Insulin resistance and hyperinsulinemia are familial traits that may precede and predict the onset of non-insulin-dependent diabetes mellitus (NIDDM). In some populations, the distribution of fasting insulin levels and measures of in vivo insulin action suggest the effects of a single major gene. We previously noted hyperinsulinemia among unaffected members of 16 large white pedigrees ascertained through two or more NIDDM siblings. To examine the hypothesis that insulin levels are determined by a single major genetic locus, we used segregation analysis to examine fasting insulin levels in 206 family members and 65 spouses who had normal glucose tolerance tests by World Health Organization criteria. Segregation analysis supported a major locus determining fasting insulin levels and segregating as an autosomal recessive allele with a frequency of 0.25. Thus, homozygotes represented 6.25% of the population, and homozygosity for the hyperinsulinemia allele elevated the mean fasting insulin level from 70.3 to 211.1 pM (11.7-35.2 microU/ml). The analysis apportioned the variance in fasting insulin as 33.1% due to the major autosomal locus, 11.4% due to polygenic inheritance, and 55.5% due to unmeasured effects. Homozygotes for the recessive allele had higher 1-h insulin levels than all others (911.7 vs. 427.2 pM [152.0 vs. 71.2 microU/ml]). We also found evidence for a major locus determining 1-h-stimulated insulin levels, with codominant inheritance as the most likely pattern in inheritance. The causal relationship between these findings and NIDDM has not been determined, and segregation of direct measures of insulin action remains to be demonstrated.(ABSTRACT TRUNCATED AT 250 WORDS)