With the advent of computerized tomography (CT), an increasing number of patients with only minimal neurological symptoms and no signs of brain herniation are found to harbor subacute or chronic extradural hematomas (EH's). The authors present the cases of 11 symptomatic but neurologically normal children with medium to large EH's managed by close observation. These EH's were discovered 4 hours to 6 days after injury; three were in the posterior fossa, seven over the frontoparietal convexity, and one in the temporal fossa. These clots were followed by serial CT scans. Nine children recovered without surgery from 4 to 18 days after injury, and all had evidence on CT of spontaneous clot resorption. Of these nine EH's, five clots displayed volume expansion from 5 to 16 days after injury before final resorption occurred. Expansion correlated with persistence or increase in symptoms, whereas resorption correlated with improvement. Two patients showed gradual uncal herniation on Days 6 and 8, respectively, presumably during the "expansile phase" of their clots. Both had emergency craniotomy and recovered without morbidity. It is hypothesized that the resorption dynamics of the subacute or chronic EH are similar to that of the chronic subdural hematoma, with predictable volume changes, and the outcome of each lesion depends on the interplay between the patient's intracranial pressure buffering capacity and the rate of volume change. If subtle signs of brain dysfunction are adopted to signal the failure of conservative treatment and the need for craniotomy, these patients may be safely, and many successfully, managed without surgery. Factors that influence outcome of medical treatment include the size, location, configuration, and the rapidity of accumulation of the clot, the presence of associated intradural lesions, the extracranial decompression of blood through skull diastases, and the age of the patient. These factors, the criteria for patient selection, and the indications for immediate operative intervention are discussed.