This study reports breath-analyzer readings and self-reports as measures of alcohol-related admission to the emergency room of San Francisco General Hospital. A 20% probability sample of patients admitted during a 60-day period was breath analyzed and interviewed. Interviews and breath samples were obtained on 75% of the sample of 2,516 patients. Twice the proportion of injury patients compared to noninjury patients had positive admission breath samples and reported drinking prior to the event. Alcohol involvement reached 41% for self-reports among injured men and over half of both men and women injured in fights or assaults reported drinking prior to the event. Self-reported alcohol use was found to be a valid measure of alcohol consumption when compared to breath-analyzer readings for the same individuals. Emergency room patients may be more likely than others to provide accurate reports of alcohol consumption if they feel that disclosure of amount and timing of drinking prior to an injury or illness could be important in their care. The data suggest that self-reports when used in conjunction with a quantifiable estimate of blood alcohol may be an appropriate method of ascertaining alcohol's involvement in emergency room cases.