Cognitive Behavioral Theory Versus Rational Emotive Behavioral Theory

Available online at www.sciencedirect.com

ScienceDirect Behavior Therapy 47 (2016) 785–803

www.elsevier.com/locate/bt

Cognitive-Behavioral Therapy: Nature and Relation to Non-Cognitive Behavioral Therapy

Lorenzo Lorenzo-Luaces John R. Keefe

Robert J. DeRubeis University of Pennsylvania

1

Since the introduction of Beck’s cognitive theory of emotional disorders, and their treatment with psychotherapy, cognitive- behavioral approaches have become the most extensively researched psychological treatment for a wide variety of disorders. Despite this, the relative contribution of cognitive to behavioral approaches to treatment are poorly understood and the mechanistic role of cognitive change in therapy is widely debated. We critically review this literature, focusing on the mechanistic role of cognitive change across cognitive and behavioral therapies for depressive and anxiety disorders.

Keywords: cognitive-behavioral therapy; cognitive theory; psychotherapy processes; depression; anxiety

THE ORIGIN OF COGNITIVE-BEHAVIORAL THERAPIES (CBTs) as a family of interventions can be traced to the advent of behavioral treatments for psychopa- thology in the 1950s and, later, the so-called “cognitive revolution” of the 1950–1960s (Dobson, 2009). Consequently, CBTs blend techniques that are emphasized in behavioral therapies (BTs) and cogni- tive therapies (CTs). However, there remains skepti- cism regarding the relative contributions of CT strategies to BT strategies in promoting symptom change within the CBTs (Longmore & Worrell, 2007). Additionally, critics have asserted that changes in thinking are not mechanisms of change in CBTs (e.g., Kazdin, 2007), calling into question whether

Correspondence regarding this article should be addressed to Lorenzo Lorenzo-Luaces, University of Pennsylvania, Department of Psychology, 3720 Walnut Street D20, Philadelphia PA 19104; e-mail: lorenzl@sas.upenn.edu.

0005-7894/© 2016 Association for Behavioral and Cognitive Therapies. Published by Elsevier Ltd. All rights reserved.

there is any kind of contribution of the “cognitive” in cognitive-behavioral therapy. Despite debate regarding their active treatment

components as well as working mechanisms, CBTs continue to be the most widely studied forms of therapy (Hofmann, Asmundson, & Beck, 2013). A uniquely appealing aspect of CBTs is that their theo- ries of therapeutic change comport well with most modern conceptualizations of psychopathology. In this review, we attempt to reconcile skepticism regarding the relative contribution of CT strategies to BT, as well as the mechanisms that account for their efficacy. First, we provide a very brief historical over- view of the origins of CBT and discuss the support for the cognitive vulnerability models to depression and anxiety disorders. We discuss methodological chal- lenges in psychotherapy research that have impeded a more thorough understanding of the relative con- tributions of cognitive to behavioral techniques. We then focus most of our discussion on research on the cognitive mechanisms of change in CT, BT, and CBTs for depression and anxiety disorders. We use the terms cognitive therapy (CT) and

cognitive techniques to refer to behaviors therapists engage in that are targeted towards changing the content or process of thoughts, inferences, inter- pretations, cognitive biases, and cognitive schemas.1

The terms “cognitive therapy” (CT) and “cognitive-behavioral therapy” (CBT) are often used interchangeably. We believe this is somewhat unfortunate in that it might be informative to reserve the term CT to a set of interventions within the broader family of CBTs that are more “purely” cognitive in nature. However, throughout the article, when we refer to findings in studies of CT or CBT, we are adhering to the label the study authors use. Additionally, we use CBTs, in plural, to refer to the family of cognitive-behavioral therapies.

 

 

786 l o re n zo – l ua c e s et a l .

