Correlation between testcross performance of lines at early and late selfing generations

In hybrid breeding programs, testcross evaluation of lines can be done during the early stages of selfing (early testing) or delayed until the lines are near-homozygous. To evaluate the usefulness of early testing, the expected genetic and phenotypic correlations between testcross performance at different selfing generations were examined. The genetic correlation (r GnGn' ) between testcross performance of S n and S n' , (n'>n) individuals or lines is equal to the square root of the ratio of their testcross genetic variances, and it is a function of the inbreeding coefficients (F) at the two selfing generations, i.e., r GnGn'=[(1+F n )/(1+F n )](0.5). The genetic correlation between testcross performance of lines and their directly descended homozygous (n'=∞) lines is 0.71 for S1; 0.87 for S2, 0.93 for S3, 0.97 for S4, 0.98 for S5, and 0.99 for S5 lines. The effectiveness of early testing is limited mainly by nongenetic effects. The square root of testcross heritability at generation n sets the upper limit on the correlation between phenotypic value at generation n and genotypic value at homozygosity. The probabilities of correctly retaining S n individuals or lines that have superior testcross performance at homozygosity (n'=∞) indicate that early testing should be effective in identifying lines with above- and below-average combining ability. However, the risk of losing lines with superior combining ability is high if strong (best 10%) selection pressure is applied during early testing. If only a small proportion of lines is retained based on testcross performance and/or if the heritability of the trait is low, selfing for two or three generations prior to testcrossing may be desirable to increase the likelihood of retaining lines that perform well at homozygosity. The theoretical results in this study support the testcross evaluation procedures for grain yield used by most maize (Zea mays L.) breeders.