Among patients with insulin-dependent diabetes mellitus (IDDM), an excess of DR3 and DR4 alleles is classically described when compared with the general population. In addition, an excess of maternal DR3 and paternal DR4 alleles among patients (DR3DR4) is observed. In order to explain these observations, two alternative hypotheses can be tested: maternal effect and parental imprinting. Maternal effect has been tested and not rejected on a sample of 416 caucasians affected with IDDM. Under this hypothesis, the children of a DR3 mother are expected to have an earlier exposure and, hence, an earlier age at onset. However, we did not observe such a difference in age at onset in this data set. Using the marker-association-segregation-chi 2 method, we have tested four hypotheses with different parental effects of two susceptibility alleles, alpha 0 and beta 0, at two different closely linked loci. Under the hypothesis that best fitted the data, the probability of being affected depended on the parental inheritance of the susceptibility alleles, suggesting parental imprinting (i.e., differential role of maternal and paternal allele), without evidence for a cis-trans effect. We conclude that parental imprinting on a specific allelic combination may explain the observations on the HLA genotypes of the patients and their relatives.