Predicting hybrid value in bread wheat: biometric modelling based on a "top-cross" design

Multi-environmental experimentation made it possible to compare 311 hybrids and their 59 parental lines for five characters: yield, two yield componen ts (number of grains per m(2) and 1000-grain weight), height and date of he ading. At all three sites, average high-parent heterosis exceeded 6%, and o ur results confirm that hybrids in general head as early as the earlier par ent does and are as high as the taller parent. For the five characters stud ied ecovalences were significantly lower in the Fl than in the parents, ind icating a higher level of homeostasis for hybrids. The experimental model, based on a "top-cross" design comprised of 55 lines used as females and fou r testers, made it possible to implement different hybrid-value prediction models. For yield, by far the most important character to predict, the pred ictor based on the general combining ability and the per se value of the pa rents, appears to be the most interesting. The prediction remains, however, imperfect as specific combining ability (SCA) has a strong impact for this character. Attempting to take account of SCA in the prediction, by using a multiplicative model for "malexfemale" interaction modelling, did not prov e convincing.