To determine whether and to what extent alcohol consumption increases the risk of injury. The case-crossover study is a new research design for assessing rare acute events (such as trauma) in relation to intermittent exposures (such as alcohol use) that have transient effects. All interviewed patients serve as their own controls, and the study design compares exposure during a "case" period, such as the 6 hours prior to injury, with exposure during a "control" period, such as the same 6-hour window on the previous day. Emergency centers of a university hospital and a private not-for-profit hospital in a small midwestern city. Three hundred fifty adults who presented with an injury within 48 hours of the event. Retrospective self-report of alcohol use in each of the 28 days and each of the 30 hours prior to injury, demographic variables, drug use prior to injury, patients' attributions regarding causes of their injury, and weather factors. Primary analysis was based on pairs of observations that were discordant for alcohol use in the 6 hours prior to injury and the same 6-hour window on the previous day. The odds ratio for one or more standard drinks vs none was 2.5 (95% confidence interval, 1.2 to 5.4); for four or more drinks vs three or fewer, 5.0 (95% confidence interval, 1.4 to 27). Alcohol use in the 6 hours prior to injury is associated with an increased risk of injury. The findings are consistent with a dose-response effect, but with this sample size no threshold of risk was found.