This study includes 124 patients with closed head injuries and with Glasgow Coma Scale (GCS) scores of less than or equal to 8, who were admitted over a 7 1/2-month period. The time at which death occurred after injury was bimodal: deaths occurred either within 48 hours or after 7 days or longer after injury. Neurological deterioration, however, occurred with equal frequency on Days 2 to 7 after injury. Patients who survived the first 48 hours and then suffered neurological deterioration did not differ from the total population in age, sex, GCS scores on admission, or pupillary reactivity, but had a much higher incidence of intracranial hematomas of all types. Deterioration occurred three times more frequently in those with hematomas than in those with diffuse brain injury. Patients who deteriorated were rarely among the 35% of those who rapidly improved in the first 48 hours (4 points or more on the GCS). Computerized tomography (CT) scans of those deteriorating (24 patients) could be divided into four categories: 1) those without new mass effect (eight cases); 2) those with new or increased hemispheric edema (six cases); 3) those with generalized edema (two cases); and 4) those with focal or lobar areas of new edema or hemorrhage (eight cases). Of the patients in coma who deteriorated, 19% had large, delayed intracerebral hematomas. In 11 of 16 cases deteriorating with new mass effect, prior compression by overlying extracerebral hematoma, disruption of brain by intracerebral hematoma, or preexisting hemispheric edema preceded the brain swelling that caused deterioration. Areas of disruption or compression on CT scan typically developed decreased attenuation 2 to 7 days after injury, but did not cause deterioration unless new mass effect accompanied the lucency appearing on CT scan. A mortality rate of 29% was achieved for the 124 cases, which were managed with early evacuation of hematomas and control of intracranial pressure. Certain methods are suggested for evaluating therapy and for comparing clinical series.