Several prospective studies have suggested that moderate alcohol consumption may offer protection against total and coronary heart disease (CHD) mortality. These studies have been criticized for failing to control for changes in drinking and the influence of comorbidity on consumption decisions. In the present study, we examined whether rates of death from all causes and from CHD were related to overall consumption as well as variability in or problems with drinking. In 1973, a drinking questionnaire was completed by 1,823 male subjects participating in a longitudinal study who were prescreened for absence of serious or chronic disease. After 12 years of follow-up per subject (21,716 man years of follow-up in all), 159 men have died, 74 from CHD. Incidence rates of overall mortality were lowest for moderate drinkers in each of three age groups. CHD death rates for moderate drinkers were similar to those of non-drinkers except in the oldest men where rates were lower for moderate drinkers. Proportional hazards models testing several measures of consumption consistently showed moderate or regular drinkers to have lower risk of death than teetotalers. Regular drinkers had lower overall and CHD mortality than lifetime abstainers. For all-cause and CHD mortality, drinking heavily in the past, ever having tried to quit drinking and having had problems with alcohol were not related to increased risk. These results lend support to the hypothesis of the beneficial effect of moderate drinking, with respect to mortality.