To understand the factors that distinguish terminally ill, incompetent patients who are given do-not-resuscitate status by physicians against the wishes of the patient's family, I reviewed the charts of 20 such patients (cases) seen in consultation by the ethics committee at the Massachusetts General Hospital over 10 years. I also compared this group of patients with 105 patients who received do-not-resuscitate status with family consent over 3 months in 1986. Socioeconomic factors were similar between cases and controls. Selected prehospitalization characteristics such as history of alcohol or substance abuse, depression, dementia, or other psychiatric diseases were rarely present in the ethics committee group. However, these patients were likely to be incontinent (90%), incapable of any self-care (80%), and have low Glasgow Coma Scale scores (mean, 9.4). The ethics committee cases were relatively young (mean, 63.6 years), were in the hospital for a long period before a do-not-resuscitate decision was reached (mean, 48.2 days), were frequently admitted to intensive care units (75%), and often received ventilatory (70%) and pressor medication (50%) support. Eighty percent of the ethics committee cases died in the hospital. The control do-not-resuscitate patients were much less sick and received much less invasive support. This research suggests that the ethics committee at the study hospital was not redefining the principles of medical ethics that underlie decisions not to resuscitate when it recommended limited care in the absence of family consent, but rather was recommending such limitations only for patients in whom it appeared further care was futile.