This paper compares alcohol consumption and casualties in probability samples of two diverse emergency room populations: San Francisco General Hospital (SFGH) (n=2516) and four hospitals representative of a nearby California county (n=3609). Both studies used similar methods and data collection instruments. Patients were breathalysed and interviewed regarding self-reported alcohol consumption 6 hours prior to the injury or illness event, usual drinking patterns and alcohol-related problems. Injuries were found to be positively associated with breathalyser readings, self-reported consumption prior to the event and more frequent heavy drinking in both samples. In the county sample injuries were also positively associated with more frequent drunkenness, symptoms of alcohol dependence and loss of control and prior alcohol-related accidents. The SFGH sample had higher rates than the county sample on all alcohol variables and both samples reported higher rates of alcohol-related problems than that found in U.S. general population surveys.