The validity of self-reported hours in which one engages in activities strenuous enough to produce sweating was assessed as a measure of physical activity. Respondents were 732 randomly selected adults between the ages of 25 and 65 years from the Boston, Massachusetts, metropolitan area who participated in a field trial of health risk appraisal instruments in 1987. A total of 68% of the men and 57% of the women in the sample were involved in sweat-inducing activities at least once per week. The correlation between the natural logarithm of reported sweat hours per week and energy expenditure measured by the Harvard Alumni Activity Survey was 0.39. Following a logarithmic transformation and adjustment for age and sex, sweat hours was significantly correlated with high density lipoprotein cholesterol (r = 0.11, p less than 0.05). However, these associations were not as strong as those found for the age- and sex-adjusted log of the Harvard Alumni Activity Survey score (r = 0.19 and r = -0.15 for high density lipoprotein cholesterol and body mass index (weight (kg)/height(m)2), respectively) and are considerably weaker than those reported in other studies using sweat episodes (days per week on which sweating occurred) as an indicator of physical activity. These results suggest that the utility of self-reported sweat hours may be limited to distinguishing active from inactive subjects in epidemiologic surveys.