From the Insulin Resistance Atherosclerosis Study (IRAS), 1173 men and women of African-American, non-Hispanic white, and Hispanic ethnicity with no history of diabetes were included in an evaluation of the cross-sectional relation of habitual dietary fat intake with insulin sensitivity (SI) as assessed by minimal-model analysis of a 12-sample, insulin-modified frequently sampled intravenous-glucose-tolerance test. Dietary intake was measured by a food-frequency interview modified to enhance sensitivity to food choices within the three ethnic groups. Percentage of energy from total fat was associated with rank of SI (SI(rank); r = -0.06, P = 0.03), with log fasting insulin (r = 0.10, P < 0.001), and with BMI (r = 0.10, P < 0.001). Multiple-linear-regression models included SI(rank) as the dependent variable, dietary fat (g/d) as the primary independent variable adjusted sequentially for total energy, other covariates, body mass index, and waist-hip circumference ratio (WHR). For all subjects combined, total fat intake was inversely related to SI(rank), but this association was not significant (P = 0.14) and was attenuated by adjustment for body mass index and WHR (P = 0.44). The association of total fat (g/d) with SI(rank) differed significantly (P < 0.01) for obese compared with nonobese individuals. Higher fat intake was associated with lower SI(rank) among obese (beta = -1.40, P = 0.03) but not among nonobese persons (beta = 0.16, P = 0.80). Among the obese (body mass index < or = 63), adjustment for body mass index largely accounted for the observed association of dietary fat with SI(rank). These findings were generally consistent for monounsaturated, polyunsaturated, and saturated fats. Among individuals already at increased risk for non-insulin-dependent diabetes mellitus because of obesity, high intake of dietary fat may worsen insulin sensitivity. This effect may be mediated by the relation of dietary fat to obesity.