A 2 1/4-year prospective study of children suffering head injury is described. Three groups of children were studied: (a) 31 children with 'severe' head injuries resulting in a post-traumatic amnesia (PTA) of at least 7 days; (b) an individually matched control group of 28 children with hospital treated orthopaedic injuries; and (c) 29 children with 'mild' head injuries resulting in a PTA exceeding 1 hour but less than 1 week. Individual psychological testing was carried out as soon as the child recovered from PTA, and then again 4 months, 1 year, and 2 1/4 years after the injury. A shortened version of the Wechsler Intelligence Scale for Children (WISC), the Neale Analysis of Reading Ability and a battery of tests of specific cognitive functions were employed. The mild head injury group had a mean level of cognitive functioning below the control group, but the lack of any recovery; during the follow-up period indicated that the intellectual impairment was not a consequence of the injury. In the severe head injury group, the presence of cognitive recovery and a 'dose-response' relationship with the degree of brain injury showed that the intellectual deficits were caused by brain damage. Some degree of cognitive impairment was common following head injuries giving rise to a PTA of at least 2 weeks. Conversely no cognitive sequelae, transient or persistent, could be detected when the PTA was less than 24 hours. The results were less consistent in the 1-day to 2-week PTA range, but the evidence suggested that a broadly defined threshold for impairment operated at about that level of severity of injury. Timed measures of visuo-spatial and visuo-motor skills tended to show more impairment than verbal skills but otherwise there was no suggestion of a specific pattern of cognitive deficit. Recovery was most rapid in the early months after injury, but substantial recovery continued for 1 year with some improvement continuing n the second year in some children, especially those with the most severe injuries. Age, sex and social class showed no significant effects on the course of recovery.