The effect of breathing therapy was evaluated in patients with hyperventilation syndrome (HVS). The diagnosis of HVS was based on the presence of several suggestive complaints occurring in the context of stress, and reproduced by voluntary hyperventilation. Organic diseases as a cause of the symptoms were excluded. Most of these patients met the criteria for an anxiety disorder. The therapy was conducted in the following sequence: (1) brief, voluntary hyperventilation to reproduce the complaints in daily life: (2) reattribution of the cause of the symptoms to hyperventilation: (3) explaining the rationale of therapy-reduction of hyperventilation by acquiring an abdominal breathing pattern, with slowing down of expiration: and (4) breathing retraining for 2 to 3 months by a physiotherapist. After breathing therapy, the sum scores of the Nijmegen Questionnaire were markedly reduced. Improvements were registered in 10 of the 16 complaints of the questionnaire. The level of anxiety evaluated by means of the State-Trait Anxiety Inventory (STAI) decreased slightly. The breathing pattern was modified significantly after breathing retraining. Mean values of inspiration and expiration time and tidal volume increased, but end-tidal CO2 concentration (FETCO2) was not significantly modified except in the group of younger women (< or = 28 years). A canonical correlation analysis relating the changes of the various complaints to the modifications of breathing variables showed that the improvement of the complaints was correlated mainly with the slowing down of breathing frequency. The favorable influence of breathing retraining on complaints thus appeared to be a consequence of its influence primarily on breathing frequency, rather than on FETCO2.