Behaviour Research and Therapy 44 (2006) 1���25 Acceptance and Commitment Therapy: Model, processes and outcomes Steven C. Hayesa, , Jason B. Luomaa, Frank W. Bondb, Akihiko Masudaa, Jason Lillisa a Department of Psychology, University of Nevada, Reno, NV 89557-0062, USA b Goldsmiths College, University of London, UK Received 4 June 2005 accepted 30 June 2005 Abstract The present article presents and reviews the model of psychopathology and treatment underlying Acceptance and Commitment Therapy (ACT). ACT is unusual in that it is linked to a comprehensive active basic research program on the nature of human language and cognition (Relational Frame Theory), echoing back to an earlier era of behavior therapy in which clinical treatments were consciously based on basic behavioral principles. The evidence from correlational, component, process of change, and outcome comparisons relevant to the model are broadly supportive, but the literature is not mature and many questions have not yet been examined. What evidence is available suggests that ACT works through different processes than active treatment comparisons, including traditional Cognitive-Behavior Therapy (CBT). There are not enough well-controlled studies to conclude that ACT is generally more effective than other active treatments across the range of problems examined, but so far the data are promising. r 2005 Elsevier Ltd. All rights reserved. Keywords: Acceptance and Commitment Therapy Relational Frame Theory Mindfulness Acceptance Mediational analysis Third generation CBT Clinical behavior analysis Contextualism Introduction The behavior therapy movement began with two key commitments. Behavior therapy was to be a field designed to (1) produce a scientifically based analysis of behavioral health problems and their treatment cast in terms of basic psychological processes, and (2) develop well-specified and empirically validated interventions for such problems. Franks and Wilson���s (1974) well-known early definition of behavior therapy shows that dual commitment clearly, asserting that behavior therapy was based on ������operationally defined learning theory and conformity to well established experimental paradigms������ (p. 7). Over the 40 years of development of behavior therapy, however, only the latter of these two commitments has been firmly kept. ARTICLE IN PRESS www.elsevier.com/locate/brat 0005-7967/$ -see front matter r 2005 Elsevier Ltd. All rights reserved. doi:10.1016/j.brat.2005.06.006 Corresponding author. Tel.: +17757463121 fax: +17757841126. E-mail address: hayes@unr.edu (S.C. Hayes).

Many methods of change in the contemporary cognitive and behavioral therapies are linked to relatively narrow clinical theories, not to basic principles derived from ������operationally defined learning theory������ or indeed any other basic science. The failure to provide an adequate basic account has reduced the scientific progressivity of the behavioral and cognitive therapies, and the overall coherence of the resulting science. Focusing merely on validation of an ever-expanding list of multi-component manuals designed to treat a dizzying array of topographically defined syndromes and sub-syndromes creates a factorial research problem that is scientifically impossible to mount. Such a ������brute force������ empirical approach makes it increasingly difficult to teach what is known or to focus on what is essential. Linkage to basic principles helps solve this problem because it allows the diversity of methods that result from clinical creativity to be distilled down to a manageable number of common core processes (cf. Harvey, Watkins, Mansell, & Shafran, 2004). It also fosters the practical value and coherence of psychology itself by allowing empirical clinical psychology to contribute the development of the discipline. The present article briefly considers why the linkage to basic principles was weakened, and examines evidence regarding the model of psychopathology and change that underlies Acceptance and Commitment Therapy (ACT, said as a single word, not as initials Hayes, Strosahl, & Wilson, 1999). ACT has followed a very different developmental approach by consciously developing a basic research program to meet the needs of modern behavioral and cognitive therapies such as ACT, in the hopes that the models that result will be more fruitful. How behavior therapy weakened its link to basic principles Behavior therapy can be divided into three generations: traditional behavior therapy, cognitive-behavior therapy (CBT), and the more recent ������third generation������ of relatively contextualistic approaches (Hayes, 2004). In the first generation of behavior therapy it was possible to keep both of its founding commitments because traditional behavior therapists drew on a large set of basic principles drawn from the basic behavioral laboratories. Even in the earliest days, however, authors of behavioral principles texts realized that these principles needed to expand beyond operant and classical conditioning principles to include those focused on human cognitive processes (Bandura, 1968). Clinicians realized that as well, and this insight was at the core of the second generation of traditional cognitive therapy and CBT (e.g., Beck, Rush, Shaw, & Emery, 1979). Unfortunately, behavior analysis was unable to supply an empirically adequate account of cognition, despite taking private events seriously. This left basic cognitive models as the only alternative, but none were as easily linked to clinical interventions as were learning theory principles. The reasons for this are complex, but time has shown that they go well beyond merely the stage of