Abstract
Maximum likelihood methods have been used to compare the fit of twenty different genetic models to experimental data on fourteen characters, each measured on two parental strains, F(1) hybrids and both backcrosses. Although variation in all characters was continuous, differentiation between the various models was meaningful, the mean likelihood ratio between the best and worst models for each character being greater than 10(4). Models with only one or two loci were adequate to account for the observed genetic variation in eleven of the fourteen characters. These results indicate that even in species without special genetic advantages, it may be possible to identify individually some of the genes responsible for naturally-occurring variation within the range of normality.