Keywords

Any account of how assessment and formulation within behavioral therapy evolved into a “second wave” of cognitive behavioral therapy is likely to focus on the period during which the two perspectives diverged. Within the conventional narrative echoed in hundreds of papers, Mahoney’s (1974) invocation of a cognitive revolution is portrayed as a discrete inflection point. However, a satisfactory understanding of how assessment and formulation evolved requires moving beyond this simplified introductory textbook account. Psychotherapy, in line with psychology as a whole, had long been trending away from strict behaviorism by this point, and Mahoney’s own writings at the time describe more of an evolution than a revolution. The approach he promoted continued along the lines of Albert Bandura’s efforts (e.g., Bandura et al., 1966) in seeking to preserve the functional framework of behaviorism but extending it to take in inferred, non-observable (at least intersubjectively) mediators of observable contingencies. The focus of behaviorism was on the contingencies between environmental events and their behavioral consequences, and so it seemed to those working from this viewpoint to be a small concession for a relatively large gain in explanatory scope to permit consideration of perceived contingencies. If Mahoney had hoped these ideas would gain currency among Skinnerians, he was disabused of this by B.F. Skinner himself. As he recounted, “My interest in beliefs, imagery, ‘perceived’ contingencies” and other ‘inner person processes’ …was deemed ‘misguided’ by Skinner, who insisted that there was no evidence whatsoever to support the ‘mentalistic speculations’ of cognitive psychology (Mahoney, 1985, p. 5). Mahoney persisted in advocating for an orderly extension of behaviorism into the cognitive domain, maintaining that cognitivism was separable from “mentalistic speculation” and that cognitive behavior modification (the name itself, perhaps, reflecting a desire for harmonious co-evolution) preserved the basic procedures of behavioral therapy, differing only in emphasizing an inferred change in cognition as a target (Mahoney & Kazdin, 1979).

In contrast, the emergence of A.T. Beck’s approach, which has come to be most closely identified with “second wave” cognitive therapy, signaled a conscious and definitive break from orthodox behaviorism. In the inaugural issue of the journal, Behavior Therapy, Beck (1970) unapologetically placed introspection at the center of the therapy he and his collaborators had been developing: “Study and analysis of the introspective data suggest that the cognitive organization, far from being a mere link in the stimulus response chain, is a quasiautonomous system in its own right. Although this system generally interacts with the environment to a large extent, it may at other times be relatively independent of the environment… Data suggest that cognitive organizations are highly active and are much more than a simple conduit between stimulus and response” (p. 194).

Notably, Beck’s own name for the therapy he helped originate was “cognitive therapy,” a term free of any implied shared credit with behaviorism. Beck is forthright in his claims; however, although he uses scientific terminology, it is evident that he is actually describing clinical rather than scientific investigation. The data he cites was gained through clinical observation, the analysis was clinical formulation, and the structures and organization were based on clinical inferences drawn from this material. Beck’s position and scientific framing reflected confidence that empirical validation of these clinically derived conclusions would inevitably follow in due course. This was premised on a wager on progress in the science behind the cognitive revolution within the emerging information processing paradigm and that this knowledge could be drawn upon to explain phenomenal experience. Much of what would unfold over the next several decades turned on whether cognitive therapy could make good on this wager.

Schwartz (1982) reviewed the points of contention between the interlocking positions from these mid-twentieth century debates centered around different views of the status of cognitions. The typical behaviorist position would be to regard cognitions as private behaviors that are subject to the same general behavioral laws (such as reinforcement) as overt behaviors. In contrast, from the standpoint of social learning theorists such as Bandura, cognitions mediate between antecedent stimuli and overt responses and can have effects on learning and behavior that cannot be explained strictly according to standard behavioral principles, although they can still be largely accommodated within the same functional framework. The cognitivist view differed from both of the preceding positions, as reflected in Beck’s stance that cognitions are distinct phenomena organized in a manner that is different from how behaviors are organized and that are governed by their own set of scientific laws. As such, distinguishing cognitivism from mediational behaviorism critically depended on establishing knowledge of these distinct laws.

Schwartz also draws on a distinction discussed by Eysenck (1972) between methodological and analytic behaviorism. The latter of these would include Skinner’s radical behaviorism, which, on grounds of parsimony, objected to affording special status to private events out of concern that the unneeded postulation of such hypothetical entities represented a conceptual slippery slope (as reflected in Skinner’s response to Mahoney’s ideas). However, the main behavioral critiques were made by methodological behaviorists, who were less embedded in a philosophy of science than the analytic behaviorists, but condemned research approaches that did not exercise sound experimental control. Macleod (1993) put this starkly: “…developing an acceptable formal science of cognitivism requires adherence to rigorous methodological constraints that are at least as severe as those imposed by behaviorism. Conversely, the endorsement of the self-report data, yielded by introspection, as an acceptable source of information concerning mental processes, must place our discipline clearly out with the boundaries of legitimate science” (p. 170). MacLeod could be counted as one of a group of critics who might be called “cognitive denialists” who argued that the cognitive revolution was not yet proven (See O’Donohue et al., 2003). They were notably unimpressed with the idea that a new paradigm was taking shape before their eyes. In keeping with idea that the terms used to designate the therapies being compared were revealing, one such critique (Beidel & Turner, 1986) suggested that the name of the approach should be hyphenated as “cognitive-behavioral therapy” given the not-yet-proven status of the cognitive approach meant that it did not warrant a full-fledged adjective, having not made good on the wager that the theoretical slippage of opening the door to introspective phenomena was justified by commensurate scientific gains.

