Browse the corpus

Metacognitive Therapy (MCT; ) is a transdiagnostic psychological treatment grounded in the Self-Regulatory Executive Function (S-REF) model (). According to this model, emotional disorders are maintained by a particular way of responding to internal experiences—such as thoughts, feelings, and memories—known as the Cognitive Attentional Syndrome (CAS). The CAS consists of perseverative thinking styles, particularly worry and rumination, attentional fixation on threat, and unhelpful coping behaviours that prolong distress. The CAS is driven by dysfunctional metacognitive beliefs (e.g., beliefs about the need to worry or the danger of thoughts). MCT aims to modify these beliefs and remove the CAS, thereby restoring adaptive self-regulation. Two meta-analyses suggest that MCT is an effective treatment for emotional disorders and may be superior to other psychotherapies (, ). MCT consists of several therapeutic components and techniques. A central intervention is Detached Mindfulness (DM), which represents the opposite mode of processing to the CAS. In DM, individuals notice an intrusion, adopt a decentred stance toward it, refrain from further conceptual processing, and allow the intrusion to occupy its own mental space without responding to it. Thoughts are experienced as transient mental events, that do not require evaluation, control, or action.

According to , (), DM comprises six essential elements: 1) Meta-awareness – noticing thoughts and mental activity as events in the mind; 2) Cognitive decentring – recognising that thoughts are not facts and do not define the self; 3) Flexible attention – the ability to shift attention freely rather than becoming fixed on the intrusion; 4) Low conceptual processing – suspending analysis, elaboration, and meaning-making in response to the intrusion; 5) Low goal-directed coping – refraining from attempts to control, neutralise, suppress, or otherwise cope with the intrusion; and 6) Self-awareness as observer – a state of awareness in which the person experiences a sense of self as an observer of mental events but distinct from them. As can be inferred from these six elements, DM is used in MCT not only to reduce the CAS but also to modify metacognitive beliefs (e.g., beliefs about the uncontrollability of thoughts), and to help patients adopt a metacognitive mode in which thoughts and feelings are experienced as mental phenomena rather than objective reality. It is also used to reduce the influence of cognitive content on the individual’s sense of self.

DM can be differentiated from many mindfulness-based interventions (see , for a full discussion). DM does not involve increasing awareness of sensory experiences or the present moment. It also does not involve attentional anchoring or the use of meditation or breathing exercises. These techniques can require sustained attentional focus on internal experiences which is incompatible with DM. Instead, DM focuses specifically on meta-awareness of thoughts, in a decentred manner, without further processing, in a way that is incompatible with the CAS. This also differentiates DM from some forms of acceptance-based approaches where imagery is used to transform thoughts. However, some metacognitive constructs specified in DM may overlap with mindfulness-based constructs (e.g., non-judging of inner experience), while others (e.g., observing and describing) are more unique to mindfulness-based approaches ().

, () described ten different methods to facilitate DM. These include exercises such as the tiger task, in which the individual imagines a tiger and observes what happens without intervening; the free-association task, where a list of words—first neutral and then including triggering words—is read aloud while the individual simply listens without actively generating or preventing images; and the suppression versus counter-suppression experiment, where DM is contrasted with thought suppression. The ten methods also include the use of metaphors, such as the train metaphor—in which thoughts are compared to trains arriving and leaving a station—and the cloud metaphor, where thoughts are compared to clouds that float by and cannot be influenced, making attempts to control or change them unnecessary.

An additional technique in MCT that is separate from but related to DM is the Attention Training Technique (ATT). ATT is a 15-minute auditory exercise with three components: selective attention, attention switching, and divided attention. It was designed to increase attentional flexibility and reduce perseverative conceptual activity. Although ATT—like DM—was originally developed as one component within the wider MCT protocol, several studies have investigated ATT as a stand-alone intervention, and two systematic reviews support its effectiveness (, ), reporting large effect sizes for anxiety and depression (). Furthermore, evidence suggests that improvements in attentional control may be a candidate mechanism of change in ATT ().

