To test whether alveolar hypoventilation and an abnormal ventilatory response to inhaled carbon dioxide explains some episodes of sudden-infant-death syndrome, we assessed ventilatory control during quiet sleep in 12 normal infants and 11 infants who had required at least two resuscitations because of prolonged apnea (greater than 20 seconds) during sleep (aborted form of the syndrome). Infants with the aborted syndrome hypoventilated during quiet sleep (alveolar partial pressure of carbon dioxide, 38.9+/-3.5 mm Hg) as compared to normal infants (35.1+/-1.9, P less than 0.01). In addition, the ventilatory response to carbon dioxide breathing during quiet sleep was impaired (mean change in minute ventilation per change in partial pressure of carbon dioxide 22.1+/-8.9, as compared to 63.1+/-19.1 ml per kilogram per minute per millimeter of mercury in controls [p less than 0.001]). Three infants with the "aborted syndrome" subsequently died during sleep at home; autopsy, done in two, revealed no apparent cause of death. We conclude that infants who have had an episode consistent with sudden-infant-death syndrome have a defect in the regulation of alveolar ventilation.