The value of hyperbaric oxygen in the treatment of acute carbon monoxide intoxication was assessed in 629 adults who had been poisoned at home in the 12 h before admission to hospital. In patients without initial impairment of consciousness (group A) the effect of 6 h of normobaric oxygen (NBO) (group A0, n = 170) was compared with that of 2 h of hyperbaric oxygen (HBO) at 2 atmospheres absolute (ATA) plus 4 h NBO (group A1, n = 173). At the 1 month follow-up 66% of A0 and 68% of A1 patients had recovered. In patients with initial impairment of consciousness the effect of one session of HBO (group B1, n = 145) was compared with that of two sessions (group B2, n = 141); all group B patients also received 4 h of NBO. At 1 month of follow-up 54% group B1 and 52% group B2 patients had recovered. The 7 patients left with neuropsychiatric sequelae (3 B1, 4 B2) and the 4 who died (2 B1, 2 B2) had all presented with coma. HBO was not useful in patients who did not lose consciousness during carbon monoxide intoxication, irrespective of their carboxyhaemoglobin level, nor were two sessions of HBO in patients who sustained only a brief loss of consciousness. The prognosis is poorest for those presenting with coma; the trial needs to be pursued in this group of patients until the power of the study is sufficient to demonstrate the value or otherwise of HBO.