When comparing the odds ratios (OR's) obtained by contrasting various HLA‐DR phenotypic classes in patients (e.g. diabetics) and controls, it is desirable to use the same reference phenotypes for all OR's. If this is done, it can be shown that the OR for, say, DR3/4, heterozygotes cannot exceed the OR's for both of the two homozygotes (DR3/3 and DR4/4), if there is one disease susceptibility locus with one normal allele and one susceptibility allele in linkage disequilibrium with two HLA‐DR alleles {HLA‐DR3 and 4) and if the action of the susceptibility gene is dominant, recessive or intermediate between dominant and recessive. In each of these cases, the OR for the heterozygotes will be intermediate between the OR's of the two homozygotes. If the two alleles at the susceptibility locus act in an overdominant way, the OR for the heterozygotes may in some cases exceed the OR's of both homozygotes. Comparison of the magnitude of two OR‐values can be reduced to investigating whether the OR value generated by comparing one phenotypic class against the other (as reference group) differs from unity. There is suggestive but not conclusive evidence that the OR of developing insulin‐dependent diabetes for the DR3/4 genotype is higher than the corresponding OR‐values for both of the two homozygotes (DR3/3 and DR4/4), indicating that the susceptibility to this disease may not be explained by a dominant, recessive or intermediate model.