The prevalence of overweight in men varies with socioeconomic and marital status. To explore the origin of these associations, we studied the effects of family income, education, and changing marital status on change in body mass index (BMI, kg/m2) over 10 years in a representative sample of US men. The subjects were 1552 white and black US men who entered the Health and Nutrition Examination Survey-I (in 1971-75) at ages 25-44 and were re-evaluated a decade later (in 1982-84). After adjusting for socio-demographic factors, baseline BMI, smoking and physical activity, the mean 10-year change in BMI was greater for men with 12 years of education (difference = 0.31 BMI units (95 percent CI 0.04-0.59) ) or with less than 12 years (difference 0.58 (0.22-0.94) ) compared with men who studied beyond 12th grade. Ten-year weight changes were also defined categorically as major weight gain (BMI change greater than or equal to +4 units) or major weight loss (BMI change less than or equal to -2 units). By multiple logistic regression analysis, the odds of experiencing major weight gain were independently associated with low family income (odds ratio (OR) = 1.8 (95 percent CI, 1.0-3.3) ) compared with favorable income, and with becoming married (OR = 3.3 (1.7-6.3) ) or remaining unmarried (OR = 2.1 (1.1-4.2) ) compared with men who were consistently married. The risk of major weight loss was independently associated with marriage ending (OR = 1.8 (1.0-3.3) ) or with remaining unmarried (OR = 2.5 (1.3-4.7) ). A US public health strategy for the prevention of men's weight gain should focus on men with little education or low family incomes and those who are unmarried. There may also be a benefit to preventive weight-gain counseling for men as they enter marriage.