Cerebral blood flow (CBF) and other physiological variables were measured repeatedly for up to 10 days after severe head injury in 102 patients, and CBF levels were related to outcome. Twenty five of the patients had a reduced CBF [mean (SD) 0.29 (0.05) ml/g/min]; 47 had a normal CBF, (0.41 (0.10) ml/g/min); and 30 had a raised CBF (0.62 (0.14) ml/g/min). Cerebral arteriovenous oxygen differences were inversely related to CBF and averaged 2.1 (0.7) mumol/ml in the group with reduced CBF, 1.9 (0.5) mumol/ml in the group with normal CBF, and 1.6 (0.4) mumol/ml in the group with raised CBF. Patients with a reduced CBF had a poorer outcome than patients with a normal or raised CBF. Mortality was highest in patients with a reduced CBF, and was 32% at three months after injury, whereas only 21% of the patients with a normal CBF and 20% of the patients with a raised CBF died. There were no differences in the type of injury, initial score on the Glasgow Coma Scale, mean intracranial pressure (ICP), highest ICP, or the amount of medical treatment required to keep the ICP less than 20 mm Hg in each group. Systemic factors did not significantly contribute to the differences in CBF among the three groups. A logistic regression model of the effect of CBF on neurological outcome was developed. When adjusted for variables which were found to be significant confounders, including age, initial Glasgow Coma Score, haemoglobin concentration, cerebral perfusion pressure and cerebral metabolic rate of oxygen, a reduced CBF remained significantly associated with an unfavourable neurological outcome.