Two Levels of Assessment

Elaborating further in his 1970 Behavior Therapy paper on what he viewed to be the subject matter of the emerging scientific basis of cognitive therapy, Beck wrote, “Introspective data indicate the existence of complex organizations of cognitive structures involved in the processes of screening external stimuli, interpreting experiences, storing and selectively recalling memories, and setting goal and plans… (p. 194)”. Here, Beck echoes the four processes Bandura had set out as being involved in social learning (attention, retention, reproduction, and motivation) but makes a notable addition: interpretation of experiences. While the other four aspects listed are important for anchoring the cognitive approach within the emerging information processing paradigm and play a role in various theoretical constructions, the interpretation of experiences, through identification and analysis of the thoughts arising in reaction to environmental events, so called automatic thoughts, was the central focus of cognitive therapy and the main target of early formal assessment efforts that grew directly from therapy practices.

The emphasis of initial cognitive therapy sessions was on helping the client to identify the reasoning errors and biases that were viewed as underpinning emotional distress. Tools such as the Daily Record of Dysfunctional Thoughts (DRDT; Beck et al., 1979) came into use to aid recording and discussing these thoughts, and early scales of depressive cognition such as the Automatic Thoughts Questionnaire (ATQ, Hollon & Kendall, 1980) can be seen as a straightforward extension of therapy-based data gathering, albeit in a more standardized form that would potentially permit systematic study. Scores on scales like the ATQ afforded a summary score of depressogenic cognition that offered the potential to compare individuals on these dimensions and to quantify change in response to treatment as therapists raised awareness of negative thought content and helped correct logical errors.

In working clinically with automatic thoughts, it became evident to Beck and his colleagues that these transient appraisals of ongoing experience (e.g., “no one at this party likes me”) were markers of more enduring, thematically relevant underlying beliefs (e.g., “if you don’t impress people with your personality, they won’t like you”). Beck et al., (1979) observed that the same beliefs seemed to recur over successive symptomatic episodes and so likely persisted in some form, representing a vulnerability for future depression (Kwon & Oei, 1994). The expression of such beliefs was taken to reflect the operation of schemas, the central mechanisms guiding information processing built up over the individual’s learning history. Similar to the schema concept in Piagetian theory (Hollon & Kriss, 1984), these served to store previously encoded knowledge, but, importantly, to also play a role in processing new information, helping to determine which information would be attended to and which would be ignored, how much importance to attach to stimuli, and how to structure information (Hollon & Kriss, 1984, p. 37). Information congruent with schematic processing would be preferentially processed relative to schema irrelevant information, whereas schema incongruent material would be ignored or at least minimized (Beck, 1987; Clark et al., 1999).

Therapy typically progressed from initial focus on challenging automatic thoughts to efforts later in therapy to bring to awareness the ongoing beliefs that appeared to give rise to these thoughts. As a counterpart to scales such as the ATQ, Weissman (Weissman, 1979; Weissman & Beck, 1978) constructed the Dysfunctional Attitude Scale (DAS) to capture the corresponding level of enduring beliefs. DAS items were written so that the operation of the sort of arbitrarily negative reasoning patterns that Beck had identified as being at the core of depression were embedded in the logic of the stated conditional beliefs (e.g., the item “If a person is indifferent to me, it means he does not like me” reflecting an arbitrary inference). Endorsement of the maladaptive beliefs were assumed to indicate a disposition to apply comparable logic when the respondent encountered similar situations in the course of their own experiences. Weissman’s stated aim was to compile a set of items that “cover most of the essential dimensions of depressogenic cognitions, even if these were confounded, overlapping, or otherwise not as clear-cut as later research might help to make them.” (pp. 63–64).

