Subtypes of the psychedelic experience have reproducible and predictable effects on depression and anxiety symptoms

Depression and anxiety are common, disabling disorders with limited response to existing treatments, prompting renewed interest in alternative approaches. Earlier clinical trials and naturalistic surveys have reported substantial and sometimes durable reductions in depression and anxiety following psychedelic-assisted experiences, with prior work implicating facets of the subjective experience—particularly mystical-type experiences and psychological insight—in mediating benefit. However, most prior analyses have used a top-down approach testing a small number of hypothesised relationships (for example mystical-type effects versus outcome), leaving unexplored the possibility that there are reproducible subtypes of psychedelic experiences that map differently onto mental health outcomes. Nikolaidis and colleagues set out to identify data-driven subtypes of the acute psychedelic experience and to test whether these subtypes predict changes in depression, anxiety and related measures. Specifically, the study asked two main questions: whether distinct subtypes (spanning mystical, insight and challenging experiences) could be derived and whether those subtypes replicate across LSD and psilocybin subsamples and show consistent relationships to symptom change. To address these aims the authors re-analysed data from a large cross-sectional online survey using hierarchical clustering applied to item-level measures of subjective experience.

The analysis used data from an anonymous internet-based cross-sectional survey collected June 2018–January 2019. Recruitment began with 14,911 clicks on study advertisements; 3,080 respondents consented and completed the survey. After excluding participants who reported using more than one psychedelic for the target experience (n=960) and restricting to respondents who indicated a change in anxiety or depression attributable to a psychedelic experience and who had complete data, the final analytic sample comprised n = 985. No prospective follow-up data were collected. The extracted text indicates the investigators controlled for potential bias by adjusting for race, age and socioeconomic status where relevant, and that detailed recruitment and data-cleaning procedures are described in the original report referenced by the authors. Subjective acute experience was assessed with multiple validated instruments: the Mystical Experiences Questionnaire (MEQ, 30 items; total scale alpha = 0.94), the Psychological Insight Questionnaire (PIQ, 23 items; alpha = 0.93), the Challenging Experiences Questionnaire (CEQ, 26 items; alpha = 0.95), and a Salient Emotions checklist (15 emotions, yes/no). Enduring effects were measured with a 14-item instrument rating meaningfulness and positive/negative change. Mental health measures included the Depression and Anxiety subscales of the DASS-21 administered in a retrospective pre/post format (three months before and after the experience), the Satisfaction With Life Scale (SWLS) before/after, and the Acceptance and Action Questionnaire-II (AAQ-II) for psychological flexibility before/after. Internal consistency reliability statistics for before and after assessments are reported in the text. Analytically the team first ran confirmatory factor analyses using Maximum Likelihood Estimation for MEQ, PIQ and CEQ and reported fit indices (e.g. TLI and RMSEA) and Cronbach's alphas for subscales. To derive subtypes they applied hierarchical agglomerative clustering (HAC) to item-level responses across MEQ, PIQ, CEQ and Salient Emotions, using Ward's linkage as implemented in R's hclust. Cluster quality was evaluated with the Silhouette index, balancing within-cluster homogeneity and between-cluster separation; the authors selected a three-cluster solution as offering a useful compromise between cluster quality and interpretability. They examined variable-wise clusters in parallel to participant clusters and visualised correlations among items. Differences between subtypes in demographics, baseline characteristics and outcomes were tested using ANOVA, ANCOVA and chi-square tests, with outcomes adjusted for demographic or participant characteristics that differed between subtypes in each sample. The clustering procedure and subsequent hypothesis tests were repeated separately within LSD and psilocybin subsamples to test replicability across drugs.

