The genetics of insulin-dependent diabetes mellitus (IDDM) is currently an area of controversy, with some investigators proposing heterogeneity within the HLA region and even the existence of non-HLA-linked susceptibility genes, and others maintaining that a simple autosomal recessive gene linked to HLA with reduced penetrance is an adequate explanation. To resolve this latter question, we report here a simple method of testing whether a single HLA-linked susceptibility gene is inherited in a recessive fashion when it is associated with two different HLA alleles, as is the case for IDDM. It is shown that if the number of DR3/DR4 heterozygotes in a diabetic population exceeds the combined sum of DR3/3 and DR4/4 homozygotes in that same diabetic population, then a recessive mode of inheritance can be rejected. The advantages of the method are that it does not depend on ratios, as do relative risk calculations, nor does it depend on control data, but is based only on studies of the diabetics themselves. With data on 193 genotyped IDDM patients, we can clearly reject the recessive mode of inheritance, since the number of heterozygotes is 68 compared with a maximum of 22 homozygotes (P < 10⁻⁴). Eight other published studies are in concordance with these results. Therefore, a nonparametric test, independent of the significance of any individual study, rejects equality of heterozygotes and homozygotes (P < 0.002) and rejects the simple recessive mode of inheritance as a direct consequence. We conclude that more complex modes of inheritance of HLA-linked IDDM susceptibility, such as two different diabetogenic alleles or multiple loci, must be entertained.