development of basic cognitive analyses at the time. After decades of relatively unsuccessful effort, this difficulty in linking behavior therapy to basic cognitive models appears more likely to be the result of a philosophical mismatch. When CBT emerged, the dominant cognitive models largely were (and remain) either mechanistic information processing approaches or organismic cognitive developmental approaches. For philosophical reasons, both are more focused on the nature and evolution of cognitive acts and their impact on other forms of action than they are on the specific contextual events that regulate these psychological events and relate them one to the other. This feature tends to limit the direct applied relevance of the basic concepts that result (Hayes & Brownstein, 1986). Let us explain. A principle like reinforcement is focused on the interface between action and its manipulable context, in effect, unifying both dependent and independent variables into a single unit. When the clinician applies such a concept to change behavior (we will we use the term ������behavior������ in this paper as it is used in behavior analysis, that is, as a term for all forms of psychological activity, both public and private, including cognition), the independent variables specified by the term can be manipulated and the effect noted. This is not, in the main, true of the cognitive concepts generated by information processing and developmental cognitive perspectives. A concept like cognitive schemas (Piaget, 1964) is focused on the organization of a specific kind of dependent variable (cognition) but it does not itself specify the contextual events that alter this variable or regulate its impact on other forms of activity. Similarly, explanations of cognition that focus on the material causality of the brain in essence shift a dependent variable from one level of analysis to another level of analysis but without providing concrete and manipulable independent variables ARTICLE IN PRESS S.C. Hayes et al. / Behaviour Research and Therapy 44 (2006) 1���25 2

that can be directly changed. Explaining cognition in terms of unfolding developmental patterns has this same ������independent variable free������ quality. As a result, principles drawn from basic theories of this kind have hard time leading directly or efficiently to the practical causes emphasized by clinicians and their pragmatic purposes. Unable to rely fully on basic cognitive accounts, as CBT was born clinicians created their own cognitive models to guide treatment development. This step allowed greater specificity about clinically relevant cognitive targets and it allowed the field to move ahead without having to wait for a basic account that could be usefully applied, but it also meant that the behavior therapy tradition was implicitly adopting a new and very different model of scientific development in which basic principles would be far less important to behavior therapy than its originators imagined. For example, CBT researchers defined and measured specific patterns of cognitions characteristic of specific forms of psychopathology (e.g., Hollon & Kendall, 1980). The terms used to describe these patterns sometimes were loosely linked to basic cognitive psychology (e.g., schemas), but often they were not (e.g., Ellis, 1962) and in either case the actual content of these cognitive processes (e.g., over- generalization Black and White thinking emotional reasoning irrational cognitions, and so on) were of little importance to basic cognitive science. These concepts were ������cognitive������ in the sense that they were about thinking as understood in common sense terms���they focused on ������thoughts.������ In the area of treatment, the relationship to basic processes was even more tenuous: cognitive disputation, empirical tests, collaborative empiricism, and so on were not methods of fundamental importance to the basic cognitive science laboratory���they were common-sense practical procedures generated clinically. The second generation of behavior therapy is now 30 years old, and the long-term impact of this second model of scientific development can be examined. The results are mixed. CBT techniques have produced impressive outcomes in many areas but it is not clear how much of this is due to what was added to traditional behavior therapy. When we look specifically at the original goal of an analysis linked to basic principles, the picture is not positive. The link between cognitive therapy and basic cognitive science continues to be weak. Looking at the array of popular techniques developed in CBT, none are known to have emerged directly from the basic cognitive science laboratories. Component analysis studies have generally failed to find support for the importance of direct cognitive change strategies, which was the common sense lynch pin of CBT (Gortner, Gollan, Dobson, & Jacobson, 1998 Jacobson et al., 1996 Zettle & Hayes, 1987). Well-known cognitive therapists have been forced to conclude that in some important areas there is ������no additive benefit to providing cognitive interventions in cognitive therapy������ (Dobson & Khatri, 2000, p. 913). The response to traditional cognitive therapy often occurs before cognitive change techniques have been implemented (Ilardi & Craighead, 1994), a finding that has still not been adequately explained. Support for the hypothesized mediators of change in CBT is weak (e.g., Burns & Spangler, 2001 Morgenstern & Longabaugh, 2000), particularly in areas that are causal and explanatory rather than descriptive (Beck & Perkins, 2001 Bieling & Kuyken, 2003). This overall picture presents an anomaly. One the one