In studies involving isozymes or restriction fragment length polymorphisms (RFLPs), correlations of parental molecular marker diversity with grain yield of maize (Zea mays L.) single-crosses have been too low to be of any predictive value. The relationship of molecular marker heterozygosity (Dij) with hybrid performance (μ ij) and combining ability was examined. For a simple genetic model involving uncorrelated parental allele frequencies and complete coverage of quantitative trait loci (QTL) by molecular markers, the correlations between μ ij and Dij were ≤0.25. μ ij and Dij were partitioned into general and specific effects. The expected correlation between specific combining ability and specific molecular marker heterozygosity is high. Expected correlations between general combining ability and general molecular marker heterozygosity are either positive or negative, depending on allele frequencies in the tester lines. Computer simulation was used to investigate a more complex but more realistic genetic model involving incomplete coverage of QTL by molecular markers. All of the following conditions are necessary for effective prediction of hybrid performance based on molecular marker heterozygosity: (1) dominance effects are strong; (2) allele frequencies at individual loci in the parental inbreds are negatively correlated; (3) trait heritability is high; (4) average parental allele frequencies vary only within a narrow range; (5) at least 30-50% of the QTL are linked to molecular markers; and (6) not more than 20-30% of the molecular markers are randomly dispersed or unlinked to QTL.