During Phase II of the Cross-National Panic Study, descriptions of the patient's last severe panic attack were collected for 1168 patients. Statistical analysis indicated that patients could be divided into two groups, characterised by the presence or absence of prominent respiratory symptoms. The two groups did not differ on demographic variables or coexisting diagnoses, but they did differ on psychopathology on entry to the study and treatment outcome. The group with prominent respiratory symptoms suffered more spontaneous panic attacks and responded to imipramine, whereas the group without prominent respiratory symptoms suffered more situational panic attacks and responded more to alprazolam. It is important to distinguish spontaneous and situational panic attacks, to aid choice of treatment.