Early generation data consisting of F1 heterosis, F1, F2 and F3 mean performances, parent-offspring regression, and F2–F3 intergeneration correlation were used to identify potentially promising spring wheat (Triticum aestivum L. em Thell) crosses in terms of yield, kernel weight and protein content. The F1 test identified one high-yielding cross out of six showing significant level of higher parent (38%) and mid-parent (70%) heterosis for yield, respectively. The top yielding cross, viz. Glenlea × NB505, in F1 was also the second highest yielding population in F2 and the best yielder in F3 based on two locations. Hence, although F2 single plant productivity measurements misplaced the rankings of some of the crosses, F1 and F3 performances indicated the relative potential of the different populations. Most crosses showed nonsignificant F2–F3 regression and correlation coefficients for yield but significant coefficients for kernel weight. For protein content highly significant F2–F3 regression and correlation coefficients were observed only for crosses involving the high protein parent. The use in F1 of weight-free selection indices involving yield, kernel weight and protein content ranked Glenlea × NB505 as the best of six populations whereas in F3 the same cross had the best aggregate merit when only yield and kernel weight were considered in the index. The inclusion of protein content in the index favored a high protein cross, Sinton × Glenlea.Key words: Wheat, Triticum aestivum, heterosis, parent-offspring regression, intergeneration correlation, selection index