These interventions can include Socratic question- ing, examining the evidence for and against beliefs, cognitive restructuring, and adopting alternative core beliefs. We use the terms behavior therapy (BT) and behavioral techniques to refer to behav- iors therapists engage in that are targeted towards a change in observable behavior, including in vivo exposure, imaginal exposure, and activity sched- uling. We use cognitive-behavioral therapies in the plural (CBTs) to refer to the family of interventions to which CT and BT belong, and in the singular, CBT, to refer to a treatment package that combines cognitive and behavioral techniques. By cognitive change, we refer to changes in the content of thoughts, inferences, interpretations, and cognitive biases. By behavioral change, we refer to changes in behavior, such as increasing the frequency of selected behaviors (e.g., approaching feared stimuli, engaging with pleasurable activities) or decreasing the frequency of other behaviors (e.g., safety behaviors). We include in our paper a discussion of issues related to the conceptualization and measure- ment of cognitive vs. behavioral interventions as well as cognitive vs. behavioral mechanisms of change and conclude with a summary and with recommen- dations for future research.

Cognitive Therapy: Nature and Relation to Behavioral Therapy

Behavioral therapies emerged in the 1950s–1960s (O’Donohue & Noll, 1995). The behavioral models emphasized maladaptive learning and self- sustaining behaviors as key to the maintenance of psychopathology. This made behavioral change the obvious target of treatment, an approach that was in stark contrast to the previously dominant psychoanalytic models. Under psychoanalysis, pathological behavior was seen to reflect dysfunc- tion in underlying psychic structures. Behavioral change was thus seen as surface-level “symptom reduction” that did not address underlying prob- lems. BTs proved very effective, particularly in the treatment of phobias and more circumscribed states of anxiety. Principles of associative learning were used to account for the efficacy of these interventions. To the behaviorists, learning had a specific meaning: an overt change in behavior (e.g., approaching a previously avoided stimulus) in the absence of symptoms (e.g., without display- ing the fear reaction). This definition avoided “mentalistic” terms. Although early behavioral models featured theoretical accounts focused on associative learning, nonassociative learning, in- cluding habituation, was also seen as important. Newer behavioral models also focus on inhibitory learning (Craske et al., 2008).

CT emerged in the context of the so-called cogni- tive revolution (Beck, 1991; O’Donohue, Ferguson, & Naugle, 2003) from the writings of Ellis (1962), who described a form of therapy known as rational- emotive therapy, and Beck (1963). The cognitive models of Ellis and Beck focused on inferential errors leading to maladaptive views of one’s self, world, and the future. According to Beck, cognitive biases and maladaptive cognitive content are the product of the activation of cognitive schemas that typically develop early in life. Unlike BTs, which were initially successful in specific phobias and circumscribed anxieties, CTs were focused on depressotypic pre- sentations and more generalized anxiety. Early in his writing, Beck recognized that his cognitive theory of psychopathology, which gave a central role to cognition in the etiology of disorder, contrasted with behavioral theories of psychopathology. In his highly cited article, “Cognitive Therapy: Nature and Relation to Behavioral Therapy,” Beck (1970) described important differences between the theories that underlie BT and CT while recognizing areas of overlap in the performance of the therapies. Similar- ities include that both therapies deal with issues in the present, are symptom-focused, and require active therapist contribution. Beck (1970) recognized differences between

behavioral and cognitive approaches. He applied the principles of his then nascent cognitive theory to account for the mechanisms of action of systematic desensitization, a BT. He concluded that the cog- nitive model “provides a greater range of concepts for explaining psychopathology as well as the mode of action of therapy.” That is, Beck made a distinction between the nature of the therapeutic interventions (i.e., cognitive vs. behavioral) and their working mechanisms in providing a cognitive account of the effects of a behavioral intervention. Beck’s paper would become one of the early reflec- tions on the relative contributions of cognitive to behavioral strategies and the relevant mechanisms of change. Although Beck has provided two up- dates to his cognitive model (Beck, 1996; Beck & Haigh, 2014), its basic tenets remain largely intact: that the distinction between different forms of psy- chopathology can be traced to differences in the locus of the cognitive pathology and that cognitive change, regardless of how this change is achieved, is integral to symptom change.

