[Clin Psychol Sci Prac 18: 105–112, 2011]
In their stimulating article, David and Montgomery (2011) underscore the importance of identifying pseudoscientific and potentially harmful psychotherapies, and observe that the criteria for empirically supported therapies neglect to take into account the theoretical support for treatment packages. They propose a framework for evidence-based therapies that includes empirical support for both the efficacy of the treatment and the theory that spawned it. I discuss several challenges to this and similar frameworks, including (a) the absence of convincing data regarding how most current therapies work, (b) the underdetermination of theory by data, making it difficult to strongly corroborate models of therapeutic change mechanisms using only a small number of studies, and (c) the risk of false negatives, viz., efficacious treatments derived from nonexistent or flawed theories. As a friendly amendment, I propose a Bayesian alternative to David and Montgomery’s classification that incorporates the role of theoretical plausibility while minimizing the aforementioned difficulties.