In a prospective study of more than 10000 Yugoslav men it was found that consumption of alcoholic beverages was inversely related to non-sudden death from coronary heart disease (CHD) and positively related to death from trauma. The consequence was an apparently U-shaped relation between alcohol consumption and death, the lowest mortality being among moderate drinkers. Excess mortality from trauma was evident only among men under 55 and only for those who reported at entry to the study that they had been drunk during the preceding week. Alcohol consumption as reported at entry was unrelated to subsequent mortality from liver cirrhosis or any form of cancer. An enlarged liver, however, was associated with higher death rates for liver cirrhosis. This raises the possibility that some of the men were heavy drinkers preceding their entry to the study but were no longer drinking heavily at the time of entry. Enlarged liver, however, was also related to hypertension and to chronic obstructive pulmonary disease and thus was not a specific indicator of alcohol abuse in this population. Recent drunkenness but not frequency of drinking was related to death from trauma and liver cirrhosis and to sudden CHD death. In short, both the pattern of drinking and the usual level of alcohol consumption appear to be related to mortality in this population.