Factor analyses indicated coherent within-domain structures for the PIQ, MEQ and CEQ with good-to-excellent internal consistency (subscale alphas and fit indices reported; for example MEQ subscales Cronbach's alpha up to 0.94). Correlation matrices showed strong within-domain correlations and relatively weaker between-domain correlations, and the variable-wise HAC solution grouped items largely by questionnaire domain (PIQ, MEQ, CEQ), with Salient Emotions distributed across these clusters as appropriate. HAC applied to participant-level item vectors produced a three-cluster solution the authors labelled descriptively as: Subtype 1 "High Scoring" (high PIQ and MEQ, moderate CEQ), Subtype 2 "Low Scoring" (moderately low PIQ and MEQ, low CEQ), and Subtype 3 "Positive Scoring" (high PIQ and MEQ, low CEQ). The three participant subtypes showed marked differences across the three variable clusters. Demographically, subtype differences were observed though mean differences were small relative to within-group variability. The High Scoring subtype had a slightly younger mean age (30.2 years) versus about 33.5 years for the other groups, and had more recent experiences on average. Drug-type differences were apparent: psilocybin was relatively more common in the Low Scoring and Positive Scoring subtypes, whereas DMT-type use (ayahuasca, N,N-DMT, 5-MeO-DMT) was most common in the High Scoring subtype. Dose-level reporting also differed: the High Scoring subtype reported a higher proportion of "High" dose experiences and fewer "Low" doses. The Positive Scoring subtype contained a higher proportion of female respondents compared with the others. Regarding mental health and other outcomes, baseline measures differed by subtype: the Low Scoring subtype had lower baseline psychological flexibility and lower anxiety and depression scores than the High Scoring subtype. In changes from before to after the experience, the Positive Scoring subtype showed the largest improvements in psychological flexibility and life satisfaction, while the Low Scoring subtype exhibited smaller improvements in psychological flexibility, satisfaction with life, anxiety and depression compared with the High Scoring and Positive Scoring subtypes. On enduring effects, the Positive Scoring subtype scored significantly higher than other subtypes on 8 of 10 reported items. Comparisons between High and Low Scoring subtypes implicated overall intensity of subjective experience (PIQ and MEQ) as positively associated with beneficial outcomes; differences between High and Positive Scoring subtypes were driven mainly by CEQ (challenging) scores, with minimal differences in reported mental health improvement between these two groups. The clustering solution and subtype–outcome relationships replicated when analyses were restricted to LSD and psilocybin subsamples. Each drug-specific HAC yielded the same three-subtype pattern. Notably, the proportion of participants falling into the Positive Scoring subtype differed by drug: Full sample 27.5%, LSD 61.4%, psilocybin 41.6%. In the LSD subsample subtype membership was less associated with age, time since experience or sex but still differed by dose level and duration; in the psilocybin subsample patterns resembled the full sample with subtype differences by age, time since experience and duration but not by dose. In both drug-specific samples, the Low Scoring subtype had the lowest baseline psychological flexibility and the smallest pre/post improvements, while the Positive and High Scoring subtypes showed larger improvements and higher enduring effect scores.

Nikolaidis and colleagues interpret their findings as evidence that acute subjective psychedelic experiences fall into reproducible subtypes that are meaningfully related to baseline characteristics and to subsequent self-reported changes in depression, anxiety and psychological flexibility. Across the full sample and in LSD and psilocybin subsamples, MEQ and PIQ scores tended to co-occur and cluster together; CEQ scores were frequently elevated alongside mystical and insight experiences, indicating that challenging experiences often occur together with intense mystical or insight phenomena. Subtypes characterised by higher MEQ/PIQ scores were more likely to show improvements in depression and anxiety and greater enduring effects, while the subtype with lower MEQ/PIQ scores showed smaller improvements. The authors note that dose and drug type help explain some subtype differences—High Scoring participants reported higher doses and were more likely to have used DMT-type compounds—yet dose alone does not fully account for outcome differences. Baseline differences also matter: Low Scoring participants tended to have lower baseline symptom severity and psychological flexibility and therefore less scope for improvement. The varying proportions of participants in the Positive Scoring subtype across drugs (LSD showing a higher proportion than psilocybin in this sample) are presented as suggestive that different classic psychedelics differ in their tendency to produce particular subjective subtypes. Key limitations are acknowledged. The study relies on retrospective self-report and a quasi-experimental one-group pretest/posttest design using "then-test" items to estimate baseline symptoms, which introduces recall and present-state biases. Recruitment specifically targeted people who reported psychological insight from a prior psychedelic experience, creating selection bias that may exclude other clinically relevant subtypes. The cross-sectional clustering approach cannot establish causal relationships between subjective experience and outcome, and some subjective domains may not have been captured by the chosen instruments. The authors recommend prospective data collection in future studies to more robustly characterise subtypes and their causal relationships to therapeutic outcomes. Overall, the paper concludes that reproducible subtypes of the subjective psychedelic experience can be identified with machine learning clustering and that these subtypes show predictable associations with changes in depression, anxiety and psychological flexibility. The investigators suggest that enhancing mystical-type and personal insight experiences may be an important target for maximising beneficial outcomes in clinical applications, while noting the need for prospective, causal designs to confirm these relationships.