Anorexia nervosa is a severe and complex disorder with incompletely known vulnerability factors. It is generally recognized that anorexia nervosa is a familial disorder, but the majority of twin studies have shown that the concordance rate for monozygotic twins is higher (on average 44%) than for dizygotic twins (on average 12.5%). This difference in concordance rates shows that genetic factors, more than common familial environment, may explain why the 'anorexia nervosa' phenotype runs in families. In order to estimate the heritability in the broad sense of anorexia nervosa according to published familial and twin studies, we first assessed the intrapair correlation between monozygotic and dizygotic twins, and secondly calculated the deviation threshold of relatives of affected probands from the relative mean. In this review, we obtained an estimation of the heritability at 0.72 according to all published controlled familial studies (six references quoted in MEDLINE(R)), and 0.71 for all published twin studies (59 references quoted in MEDLINE(R)). This estimation is close to the ones previously proposed, between 0. 5 and 0.8. Familial and twin studies may also help to define the boundaries of the phenotype, shedding light on the complex relationship between anorexia nervosa on the one hand, and bulimia nervosa, mood disorders, and alcoholism on the other. Demonstrating the importance of genetic factors in anorexia nervosa, and more specifically for anorexia of the restrictive type, requires not only prospective and adoption studies (which are still lacking), but also genetic polymorphisms analyses, which began very recently.