Patients who experienced panic attacks, with or without avoidance, were treated for two weeks with either training in controlled breathing or a placebo treatment. Subsequently, both groups received a limited period of conventional anxiety treatments, most commonly in vivo exposure. Patients were subdivided into ‘hyperventilators' and ‘non-hyperventilators' on the basis of the conventional provocation test. Observer ratings of anxiety showed a greater improvement for the group that received breathing training, but there was no evidence that ‘hyperventilators' benefited more from respiratory training than ‘non-hyperventilators'. Self-report measures of anxiety, avoidance, and depression/dysphoria showed no difference between treatments. These findings suggest that training in controlled breathing is not of specific benefit for those identified as ‘hyperventilators' by the provocation test, but that it may have a non-specific effect in the treatment of patients with panic attacks.