IntroductionIn much animal production, commercial animals are crossbreds from closed lines or breeds under long‐term directional selection. Therefore it is desirable to be able to predict heterosis gains over the generations as it is done for genetic progress under within‐line selection. However, heterosis is the phenotypic expression of a complex phenomenon which may involve several types of genetic effects like dominance and epistasis. In animal breeding, basic quantitative genetics theory indicates that heterosis should be proportional to (squared) differences in gene frequency between populations (e.g. F alconer and M acK ay 1996), and it has been found approximately correct, so it is commonly used for planning crosses. Under that type of heterosis, however, selection towards the same objective in two populations should bring gene frequencies closer, and therefore it should eventually decrease heterosis. On the other hand, reciprocal recurrent selection designed to increase genetic distance between lines should eventually achieve maximum heterosis (O llivier 1982). Some experiments reviewed by brun (1982) have already compared genetic progress under within‐line and reciprocal recurrent selection, but they did not focus on comparing the trend of heterosis with generations between the two selection methods. Also, heterosis was monitored periodically in some selection experiments on poultry, and results were reviewed by F airfull (1990). They were somewhat contradictory, but they generally failed to relate genetic progress to loss of heterosis under within‐line selection. Moreover, in commercial production, as purebreds and crossbreds are not contemporaries and are generally maintained under very different management systems, estimations of heterosis and of the evolution of crossbred advantage over the generations may be difficult to obtain.Using the Japanese quail as an experimental animal, the present work was initiated specifically to follow the changes in heterosis brought about by selection for a single heterotic trait, early egg production (M invielleet al. 1995). For that purpose, two selection methods expected to have opposite effects on heterosis, directional within‐line (or individual) selection and reciprocal recurrent selection, were applied for 13 generations in four quail lines started from two different origins, and trends of heterosis in the selected character and in weight and egg traits were evaluated.