Many investigators have attempted to confirm the prediction that increased levels of heterozygosity entail greater developmental stability, manifesting itself through decreased phenotypical variation. The evidence presented so far is equivocal. The predicted relationship has been found in some morphological studies, but not in others. I propose that the variability of a character should be seen as different from the character itself. For most morphological characters, natural selection promotes strong canalization of development but, to facilitate responses to environmental changes, the organism needs to retain malleability of physiological and behavioral traits. These different types of selection should lead to distinct genetic architectures for these phenotypes. I report on the results of a diallel cross between four inbred mouse strains. Qualitatively different genetic architectures were in fact revealed for variation in behaviors in the open-field. In a second study, variances of inbred and hybrid populations for hippocampal morphometry were studied. Again, hybrids were not always less variable than inbreds and sometimes even more variable. It follows that there exists no one-to-one relation between heterozygosity and developmental stability.