In winter wheat (Triticum aestivum L.), the development of a methodology to estimate genetic divergence between parental lines, when combined with knowledge of parental performance, could be beneficial in the prediction of bulk progeny performance. The objective of this study was to relate F2 heterosis for grain yield and its components in 116 crosses to two independent estimates of genetic divergence among 28 parental genotypes of diverse origins. Genetic divergence between parents was estimated from (a) pedigree relationships (coefficients of kinship) determined without experimentation, and (b) quantitative traits measured in two years of field experimentation in Kansas and North Carolina, USA. These distances, designated (1 -r) and G, respectively, provided ample differentiation among the parents. The 116 F2 bulks were evaluated at four locations in Kansas and North Carolina in one year. Significant rank correlations of 0.46 (P = 0.01) and 0.44 (P = 0.01) were observed between G and grain yield and kernel number heterosis, respectively. Although (1 -r) was poorly associated with grain yield heterosis, G and midparent performance combined to account for 50% of the variation in F2 yields among crosses when (1 -r) was above the median value, whereas they accounted for only 9% of the variation among crosses when (1-r) was below the median. Midparent and (1 -r) had equal effects on F2 grain yield (R (2)= 0.40) when G was greater than the median value. A breeding strategy is proposed whereby parents are first selected on the basis of performance per se and, subsequently, crosses are made between genetically divergent parents that have both large quantitative (G) and pedigree divergence (1 -r).