The present study investigates the differential effectiveness of three treatment packages for agoraphobia. Patients suffering from panic disorder with agoraphobia (DSM-III-R) received one of three treatments: Breathing Retraining with Cognitive Restructuring (BRCR), graded self-exposure in vivo (EXP), or a combination of BRCR and EXP. Treatments consisted of 8 sessions. Assessment consisted of self-report measures for panic, phobic anxiety and avoidance, depression, general anxiety, somatic complaints and fear of body sensations, and of two respiratory measures (respiratory rate and alveolar pCO2). The treatments resulted in a reduction in symptomatology on all self-report measures, except panic frequency, and in a decrease in respiratory rate. There was no evidence for a differential efficacy for any of the treatments on any of the variables. Contrary to expectation, and at odds with findings from earlier studies, BRCR had no significant effect on panic frequency. A detailed comparison of sample characteristics of patients in our study and previous studies, did not yield insight into possible causes for the failure to replicate earlier results. The limited effectiveness of breathing retraining in reducing panic, as observed in the present study, leads us to conclude that the role of hyperventilation in panic is less important than previously thought.