hand, most empirical clinicians agree that traditional behavior therapy was simply not adequate and that better methods of dealing with thoughts and feelings were needed. CBT is widely understood to have been a step forward in freeing up the behavior therapy tradition to work directly with cognition, and the outcomes for CBT protocols are generally quite good compared to work outside of behavior therapy writ large. On the other hand, the core conception of traditional cognitive and CBT���that direct cognitive change is necessary for clinical improvement���is still not well supported, and there is scant evidence that traditional CBT is bringing together basic and applied analyses into a more scientifically coherent and useful discipline. Taking another road: ACT, RFT, and contextual behavior analysis This is a time of upheaval in behavioral and cognitive therapy, particularly due to the rapid rise of acceptance and mindfulness-based interventions. These third generation approaches have been defined as follows (Hayes, 2004): Grounded in an empirical, principle-focused approach, the third wave of behavioral and cognitive therapy is particularly sensitive to the context and functions of psychological phenomena, not just their form, and ARTICLE IN PRESS S.C. Hayes et al. / Behaviour Research and Therapy 44 (2006) 1���25 3

thus tends to emphasize contextual and experiential change strategies in addition to more direct and didactic ones. These treatments tend to seek the construction of broad, flexible and effective repertoires over an eliminative approach to narrowly defined problems, and to emphasize the relevance of the issues they examine for clinicians as well as clients (p. 658). Examples of third wave CBT interventions include ACT, dialectical behavior therapy (DBT Linehan, 1993), mindfulness-based cognitive therapy (MBCT Segal, Williams, & Teasdale, 2001), and meta-cognitive approaches (Wells, 2000), among several others. Rather than focusing on changing psychological events directly these interventions seek to change the function of those events and the individual���s relationship to them through strategies such as mindfulness, acceptance, or cognitive defusion (Teasdale, 2003). Third generation approaches are emerging both within more behavioral and more cognitive wings of CBT, which is part of what justifies thinking of these changes in generational terms. In times of transition, basic assumptions and strategies can be more easily considered. In the present article, it is our purpose to characterize ACT and its underlying theory, to summarize the data available, and to begin to contrast ACT with traditional CBT. While ACT is part of current developments, it is also distinct in the particular development path it has followed, particularly in its attempt to develop the basic behavioral laboratory itself so as to generate more clinically adequate basic principles. The second generation broke away from the first because operant and classical learning principles were not alone adequate to account for human cognition. ACT is the applied extension of a 20 year long attempt to create a modern form of behavior analysis that could overcome this challenge by adding the principles needed to account for cognition from a functional contextual or behavior analytic point of view. The core conviction of this effort is that behavior therapy indeed needs to deal more effectively with cognition but that a contextualistic theory of cognition would be more likely to lead to accomplishment of its practical goals while retaining the original basic science commitments of the behavior therapy tradition. Clinicians always reside in the context surrounding the actions of clients and thus can only have an impact on these actions by manipulating contextual variables. These variables are specified in the case of traditional behavioral principles, but they are not in traditional mechanistic, organicist, or clinical theories of cognition. ACT embraces a contextualistic philosophy of science, a basic theory of language and cognition, and an applied theory of psychopathology and psychological change. We will very briefly describe each of these before proceeding to the empirical results of this approach. ACT philosophical roots Expanding behavior analysis to create a contextual theory of cognition required considerable philosophical clarity about the kind of contextualistic approach that might be adequate to that task. ACT is rooted in the pragmatic philosophy of functional contextualism (Biglan & Hayes, 1996 Hayes, 1993 Hayes & Brownstein, 1986 Hayes, Hayes, & Reese, 1988), a specific variety of contextualism that has as its goal the prediction and influence of events, with precision, scope and depth (Hayes, 1993). Contextualism views psychological events as ongoing actions of the whole organism interacting in and with historically and situationally defined contexts. These actions are whole events that can only be broken up for pragmatic purposes, not ontologically. Because goals specify how to apply the pragmatic truth criterion of contextualism (Hayes, Hayes, Reese, & Sarbin, 1993), functional contextualism differs from other varieties of contextualism that have other goals. ACT thus shares common philosophical roots with constructivism, narrative psychology, dramaturgy, social constructionism, feminist psychology, Marxist psychology, and other contextualistic approaches, but its unique goals leads to different qualities and different empirical results than these more descriptive forms of contextualism, seeking as they do a personal appreciation of the complexity of the whole (Hayes, 1993) rather than prediction and influence per se. ACT itself reflects its philosophical roots in several ways. ACT emphasizes workability as a truth criterion, and chosen values as the necessary precursor to the assessment of workability because values specify the criteria for the application of workability. Its causal analyses are limited to events that are directly manipulable, and thus it has a consciously contextualistic focus. From such a perspective, thoughts and feelings do not cause other actions, except as regulated by context (Biglan & Hayes, 1996 Hayes & ARTICLE IN PRESS S.C. Hayes et al. / Behaviour Research and Therapy 44 (2006) 1���25 4

