Prediction of the means and genetic variances in segregating generations could help to assess the breeding potential of base populations. In this study, we investigated whether the testcross (TC) means and variances of F3 progenies from F1 crosses in European maize can be predicted from the TC means of their parents and F1 crosses and four measures of parental genetic divergence: genetic distance (GD) determined by 194 RFLP or 691 AFLP(TM) (1) markers, mid-parent heterosis (MPH), and absolute difference between the TC means of parents (∣P1-P2∣). The experimental materials comprised six sets of crosses; each set consisted of four elite inbreds from the flint or dent germplasm and the six possible F1 crosses between them, which were evaluated for mid-parent heterosis. Testcross progenies of these materials and 20 random F3 plants per F1 cross were produced with a single-cross tester from the opposite heterotic group and evaluated in two environments. The characters studied were plant height, dry matter content and grain yield. The genetic distance between parent lines ranged between 0.17 and 0.70 for RFLPs and between 0.14 and 0.57 for AFLPs in the six sets. Testcross-means of parents, F1 crosses, and F3 populations averaged across the six crosses in a particular set generally agreed well for all three traits. Bartlett's test revealed heterogeneous TC variances among the six crosses in all sets for plant height, in four sets for grain yield and in five sets for dry matter content. Correlations among the TC means of the parents, F1 crosses, and F3 populations were highly significant and positive for all traits. Estimates of the TC variance among F3 progenies for the 36 crosses showed only low correlations with the four measures of parental genetic divergence for all traits. The results demonstrated that for our material, the TC means of the parents or the parental F1 cross can be used as predictors for the TC means of F3 populations. However, the prediction of the TC variance remains an unsolved problem.