Cognitive Vulnerability to Depression and Anxiety

Basic research supports the notion that cognitive vulnerabilities confer risk to the onset and main- tenance of psychopathology (see Mathews & MacLeod, 2005). Attentional biases to threatening

 

 

787cb t: n at ur e an d r el at ion to non -c bt

stimuli, along with overestimation of threat, have been implicated in the etiology of anxiety disor- ders (Bar-Haim, Lamy, Pergamin, Bakermans- Kranenburg, & Van Ijzendoorn, 2007). Biases associated with depression include difficulties dis- engaging from negative material, sustained or sym- metrical attention to negative, relative to positive, stimuli (Kircanski & Gotlib, 2015), negative biases in the appraisal of life events (Mehu & Scherer, 2015), symmetric memory for negative vs. neutral or positive information (Kircanski & Gotlib, 2015), and negative schemas about the self that foster maladaptive and negative thinking (Beck & Haigh, 2014). Overall, existing research is supportive of cog-

nitive vulnerability models of affective disorders. For example, cross-cultural research consistently suggests that, on average, healthy individuals have a bias towards optimistic thinking that is not found in individuals who are depressed and who, instead, have a bias towards more negative thinking (Mezulis, Abramson, Hyde, & Hankin, 2004). Similarly, in a meta-analytic review of 172 studies examining biases towards threatening stimuli, Bar-Haim et al. (2007) found that anxious partic- ipants are biased to attend to threatening stimuli, relative to nonanxious participants (d = 0.45). The causal role of these cognitive vulnerabilities, par- ticularly in depression, has been questioned partly because most of the early research on this matter was correlational in nature (see Ingram et al., 2006). Findings from prospective studies, however, also support cognitive models. For example, daily fluctuations in negative automatic thoughts have been found to predict subsequent negative mood, even controlling for prior levelsof automatic thoughts (Wenze, Gunthert, & Forand, 2007; Wenze et al., 2010). Negative dysfunctional attitudes also pre- dict depressed mood following a stressor (Hankin, Abramson, Miller, & Haeffel, 2004). In one study, participants who were classified as being at high cognitive risk were almost 7 times more likely to report a major depressive episode at 2.5 years follow-up, relative to those at low risk (Alloy et al., 2006). Although prospective studies provide a stronger

level of evidence for causality than correlational studies, findings from these studies are still subject to third variable confounds, making experimental designs preferable. Relatively few experiments manipulating cognitions and assessing the effects of the manipulation on mood have been conducted. The results of these experiments, however, are consistent with models of cognitive vulnerability (see Mathews & MacLeod, 2005). For example, in a series of experiments, Mathews and Mackintosh

(2000) reported that inducing bias in the interpre- tation of ambiguous information as threatening leads to increases in state anxiety. In another study, MacLeod et al. (2002) manipulated attention to emotionally negative information. After a stressor task, participants who had had their attention manipulated towards negatively valenced stimuli showed greater anxiety and depression than par- ticipants in the control group. If cognitive biases increase the risk for depression

and anxiety states, it follows that strategies that address these biases should result in a reduction of risk. This hypothesis has support in basic research on emotion regulation. Webb, Miles, and Sheeran (2012) conducted a meta-analysis of 306 compar- isons of emotion regulation strategies. Strategies that focused on cognitive change were estimated to be the most consistently effective ways of regulating emotions (d = 0.36). Strategies aimed at helping individuals adopt more rational perspectives, as is encouraged in CT, were associated with the largest effect (d = 0.45). Providing even more support for cognitive theories, studies that examine the bio- logical vulnerabilities to negative emotional states suggest that, at the phenomenological level, biolog- ical vulnerabilities render individuals more likely to experience negative emotional states by interfering with their abilities to engage in cognitive reapprais- al strategies (Firk, Siep, & Markus, 2013; Lemogne et al., 2011). More research is needed that characterizes more

precisely the nature of the cognitive biases impli- cated in depression and anxiety, especially research that is experimental. The relationship between affective disorders and cognition is bidirectional, which must also be accounted for in theories of psychopathology. However, given the amount of evidence and the dearth of competing explanations, it can be safely asserted that the cognitive model is a valid characterization of the etiology of affective disorders. Thus, one would expect considerable support for the hypothesis that change in cognition mediates symptom change in the context of psy- chotherapy. Instead, the literature contains ques- tions about whether “we need to challenge thoughts in cognitive behavior therapy?” (Longmore & Worrell, 2007) and assertions such as “whatever may be the basis of changes with CT, it does not seem to be the cognitions as originally proposed” (Kazdin, 2007). Why is this so?

