TESTCROSS SELECTION PRIOR TO FURTHER INBREEDING IN MAIZE - MEAN PERFORMANCE AND REALIZED GENETIC VARIANCE

Testcross evaluation of maize (Zea mays L.) families is usually done a fter, but not prior to, several generations of selfing. The objectives of this study were to (i) evaluate the usefulness of one cycle of tes tcross selection and recombination of F-3 families prior to further se lfing and (ii) develop a method for estimating realized genetic varian ce. In each of two biparental maize populations (POP1 and POP2), five Cycle 0 F-3 families that had the highest yields when testcrossed to B 73 were recombined to form Cycle 1. Testcrosses of Cycle 1, Cycle 0, a nd the five selected Cycle 0 F-3 families were evaluated at five locat ions in 1993. Selection differentials (S) for yield were 7 to 8% of th e corresponding Cycle 0 means. Observed response to selection (R) expr essed as a proportion of the Cycle 0 mean was 7% in POP1 and 5% in POP 2. in each population, the five highest-yielding Cycle 1 testcrosses o utyielded the five highest-yielding Cycle 0 testcrosses by 8 to 9%. Te stcross selection would have added 2 yr to the time required for inbre d development, but average gains per year for yield were 0.9% larger w ith than without selection and recombination of the best five F-3 fami lies in Cycle 0. Realized genetic variance was estimated as (R)<(sigma )over cap>(2)(G) = SR/i(2), where i was the standardized selection dif ferential with truncation selection. There was no evidence that geneti c variance decreased after one cycle of testcross selection. In experi ments that compare several cycles of selection, realized genetic varia nce can be estimated with bulks of all families and bulks of the selec ted families instead of estimating genetic variance with large numbers of individual families in each cycle.