Acceptance as a possible link between past psychedelic experiences and psychological flexibility

Participants were recruited via the online platform Prolific. From 2500 completers of a pre-screen, 840 met inclusion criteria (≥18 years and at least one meaningful experience with a classical psychedelic), 701 completed the main survey, and after exclusions for inconsistent reports or use of other drugs the final sample comprised N = 629. Data collection followed GDPR and Helsinki principles and received institutional ethics approval. The survey collected demographics, detailed retrospective measures of a personally meaningful psychedelic experience, lifetime use frequency and dose estimates, and validated self-report scales of psychological flexibility/inflexibility and well-/ill-being. Acute features of the chosen experience were assessed with the MEQ30 (mystical-type phenomena) and the PIQ (psychological insights, with Avoidance and Maladaptive Patterns (AMP) and Goals and Adaptive Patterns (GAP) subscales). Frequency measures included cumulative lifetime frequency, time since last use (months), average frequency (0–8 scale from never to daily) and average dose (omitted from some models because 135 participants responded "I don't know"). Psychological flexibility and inflexibility were measured with the MPFI-24, yielding 12 subscale mean scores. Well-being was assessed with the WEMWBS and Peace of Mind Scale (PoMS); ill-being with the PHQ-9 (depression) and GAD-7 (anxiety). Two network models were estimated: an 18-node Acute Model (MEQ30, PIQ, the 12 MPFI subscales, WEMWBS, PoMS, PHQ-9, GAD-7) and a 19-node Frequency Model (replacing acute measures with frequency variables). Variables deviated from multivariate normality, so nonparanormal transformations were applied. Networks were estimated as Gaussian Graphical Models (partial correlation networks) using the mgm package with graphical LASSO regularisation and EBIC model selection (gamma = 0.5). Predictability (R2), centrality (strength), and stability were assessed. Edge-weight CIs were obtained via non-parametric bootstrap (1000 samples) and centrality stability via case-dropping bootstraps (2500 samples). Exploratory graph analysis (EGA) with the Walktrap algorithm and bootEGA evaluated community structure and cluster stability. Mediation analyses departed from the preregistration by using an aggregated PF score (total MPFI mean), for comparability with prior literature and clearer interpretation. Mediation was conducted in JASP using standardised estimates and 1000 bootstrap replications to obtain bias-corrected 95% CIs. The Acute mediation model used MEQ and PIQ means as predictors, aggregate PF as mediator, and PoMS, WEMWBS, GAD-7 and PHQ-9 as outcomes, with age and gender as covariates (analyses limited to males and females due to small "other" gender n = 5). The Frequency mediation model substituted frequency variables for the acute predictors. Significance was set at 0.05.

Sample and descriptive information: After exclusions the analytic sample comprised N = 629. All variables had no missing data. Descriptive tables of means, internal consistencies and bivariate correlations were reported in the extracted text (tables not reproduced here). Network structure and key edges (Acute Model): Psychological flexibility components clustered together, as did most psychological inflexibility components, although Experiential Avoidance did not show expected associations with other inflexibility nodes. Psychological insight and mystical-type features were strongly associated with each other (partial association 0.47). Psychological insight was the only experience-related node that showed a clear positive partial association with a PF component, specifically Acceptance (edge weight 0.11). Well-being and ill-being nodes had strong intra-cluster associations: peace of mind with mental well-being (0.51), and depression with anxiety (0.55). Mental well-being inversely associated with depression (-0.14), and peace of mind inversely associated with anxiety (-0.18). Mental well-being related positively to Committed Action (0.09) and Values (0.07); peace of mind related to Defusion (0.06) and inversely to Fusion (-0.12). Anxiety related to Fusion (0.08); depression related to Lack of Awareness (0.11) and Inaction (0.09). Frequency Model network: Psychological flexibility and inflexibility again formed cohesive clusters with the same exception for Experiential Avoidance. Cumulative frequency and time since last use were inversely correlated (-0.20); average frequency did not associate with other frequency measures and showed no direct associations with PF/PI components after regularisation. Well-being/ill-being interrelations mirrored the Acute Model (peace of mind—mental well-being 0.51; depression—anxiety 0.57). Associations between PF components and well-being were similar to those in the Acute Model. Centrality and stability: Strength centrality identified Self-as-Context, Cognitive Fusion, Values, Inaction, Committed Action and Acceptance among the most central PF/PI nodes in the Acute Model, with raw strength values listed (Self-as-Context 0.93; Cognitive Fusion 0.85; Values 0.85; Inaction 0.80; Committed Action 0.75; Acceptance 0.68). Depression, peace of mind, anxiety and mental well-being were the most central well-being/ill-being nodes. Psychological insight (0.58) was more central than mystical experience (0.47) among experience-related nodes. The Frequency Model showed similar centrality ordering. Node-strength stability was robust, with a correlation stability coefficient (CS) of 0.75 for both models. Bootstrapped edge-weight CIs were relatively narrow, indicating stable edge estimates. Exploratory graph analysis (EGA): The Acute Model EGA yielded four communities: (1) psychological flexibility, (2) features of the past psychedelic experience (mystical-type, psychological insight), (3) well-being (mental well-being, peace of mind), and (4) ill-being (depression, anxiety) combined with psychological inflexibility. The Frequency Model produced four clusters as well: (1) psychological flexibility and average use, (2) cumulative frequency and time since last use, (3) well-being, and (4) ill-being with psychological inflexibility. Bootstrapped EGA solutions matched the original four-cluster solution in approximately 89% (Acute) and 85% (Frequency) of iterations. Some nodes showed unstable cluster assignment—ill-being nodes, Experiential Avoidance, and average use were assigned to their clusters only 48%–70% of the time—suggesting cautious interpretation of certain cluster findings. Mediation analyses (Acute Model): Aggregate psychological flexibility mediated associations between psychological insight and outcomes. No significant direct or indirect effects were found for mystical-type experiences. Psychological insight showed significant direct positive associations with depression (standardised E = 0.26, SE = 0.05, p < .001) and anxiety (E = 0.27, SE = 0.05, p < .001). Significant indirect effects of psychological insight via PF were observed for depression (E = -0.09, SE = 0.02, p < .001), anxiety (E = -0.08, SE = 0.02, p < .001), mental well-being (E = 0.14, SE = 0.03, p < .001) and peace of mind (E = 0.12, SE = 0.02, p < .001). Age showed a very small positive association with PF. Post hoc PIQ-subscale mediation: When separating PIQ into Goals and Adaptive Patterns (GAP) and Avoidance and Maladaptive Patterns (AMP), GAP predicted higher mental well-being and peace of mind (E = 0.23 and 0.24 respectively) and lower anxiety (E = -0.16), with significant indirect effects through PF for all outcomes (well-being and ill-being). AMP predicted lower mental well-being and peace of mind (E = -0.23 and -0.30) and higher depression and anxiety (E = 0.27 and 0.40); no significant indirect effects via PF were found for AMP. Mediation analyses (Frequency Model): No significant direct effects of cumulative frequency, time since last use, or average frequency on outcomes were detected. Average frequency showed small but significant indirect effects mediated by PF for depression (E = -0.04, SE = 0.01, p < .001), anxiety (E = -0.03, SE = 0.01, p < .001), mental well-being (E = 0.07, SE = 0.02, p < .001), and peace of mind (E = 0.06, SE = 0.02, p < .001). Age again showed small associations with PF and cumulative frequency.

