Common versus specific factors in psychotherapy: opening the black box

Summary

Do psychotherapies work primarily through the specific factors described in treatment manuals, or do they work through common factors? In attempting to unpack this ongoing debate between specific and common factors, we highlight limitations in the existing evidence base and the power battles and competing paradigms that influence the literature. The dichotomy is much less than it might first appear. Most specific factor theorists now concede that common factors have importance, whereas the common factor theorists produce increasingly tight definitions of bona fide therapy. Although specific factors might have been overplayed in psychotherapy research, some are effective for particular conditions. We argue that continuing to espouse common factors with little evidence or endless head-to-head comparative studies of different psychotherapies will not move the field forward. Rather than continuing the debate, research needs to encompass new psychotherapies such as e-therapies, transdiagnostic treatments, psychotherapy component studies, and findings from neurobiology to elucidate the effective process components of psychotherapy.

Introduction

The aim of this Review is to discuss a pivotal debate from psychotherapy research: do psychological therapies work primarily through the specific factors that are described in treatment manuals (eg, cognitive restructuring in depression, or exposure in anxiety disorders), or do they work primarily through factors that are common across most therapies (eg, positive working alliance and expectation)? Empirically, the debate between specific versus common factors centres on two findings arising in repeated meta-analyses. First, psychotherapies are more effective than unstructured interactions and more effective than nothing,1, 2 and second, specific psychotherapies such as cognitive behavioural therapy (CBT), interpersonal therapy, mindfulness, and acceptance and commitment therapy generally do not differ in effectiveness (panel 1).11, 12 These findings are exemplified by a network meta-analysis¹³ of seven psychotherapeutic interventions for patients with depression. The investigators reported that, although each intervention was better than a waitlist control, the relative effects of different psychotherapeutic interventions on depressive symptoms were similar. This evidence suggests that factors shared across psychological therapies (ie, common factors) might be the major therapeutic mechanism. The apparent importance of common factors has been characterised as the dodo bird effect, after the line in Alice in Wonderland (“At last the Dodo said, ‘Everybody has won, and all must have prizes’”).¹⁴

A highly cited 2002 review¹⁵ of 17 meta-analyses comparing active treatments with each other showed small, non-significant differences in outcome, which diminished further after the substantial effects of researcher allegiance were controlled for. However, these and similar findings have been questioned by those of other meta-analyses. Tolin,¹⁶ for example, reported that CBT was associated with a significant advantage over other therapies, at least in patients with anxiety or depressive disorders. Similarly, Hofmann and colleagues¹⁷ showed evidence for treatment specificity in a review of 269 meta-analytic studies examining CBT for different psychological problems. These apparently contradictory findings have led to two broad groupings of psychotherapy theorists: those emphasising relationship, patient expectancy, and process (common factors), and those emphasising procedural techniques (specific factors). As observed by Norcross,¹⁸ “The culture wars in psychotherapy dramatically pit the treatment method against the therapy relationship”.

Common factors should not be considered non-specific in the sense of being unintended benefits of therapy (like the placebo effect). Indeed, most common factor theorists consider these factors to collectively shape a theoretical model about the mechanisms of change in psychotherapy (panel 2).²⁰