Contribution
The paper sets out ACT’s model of psychopathology and of change in one place, with its philosophical root (functional contextualism) and its basic-science root (Relational Frame Theory). It is the first review, by its own account (p. 18), to ask whether two decades spent building that basic science have produced “a progressive model of psychopathology and its treatment”. It answers with a review of four kinds of evidence: correlational, experimental component, mediational, and outcome.
Key insight
Behaviour therapy promised both validated treatments and treatments built on basic principles. Cognitive behaviour therapy kept only the first, because its cognitive concepts named things to change without naming the contexts a clinician can manipulate. ACT tries to keep both. Its basic theory says thoughts and feelings do not cause action except “as regulated by context” (p. 5). So the clinical target is the context that links them to action, not their content. Psychological problems are the narrowing that follows when verbal rules and the avoidance of private experience govern behaviour more than its actual consequences do.
Assumptions
- Functional contextualism (p. 5). The goal is “the prediction and influence of events, with precision, scope and depth”. Truth is workability, and causal analysis is limited to manipulable events.
- RFT (pp. 5–6). Human language and cognition is “the learned ability to arbitrarily relate events, mutually and in combination, and to change the functions of events based on these relations”. Four implications follow for psychopathology (p. 6): (1) the normal verbal processes of problem solving underlie it and “cannot be eliminated”; (2) context controls the content and impact of cognitive networks; (3) networks are historical, so logical restriction or elimination “is generally not psychologically sound”; (4) direct change attempts on a node “tend to elaborate the network in that area and increase its functional importance”.
- Self-report measures of process. The AAQ is the main process measure; the authors note it is not a pure measure of experiential avoidance but “a more general measure of several ACT processes” (p. 9).
Key results
- The inflexibility model (pp. 6–8). Cognitive fusion is “excessive or improper regulation of behavior by verbal processes”. It is fostered by contexts of literality, reason-giving and emotional control. It supports experiential avoidance, defined with the 1996 paper (LIT-551) as “the attempt to alter the form, frequency, or situational sensitivity of private events even when doing so causes behavioral harm”. Avoidance is “due to the natural effects of human language”, then amplified by a culture of “feeling good”. It makes avoided events more salient, so it narrows the repertoire. The conceptualized past, future and self gain control, and values “take a backseat”.
- The six processes (pp. 7–9). Acceptance (the alternative to experiential avoidance); cognitive defusion (changing a thought’s functions, not its form, e.g. “I am having the thought that I am no good”); being present; self as context (from deictic frames I–you, here–there, now–then); values (“chosen qualities of purposive action that can never be obtained as an object”); committed action. The first four “provide a workable behavioral definition of mindfulness” (p. 9).
- Correlational evidence (pp. 9–11), in the manuscript’s numbers. 21 studies, 4,721 participants, 52 correlations between the AAQ and outcomes. The overall effect is printed “.030 (95% confidence interval: 0.29 – 0.32)”, which cannot be right as printed (see Corrections). Sub-analyses: GHQ r = −.40, BDI −.50, SCL-90-R −.53, anxiety measures −.49. Bond and Bunce’s two-wave panel: the AAQ predicted later GHQ and job performance, and not the reverse.
- Component studies (pp. 11–13). Defusion by word repetition lowered the discomfort and believability of negative self-referential thoughts more than distraction or thought control (N = 8, alternating treatments). Acceptance and defusion raised cold-pressor tolerance, though not reported pain (N = 32). In CO₂-inhalation studies, acceptance instructions reduced avoidance and raised willingness to return (Eifert & Heffner, N = 60; Levitt et al., N = 60, panic disorder).
- Mediation (pp. 13–15). Process measures met Baron–Kenny or MacKinnon criteria in worksite stress (AAQ, N = 90), diabetes self-management (AADQ, N = 81), smoking (AIS, N = 67; one-year abstinence 35% vs 15% for nicotine replacement), counsellor stigma and burnout (SAB, N = 93), and psychosis (believability, N = 29). A new reanalysis of Zettle and Hayes (1986, N = 18): mid-treatment believability of depressive thoughts (ATQ-B) mediated ACT’s advantage over cognitive therapy on the BDI and the Hamilton.
- Outcomes (pp. 17–18, Table 2). Weighted Cohen’s d: within ACT 1.11 at post (N = 253) and 1.26 at follow-up (N = 210). Against manualized active treatments, .49 at post (N = 322) and .69 at follow-up (N = 218). Against wait list, TAU, education or placebo, .97 and 1.58. Against CT/CBT alone, .73 and .83, on N = 96 and 39.
Claims
| id | claim | strength | support |
|---|---|---|---|
| C1 | Psychopathology is, primarily, psychological inflexibility produced by fusion and experiential avoidance | moderate as model, weak as causal claim | theory (RFT); correlational and mediational evidence summarized |
| C2 | The AAQ (inflexibility) correlates moderately to strongly with distress and disorder | moderate | meta-analysis of 21 studies in this manuscript; self-report on both sides |
| C3 | Flexibility predicts later mental health and job performance, not the reverse | weak–moderate | one two-wave panel (Bond & Bunce 2003) |
| C4 | Acceptance and defusion components are psychologically active | moderate | small randomized analogue studies (N = 8–60) |
| C5 | ACT’s effects are mediated by the processes it targets | weak–moderate | seven studies, mostly self-report, mediators often measured after outcomes improved (the authors’ caveat) |
| C6 | ACT works through different processes from CT/CBT | weak | four small comparisons, N = 96 at post; allegiance bias acknowledged |
| C7 | ACT is more effective than other active treatments | not claimed | the authors say there are “not enough well-controlled studies” |
Concepts
- psychological flexibility: “the ability to contact the present moment more fully as a conscious human being, and to change or persist in behavior when doing so serves valued ends” (p. 7).
