The association of serum total and high density lipoprotein cholesterol values with 15 year mortality was examined in a cohort of 10 059 Israeli male civil servants and municipal employees aged 40 and above. In 618 of 1664 deaths in the cohort (37%) coronary heart disease was documented as the cause of death. Risk of mortality was analysed by quintiles. Neither total mortality nor coronary heart disease mortality rose with serum cholesterol concentrations up to 5–6 mmol/1 (216 mg/ 100 ml), representing 60% of the sample. Rates rose appreciably only in the highest quintile (cholesterol concentration >6.2 mmol/1; >241 mg/100 ml). High density lipoprotein cholesterol was similarly, although inversely, associated with total mortality when expressed as a percentage of total cholesterol. The inverse association of high density lipoprotein cholesterol with coronary heart disease mortality was, in contrast, continuous. These data support the hypothesis that over most of the range of cholesterol values coronary mortality risk and total mortality risk are nearly independent of total cholesterol and most probably independent of low density lipoprotein cholesterol values. In multivariate analysis a low concentration of high density lipoprotein cholesterol appeared to be more predictive of mortality than a high concentration of total cholesterol. The latter was very weakly related to mortality from all causes after multivariate adjustment. It is concluded that the findings of this and other major epidemiological studies support the notion of a.