Publication of Beck’s (1976) Cognitive Therapy and the Emotional Disorders crystallized the theory first set out with reference to depression and laid the groundwork for the extension of both the psychotherapy and the accompanying research methods and instruments to the broad spectrum of clinical psychology phenomena. In this book, Beck maintained that cognitive content could provide an essential basis for distinguishing between diagnostic categories, a view formalized as the cognitive content specificity hypothesis (Beck et al., 1987a; Clark et al., 1989; Baranoff & Oei, 2015). The content specificity hypothesis provided the seeds for thinking systematically about other disorders with the same approach that had been applied to depression, and, accordingly, self-statement questionnaires similar to the ATQ but encompassing the distinctive expressed cognitive content of other disorders would soon follow (e.g., the Agoraphobic Cognitions Questionnaire; Chambless et al., 1984). Likewise, scales measuring more enduring constructs also tracked the extension of CBT to different disorders, with the DAS joined, for example, by the Anxiety Sensitivity Index (Reiss et al., 1986) for panic disorder and, somewhat later, the Obsessive Beliefs Questionnaire (OBQ; Obsessive Compulsive Cognitions Working Group, 2003) for obsessive-compulsive disorder. Items found on scales in this category typically reflect beliefs concerning experiences salient to the disorder and so often adopt an if-then (if a salient trigger occurs, then an expected consequence will follow) format, as shown in the ASI item, “When my chest feels tight, I get scared that I won’t be able to breathe properly” (indicative items from the DAS and OBQ include “If I fail at my work, then I am a failure as a person” and “If I do not control my thoughts, I will be punished,” respectively), hearkening back to Mahoney’s category of self-perceived contingencies.

Targets of Self Report Assessment

The disagreement between the different behavioral and cognitive positions over the admissibility of self report related to what evidence such reports were understood to provide. Self-report and introspection were never categorically proscribed, and none of the theoretical stances would take issue with the idea that a contemporaneous report of subjective thought was likely a report of something that seemed real to the person providing it. Indeed, it would be mistaken to assume that self-report was absent from behavioral approaches. This was duly noted as seemingly hypocritical by various commentators as at odds with the professed distrust of self report among self-identified behaviorists. For example, Bergin (1970) remarked, “It is difficult…to imagine how desensitization can be considered to be a ‘behavioral’ procedure in any definitive sense. In employing the technique, the initial diagnostic evaluation relies chiefly upon introspective reports in interviews and personality inventories or fear surveys” (Bergin, 1970, p. 206; see also Breger & McGaugh, 1965). Behaviorists would likely fail to perceive any inconsistency as long as the verbalization was not regarded as offering proof of any particular internal phenomenon. Those working more from a social learning perspective would also have considered themselves behaviorists at that time but defined their approach to assessment mainly in contrast to the more traditional structural approaches based on personality types and including the psychodynamic tradition. Accordingly, Goldfried and Kent’s (1972) account of behavioral assessment echoed Mischel’s arguments for the centrality of the situational context rather than structural fixed traits as determinants of behavior (e.g., Mischel, 1973), “The techniques associated with behavioral assessment include the observation of individuals in naturalistic situations, the creation of an experimental analogue of real-life situations via role playing, and the utilization of the individual’s self-reported responses to given situations” (p. 412). They go on to draw on Goodenough’s (1949) distinction between signs and samples as targets of assessment: “The sign approach assumes that the response may best be construed as an indirect manifestation of some underlying personality characteristic. The sample approach, on the other hand, assumes that the test behavior constitutes a subset of the actual behaviors of interest. Whereas traditional personality tests have typically taken the sign approach to interpretation, behavioral procedures approach test interpretation with the sample orientation” (p. 413).

The shift from the social learning position, consistent with Mahoney’s “self-perceived contingencies,” to the “cognitivist” position, as stipulated in Schwartz’s analysis, is a subtle one. It is reflected in the move from self-perceived contingencies—that is, the subjective association of particular events with particular contextual factors—to framing beliefs as reflective of reasoning processes, and explainable in terms of information processing concepts instead of merely contingent co-occurrence. This subtle shift is discernible in the wording of items, which go from referring to the contexts within which thoughts and behaviors occur (“when my heart races, I think I will have a heart attack”) to reflecting a particular understanding of the world (“if my heart is racing, it’s a sign of a heart attack”). Concerns about what these scales could be understood to be capturing relative to actual thought processes can be found in the literature, although this is fairly rare relative to the burgeoning amount of research these scales would enable. Though seldom articulated in the literature, there are indirect indications that the salience of concerns about the veridicality of self report was heightened by the publication of Nisbett and Wilson’s (1977) “Telling More Than We Can Know: Verbal Reports on Mental Processes,” which documented in detail potential pitfalls of taking first-hand verbal accounts of thinking at face value. Writing some years later, rather than viewing these reports as samples as Goldfried and Kent (1972) had argued, Dobson and Segal (1992) stated that “self report assessment procedures are commonly interpreted as signs of behavior or other responses; that is, although they do not directly enable observation of real-life processes, they approximate these processes (p. 279).” Likewise, with regard to the resemblance of questionnaire items to the targeted mental processes, Glass and Arnkoff (1982) noted that “it is unlikely that people have precisely the thoughts that they endorse on a questionnaire, because actual thought processes are probably highly idiosyncratic, automatic, not in the form of complete sentences, and heavily based in imagery and not just language (pp. 51–52).” Finally, Glass and Arnkoff (1997) echo Nisbett and Wilson’s concerns in considering how to construe the thought frequency estimates often required on self-report scales, suggesting that respondents might more realistically be (1) gauging the impact, salience, or importance of the thought and on this basis inferring that a particularly pertinent thought must have occurred frequently; (2) translating from idiosyncratic or fragmented actual thoughts to the grammatically complete sentences on the inventories; (3) translating affective experience into a language-based self-statement format; or (4) conveying that the item on the questionnaire matches their self-concept, indicating ‘that’s like me’ by endorsing the item. This general understanding of what respondents are conveying in answering self report questions is broadly in line with the conclusions of formal research into the cognitive underpinning of questionnaire and survey responses (e.g., Schwarz, 1999).

