A recent double-blind, placebo-controlled trial (Barlow et al., 2000 JAMA. 283:2529-2536) examined separate and synergistic effects of psychological and pharmacological treatments for panic disorder. One finding warranting further investigation involved relatively high relapse rates of participants who received cognitive-behavioral therapy (CBT) + imipramine when compared with those receiving CBT + placebo. In this article, we investigate why CBT was less effective in protecting against relapse for individuals in the active drug condition. We hypothesized that participants correctly deduced treatment assignments and, for those taking imipramine, this was associated with the belief that they were no longer taking active drug after discontinuation, accounting for increased relapse rates. Contrary to hypothesis, there were no group differences in frequencies of guessing drug or placebo, nor were specific beliefs about taking drug or placebo differentially associated with relapse. Other possible reasons for differential relapse rates and treatment implications are discussed.