In contrast to the hypothesis that endogenous testosterone decreases plasma high-density lipoprotein (HDL) cholesterol levels, many, but not all, studies have reported a positive correlation between plasma total testosterone and HDL cholesterol. We examined behavioral correlates of plasma testosterone and estradiol and the relationships between these sex hormones and plasma lipoproteins, in middle-aged Japanese men. Plasma, lipids, including HDL subfractions, total and free testosterone, and total estradiol were determined with 313 men aged 50-54 years who received a preretirement health examination at the Self-Defence Forces Fukuoka Hospital from January to June in 1992. Body mass index and waist-hip ratio were also measured. Smoking habit, alcohol use, and physical activity were ascertained by a self-administered questionnaire. Obesity, especially waist-hip ratio, was a strong correlate of both total and free testosterone, but not of estradiol. Smoking was associated with elevated levels of testosterone without a dose-effect relation. Neither alcohol use nor physical activity was associated with total or free testosterone, but plasma estradiol levels were higher among current alcohol drinkers. HDL and HDL2 cholesterol were unrelated to either total or free testosterone in the univariate analysis, but negatively associated with free, not total, testosterone after adjustment for obesity. HDL and HDL2 cholesterol also were positively associated with estradiol regardless of adjustment for obesity and other covariates. These findings add to evidence for a hypothesis that high levels of endogenous testosterone and low estradiol levels may cause a decrease in plasma HDL cholesterol, thereby being linked with atherosclerosis in middle aged men.