We assessed the validity of death certificate data for alcohol-related mortality in a population-based follow-up study of young, male US Army veterans. In a random sample of more than 18,000 men who entered the service between 1965 and 1971, there were 446 postservice deaths through Dec 31, 1983. For 426 of these deaths, we obtained both the death certificate and all other available medical and legal records pertaining to cause of death. A nosologist recorded each death certificate in accordance with the ninth revision of the International Classification of Diseases. A medical panel, without having access to the death certificates, assigned underlying and contributory causes of death on the basis of a review of only the medical and legal records. The panel recorded 133 alcohol-related deaths, or more than six times the number (21 deaths) determined by the original death certifiers. Omission of elevated blood alcohol levels in deaths due to injury accounts for most of the underreporting of alcohol-related mortality on the death certificates. Our findings suggest that death certificate data grossly underestimate the contribution of alcohol to mortality, especially in the area of injury, and the validity of official vital statistics for alcohol-related deaths would be enhanced if death certifiers incorporated all available antemortem and postmortem diagnostic information.