- psychological inflexibility: “an inability to persist or change in the service of long term valued ends” (p. 6).
- cognitive fusion: “excessive or improper regulation of behavior by verbal processes, such as rules and derived relational networks” (p. 6).
- experiential avoidance: as in LIT-551, quoted above.
- self as context: a sense of self as “a locus or perspective”, arising from deictic relations, which “is a context for verbal knowing, not the content of that knowing” (p. 8).
- values: chosen qualities of action, instantiated but never achieved; goals, by contrast, can be achieved (pp. 8–9).
- third wave: behavioural and cognitive therapies “particularly sensitive to the context and functions of psychological phenomena, not just their form” (p. 4, quoting Hayes 2004).
Connections
- Experiential avoidance (LIT-551), unread. Its definition is
quoted here, and its construct becomes one process of six. That is the
extendsrelation on the LIT. - RFT (LIT-549 unread; LIT-541 read). The basic theory ACT is “the applied extension” of (p. 5). The read 2004 paper states RFT more fully and gives RFT readings of three of the six processes.
- Kashdan and Rottenberg (LIT-553), read. They widen “psychological flexibility” beyond ACT and cite this paper’s meta-analysis, giving it as 32 studies and r = .42, which is the published version, not this manuscript.
- Macri and Rogge (LIT-543), unread. A 2024 meta-analysis of the six processes as mechanisms. It is the test of C5 with eighteen more years of trials.
- Salkovskis, Sighvatsson and Sigurdsson (LIT-538), filed by another agent, argue that a treatment’s success is not evidence for its theory, except when changing the theory’s crucial factor fails to help. This review’s mediational evidence is the kind of test that argument calls for.
- Self-determination theory (LIT-558, LIT-559). Deci and Ryan’s “rigid behavior patterns” that keep people “from dealing with their inner experiences” and “persist into new situations in which they are not needed” (LIT-558, p. 250) describe experiential avoidance and inflexibility, but SDT traces them to thwarted psychological needs, not to fusion. ACT’s values, “chosen qualities of purposive action”, correspond to SDT’s identified and integrated regulation (LIT-559, pp. 72–73). This reading records no ACT distinction between values held from guilt or approval and values held volitionally, which is SDT’s line between introjection and identification. The two transdiagnostic claims, THEORY-050 and THEORY-052, are compatible.
Bearing on the record
- A candidate THEORY: “psychological inflexibility, the dominance of verbal rules and experiential avoidance over direct contingencies, is a transdiagnostic process in psychopathology, and treatments that reduce it reduce distress”. Sources: this paper, LIT-553, and LIT-543 once read. Status Proposed until the meta-analysis is read, since the evidence here is early and mostly self-report. See the batch report.
- Against the network model (LIT-547). This model puts one functional process behind many syndromes. Borsboom’s network theory denies most disorders a common cause. The two make different predictions about whether reducing one process moves all symptoms. Neither paper addresses the other.
- No ML instruction. Nothing here belongs in the anthology.
- The model is filed as THEORY-050, Proposed. Self-as-context is THEORY-044 (the account) and THEORY-051 (the treatment claim).
Limitations
- The manuscript, not the article. See Corrections.
- Allegiance. All five authors are ACT developers or researchers, and most cited trials are by ACT researchers; the authors note “bias is possible” (p. 18).
- Self-report processes. The authors list the weaknesses themselves (p. 17): process measures without published psychometrics, measured after outcomes began to improve, and high in face validity.
- Small trials. Most outcome studies are small and short, which the authors put down to the programme’s stage (p. 18).
- The model is not tested as a model. The six processes are reviewed one study at a time; no study tests the whole structure in Figure 1.
Open questions
- Do the six processes separate empirically, or is there one inflexibility factor? The AAQ’s single general factor (p. 9) leaves this open.
- Does inflexibility precede disorder, or follow it? One two-wave panel is the only evidence here.
- Does reducing fusion or avoidance in the laboratory produce the clinical changes? The authors call this the translational test the programme still needs (pp. 18–19).
Corrections
- This is an early manuscript. It has unfilled citations: “trained perspective taking (xx)” and “Barnes-Holmes. Xxx 2003” (p. 6), “Sobel’s method (19xx)” (p. 15), “Brown xxx” and “Baer xxx” (p. 11), “Herbert xxx 2004” (p. 18). Its correlational meta-analysis has 21 studies and 4,721 participants. Kashdan and Rottenberg, citing the published article, give “a meta-analysis of 32 studies” with a mean correlation of .42. The published numbers were not seen.
- “.030 (95% confidence interval: 0.29 – 0.32)” (p. 10) is impossible as printed: the point estimate lies outside its own interval. The text around it calls the relation “moderate”, so “.30” is the likely intended value. It is the manuscript’s typo.
- A copied statistic. In the Zettle and Hayes reanalysis (p. 15), the mid-treatment comparisons for the BDI and the ATQ-B are both printed “F (1, 16) = 2.81, p = .12”. The text says the first did not differ and the second did, so the ATQ-B statistic was evidently copied from the BDI line.
- Repository metadata. The GSU record lists the first author as “Hayes, Stephen C.”; the manuscript and Crossref have Steven C. Hayes.