Patients with clinically diagnosed agoraphobia to whom it had been demonstrated that a standard hyperventilation provocation could produce their feared symptoms were alternately allocated to two treatment groups: 7 patients were treated with two sessions of breathing retraining followed by 7 weekly sessions of real-life exposure to the feared situation, and 5 patients were treated with 9 weekly sessions of real-life exposure without breathing retraining. At discharge and at 1 month's follow-up the two groups of patients showed similar degrees of improvement in frequency of panic attacks and other psychophysiological scores. At 6 months' follow-up the patients treated with real-life exposure alone were beginning to show a fall-off in relearned adaptive behaviour (learning decrement), whereas those given breathing retraining showed further improvement. This suggests that patients pretreated with breathing retraining are less likely to need further treatment.