The studies of HLA in diabetes mellitus have strongly supported the subdivision into (at least) two genetically distinct groups: The juvenile onset (or perhaps better the insulin dependent) type DM, which shows a marked association to the HLA system, and the maturity onset type with a weak or no association to HLA. The HLA system has provided tools to test specific genetic models for the insulin dependent DM. A model involving one disease susceptibility locus closely linked to the HLA-loci and with a recessive susceptibility allele in linkage disequilibrium with some of the HLA factors has been proposed by others. This model cannot be totally rejected on the basis of the available data, but it is made less likely, especially because it leads to very low estimates of the penetrance which is discrepant from independent estimates, based both on studies on monozygotic twins and other families. In addition suggestive but not yet conclusive evidence of an excessive risk for HLA-Dw3/Dw4 heterozygotes is in conflict with this model.