Analysis of the Fifth Genetic Analysis Workshop (GAW5) insulin-dependent diabetes mellitus (IDDM) data leads to the following conclusions: 1) With a maximum-likelihood affected sib pair method, there is strong evidence for linkage with HLA and no evidence for linkage with INS, Gm, or Km. 2) Susceptibility as defined by HLA genotypes is very complex. Each DR allele has a unique susceptibility, and DR3 and DR4 haplotype associations for DR 3/4 genotypes are different from those for 3/X and 4/X. 3) Risk is substantially higher in sibships with an affected father compared to those with an affected mother. This excess cannot be attributed to transmission distortion at HLA.