The importance of delayed or secondary brain insults in the eventual outcome of closed-head trauma has been documented in experimental models. To understand this phenomenon in the clinical setting, we studied a series of head-injured patients in whom multiple cranial computed tomographic (CT) scans were obtained. Patients whose follow-up CT studies revealed new intracranial lesions or worsening, compared with admission findings, were considered to have delayed cerebral injury. One hundred forty-nine (44.5%) of 337 consecutively studied patients developed delayed brain injury. There were highly significant associations (P < 0.001) between the appearance of delayed cerebral insults and the severity of the initial brain injury, the need for cardiopulmonary resuscitation in the field, the presence of coagulopathy at admission, and subdural hematoma on the initial CT scan. In addition, delayed injury was associated (P < 0.001) with higher mortality, slowed recovery, and poorer outcome at 6 months. Delayed brain injury was not significantly associated with patient age, sex, injury mechanism, associated injury, the need for endotracheal intubation in the field, early talking, CT abnormality other than intracranial hematoma, or type of residual neurological deficits. We used multiple regression analysis to explore the relationship between severity of injury, delayed insults, and outcome. As expected, the severity of the initial brain trauma contributed significantly to neurological outcome. The presence of delayed cerebral injury makes the outcome dramatically worse for each category of initial injury severity. The relationship between initial and secondary brain injury is discussed.