Anorexia nervosa is a severe eating disorder characterised by restricted eating, the relentless pursuit of thinness and obsessive fears of being fat. The involved risk factors are probably numerous, but the existence of a genetic vulnerability has been proposed for decades. The heritability in the broad sense is computed on the basis of aggregation studies, treated twin samples and twin studies from the general population. Many difficulties make this heritability estimation problematic, but the convergence of the results (from family studies and two types of twin studies) gives the most convincing evidence in favour of a major role of genetics in the vulnerability to anorexia nervosa, with a heritability around 70%. Regarding the analysis of candidate genes, the most frequently studied is the 5-HT(2A) gene, with positive and negative results. We thus propose a meta-analysis showing that a large heterogeneity between samples exists, but the main effect of the -1438A allele persists even when extracting this contaminating effect (p=0.003). Furthermore, the absence of significant correlation between odds ratio and time after first publication of each sample, and size of each sample, is in accordance with the fact that the A allele is a risk factor. In order to explain the high heterogeneity between the nine studies yet performed, an alternative explanation such as a "modifying the phenotype" effect is proposed.