The relationship of low serum cholesterol and mortality was examined in data from the NHANES I Epidemiologic Followup Study (NHEFS) for 10,295 persons aged 35-74, 5833 women with 1281 deaths and 4462 men with 1748 deaths (mean (followup = 14.1 years). Serum cholesterol below 4.1 mmol/l was associated with increased risk of death in comparison with serum cholesterol of 4.1-5.1 mmol/l (relative risk (RR) for women = 1.7, 95% confidence interval (CI) = (1.2, 2.3); for men RR = 1.4, CI = (1.1, 1.7)). However, the low serum cholesterol-mortality relationship was modified by time, age, and among older persons, activity level. The low serum cholesterol-mortality association was strongest in the first 10 years of followup. Moreover, this relationship occurred primarily among older persons (RR for low serum cholesterol for women 35-59 = 1.0 (0.6, 1.8), for women 70-74, RR = 2.1 (1.2, 3.7); RR for low serum cholesterol for men 35-59 = 1.2 (0.8, 2.0), for men 70-74, RR = 1.9 (1.3, 2.7)). Among older persons, however, the low serum cholesterol-mortality association was confined only to those with low activity at baseline. Factors related to underlying health status, rather than a mortality-enhancing effect of low cholesterol, likely accounts for the excess risk of death among persons with low cholesterol. The observed low cholesterol-mortality association therefore should not discourage public health programs directed at lowering serum cholesterol.