Background: Beliefs about the controllability of a disorder may be relevant in the causation, maintenance and treatment of disorders. We investigated whether congruence between patients' beliefs about controllability of a panic disorder and the type of treatment provided predicted outcome. Methods: The differential effectiveness of cognitive therapy and antidepressant treatment (paroxetine or clomipramine) was investigated in a sample of 129 panic disorder patients in a 12-week, pretest posttest placebo-controlled study. Panic frequency, agoraphobic avoidance, anxiety, depression, and disability were measured with various validated interviewer and self-report measures. Beliefs about controllability were measured with the Multidimensional Anxiety Locus of Control Scale measuring an internal, chance, therapist and medication locus of control. In order to analyze aptitude-treatment interactions a new strategy called the Regression Trunk Approach was used in addition to classical hierarchical multiple regression analysis. Results: Using the Regression Trunk Approach we found that locus of control orientation (LOC) predicted the differential effectiveness of cognitive therapy. Those patients with a medium internal LOC who received cognitive therapy performed significantly better than all patients who received a placebo pill on 8 of the 10 outcome variables. We did not find a differential LOC effect for antidepressant treatment. No evidence for aptitude-treatment interactions using hierarchical multiple regression analysis was found. Conclusions: Moderately strong beliefs about self-control of panic disorder congruent with the cognitive intervention provided seem to moderate treatment effectiveness. Future studies must be more attentive to the nonlinear effects of patient characteristics on the outcome of different types of treatments.