It’s Complicated Previously, we (Lorenzo-Luaces, German, & DeRubeis, 2015) have argued that disagreement among commentators (e.g., Kazdin, 2007; Longmore & Worrell, 2007) regarding the role of cognitive

 

 

788 l o re n zo – l ua c e s et a l .

change in promoting symptom change in psycho- therapy for depression emerges from different assumptions regarding the inferences that can be drawn from treatment studies. Below we review some of these issues in psychotherapies for depres- sion and anxiety.

experimental designs: additive and dismantling studies When two treatment packages are very different (e.g., psychoanalysis vs. exposure and response- prevention), it is easy and perhaps even appropriate to interpret findings from studies comparing treat- ment packages to reflect the relative efficacy of specific therapy procedures (e.g., analyzing trans- ference vs. engaging in exposure). However, when assessing treatments, such as CBT or eye-movement desensitization and reprocessing (EMDR), which combine multiple active and overlapping elements, in this case exposure and cognitive restructuring (Tolin, 2014), it becomes more difficult to extrap- olate conclusions about therapy procedures from outcome data. In lieu of tightly controlled basic research, such

as the research on emotion regulation strategies (Webb et al., 2012), researchers have used compo- nent studies as a way of addressing questions about the differential utility of treatment elements. These component studies are often referred to as if they represent a single class of study design, but there are at least two different types of study designs, additive and dismantling designs, that fall under this rubric. As described by Bell, Marcus, and Goodlad (2013), they address different kinds of questions. In additive component studies, in one condition a component is added to and compared with an already-existing, simpler treatment. Butler, Cullington, Munby, Amies, and Gelder (1984) provide an early example of such a study. They examined the value of adding anxiety management to exposure for social anxiety by comparing the combined treatment to exposure only as well as to exposure plus a nonspecific filler. Their findings suggested that adding anxiety man- agement to exposure improved treatment outcomes. In dismantling designs, at least one component of a multicomponent treatment package is removed from the treatment and compared to the full treatment package or to the other components. For example, Foa, Steketee, Grayson, Turner, and Latimer (1984) dismantled exposure and response-prevention (ERP) and compared its effects with the effects of exposure only and response prevention only. Their findings suggested that ERP was superior to either of its single components and that, for contamination fears, ex- posure alone may be more effective than response- prevention alone. Bell et al. (2013) conducted a

meta-analytic review of components studies and concluded that it is uncommon, in studies that have used dismantling designs, for one component of a treatment to outperform another. However, in studies that have used additive designs, adding one treatment component to another enhances positive therapeutic outcomes, particularly in the longer term. A meta-analysis by Adams et al. (2015) addressed

the contribution of CT to BT and did not find an added benefit of CT to BT and CBT packages. However, this meta-analysis did not differentiate between additive and dismantling designs. Al- though, on the face of it the component studies in the meta-analysis by Adams et al. seem like they provide very conclusive answers about the superi- ority or equipotency of CT and BT, component studies, as they have been conducted and inter- preted, have been extremely problematic. Summar- ily reviewing some of the limitations of component studies, Bell et al. (2013) stated:

… Null results [in component studies] do not directly address the issue of specific versus common factors because there is no group that received only common treatment components. […] Component designs may also under- estimate the contributions of the component. Rehm (2009) suggested that because much improvement typically occurs in the early stages of therapy, whichever component is presented first will appear to be the most effective. Thus, the dismantled component (which is never introduced) is likely to appear unnecessary. Component studies are also likely to be statistically underpowered (Kazdin & Whitley, 2003) to detect the relatively small effect sizes that are likely to occur with these types of designs. […]. [A] two-group component study with a presumed effect size of .24 (half the treatment vs. placebo effect size) would require over 250 patients in each condition to have a power of .80. Even Kazdin and Whitley’s (2003) higher estimate of an effect size of .45 for additive design studies would require 78 patients in each condition. In contrast, the average sample size for the studies included in the present meta-analysis was 23 participants in each condition, which would require a large effect size of .84 to have a power of .80.