The findings from ATT suggest that individual techniques within MCT may be effective when delivered on their own, which has encouraged growing interest in evaluating DM as a stand-alone intervention. Examining single components of MCT is important both for understanding the therapeutic mechanisms of the full treatment and for developing brief, transdiagnostic metacognitive interventions to alleviate distress. The present paper therefore aimed to conduct, to our knowledge, the first systematic review and meta-analysis of DM delivered as a stand-alone intervention. The key question addressed in this initial study was whether DM, when given alone, produces beneficial effects and, if so, how strong these effects are.

The review was carried out based on PRISMA guidelines. Although a protocol was not pre-registered, the review was guided by decisions made in advance regarding eligibility criteria, search strategy, data extraction, the use of meta-analysis where studies were sufficiently homogeneous and using random-effects models to address the anticipated levels of heterogeneity. Other aspects of the review, such as analysing potential causes of heterogeneity, were developed during the review process.

In May 2025, a search was made of PubMed, Web of Science, and Scopus using the exact phrase “Detached Mindfulness”. Searches in Web of Science and Scopus were also restricted to studies published in English. PubMed: “Detached Mindfulness” [Title/Abstract]. No date, language or other filters were applied. Scopus: TITLE-ABS (“Detached Mindfulness”). This was limited to English-language publications but no other limitations were applied. Web of Science: TS (“Detached Mindfulness”). This was limited to English-language publications but no other limitations were applied. Additionally, the first 200 results for the same phrase in Google Scholar were examined. The search string for Google Scholar was “Detached Mindfulness”. All search results were exported to Rayyan for duplicate removal and screening. Grey literature sources (e.g., conference abstracts, posters, dissertations, and preprints) and clinical trial registries were not systematically searched as the review focused on peer-reviewed empirical studies to ensure methodological quality and allow study appraisal. The review was restricted to English-language publications to ensure that both authors could assess and extract data from all included studies in a language they fully understand.

Titles and abstracts were independently screened by both authors using the eligibility criteria, with discrepancies resolved through discussion. Full-text articles of potentially eligible studies were subsequently assessed using the same criteria, and disagreements were similarly resolved through discussion.

The following information was extracted from all articles by each author independently, where available: Study details, including year of publication, country where the study was conducted, and journal of publication. Study design, including trial or experimental type. Sample characteristics, including sample size, control group, recruitment method, population, eligibility criteria, age, and gender. Intervention details, including how DM and any control interventions were delivered, number of sessions, and session length. Outcome measures, including primary and secondary outcomes, and dropouts. Results, including within- and between-group changes on outcome measures.

The methodological quality of the included studies was assessed using an adapted checklist drawn from the National Institutes of Health’s Quality Assessment Tool for Before–After (Pre–Post) Studies With No Control Group (). The tool was modified to accommodate both experimental and intervention designs and to evaluate fidelity to Metacognitive Theory as well as reporting transparency. A customised tool was used as the review included heterogeneous study designs (i.e. randomised controlled trials, open trials, and experimental studies). Design-specific tools such as Cochrane RoB 2 (optimised for randomised controlled trials; ), ROBINS-I (developed for non-randomised studies of interventions; ), and JBI critical appraisal tools (tailored to specific study designs; ) were therefore not used, as their design and scoring frameworks are less suitable for synthesising evidence across diverse empirical approaches. The adapted checklist comprised ten domains assessing: 1) Clarity of research objectives, 2) Adequacy of population description, 3) Sample size justification, 4) Clarity of intervention description and replicability, 5) Theoretical fidelity of the DM intervention, 6) Outcome measures, 7) Timing of measurements, 8) Data completeness, 9) Appropriateness and adequacy of statistical methods, and 10) Reporting transparency.

Each domain was rated on a 3-point scale, scoring 1 if the criterion was met, 0.5 if it was partially met or unclear, and 0 if it was not met. Total scores reflected overall methodological quality. The two reviewers independently rated all studies, and discrepancies were resolved through discussion.

Form of synthesis was decided based on homogeneity of design, population, and outcome measures. The clinical trials were judged to have sufficient homogeneity to warrant quantitative synthesis using meta-analysis, whereas for the experimental studies a narrative synthesis was used as designs and outcomes were heterogeneous.