Bolstering justification for the use of self-reports, Ericsson and Simon (1980), in their paper “Verbal reports as data,” had offered a counterpoint to Nisbett and Wilson and outlined the circumstances under which verbal reports could be considered more or less reliable. They contended that verbal reports are more reliable where they are direct accounts of what can be attended to in short term memory and less reliable the more respondents are required to attend to and report on information that would not otherwise be attended to. The predominant endorsement format mainly used in self report scales, which requires respondents to endorse preset items with a fixed response format, often calling upon respondents to retrospect or to construct hypotheticals, is clearly open to criticism from the standpoint of these criteria. In contrast, production methods, such the Articulated Thoughts in Simulated Situations paradigm (ATSS; Davison et al., 1983), which, would provide respondents with a general prompt (e.g., a hypothetical social predicament) and ask the respondent to think aloud, fare better relative to the Ericsson and Simon framework. Material from such methods were encouraging to the extent that they produced similar content to what was included on endorsement format questionnaires. Less encouraging was the lack of concordance between production and endorsement methods (e.g., Heimberg et al., 1990), aside from the greater practical challenges of employing production methods compared to endorsement methods in routine practice and research.

However, predominantly, everyday use of self report largely took for granted that reports of subjective processes, broadly speaking, were largely veridical. Indeed, beyond the question of whether behaviorists themselves practiced what they preached when it came to self-report, cognitivists regarded the proscription against self report as counterproductive and an inhibitor of progress: “Where cognition has not been demonstrated to be important, it has often not been researched. This may be part of our behavioristic legacy. By defining cognitions as irrelevant, they have remained unexamined (Mahoney, 1977, p. 11).” The subsequent exponential growth in use of self-report was a testament to the resonance of this sentiment. Lawyer and Smitherman (2004), using Lang’s (1979) three response systems framework for anxiety (i.e., behavior, subjective report, and physiology) as a frame of reference, traced a discernible change in assessment approach, with the appearance of studies assessing more than one response system reaching a peak in the late 1970s and then declining steadily, overtaken by an increase in single-system assessment with an exclusive focus on self-report methodology. These mono-method self report studies had always predominated, but by 2002 they represented 97.8% of research reports of anxiety disorders, compared to 85.5% during the 1970s, a trend tracking the rise of CBT for anxiety (e.g., Beck et al., 1985). A comprehensive review taking in the previous two decades by a prominent researcher in the area by the late 1990s stated that assessment had lagged behind progress in other areas and noted with concern the nearly exclusive reliance on retrospective self report endorsement based scales (D.A. Clark, 1997).

Theoretical Critiques

For their part, with the proliferation of research centered on self-report, Skinnerians may have felt vindicated in their warnings of the slippery slope that would be set in motion by opening the door to theories relying on hypothetical mental entities (e.g., Hayes & Brownstein, 1986). Moreover, the fact that most of these studies assessed only self-report and no other response modes bore out the more fundamental concern that empirical findings of this sort were highly susceptible to tautological inferences in the absence of corroborating or validating information ascertained through separate response systems, emboldening claims that the cognitive revolution was a hollow victory (e.g., Beidel & Turner, 1986; MacLeod, 1993). Whether fairly or not, the numerical dominance of self-report research made it possible for critics to largely overlook the more modest but steady growth of a body of experimental findings using both new paradigms and also energetically adapting advances in experimental cognitive psychology and social cognition. By 1988, enough experimental evidence had accumulated to underpin an influential volume (Williams et al., 1988) that summarized the body of evidence of laboratory research into the cognitive approach to emotional disorders and advanced a theory synthesizing this evidence with regard to depression and anxiety.

If the main limitation of the first wave was allegiance to an elegant but restrictive theory with limited applicability to the problems that needed to be solved, the second wave faced the opposite criticism, that is, that there was an abundance of applicability but a lagging theoretical basis. Efforts to redress this imbalance took the form of conceptual frameworks that sought to bridge the large body of self-report findings with the more gradually accumulating knowledge base of experimental data to signpost where the field was heading in anticipation of an expected convergence of the two streams. Hollon and Bemis (1981) drew a parallel between the two types of self-report constructs (automatic thoughts and enduring beliefs) and the distinction drawn within the information processing paradigm between surface and deep cognition. Hollon and Kriss (1984) then proposed a taxonomy of cognitive structures, products, and processes. Products are the thoughts, images, self-statements or internal dialogue that represent output from the information processing system, and cognitive propositions that are the content of underlying beliefs or schemas. Previously encoded enduring beliefs are considered to be reflections of schema structure in propositional form. In contrast, cognitive products are the conscious outputs of the information processing system, and include momentary cognitions (e.g., automatic thoughts). Enduring beliefs when they have been retrieved and are in a person’s awareness are also considered to be cognitive products (as such, Ingram and Kendall (1986) distinguish between stored and accessed beliefs in their similar taxonomy).

