To learn more about the causes of nonfatal construction worker injuries, and to identify injury cases for further work-site investigations or prevention programs, an emergency department-based surveillance program was established. Construction workers with work-related injuries or illnesses were identified by reviewing the medical records of all patients treated at the George Washington University Emergency Department between November 1, 1990 and November 31, 1992. Information regarding the worker, the injury, and the injury circumstances were abstracted from medical records. Information was obtained on 592 injured construction workers from numerous trades. Lacerations were the most commonly treated injuries among these workers, followed by strains and sprains, contusions, and eye injuries. Injuries were most commonly caused by sharp objects (n = 155, 26%), falls (n = 106, 18%), and falling objects (n = 70, 12%). Thirty-five percent of injuries were to the hands, wrists, or fingers. Among the twenty-eight injuries severe enough to require hospital admission, eighteen (64%) were caused by falls. Laborers and Hispanic workers were overrepresented among these severe cases. Emergency Department records were a useful surveillance tool for the initial identification and description of work-related injuries. Although E codes were not that useful for formulating prevention strategies, detailed review of injury circumstances from Emergency Department records was valuable and has helped to establish priorities for prevention activities.