The component methodology evidenced a surge in popularity following a landmark study by Jacobson et al. (1996). Jacobson et al. conjectured that the full CBT for depression package could be divided into three components: (1) behavioral activation (BA); (2) challenging automatic thoughts (ATs); and (3) modifying core beliefs (CBs). To compare the relative efficacy of these procedures, and perceiving limitations in prior work suggesting that CT for depression was superior to BT (Shaw 1977), Jacobson et al. randomized participants to three conditions lasting a maximum of 20 sessions: (1) 100% BA; (2) a condition that could use all the elements of BA and could include AT work; and (3) a condition that could use all of the elements of

 

 

789cb t: n at ur e an d r el at ion to non -c bt

BA, work on ATs, with a required minimum of 8 sessions devoted to CB work. In this study, across various metrics, no statistically significant between- conditon differences in outcome were reported. These findings have generally been misinterepreted as indicating that BA is the active component in CBT for depression and that the interventions provided in the cognitive components were inert, thus presenting a major challenge to Beck’s cogni- tive theory (Longmore & Worrell, 2007). Data from assessments of therapists’ adherence

speak to the construct validity of the experiment by Jacobson et al. (1996). Although, in terms of the absolute frequency with which techniques were conducted, behavioral work was nearly as frequent in the CB condition as they were in BA, the relative frequency of BA procedures was greater in BA than in AT and CB. In the follow-up analysis of the trial, Jacobson and colleagues went on further to note:

… by definition, participants in the BA condition received more BA than did those in the other treatment conditions. Although one might be tempted to infer from this study that cognitive interventions are nonessential, our study does not directly address the validity of such an interpretation. All we can conclude is that adding cognitive interventions to BA is no more effective than using that time to add more BA. (Gortner, Gollan, Dobson, & Jacobson, 1998, p. 381, emphasis added)

In other words, if the findings from Jacobson et al. (1996) are taken to mean that CT procedures are inert, a corollary emerges that was not tested in their design: that a BA condition that allowed only 6 or 7 sessions (one-third of the 20 sessions) should yield outcomes equivalent to a 20-session course of BA. Thus, per Gortner et al. (1998), the only hypotheses regarding behavioral treatment of de- pression that could have been tested with the study design were that BA is: (a) superior to, or (b) not inferior to, a cognitive-behavioral treatment. As there were no significant differences between the treatments, the most that can be said is that the BA condition, in the context of a study with low power, was not shown to be inferior to treatments that divided time between cognitive components and behavioral ones. Very few dismantling studies have directly

compared “pure” cognitive and behavioral inter- ventions. The handful of studies that have com- pared purely behavioral (e.g., activity scheduling) to purely cognitive (e.g., cognitive restructuring) treatments for depression have tended to find little if any difference in the acute effects of cognitive versus behavioral treatments (Mazzucchelli, Kane, & Rees, 2009). In one study comparing BA to CT (Dimidjian et al., 2006), although there were no statistically significant differences between the two

treatments, BA was more effective than CT or medications for individuals who were severely depressed. In the BA condition, 76% of more severely depressed participants met criteria for re- sponse or remission, compared to 48% of patients in CT and 49% in antidepressant medications. Among the less severely depressed patients, response rates on the BDI were 56% in CT, 60% in BA, and 40% in ADM. Coffman et al. (2007) identified patients from the Dimidjian et al. trial who exhibited a pattern of “extreme non-response” (ENR) in CT (approximately one-fourth of those assigned to CT) and noted that none of the patients assigned to BA evidenced an ENR. At baseline, the CT ENR patients were more severely depressed, evidenced more functional impairment, and reported more problems with their primary support group. They interpreted these findings to suggest that, relative to CT and medications, BA may be particularly effective for patients with severe depression that is accompanied by interpersonal dysfunction. However, it should be noted that the advantage of BA over CT dissipated entirely across the trial’s 2-year follow-up (Dobson et al., 2008). Moreover, to our knowledge, the findings of Dimidjian et al. (2006) have not been replicated. An attempt to replicate Coffman et al.’s ENR findings in a separate sample of depressed patients (Koenig, Jarrett, Gallop, Barrett, & Thase, 2014) treated with CT found a low (6%) rate of nonresponse, and severity, functional impairment, and interpersonal problems were not good predictors of nonresponse. Finally, Webb et al. (2013) found that the therapists in the Dimidjian et al. (2006) trial implemented CT with a relatively more behavioral than cognitive focus, compared to therapists from other CT trials. The effect of cognitive change strategies in Beck’s

