Although cognitive-behavioral therapy (CBT) represented a breakthrough in efficacious treatment of panic disorder, a substantial minority of patients fail to respond. Identifying which individuals benefit from CBT and which do not—and why—facilitates efforts to tailor and develop treatments for those not currently being helped. Nevertheless, few replicated predictors of outcome have been identified, and little is understood about the mechanisms of known predictors. While the focus of Chapters 1 and 2 is on elucidating an established predictor of poor outcome—hostile resistance (combative behavior directed at therapists)—in Chapter 3 we seek to identify novel predictors. Specifically, in Chapter 1 we employed a mixed-method approach to understand which patients became hostilely resistant in CBT and why. Compared to their never-hostile counterparts, hostilely resistant patients had more personality pathology, found treatment less credible, and had lower expectancy for improvement. However, grounded theory analyses revealed that, while patient characteristics are important, most often patient factors interacted with therapist failures (particularly of empathy) and treatment factors to produce hostility. In Chapter 2, we examined how hostile resistance translates into worse outcomes by evaluating two putative mediators—therapist adherence and working alliance. Results indicated that neither accounted for hostility’s association with attrition or worse symptomatic improvement, nor were these factors independently predictive of CBT outcome. In Chapter 3 we sought to identify novel prognostic and prescriptive factors by testing hypotheses drawn from clinicians’ observations. Of 17 clinical intuitions about barriers to treatment progress examined, none were empirically borne out; in fact, clinicians’ agreement on a putative barrier was inversely related to the strength of a predictor and, contrary to clinicians’ beliefs, dissociation during panic predicted improved response. Moderation analyses revealed that CBT outperformed psychodynamic therapy for patients with posttraumatic-stress disorder and less severe panic. This research adds to the breadth and depth of the outcome prediction literature, testing a wider range of variables and elucidating the origins of a known predictor. Taken together, findings point to the importance of training in common factors, flexible applications of treatment manuals, therapists’ management of powerful emotional reactions, and skepticism of empirically untested clinical intuitions.