We have carried out two independent family studies in low-income Mexican-Americans from San Antonio, Texas. In the first study, probands were ascertained at random without regard to any medical condition (658 examined individuals from 50 families), and in the second study, probands were subjects with type II diabetes identified in a prior epidemiological survey (523 examined individuals from 29 families). Pedigrees ranging in size from 2 to 45 family members (median 11) in the first study and from 2 to 50 family members (median 12) in the second study were examined. Diabetes was diagnosed according to World Health Organization criteria. In both sets of families, segregation analyses revealed support for a major gene with an autosomal dominant mode of inheritance influencing early age of onset of diabetes. Non-Mendelian inheritance was rejected in both data sets. Individuals with the early age of onset allele had a mean age of diabetes onset of 51 years in the first data set and 60 years in the second data set. In the first data set, the major gene accounted for approximately 70% of the phenotypic variance in age of onset of diabetes, and there were no residual family effects once the major gene effect was taken into account. In the second data set, the major gene accounted for approximately 50% of the phenotypic variance, and residual family effects were statistically significant. Linkage analyses were performed with 11 candidate genes, and tight linkage with diabetes was rejected for Rh blood group, glucose transporter 2, fatty acid-binding protein, tumor necrosis factor beta, glucokinase, and lipoprotein lipase. A logarithm of odds (LOD) score of 0.92 at a recombination fraction of 0.05 was observed for insulin receptor substrate 1. This LOD score corresponds to a chi2 of 4.24 (P = 0.039).