Distributive properties of the second cycle hybrids that are produced by inter-crossing the recombinant inbreds extractable from the F2 of a cross between two pure breeding lines can be predicted from the early generations of the original cross. Hence the frequency of such hybrids that will outperform the extreme recombinant inbreds or the original F1 can be predicted. Basic generations and triple test cross families provide the most reliable estimates of the predictors and therefore should be used whenever possible although, in the presence of linkage, randomly mated F2's may give improved predictions. Simpler experiments consisting of the parental varieties and their F1 and F2 generations, however, provide all the information that is likely to be necessary for most practical purposes. The predictive power of the new approach is demonstrated on material extracted from the cross of varieties 1 and 5 of Nicotiana rustica. The predictors were estimated from the means and variances of V1, V5, F1 and F2 raised in six environments between 1973 and 1983. The predicted frequencies of second cycle F1's which outperform the extreme recombinant inbred lines derived from this cross are compared with those observed among 190 second cycle hybrids in a diallel between 20 recombinant inbreds derived from the same cross.