Inhalations of high concentrations of carbon dioxide (CO2) reliably produce panic attacks in patients with panic disorder. The present study evaluated whether cognitive-behavioral treatment (CBT) for panic disorder would extinguish CO2-induced panic and whether changes in panic and arousal-related cognitions were associated with the induction of panic. Patients with panic disorder (N = 54) were assigned to 1 of 3 experimental conditions: CBT with respiratory training (CBT-R), CBT without respiratory training (CBT), or delayed treatment. Participants received 5 repeated vital-capacity inhalations of 35% CO2/65% O2 prior to and following either 12 treatment sessions or a 12-week waiting period. During pretreatment assessments, 74% of patients experienced a panic attack during at least 1 inhalation. At posttreatment, only 20% of treated participants (CBT-R = 19%, CBT = 22%), compared with 64% of untreated participants, panicked. Forty-four percent of treated participants, compared with 0% of untreated participants, reported no anxiety during all posttreatment inhalations. Anxiety sensitivity as well as panic appraisals regarding the likelihood of panic and self-efficacy with coping with panic were significantly related to fearful responding to the CO2 challenge.