates to compare expected gains. Estimates of phenotypic variance for yield of SFS wheat families (o(SFS)2) were small compared with estimates for S1 families (o(S1)2) due to smaller error variances. Phenotypic variances of SHS soybean families (o(SHS)2) tended to be smaller than those of S1 families for seed yield and size but not for seed protein and oil concentrations. As with the wheat populations, error variances tended to be smaller for SHS families. The expected relative efficiencies of S1 family selection vs. SFS or SHS family selection are 1.14 (sigma(SFS)/sigma(S1)) in wheat and 2.29 (sigma(SHS)/sigma(S1)) in soybean. Substituting phenotypic standard deviations into each formula resulted in no consistent advantage to using S1 family selection with the exception of seed protein and oil in soybean. In conclusion, selection among SHS or SFS families should be of greatest benefit for productivity traits that require larger seed quantities for testing. These selection units allow larger plot sizes and/or increased replications, thereby improving the precision of entry-mean estimates and reducing phenotypic variance.