The authors have investigated two commonly used methods of assessing neurological status in patients with mild head injury to determine whether they can predict intracranial damage. Of 686 such patients with cranial computed tomography (CT) scans, scan results were recorded, along with total and motor components of the Glasgow Coma Scale (GCS) and the Reaction Level Scale (RLS85). Despite relatively normal admission neurological examinations, 127 of the 689 patients (18.4%) had intracranial lesions, and 38 (5.5%) required surgery. There was no significant difference in distribution of the GCS in patients with and without intracranial lesions. The RLS85 was superior to the GCS in predicting intracranial pathology, and a significant association between RLS85 and lesions on CT scanning was noted. However, even this test was normal in 19 patients found to have intracranial pathology, including nine who required surgery. The authors conclude that a normal or near-normal mental status examination in a head-injured patient on arrival at the emergency room is inadequate to exclude a potentially serious intracranial lesion. It is unlikely that further refinements in the clinical evaluation will result in diagnostic accuracy comparable with that of CT scanning. Accordingly, we recommend that any patient who has suffered a loss of consciousness or amnesia following head injury have an urgent cranial CT scan.