C(B)T for depression has also been questioned, on the basis that much of symptom change occurs early in treatment (Ilardi & Craighead, 1994). However, it has been shown in several studies that CBT therapists use cognitive change techniques as early as session one (Braun, Strunk, Sasso, & Cooper, 2015; Conklin & Strunk, 2015; Strunk, Brotman, & DeRubeis, 2010). In fact, at least one therapy manual (i.e., Muñoz & Miranda, 1986) addresses cognitive change exclusively for several sessions before addressing behavior. In anxiety disorders, among the studies included

in the meta-analyses of Adams et al. (2015) and Bell et al. (2013), only eight studies (Barlow, Rapee, Brown, 1992; Borkovec, Newman, Pincus, & Lytle, 2002; Emmelkamp & Beens, 1991; Marks, Lovell, Noshirvani, Livanou, & Thrasher, 1998; Mattick, Peters, & Clarke, 1989, Szymanski, & O’Donohue, 1995; White, Keenan, & Brooks, 1992; Williams,

 

 

790 l o re n zo – l ua c e s et a l .

& Falbo, 1996) compared a purely cognitive treatment condition to a purely behavioral condi- tion. Among these eight studies, we conducted a random effects meta-analysis (settings as per IntHout, Ioannidis, & Borm, 2014). There was no statistically significant difference in end-state pri- mary outcomes between “pure” CT and “pure” BT (higher values indicate superiority for CT; g = 0.010 [95% CI: -0.203 to 0.222], SE = 0.090, t = 0.106, p = 0.919; see Fig. 1). There was minimal heterogeneity between effect sizes included in the meta-analysis (Cochrane’s Q = 3.371, df = 7, p = 0.849; 14.90% heterogeneity), consistent with an account that trial-level findings were generally equivocal, with few meaningful between-trial dif- ferences in effects. The results of these studies suggest that CTs can be as effective as BTs in the treatment of anxiety disorders. Although exposur- e-based treatments are considered the mainstay of CBTs for anxiety, other meta-analytic reviews also suggest that ERP, CBT, and CT are about equally effective across a range of anxiety disorders (Norton & Price, 2007; Ougrin, 2011). This led Arch and Craske (2008) to propose that cognitive restructuring is a form of exposure whose effects are possibly cognitively mediated. One exception to the pattern of equivalence in CT and BT is that, for OCD, ERP appears to be more effective than CT

FIGURE 1 Meta-analytic plot of the comparative efficacy of “pure” C Adams et al. (2015) and Bell et al. (2013). Positive values indicate a super effect size of the end-scores of a trial primary outcome measure, sele prepared by LLL. No notable changes in effect size or statistical significanc meta-regressing the Hedge’s g of the pre-treatment score differences bet not detect the presence of a significantly asymmetrical funnel plot pot Copas’ (2010) test of publication bias proffered a similar between-grou

