It has been postulated that the positive effects of physical activity on high density lipoprotein cholesterol (HDL-C) and HDL-C subfraction 2 (HDL-C2) are mediated through insulin action because increased activity lowers insulin levels and lower insulin levels are associated with higher HDL-C. These relations were evaluated in a rural population of Hispanic and non-Hispanic white (NHW) adults in Colorado. Included were 138 men and 152 women with normal glucose tolerance confirmed by World Health Organization criteria. Total physical activity was assessed by 7-day recall interviews. No significant associations were observed among women. Among men, activity was inversely associated with fasting insulin (r = -0.17, p less than 0.05). From analysis of covariance models including the interaction term activity x ethnicity, total HDL-C was 43.4 mg/dl (95% confidence interval [CI] = 39.1, 47.7) in the low tertile of activity and 50.4 mg/dl (95% CI = 46.3, 54.5) in the high tertile for NHW men, after adjustment for fasting insulin, fasting glucose, body mass index (BMI), waist to hip ratio (WHR), and age. For Hispanic men, adjusted HDL-C was 43.4 mg/dl (95% CI = 38.6, 48.2) and 49.1 mg/dl (95% CI = 44.0, 54.2) in the low and high tertiles, respectively. Adjusted HDL-C2 levels were 52% higher in the most compared with the least active NHW men, whereas there was no difference by activity for Hispanic men. Higher adjusted mean levels of HDL-C3 were observed for the high compared with the low activity tertile in both ethnic groups. Ethnicity-specific models showed that for NHW men, activity explained 12% (p = 0.01), fasting insulin explained 5% (p = 0.05), and BMI explained 6% (p = 0.04) of the variability in total HDL-C, after adjustment for fasting glucose, WHR, and age. These models confirmed that effects of insulin and body fat did not explain the observed associations between activity and total HDL-C and its subfractions.