Both defective insulin secretion and insulin resistance have been reported in relatives of type 2 diabetic subjects. We tested 120 members of 26 families with a type 2 diabetic sibling pair with a tolbutamide-modified, frequently sampled iv glucose tolerance test to determine the insulin sensitivity index (S_I) and acute insulin response to glucose (AIR_glucose). A measure ofβ -cell compensation for insulin sensitivity was calculated as the product S_I × AIR_glucose, based on the demonstrated hyperbolic relationship between insulin sensitivity and insulin secretion. A percentile score for this compensation was assigned based on published values. Of the 120 family members, 26 had previously diagnosed impaired glucose tolerance on oral testing, and 94 had normal glucose tolerance tests. As a group, family members showed a significantly lower S_I × AIR_glucose than a similar, previously reported, control population, even when impaired glucose tolerance members were excluded. We performed a multivariate analysis of diabetes status, S_I, AIR_glucose, and to estimate the heritability of each trait and the genetic and environmental correlations between traits. We estimated the heritability of S_I × AIR_glucose to be 67± 3% when all members were included and 70 ± 4% when only normal glucose tolerance members were considered. Both AIR_glucose and S_I were also familial, albeit with lower heritabilities (38 ± 1% and 38 ± 2%, respectively, for all family members). Both S_I × AIR_glucose and S_I showed strong negative genetic correlations with diabetes (−85 ± 3% and −87 ± 2%, respectively, all family members), whereas AIR_glucose did not correlate with diabetes. We conclude that insulin secretion, as measured by S_I × AIR_glucose, is decreased in nondiabetic members of familial type 2 diabetic kindreds, that S_I × AIR_glucose in these high risk families is highly heritable, and that the same polygenes may determine diabetes status and a low S_I × AIR_glucose. Our data suggest that insulin secretion, when expressed as an index normalized for insulin sensitivity, is more familial than either insulin sensitivity or first phase insulin secretion alone and may be a very useful trait for identifying genetic predisposition to type 2 diabetes.