Disagreements between cognitive and behavioral positions had settled into well-worn avenues. When Coyne and colleagues (Coyne, 1982; Coyne & Gotlib, 1983) ultimately succeeded in shifting the discourse in a landmark series of critiques, not being identified with either of the established positions was likely an advantage. Coyne (1982) began by expressing a fundamental skepticism of the possibility of parsing psychological phenomena into discrete elements. In common with behavioral critiques, Coyne went on to find fault with arbitrarily taking cognition to be the initiating factor among mutually interacting elements of behavior, cognition, emotions, and the environment: “We do not need to construct linear theoretical sequences to understand the role of cognition in behavior. Rather, we should recognize the arbitrariness of any punctuation of what is basically a circular sequence, and we should reject questions about whether cognition ultimately causes affect and behavior as the product of conceptual confusion (p. 8).” Coyne and Gotlib (1983) then went on to review the relevant literature for cognitive models of depression, finding the results commonly assumed to support the model deficient for substantiating the strong causal claims that had been set forth. This prompted an exchange with Segal and Shaw (1986; Coyne & Gotlib, 1986) covering a wide range of topics, but of particular relevance to the present focus, aside from questioning the implied primacy of cognition, was the argument on Coyne’s part that correlations between depression symptom scales and scales of putative depressive cognitions which had been taken as validity evidence regarding causality amounted instead to repeated sampling from the same pool of negative utterances among respondents who could have found themselves in such a position due to a complex combination of mutually dependent psychological and environmental factors unfolding over time.

It was against this background that Segal (1988) undertook an exhaustive review of the schema concept in cognitive therapy and its operationalization in terms of self-report, particularly with the DAS. Echoing Coyne and colleagues and several notable behaviorist critiques, he acknowledged that “the strategy of relying on negative self-reports to validate a construct whose operation is intended to explain these self-reports becomes increasingly circular unless additional external referents can be provided to demonstrate schematic processing” (p. 147). With reference to the cognitive conceptual frameworks described above, Segal argued that self-report scales like the DAS can only represent content, whereas structural concepts such as schemas require a means of capturing functional relations that is not possible solely with reference to content. Totaling scores on scales such as the DAS can only reflect the degree to which the beliefs of interest are present. How schemas are defined in the literature

emphasizes the functional aspects of the interrelation among self-descriptors, whereas questionnaire data are capable, at most, of providing evidence for a descriptive definition of schema. Such evidence is usually in the form of endorsement patterns for clusters of attitudes and beliefs. Using the DAS to argue for the existence of an organized self-structure involves assuming function on the basis of a description of interrelated attitudes and beliefs. Although it is possible that this interrelation may reflect a type of structure or personal organization in a broader sense, it cannot be accepted in support of the functional linkage between elements in a self-structure. (p. 153)

Segal’s (1988) review is a reasonable candidate for marking the end of the initial developmental arc of the second wave. The general tenor of where things stood is probably best gleaned from the views expressed by prominent scholars who were not strongly identified with either the behavioral or cognitive camps. In this connection, Bellack (1992) wrote:

The literature is filled with cognitive-sounding terms, such as attributions, schemata, and self-efficacy, but the theoretical and empirical underpinnings of these concepts are tenuous, at best. Moreover, in the rush to develop clinical models and techniques, basic research and theory on information processing and cognition have all too frequently been ignored. (Bellack, 1992, p. 384)

Similarly, Foa and Kozak (1997) would note that the central concepts of the model were still largely underpinned by clinical data:

Interestingly, while cognitive therapy embraced some of the terms of cognitive psychology (e.g., schemas), the theory that informed the practice of cognitive therapy was derived primarily from clinical observations, not experimental psychology or research in psychopathology. Indeed, the fundamental theoretical constructs of cognitive therapy, such as self-efficacy, cognitive distortions, and automatic thoughts, are based mainly on clients’ introspections. (p. 606)

It would appear from these assessments that, at least for the time being, the wager on a separate theoretical structure for knowledge derived from introspections had not been fulfilled.

