We examined the physical activity and other life-style characteristics of 16,936 Harvard alumni, aged 35 to 74, for relations to rates of mortality from all causes and for influences on length of life. A total of 1413 alumni died during 12 to 16 years of follow-up (1962 to 1978). Exercise reported as walking, stair climbing, and sports play related inversely to total mortality, primarily to death due to cardiovascular or respiratory causes. Death rates declined steadily as energy expended on such activity increased from less than 500 to 3500 kcal per week, beyond which rates increased slightly. Rates were one quarter to one third lower among alumni expending 2000 or more kcal during exercise per week than among less active men. With or without consideration of hypertension, cigarette smoking, extremes or gains in body weight, or early parental death, alumni mortality rates were significantly lower among the physically active. Relative risks of death for individuals were highest among cigarette smokers and men with hypertension, and attributable risks in the community were highest among smokers and sedentary men. By the age of 80, the amount of additional life attributable to adequate exercise, as compared with sedentariness, was one to more than two years.