For the clinical trials, meta-analyses were conducted on both primary and secondary outcomes. All analyses were performed using R (version 4.5.2) with the metafor package (version 4.8.0; ). Standardized mean change (Hedges’ g ) values were calculated for pre–post differences within each study, using an assumed pre–post correlation of r = 0.5. Random-effects models with restricted maximum likelihood (REML) estimation were applied to account for between-study variability. Heterogeneity was evaluated using Q , I ², and τ² statistics. Owing to the small number of studies, formal statistical analyses of heterogeneity (e.g. subgroup analysis) were not used, instead study differences were examined descriptively. Sensitivity analyses were performed across alternative assumed pre–post correlations ( r = 0.3, 0.5, 0.7) to assess robustness. Forest plots were used to show individual and pooled effect sizes visually. As fewer than ten studies were included in each analysis, funnel plot inspection and formal tests for publication bias (e.g., Egger’s regression) were not carried out, consistent with Cochrane recommendations that such methods are unreliable in small meta-analyses.

In studies not included in the meta-analyses, effect sizes were extracted when reported, otherwise, where possible, standardised effect sizes ( g , r or phi) were calculated from significant outcomes using R with the metafor or esc packages () to provide a standardised measure of effect size across studies. Effect sizes were interpreted using conventional guidelines: for g , small ( g = 0.20), medium ( g = 0.50), and large ( g ≥ 0.80) for r and phi, 0.10 small, 0.30 medium and ≥ 0.50 large.

For the clinical meta-analyses, assessment of reporting bias using funnel plots or statistical tests was not conducted due to the small number of studies, consistent with Cochrane recommendations. For the experimental studies, which were synthesised narratively, statistical assessment of reporting bias was not applicable. Although comprehensive searches were conducted, the inclusion of published studies only, means that reporting bias cannot be ruled out.

A PRISMA diagram showing the search results and selection process is shown in . As shown, the searches identified 152 articles, of which 82 duplicates were removed using Rayyan. The remaining 70 records were screened, and 39 were excluded after examination of the title and/or abstract, as they clearly did not meet the eligibility criteria for inclusion. The remaining 31 articles were screened using the full text. Of these, 17 were excluded for the following reasons: 1) The article was not written in English; 2) DM was combined with other interventions (e.g., challenging metacognitive beliefs or medication); 3) The study employed mindfulness rather than DM; 4) The results were not interpretable, as the authors appeared to have combined their experimental and control groups; 5) The article was not published in a peer-reviewed journal. presents the individual studies excluded at the full-text stage along with reason of exclusion. A total of 14 articles met the inclusion criteria. Summary descriptive information of included studies is shown in .

Fourteen articles were included in the review, corresponding to twelve distinct participant samples, as three articles (–) presented different outcomes derived from the same underlying study population. Across these twelve samples, the number of participants receiving DM (excluding control conditions) ranged from 8 to 50, with a total of 256 participants. All participants were aged 17 or over. Of the twelve samples, three were clinical trials (two open trials and one randomised controlled trial), two of which treated patients with obsessive–compulsive disorder (OCD) (; n = 17; ; n = 20 in the DM condition), and one treated patients with panic disorder (; n = 11). Of the remaining nine samples, seven experimentally induced symptoms or triggers to examine the immediate effects of DM on specific outcome measures. One induction study involved a clinical sample of patients with alcohol use disorder (; n = 8), while the remaining six used non-clinical samples, with DM sample sizes ranging from 12 to 30 participants. The final two samples examined the effects of DM on pre-existing symptoms rather than experimentally induced states, specifically self-critical rumination (; n = 23) and negative repetitive thinking (; n = 50 in the DM condition), in non-clinical populations.

Scores on the quality checklist across the 14 included studies ranged from 6.0 to 9.0 on a 0–10 scale, with higher scores indicating better methodological quality. Most studies demonstrated clear research objectives, adequate descriptions of intervention procedures, good theoretical fidelity to DM principles, and appropriately timed pre- and post-intervention assessments. All studies provided basic descriptions of their participant samples; however, clinical characteristics were often absent or only partially reported, particularly in experimental studies. Although outcome measures generally had good face validity, several studies included measures that had not been validated. With one exception, studies employed appropriate statistical analyses; however, reporting of key statistical assumptions (e.g., normality, presence of outliers, or handling of non-normal data) was frequently absent. Most studies did not report a sample size justification or power analysis. In addition, incomplete reporting of missing data and attrition was common, resulting in reduced scores for data completeness. Although outcomes were generally reported transparently, two studies were only given a. 5 score on this item for not providing pre- and post-intervention descriptive statistics that would have allowed calculation of standardised effect sizes. A summary of the quality assessments is presented in .