Second Wave Formulation

Cognitive therapy as it developed in the late 1970s and 1980s, compared to behavioral therapy, which was largely headquartered in academic departments, became much more aligned with the general mental health research and funding system largely defined by psychiatry (see, e.g., Barlow & Carl, 2010). This promoted an emphasis on diagnosis specific manualized treatments for putatively discrete problems geared toward conclusive testing of efficacy through large scale randomized controlled trials (see Hallam, 2013). In parallel with theoretical critiques such as Coyne’s (1982; Coyne & Gotlib, 1983, 1986), doubts about this strategic direction began to be articulated, which, in retrospect, can be seen to be the seeds of what grew into the third wave. Jacobson (1997) argued that the field had borne a cost in distancing itself from smaller scale research that was amenable to testing functional relations between variables of interest in favor of a structural approach where a problem such as depression was viewed as being characterized by particular features (e.g., depressogenic schemas) that are universal across people who fall into the category defined by the problem:

Even defining the problem as ‘depression’ which unites behaviors by their topography rather than their function, is to be guilty of formal or structural – as opposed to functional – thinking. (p. 438)

Jacobson’s argument reiterated a frequently stated behavioral criticism of essentialist accounts of phenomena like depression while also strongly echoing Coyne’s earlier analysis (Coyne, 1982): “attention to the presumptive cognitive structures that presumably cause depression redirects our attention away from those things we can see, hear, and influence directly: the social context of the depressed individual (p. 440).” In Jacobson’s portrayal, CBT’s emphasis on manualization came at a substantial cost, as it meant that it was mainly rule (i.e., technique) based rather than circumstance based, necessitating an abrupt shift from rules to contingencies once therapy ends if any gains are to be sustained once the client is left to continue on their own within their normal life context.

And yet, Jacobson’s own analysis might be viewed as too categorical. While not at the forefront of the diagnosis based manualized approaches of the time, the functional elements of cognitive therapy treatment packages became more discernible once the therapy model was no longer centered solely on depression, which, arguably, lends itself less to functional analysis compared to anxiety disorders, where manifest behavior contingencies are more salient aspects of the typical presenting problem. For example, in D. M. Clark’s (1986) catastrophic misinterpretation model of panic, the misinterpretation serves the function of establishing a vicious cycle that escalates the expectancy of an imminent catastrophe; the precise content of the misinterpretation is secondary and is not assumed to be universal to those with the relevant diagnosis. The therapy based on this panic model and similar models of social anxiety and PTSD call for the identification of maladaptive coping behaviors (avoidance, safety-seeking behaviors) that are reinforcing in the short term but preclude needed changes in beliefs. Here, again, it is the function of the behaviors rather than their specific content that is the critical aspect. It is important to note that the central role of functional elements of these models was not simply fortuitous, but rather reflects the tradition of single case functional analysis promoted by Monte Shapiro (e.g., Shapiro, 1957) at the Institute of Psychiatry in London, where much of this work originated or can be traced back to (see Hallam, 2013, pp. 69–70).

At the same time, it is prudent to avoid presuming, counterfactually, that the same or better progress would have been assured without the conceptual scaffolding provided by the diagnostic approach had purely behavioral approaches held sway. Indeed, it stands to reason that with non-mutually exclusive diagnostic groups, evidence for cross-category transdiagnostic mechanisms, if present, will likely emerge over the course of time. Accordingly, transdiagnostic frameworks built on the gradually accumulating experimental evidence (Harvey et al., 2004) and based on psychometric approaches (Brown et al., 1998) began to appear, confirming, extending, and complementing the theories underlying the single disorder treatment models. These included both structural and functional constructs, for example, with respect to anxiety, overestimation of the likelihood of occurrence of negative events (biased expectancy reasoning; Butler & Mathews, 1983; MacLeod et al., 1997; Rachman & Hodgson, 1980), emotion-driven safety-seeking behaviors (Ferster, 1973; Salkovskis et al., 1996), and avoidance of feared and risky stimuli (Barlow et al., 2002). Bergin had anticipated and promoted such an approach: “There is no magic in either the terms ‘behavior therapy’ or ‘cognitive therapy,’ but there is progress in dimensionalizing given psychological phenomena of interest and designing interventions which have relatively unique relevance to them” (Bergin, 1970, p. 207). The idea that research involving non-mutually exclusive categories will naturally evolve to take in transdiagnostic concepts is supported by the fact that psychiatry, without the benefit of a particularly prominent behavioral tradition or reliance on functional analysis, has adopted the Research Domain Criteria (RDoC; Insel et al., 2010) framework, moving away from its longstanding emphasis on disorder-based organization of knowledge of psychological problems toward understanding basic empirical constructs conceived of as dimensions rather than categories.

Still, despite increasingly prominent elements of functional concepts within mainstream second wave cognitive therapy, it should be acknowledged that this has not been pursued consciously and is mainly discernible in retrospect. A more strategic balance between functional and structural perspectives in training of the models is likely to be beneficial—specifically, in areas in which the structural approach has proven itself to be nonoptimal. One such area is case formulation. There is no doubt that formulation, often called case conceptualization in cognitive therapy, is regarded as being of central importance: “When one asks a novice cognitive therapist how they would handle a specific clinical problem […] they usually can give a variety of techniques […]. Rarely does the novice address the most important step – conceptualization” (Beck et al., 1985, p. 181). However, it is, perhaps, telling that formulation is not included in the Cognitive Therapy Scale (Blackburn et al., 2001), commonly used to assess treatment fidelity in both routine practice and in clinical trials of cognitive therapy. There is a paradox at the heart of this state of affairs stemming from the contradiction between the desire for faithfulness to a pre-stipulated evidence based therapy approach as set against the need to adapt such broadly defined approaches to individual circumstances, particularly where, as is more often than not the case, an individual presenting for therapy does not resemble the prototype implied by the manualized version of the therapy or does not belong to the demographic of the corresponding trial sample.