(Fisher & Wells, 2005; McLean et al., 2001; but see Öst, Havnen, Hansen, & Kvale, 2015). Another exception is that, in social anxiety, the CT com- ponent appears to add to the efficacy of exposure (Hofmann, 2004; Mayo-Wilson et al., 2014; Ougrin, 2011; but see Chambless & Gillis, 1993). It has been suggested that, because with some

anxiety disorders cognitive techniques may add little or nothing to the efficacy of BTs, exposure and other BTs are best conducted without the questioning of beliefs or the provision of other CT procedures (e.g., Arch & Craske, 2008). Indeed, a favored approach to the treatment of simple phobias has been and continues to be one that relies primarily or only on BT techniques (Wolitzky-Taylor, Horowitz, Powers, & Telch, 2008; but see Choy, Fyer, & Lipsitz, 2007). Some BTs, like the ones focused on relaxation, do not even directly address feared stimuli which may be taken to call into question the need to engage in cognitive work. There are at least two things to keep in mind regarding these comparisons. First, there is no evidence that these therapies lead to greater symptom reduction than CT (e.g., Borkevec et al., 2002; Mayo-Wilson et al., 2014). Second, the fact that these therapies, which do not directly address thinking, are effective does not directly inform about their mechanisms. Recall that Beck (1970) accounted for the efficacy of

T compared to “pure” BT in anxiety disorder RCTs identified by iority of CT over BT. Hedge’s g was calculated as a between-groups cted by RJD and JRK from a results-blinded list of trial outcomes e resulted from controlling for pre-treatment severity differences by ween treatments (analyses available upon request). Egger’s test did entially indicative of publication bias (p = 0.169), and Henmi and ps effect estimate (g = 0.017).

 

 

791cb t: n at ur e an d r el at ion to non -c bt

behavioral therapies, giving the specific example of desensitization, in cognitive terms.

techniques vs. mechanisms It appears to be a frequent misunderstanding of comparative psychotherapy research, including com- ponent trials, that outcome studies inform about the mechanisms of treatments. Various authors have cautioned against conflating the efficacy of thera- peutic procedures with their mechanisms of action (e.g., Hofmann, 2008; Lorenzo-Luaces et al., 2015), and the separation between therapeutic procedures and mechanisms has long been recognized (e.g., Beck, 1970; Foa & Rauch, 2004). Jacobson et al. (1996) explained it this way:

Of course, it is also possible that BA-focused treatments are more effective ways of changing the way people think than treatments that explicitly attempt to alter thinking. Perhaps the exposure to naturally reinforcing contingencies produces changes in thinking more effectively than the explicitly cognitive interventions do. (p. 303)

Although it may be tempting to assume that BT and CT produce symptom change via different theorized mechanisms (see DeRubeis, Brotman, & Gibbons, 2005), they may also do so by the same mechanisms (Hofmann, 2008). It is possible that both treatments work because they change think- ing. A pattern of findings that would be consistent with this hypothesis is that both therapies change cognition and that the cognitive changes lead to symptom change irrespective of how the cognitive changes are achieved. This latter point is one of contention in the research literature because some have erroneously assumed that a cognitive theory of change implies that change in cognition leads to symptom change uniquely in CBTs (see Hollon, DeRubeis, & Evans, 1987). To the contrary, cogni- tive theories highlight the mechanistic role of cog- nition in psychopathology (Lorenzo-Luaces et al., 2015). Given that findings from comparative outcome

studies can, at best, provide food for thought about mechanisms of change, what is needed is more research on the psychological changes that account for symptom change in psychotherapy. In the con- text of component designs (e.g., Hofmann, 2004), as well as in other kinds of randomized compari- sons, investigations of the mediation of the effects of psychotherapy promise to advance the under- standing of the workings of psychotherapy.

temporality One of the greatest challenge to our understanding of how BT and CT work is that most studies that explore these questions have been unable to rule out

reverse causality. Inferences about causality rest on the ascertainment of the correct temporal relation of the criterion and predictor variables. One must have confidence that change on the predictor variable preceded the change in the criterion. This has sometimes been interpreted to mean that pre- post changes in an outcome measure, regressed on an index of early change, establishes temporal precedence. However, it frequently is the case that substantial portions of pre to post changes in both proposed mediator and the outcome occur early in treatment, making early measurement of the medi- ator variable a crucial step in establishing causality. Changes in the criterion variable must be assessed subsequent to the assessment of change in the mediator if the aim is to rule out reverse causality. Only a small minority of tests of relations between symptom changes and proposed mediators of those changes has conformed to this pattern.