Krabbe and colleagues interpret their findings as indicating that the quality of a past psychedelic experience—specifically psychological insight—relates to a particular component of psychological flexibility, Acceptance, when controlling for other components. The absence of an association between mystical-type features and PF in the multivariate network contrasts with some earlier studies, and the authors note that Acceptance aligns with therapy principles emphasising ‘‘letting go’’ and with Acceptance and Commitment Therapy approaches used in psychedelic-assisted psychotherapy. The cross-sectional design means the direction of effect cannot be determined; high trait acceptance could equally predispose individuals to report more insights or to derive more benefit from psychedelic experiences. Self-as-Context emerged as a highly central node within the PF cluster, consistent with theoretical work that positions perspective-taking self processes as foundational for other PF components. The authors discuss plausible pathways in which Acceptance could bridge psychological insight to Values and Committed Action, thereby supporting value-congruent behaviour and improved well-being—an interpretation compatible with the Accept, Connect, Embody (ACE) model proposed in psilocybin therapy literature. However, no direct links were found between frequency of past psychedelic use and PF after regularisation; average frequency related weakly to some PF components at the bivariate level but not in the controlled network. The community detection and mediation results suggest that PF mediates relationships between certain experience features or use patterns and well-being/ill-being. Notably, higher psychological insight was unexpectedly associated with greater depression and anxiety in direct effects; post hoc analyses attributed this to the PIQ subscale indexing Avoidance and Maladaptive Patterns (AMP), whereas insights into Goals and Adaptive Patterns (GAP) were positively associated with well-being and negatively with ill-being. The authors propose that mere awareness of maladaptive patterns without acceptance or flexible responding may worsen distress, echoing mindfulness literature where increased awareness of difficult feelings can sometimes exacerbate problems. Limitations acknowledged by the investigators include the cross-sectional and retrospective design (recall and availability biases, peak-end effects), potential confounding by trait-level characteristics (for example trait acceptance), absence of attention checks in the online data collection, and issues of power and node selection inherent to network analyses. The EGA indicated some instability in cluster assignments for ill-being and Experiential Avoidance, so those specific findings warrant caution. The authors recommend future longitudinal and experimental studies, use of multidimensional PF measures rather than unidimensional scales like the AAQ-II, and replication with larger samples to clarify causal mechanisms and the roles of individual PF components in psychedelic therapy.

The study concludes that psychological flexibility, considered dimensionally, may underlie long-term mental health benefits associated with psychedelic experiences, but that it is component-specific relationships—particularly between psychological insight and Acceptance—rather than overall frequency of use that appear most relevant. Psychological flexibility mediated associations between psychological insight (and, weakly, average frequency) and both well-being and ill-being. The results therefore support treating PF as a multidimensional construct in future psychedelic research and clinical work to better identify which specific components are implicated in therapeutic outcomes.