The latest notable efforts to approach formulation systematically in second wave behavior therapy recognize the need to grapple with this central conundrum. However, deference is still by default given to the standardized protocol: “if theory and research about a specific disorder map onto a client’s presenting issues, the disorder- specific model has primacy due to its empirical support. Only by taking the time to describe the presenting issues and develop simpler explanatory models of the main presenting issues can the therapist establish which disorder specific models fit” (Kuyken et al., 2011, p. 214). In other words, for a given problem, variation from the normative model in the form of added complexity needs to be, in effect, partialled out so as to enable discerning the signal within the noise that corresponds to the embedded specific model. With regard to the complexity posed by multiple concurrent problems, “it is not always obvious which protocol to select for those clients with comorbid presentations or for those presentations that do not fit a particular model. How does a cognitive therapist choose from this vast array of choices? Case conceptualization helps the therapist select, focus, and sequence interventions” (Kuyken et al., 2011, p. 7). Here, separate disorders are regarded as being discrete, an idealized picture that does not correspond to comorbidity as more typically encountered, and there is no mention of the potential of formulation identifying common threads running between problems.

Persons, who has a had a longstanding focus on case formulation (Persons & Bertagnolli, 1999; Persons, 2008), described a similar set of assumptions with respect to predefined treatment packages designed for different diagnostic categories. Persons explicitly distinguished the approach she recommends from others that are based on functional assessment (Persons & Davidson, 2010, p. 173). Rather, in Persons’ account, formulating is a process of informed pattern matching. The drawbacks of the broad assumptions that underlie fixed diagnoses are acknowledged but are seen to be outweighed by the benefits of matching to published protocols, which bring with them evidence based sets of techniques and connect the presenting problem to the entirety of empirical literature. Persons and Davidson then continue:

Whenever possible, the case formulation is based on an empirically supported ‘nomothetic,’ or general formulation. The therapist’s task is to translate from nomothetic knowledge to idiographic practice, where an ‘idiographic’ formulation and treatment plan describe the causes of symptoms or disorders and the plan for treating them in a particular individual. (p. 175)

Skill in formulation equates to the ability, based on knowledge of the literature, to particularize the relevant nomothetic knowledge of etiological mechanisms so as to craft an individualized plan. This can also include categorization based on theory rather than solely on diagnosis. Persons offers a vignette of a client who is provided a more fitting therapy plan when it is discovered she “holds” a “subjugation schema” in the context of Young and colleagues’ schema therapy approach (Young et al., 2003), in doing so applying the sort of fixed trait approach Goldfried and Kent (1972) had argued behavior therapy was moving past.

It is instructive that when research is carried out on formulation from the mainstream CBT approach, a common practice is to benchmark against a group of experts (Kendjelic & Eells, 2007) or a specific expert (e.g., Judy Beck in the case Kuyken et al., 2005). This can be seen as a resort to personifying authority as a proxy for the authority otherwise conferred by implementing an evidence based model. However, research into the reliability of formulations based on expertise is uneven at best (Bieling & Kuyken, 2003). If, as previously argued, substantial elements of particular treatments are underpinned by functional mechanisms, many of the experimental findings incorporate the logic of functional analysis, and, as Jacobson (1997) has pointed out, functional analysis principles are often drawn upon in preparing a client for autonomously applying knowledge gained in therapy within their individual living context, the case for making functional analytic principles salient within formulation seems self-evident and, in fact, should arguably be a priority. Toward this end, Hallam has recently proposed a comprehensive approach to implementing functional analysis within individual case formulation using a diagrammatic approach (Hallam, 2013).

Implications for the Third Wave

In this chapter, I have aimed to convey an alternative to the introductory psychology textbook template in offering a retrospective account of the how of the first wave of behavior therapy passed into the second wave, as a counterpoint to the familiar narrative according to which the shiny new innovation sweeps away hidebound and blinkered old ways of doing things, what the late Scott Lilienfeld called “breakthrough-ism” (Lilienfeld, 2017).