Brownstein, 1986). Therefore, it is possible to go beyond attempting to change thoughts or feelings so as to change overt behavior, to changing the context that causally links these psychological domains. ACT theoretical roots Nearly a decade and a half passed between the earliest randomized trials on Comprehensive Distancing (the early form of ACT, Zettle & Hayes, 1986) and those in the modern era (e.g., Bond & Bunce, 2000). In that interval, the basic theory of human language and cognition underlying ACT, Relational Frame Theory (RFT Hayes, Barnes-Holmes, & Roche, 2001) was developed into a comprehensive basic experimental research program used to guide the development of ACT itself. RFT has become one of the most actively researched basic behavior analytic theories of human behavior, with over 70 empirical studies focused on it tenets. RFT is initially technically difficult, and it requires mastery of several new terms, but fortunately only a broad overview is necessary in the present context. According to RFT, the core of human language and cognition is the learned and contextually controlled ability to arbitrarily relate events mutually and in combination, and to change the functions of specific events based on their relations to others. For example, very young children will know that a nickel is larger than a dime by physical size, but not until later will the child understand that a nickel is smaller than a dime by social attribution. In addition to being arbitrarily applicable (a nickel is ������smaller������ than a dime merely by social convention), this more psychologically complex relation is mutual (e.g., if a nickel is smaller than a dime, a dime is bigger than a nickel), combinatorial (e.g., if a penny is smaller than a nickel and a nickel is smaller than a dime then a penny is smaller than a dime), and alters the function of related events (if a nickel has been used to buy candy a dime will now be preferred even if it has never actually been used before). The applied implications of RFT derive from several sources, but three critical features are that: (1) human cognition is a specific kind of learned behavior. For example, RFT researchers have recently shown that arbitrarily applicable comparative relations (the nickel and dime situation just mentioned) can be trained as an overarching operant in young children (Barnes-Holmes, Barnes-Holmes, Smeets, Strand, & Friman, 2004 Berens & Hayes, under submission) (2) cognition alters the effects of other behavioral processes (e.g., Dymond & Barnes, 1995). For example, a person who has been shocked in the presence of B and who learns that B is smaller than C, will show a greater emotional response to C than to B, even though B was directly paired with shock (Dougher, Hamilton, Fink, & Harrington, under submission), and (3) cognitive relations and cognitive functions are regulated by different contextual features of a situation (e.g., Wulfert & Hayes, 1988). The primary implications of RFT in the area of psychopathology and psychotherapy extend from the three important features just described (Hayes et al., 2001): (1) verbal problem solving and reasoning is based on some of the same cognitive processes that can lead to psychopathology, and thus it is not practically viable to eliminate these processes, (2) much as extinction inhibits but does not eliminate learned responding, the common sense idea that cognitive networks can be logically restricted or eliminated is generally not psychologically sound because these networks are the reflection of historical learning processes (3) direct change attempts focused on key nodes in cognitive networks creates a context that tends to elaborate the network in that area and increase the functional importance of these nodes, and (4) since the content and the impact of cognitive networks are controlled by distinct contextual features, it is possible to reduce the impact of negative cognitions whether or not they continue to occur in a particular form. Taken together, these four implications mean that it is often neither wise nor necessary to focus primarily on the content of cognitive networks in clinical intervention. Fortunately, the theory suggests that it is quite possible instead to focus on their functions. RFT has proven itself successful so far in modeling higher cognition in a number of areas (Hayes et al., 2001). For example, RFT researchers have successfully modeled analogy and metaphor (Stewart, Barnes- Holmes, & Roche, 2004), and shown that relational frames produce semantic priming (e.g., Hayes & Bissett, 1998). Neurobiological measures tell the same story. Complex RFT tasks generate pre-frontal activation (Barnes-Holmes, Regan, et al., in press) as would be expected based on cognitive research on relational reasoning (Waltz et al., 1999). Similarly, brain activation patterns show that priming within arbitrary stimulus relations in RFT tasks is relational, not merely associative (Barnes-Holmes et al., 2005). ARTICLE IN PRESS S.C. Hayes et al. / Behaviour Research and Therapy 44 (2006) 1���25 5