The authors analysed the serial computerized tomography (CT) findings in a large series of severely head injured patients in order to assess the variability in gross intracranial pathology through the acute posttraumatic period and determine the most common patterns of CT change. A second aim was to compare the prognostic significance of the different CT diagnostic categories used in the study (Traumatic Coma Data Bank CT pathological classification) when gleaned either from the initial (postadmission) or the control CT scans, and determine the extent to which having a second CT scan provides more prognostic information than only one scan. 92 patients (13.3% of the total population) died soon after injury. Of the 587 who survived long enough to have at least one control CT scan 23.6% developed new diffuse brain swelling, and 20.9% new focal mass lesions most of which had to be evacuated. The relative risk for requiring a delayed operation as related to the diagnostic category established by using the initial CT scans was by decreasing order: diffuse injury IV (30.7%), diffuse injury III (30.5%), non evacuated mass (20%), evacuated mass (20.2%), diffuse injury II (12.1%), and diffuse injury I (8.6%). Overall, 51.2% of the patients developed significant CT changes (for worse or better) occurring either spontaneously or following surgery, and their final outcomes were more closely related to the control than to the initial CT diagnoses. In fact, the final outcome was more accurately predicted by using the control CT scans (81.2% of the cases) than by using the initial CT scans (71.5% of the cases only). Since the majority of relevant CT changes developed within 48 hours after injury a pathological categorization made by using an early control CT scan seems to be most useful for prognostic purposes. Prognosis associated with the CT pathological categories used in the study was similar independently of the moment of the acute posttraumatic period at which diagnoses were made.