A quantitative textual analysis of the subjective effects of ayahuasca in naive users with and without depression

Cruz and colleagues situate their work in the context of growing clinical interest in ayahuasca, a traditional brew containing N,N-DMT and beta-carboline alkaloids, which has been investigated for potential therapeutic effects in mood and substance-use disorders. Earlier clinical trials have reported reductions in depression severity after ayahuasca dosing, and qualitative studies have described a range of subjective phenomena—visual visions, somatic purging, insights, and spiritual or cultural themes—but no prior study has applied quantitative textual-analysis tools to first-time ayahuasca users to compare clinical and non-clinical groups. The current study therefore aims to quantitatively characterise the self-reported acute experiences of ayahuasca-naïve individuals with treatment-resistant depression and healthy controls. Using transcribed, open-ended reports collected shortly after a single dosing session in a controlled setting, the investigators apply automated textual-analytic methods to identify recurring thematic clusters and to examine associations between those clusters and sociodemographic or clinical variables. The work seeks to illuminate how ‘set’ (pre-existing psychological state) and setting may shape subjective experience, with implications for therapeutic applications of ayahuasca.

The data derive from a subsample of participants in a previously reported randomised, double-blind, placebo-controlled clinical trial of ayahuasca in adults 18–60 years old with treatment-resistant depression. For the present analysis only participants who received ayahuasca were included: nine patients with treatment-resistant depression and 20 healthy individuals, all naïve to ayahuasca and other classical psychedelics. Exclusion criteria included current medical conditions, pregnancy, neurological disease, bipolar disorder or schizophrenia, history of mania/hypomania, and substance abuse. The extracted text does not clearly report recruitment dates or additional trial-site details beyond affiliation with a university hospital setting. Each participant received a single oral dose of ayahuasca calculated as 1 mL/kg from a single batch whose mean concentrations (± SD) were reported for N,N-DMT (0.36 ± 0.01 mg/mL), harmine (1.86 ± 0.11 mg/mL), harmaline (0.24 ± 0.03 mg/mL), and tetrahydroharmine (1.20 ± 0.05 mg/mL). The placebo was a liquid designed to mimic colour and flavour. During dosing, participants were instructed to remain silent with eyes closed and were provided a predefined music playlist; two researchers were available in an adjacent room for support. Primary and secondary clinical outcomes reported in the parent trial were changes in HAM-D and MADRS scores at specified post-dosing days, but the current paper focuses on open-ended experiential reports collected approximately 4 hours after ingestion, once acute effects had subsided. Interviews began with the prompt “Can you please freely describe your experience?” and lasted on average 13.7 minutes. Reports were audio-recorded, transcribed, proofread, and cleaned of certain punctuation to facilitate automated analysis. The textual corpus was built in Brazilian Portuguese and later translated for reporting. Sociodemographic and clinical variables were coded and dichotomised for association testing; missing values were treated as null. Quantitative textual analysis used IRaMuTeQ 0.7 alpha 2 and R 3.1.2. Methods included Descending Hierarchical Analysis (DHA, Reinert method), Specificities analysis, and Correspondence Factor Analysis (CFA). Texts were segmented (default 40 words per text segment), and DHA partitioned the corpus into clusters based on word frequency and association. Inclusion criteria for words in clusters required frequency above the corpus mean and a chi-square with the cluster greater than 3.84 (p < 0.05, df = 1). The authors analysed active word forms (nouns, verbs, adjectives, adverbs, pronouns, and unrecognised forms). Cramer's V was used to report effect sizes for word–cluster associations and α was set at 0.001 to limit type I error. The text indicates ethical approval and compliance with relevant standards and the Helsinki declaration, but the extraction truncates the full ethics statement.