The crosscurrents between the first and second waves are complex, with issues of assessment and formulation prominent in the mix. From one perspective, there is a basis for arguing that the behaviorist forebodings of unfettered proliferation of hypothetical constructs and tautological reasoning have, to a certain extent, been borne out, and this has led to obscuration rather than clarification. The counterargument to this boils down to an insistence that it was important to try, as it did not appear to be that better solutions were possible within the strictures of the behaviorist position, and human suffering was at stake. A final quote from Mahoney (1977) conveys this sentiment:

[O]ne can argue that the cognitive theorist has been obsessed with cognition. Virtually nothing has escaped the speculation of possible cognitive influence. It may be the case that the cognitive theorist has overgeneralized. We will not know, however, until we have looked. (p. 11)

Having looked, we now have an idea of the terrain, and rather than a sharp boundary between the knowable and the unknowable, the literature instead appears to form of a gradient, marked by discernible indicators of what helps and what hinders useful inferences. The phase of diagnosis based manualized therapy and efficacy tests, which has not yet ended by any means, but is being supplemented by idiographic methods, was part of an overall zeitgeist in the field. If the arguments reviewed here have merit, the importance of identifying functional relations has been reinforced and can be drawn upon with a renewed sense of determination to revisit some of the gaps in the knowledge base, most obviously with regard to case formulation.

What is the introductory psychology textbook narrative taking shape for describing how the second wave is passing into the third? As was the case with the notion of the cognitive revolution, cited thousands of times, there is tacit agreement on a script that can be routinely resorted to for manuscript introductions. It portrays second wave therapies as tied up in logical disputation, whereas, in contrast, the third wave is inherently metacognitive and offers mindful respite to clients weary of being harangued by endless debate. Those identified with the second wave might, with considerable justification (e.g., Bernstein et al., 2015) point out that distancing and decentering as alternatives to disputation have always been a part of cognitive therapy and that metacognition is a logical extension to garden variety cognitive therapy. Teasdale (1999, p. 146) made a valid distinction between metacognitive knowledge (knowing that thoughts are not necessarily always accurate) and the metacognitive insight gained through mindfulness (experiencing thoughts as events in the field of awareness, rather than as direct readouts on reality). However, this would be a rather subtle distinction on which to base a claim that a qualitative shift has taken place.

Each wave has been called upon to substantiate this sort of theoretical claim or be susceptible to the judgment that such distinctions are merely rhetorical devices useful, at best, as therapeutic heuristics. We have seen that self-report is a main focus of contention. Assessment methods and psychometrics have been key elements of the relevant academic debates, and the track record of third wave approaches in this regard has been problematic (e.g., Doorley et al., 2020). In light of this and of the fact that the third wave has arguably been with us for the better part of two decades during which justification for the proliferation of concepts could have been justified with evidence, it is not too late to take account of why perfectly reasonable and well understood scales of metacognition, such as the Thought Action Fusion Scale (Shafran et al., 1996), Metacognitions Questionnaire (Wells & Cartwright-Hatton, 2004), and Thought Control Questionnaire (Wells & Davies, 1994), should not be a starting point for assessment efforts and either incorporated or expressly improved upon. A similar argument can be made with regard to the second main tenet of acceptance and commitment therapy (ACT; Hayes, 2004), valued living. Firstly, it is not clear what relationship the construct of values within ACT has to the study of values within psychology more broadly (e.g., Schwartz, 2012). In addition, although Schwartz argues that values are not just reducible to beliefs, it is prudent to at least consider whether established second wave scales already tap into similar conceptual ground. In a recent reanalysis of the dimensionality of the Dysfunctional Attitude Scale, Brown, et al. (in preparation) reported factors reflecting the dimensions of high standards, worthiness, importance of being accepted, and imperatives (deontological morality), which would be difficult to argue are any less rooted in values than the newer third wave scales designed to operationalize committed action.

The progression through the waves can be viewed through different lenses, one of the most important ones being the relationship of therapy to theory. As noted, when ambitious claims are made about the scientific basis of a therapeutic approach, critics, in the absence of undeniable evidence to the contrary, will assert that the claims are merely rhetorical (O’Donohue et al., 2003) and not scientifically verified. Assessment methods are direct expressions of the mission of a therapeutic approach and so are the natural focus of potential critics. The proliferation of self report based research was initially regarded as a sign of the vibrancy and generativity of the second wave but was soon subject to critical examination from both adherents and detractors. In this regard, critics of ACT (e.g., O’Donohue et al., 2015) have noted that the failure to conclusively demonstrate efficacy through direct tests of theoretical mechanism creates a temptation to resort to a score keeping approach, in which theoretical validity is claimed on the basis of the sheer number of putatively supportive results. Third wave advocates lay claim to Skinner’s mantle, and so would do well to heed his admonitions against rapidly proliferating research:

That a theory generates research does not prove its value unless the research is valuable. Much useless experimentation results from theories, and much energy and skill are absorbed by them. Most theories are eventually overthrown, and the greater part of the associated research is discarded. (Skinner, 1950, p. 71, as quoted in Chiesa, 1992)

Summary and Conclusion

The role of subjective experience is larger than what can be accommodated by any one theoretical position. The second wave made grappling with this conundrum central to its mission. In the short run, the prevailing judgment has been that its ambitious goal has not been realized. However, the effort has generated a large body of evidence that may yet bear fruit over the longer term.