All studies grounded their procedures in the metacognitive theory of DM or in techniques specifically outlined by Wells , () for inducing DM. However, the format and length of the DM intervention varied across studies. In the three clinical trials, DM was delivered across four sessions and embedded within a metacognitive understanding of the particular disorder. These trials began with psychoeducation framed from this perspective, followed by several structured DM techniques, guided practice applied to disorder-specific intrusive thoughts, and homework assignments to consolidate DM skills. The other studies used briefer DM interventions. The seven studies that induced symptoms used single-session DM interventions, whereas the two studies examining pre-existing symptoms provided an initial DM instruction or session followed by several days of practice—either for one week whenever they experienced a trigger () or for five days, three times per day ().

Across studies, DM was induced in a range of different ways. One study () p 542 used only a single spoken instruction (“acknowledge the presence of it in your mind and let it occupy its own mental space and possess its own behaviour. Do not interfere with the intrusion or try to control it.”). In contrast, most studies employed at least one metaphor or a technique outlined by Wells (). Three techniques specified in more than one study were the Tiger Task, the Word Association Task, and the Suppression/Counter-suppression exercise.

Chittaro and Vianello () used a technique for inducing DM described by Wells (), in which individuals imagine their thoughts as written on clouds, watching them move by without engagement. As discussed earlier, Wells , () notes that this version still involves a degree of active engagement with thoughts so does not represent pure DM. Although, as explained earlier, use of this technique was included in the review, this intervention was marked in the Quality Checklist as having only partial fidelity to how DM is construed in Metacognitive Theory and the results were examined accordingly. Wells , () explained that cloud imagery can be used as a metaphor for pure DM. This was the approach used by Hansmeier et al. (), who asked participants to imagine thoughts as clouds passing across the sky, part of the atmosphere’s natural self-regulation and not something they could influence. This intervention was therefore rated as having full fidelity to DM as conceptualised in Metacogntive Theory.

Bolzenkötter et al. () instructed participants to use imagery metaphors with thoughts represented as clouds in the sky, leaves on a river, or trains at a station. Participants were told to watch the ebb and flow of their thoughts without engaging with them. Bolzenkotter et al. reported that the audio transcript used in their study was adapted from a mindfulness exercise (“leaves on a river”; ) and that the same script was also applied to two DM metaphors—clouds in the sky and trains at a station. Although they stated that key characteristics of DM were incorporated in the exercise, the article did not provide sufficient detail as to how the original mindfulness exercise was adapted. We contacted the author to request clarification, but no response was received at the time this article was submitted. Accordingly, the degree of adherence to DM principles remains uncertain, particularly as the original exercise came from a different therapeutic modality. The study was nonetheless included because the article stated that the procedure was faithful to key aspects of DM, but this uncertainty in DM fidelity is taken into account when interpreting its results.

In most studies the DM intervention was delivered verbally-in person or by audio recording, one used an animated video, and one provided written instructions only. Some studies required participants to practice the technique after instruction, whereas others presented the instruction or metaphor without guided practice.

Only four studies assessed either the difficulty of DM or how well participants were able to implement it. Akdağ et al. () delivered DM digitally using two videos (instructional and practice) and stated that they drew on four DM techniques described by Wells. They reported that participants rated DM as significantly more difficult than the two comparison techniques (ATT and slow breathing). Bolzenkötter et al. () asked participants to practice DM over several days using audio-based exercises and to rate how well they were able to implement the technique; average ratings indicated moderate-to-good implementation ability. Rupp et al. (), using Ecological Momentary Assessment (EMA), assessed the perceived difficulty of applying DM in daily life. On a 0–6 Likert scale (0 = not at all; 6 = very much), the mean difficulty rating for DM was 2.68, indicating low-to-moderate difficulty; the corresponding rating for cognitive restructuring did not differ significantly. Finally, Chittaro and Vianello () compared a cloud-imagery task (see earlier discussion for limitations of this method) with two non-DM techniques and found that the cloud task was rated as more difficult than both comparison conditions.