A prospective study was undertaken of the effects of applying a more selective admission policy to a group of 6685 accident and emergency attenders with head injuries. The efficacy of such a policy was assessed by comparison with 5768 head injury attenders subject to an orthodox admission policy and collected retrospectively. Epidemiological characteristics of both study groups, of patients not admitted but who would previously have been admitted, and of patients admitted because of head injury alone during the prospective study are detailed. The more selective policy was no worse than current practice in terms of immediate morbidity and mortality. Survival of patients with post-traumatic intracranial haematomas was more likely in the prospective group, as was the diagnosis and treatment of such lesions while the patients were alive. Patients admitted because of head injury alone were reduced to one-third of their expected number and all admissions, from among those attenders with head injury in the prospective study, were reduced by half. Adopting such a policy nationally could save 11,000,000 pounds annually.