The textual corpus comprised 869 text segments (TS), 2,968 distinct words occurring 29,029 times (mean occurrences per TS = 33.41). Active word forms numbered 1,829, of which 691 had a frequency ≥ 3. DHA retained 75.03% of the corpus for cluster analysis and produced five thematic clusters. Cluster labels and relative representation (as presented in the paper) were: Cluster 1 "Alterations in the state of consciousness" (~15.2% of TS), Cluster 2 "Cognitive changes" (~14.6% of TS), Cluster 3 "Somatic alterations" (~18.1% of TS), Cluster 4 "Auditory experiences" (~29.4% of TS), and Cluster 5 "Visual perceptual content" (~22.1% of TS). For the five most associated words per cluster, Cramer's V indicated associations ranging from small to medium effect sizes and, in Cluster 2, larger effects; exact Cramer's V values are not provided in the extracted text. Cluster content and notable associations were reported as follows. Cluster 1 included words tied to onset and phenomenology of altered consciousness, with strong co-occurrence of "eye" with "open" and "close", and terms such as "sleep", "slow", "nap", "dream", and temporal words like "hour". Middle-aged participants showed a small association with this cluster. Cluster 2 grouped cognitive-process words—"remember", "question", "look", "understand"—and comparative terms such as "same", alongside perceptual qualifiers like "blurred" and "strange"; small associations were observed with male gender, higher income, and past cannabis use. Cluster 4 was dominated by auditory and music-related vocabulary—"music", "speak", "listen", "sound", "silence", "voice"—and captured both positive and negative reactions to music (for example, "relax" versus "bother"). Female participants and those reporting no religion were more strongly associated with this cluster. Cluster 3 contained somatic and often aversive bodily terms—"vomit", "pain", "bathroom", "hungry", "stomach", "breathe", "dizziness"—and was uniquely associated with a diagnosis of depression, suggesting greater reporting of negative somatic experiences among depressed participants. Cluster 5 encompassed visual content with central words such as "see" linked to "colour" and specific colours, and human-like figures (e.g. "puppet", "woman") with actions like "dance" or "jump"; this cluster showed neutral-to-positive valence (e.g. links between "see" and "beautiful") and was associated with participants without depression, as well as with low income, living alone, and non-tobacco use. Correspondence Factor Analysis indicated that Clusters 1, 2, and 4 were closely related in a common area of the factor space, whereas Clusters 3 and 5 occupied more isolated positions. The Specificities analysis mapped co-occurrence communities among words, but specific numeric indices from that analysis are not reported in the extracted text.

Cruz and colleagues interpret the five clusters as reflecting core axes of the acute ayahuasca experience in naïve users: altered states of consciousness, cognitive processing changes, somatic/purgative responses, auditory/music- and speech-related elements, and visual perceptual content. The prominence of visual phenomena—participants described vivid colours, human-like figures and motion—and the reported ability to modulate visions by opening or closing the eyes are highlighted as central to the altered-state domain. Cognitive-cluster content suggested attempts to compare or reframe perceived reality, while the auditory cluster underscored the importance of music and spoken guidance in shaping experience. A key finding is the differential pattern in participants with depression: somatic, often aversive, bodily sensations (Cluster 3) were more strongly associated with depressed individuals, whereas more neutral-to-positive visual content (Cluster 5) predominated among healthy participants. The authors note that this could reflect greater somatic symptomatology in depression or cognitive biases affecting recall and reporting. Small effect sizes for most sociodemographic associations are emphasised, and the CFA result that cognitive, consciousness-alteration and auditory clusters are closely related is reported as consistent with interlinked dimensions of the psychedelic experience. The discussion positions these results relative to prior qualitative and psychometric work, noting overlap with previously identified thematic domains (visions, corporal experiences) and suggesting parallels with structured measures such as the 5D-ASC (for example, a Vigilance Reduction factor that might correspond to Cluster 1). The authors also observe that this study, unlike some previous reports, identified a distinct auditory cluster and did not retrieve a substantial corpus of emotional or spiritual/transcendental themes; they attribute this partly to the clinical, university-hospital setting used here rather than ritualistic or psychotherapeutic environments. Limitations the authors acknowledge include the small sample size, which limits statistical power and generalisability, and the inability within this dataset to link specific depressive symptom profiles to particular experiential patterns. The authors recommend future research combining subjective verbal reports with physiological measures, exploring psychometric correlates of verbal-report dimensions, and conducting studies in different settings to clarify the influence of contextual factors on subjective experience. The extracted text does not present further conclusions beyond the discussion, and no separate Conclusion section is present in the provided material.