A longitudinal study of one hundred consecutive admissions to the Royal Adelaide Hospital for carbon monoxide poisoning was conducted from 1986 to 1989. Twenty-five patients left hospital with persistent symptoms and signs of this poisoning. Five subsequently recovered. Twenty-four other patients, who were well when they left hospital, did not attend for a review one month after discharge. Extensive neuropsychiatric testing at this time showed 32% (24 of 76) had obvious sequelae of their exposure. Overall, the frequency of neuropsychiatric sequelae in the patients who only received oxygen at atmospheric pressure was 63% (N = 8) on discharge and 67% (N = 6) on one month follow-up. The frequency of sequelae among those who were given one hyperbaric oxygen treatment only was 46% (N = 24) on discharge and 50% (N = 20) on one month follow-up. In contrast, the frequency of sequelae in patients who had two or more hyperbaric oxygen treatments was only 13% (N = 68) on discharge (P less than 0.005) and 18% (N = 50) on follow-up (P less than 0.005). the frequency of sequelae was also significantly greater if hyperbaric oxygen was delayed (P less than 0.05). No markers of severe poisoning could be identified.