The computerized tomographic (CT) scan has revolutionized the management of cerebral trauma. Nevertheless, visualization of traumatically induced lesions of the brain stem by the CT scanner remains difficult. Seven patients with autopsy or CT evidence of brain stem hemorrhage were identified over a 1-year period. In six of these patients, brain stem hemorrhage could be defined by CT scan. As part of a prospective study of CT changes after head injury, we performed serial CT scans on six of the seven patients. Clinical experience shows that timing is important for identification of these lesions and that inability to visualize brain stem hematomas may occur because of the development of hematomas after CT scanning, evolution of hemorrhagic lesions that makes them isodense with the surrounding brain stem, patient movement, and technical factors such as the partial volume effect. Experimental injection of fresh blood into the pons and midbrain of cadavers shows that lesions as small as 0.1 ml in volume may be visualized by ex vivo thin section CT scanning techniques. However, the character and anatomical configuration of the hemorrhage may be as important in determining CT visualization as is the volume of the hemorrhage. For example, a hematoma displacing the brain parenchyma was visualized, but a similar-sized small hemorrhage that had diffused through the brain stem tissues was not. Although many of the experimentally placed lesions extended over a rostral-caudal length of 15 mm or more in the brain stem, no lesion was seen on more than three thin section scans. This is explained by the presence of lesions that, although extensive in a rostral-caudal direction, had relatively small cross sectional areas available for identification by the CT scanner. The small size of traumatic lesions of the brain stem and their proximity to bony structures at the base of the skull are not insurmountable obstacles to visualization of brain stem hemorrhages. Serial scanning and the application of thin section computed tomography will lead to identification of most of these lesions.