Certain maize (Zea mays L.) inbred lines are more successful than others in forming elite hybrids. This study was conducted to determine whether epistatic interactions play a significant role in hybrid performance. Statistical epistasis was measured with a modified generation means model using testcrosses. Six progeny generations (P1, P2, F1, F2, and a backcross from the F1 to each parent) were produced for all possible hybrids of a five-parent diallel in both the Iowa Stiff Stalk Synthetic (BSSS) and non-BSSS heterotic groups. Two testers were hybridized to each of the 10 possible hybrid progeny sets in both groups. Each testcross progeny set was evaluated in 10 environments. The nonepistatic model accounted for a large amount of the variation in generation means and fit the data well. Of the 40 maize testcross progeny sets studied, five resulted in a significant epistatic effect for grain yield. Four of the significant epistatic effects showed evidence of linkage, while one was due to unlinked epistatic effects. Our results suggest that parents in a hybrid cross need to be significantly different and testers need to bring out those differences to detect epistasis better by means of testcross generation means.