Insulin-dependent diabetes mellitus (IDDM) HLA class II DRB1-DQA1-DQB1 data from four populations (Norwegian, Sardinian, Mexican American, and Taiwanese) have been analyzed to detect the amino acids involved in the disease process. The combination of sites DRB1#67 and 86; DQA1#47; and DQB1#9, 26, 57, and 70 predicts the IDDM component in these four populations, when the results and criteria of the haplotype method for amino acids, developed in the companion paper in this issue of the Journal, are used. The following sites, either individually, or in various combinations, previously have been suggested as IDDM components: DRB1#57, 70, 71, and 86; DQA1#52; and DQB1#13, 45, and 57 (DQB1#13 and 45 correlates 100% with DQB1#9 and 26). We propose that DQA1#47 is a better predictor of IDDM than is the previously suggested DQA1#52, and we add DRB1#67 and DQB1#70 to the HLA DR-DQ IDDM amino acids. We do not claim to have identified all HLA DR-DQ amino acids-or highly correlated sites-involved in IDDM. The frequencies and predisposing/protective effects of the haplotypes defined by these seven sites have been compared, and the effects on IDDM are consistent across the populations. The strongest susceptible effects came from haplotypes DRB1 *0301/DQA1 *0501/ DQB1*0201 and DRB1*0401-5-7-8/DQA1*0301/ DQB1*0302. The number of strong protective haplotypes observed was larger than the number of susceptible ones; some of the predisposing haplotypes were present in only one or two populations. Although the sites under consideration do not necessarily have a functional involvement in IDDM, they should be highly associated with such sites and should prove to be useful in risk assessment.