Ayahuasca-Inspired DMT/HAR Formulation Reduces Brain Differentiation Between Self and Other Faces

Earlier research indicates that serotonergic psychedelics profoundly alter visual perception and self-referential processing. Electrophysiological work has shown modulation of early visual event-related potentials (ERPs) such as the P1 (50–100 ms) and N170 (face-sensitive, ~160–190 ms), while higher-level ERP markers like the P300 have been linked to attentional allocation to self-relevant stimuli. However, how psychedelics influence the neural distinction between self and other in face recognition—an important testbed for self-referential cognition—remains understudied. Suay and colleagues set out to fill that gap by testing an ayahuasca-inspired formulation combining intranasal DMT with buccal harmine (DMT/HAR). Using a within-subject, double-blind, placebo-controlled design, they examined EEG responses to Self, Familiar (celebrity), and Unknown faces under DMT/HAR, harmine alone (HAR), and placebo (PLA). The primary aim was to determine whether DMT/HAR alters early sensory (P1), structural face-encoding (N170), and late self-referential (P300) ERPs—and specifically whether DMT/HAR reduces neural differentiation between self and other faces, a candidate neural substrate of ego-dissolution phenomena.

Design and participants: The study used a within-subject, double-blind, placebo-controlled design with three counterbalanced sessions (DMT/HAR, HAR, PLA) separated by at least two weeks. Thirty-six right-handed healthy male volunteers were enrolled; five withdrew and one dataset was lost to technical problems, leaving 30 participants for analysis. Inclusion criteria included no personal or family history of psychotic or bipolar disorders, normal vision and hearing, fluency in German, and limited prior hallucinogen exposure (≤15 lifetime uses). Ethical approval and informed consent procedures were reported. Drug administration and timing: In the DMT/HAR condition participants received a 100 mg buccal harmine orodispersible tablet followed, after 30 minutes, by repeated intranasal DMT administrations of 10 mg at 15-minute intervals until a cumulative DMT dose of 50 mg was reached. The EEG experiment started 60 minutes after the first DMT administration, at peak drug effect. In the HAR condition participants received 100 mg harmine plus a placebo nasal spray; in PLA both tablet and spray were inert but matched for taste and appearance. Blinding was maintained until EEG preprocessing and initial statistical models were complete. Task, stimuli and behavioural measures: During each session participants performed a 30-minute visual oddball task while EEG was recorded. Stimuli comprised 396 greyscale face images split into Self, Familiar (participant-specific celebrity images), Unknown (Chicago Face Database), and scrambled-target images (36 targets). Each face category contributed 120 trials. Images were standardised for contrast, luminance and spatial frequency; an ellipse mask removed hair and outer features. Participants fixated centrally and pressed a button only when scrambled targets appeared (target probability 8–12%); mean target detection accuracy was 98%. EEG acquisition and preprocessing: Continuous EEG was recorded with a 64-channel BioSemi ActiveTwo system plus additional electrodes for EOG, ECG and mastoids; data were sampled at 2048 Hz with 0.1 Hz high-pass and 100 Hz low-pass filters. Preprocessing used EEGLAB with ICA-based artifact removal (infomax ICA, CORRMAP templates), filtering and resampling steps to improve decomposition. Data were epoched −200 to 800 ms, baseline corrected, and trials with non-stereotypical artifacts or amplitudes exceeding ±100 µV were excluded. The proportion of interpolated electrodes was kept below 10% per participant. The proportion of removed epochs averaged roughly 5–6% across conditions and categories, with no significant differences between conditions. Statistical analysis: Subjective effects (11D-ASC) were analysed with repeated-measures ANOVAs and corrected post-hoc tests. EEG analyses used a data-driven cluster-based nonparametric permutation approach: initial MANOVA-style F-tests (Monte Carlo, 1000 iterations) across all electrodes and the 0–600 ms window identified drug effects, followed by pairwise cluster t-tests (1000 iterations) when appropriate. Interaction tests examined drug-by-face effects using subtraction waveforms (e.g. Self−Unknown) with one-tailed cluster tests for negative clusters. To relate EEG and experience, Spearman correlations were computed between DMT−PLA difference scores in ERP components (P1: OZ/O1/O2, 90–150 ms; N170: P8/PO8, 160–190 ms; P300: Pz/CPz, 350–450 ms) and 11D-ASC scales, ensuring correlations reflected drug-induced changes rather than baseline differences.

Subjective effects: Repeated-measures analyses showed a robust main effect of drug condition across all 11D-ASC subscales (all p < 0.001). Post-hoc tests indicated that the DMT/HAR condition produced significantly greater scores than both PLA and HAR on measures including Oceanic Boundlessness, Anxious Ego Dissolution, Elementary and Complex Imagery, Spiritual Experience and others (all p < 0.05). No significant differences emerged between PLA and HAR (all p > 0.05). Behavioural performance and preprocessing: Participants maintained high target-detection performance during the task (mean 98% accuracy). Epoch rejection rates averaged about 5–6% and did not differ by condition or stimulus category. ERPs to self faces: Cluster-based MANOVA identified two positive clusters for self-face ERPs at 140–199 ms and 207–600 ms (both p < 0.001), indicating drug-related differences. Pairwise comparisons showed that DMT/HAR increased posterior-occipital responses at 140–200 ms relative to PLA (p = 0.022) and HAR (p < 0.001). Concurrently, DMT/HAR produced reductions in later N170–P300 waves compared to the other conditions. DMT/HAR also increased frontal responses between 200–600 ms compared with PLA (clusters p < 0.025) and HAR (p < 0.001). No significant differences were observed between HAR and PLA for self faces. ERPs to familiar faces: For familiar faces, two positive clusters were found between 80–400 ms (p < 0.001) and 400–580 ms (p = 0.017). DMT/HAR increased posterior-occipital responses in the 90–200 ms window compared with PLA (p = 0.013) and HAR (p = 0.018), reflecting altered P1–N170 dynamics. DMT/HAR increased frontal responses from 200–600 ms relative to HAR (clusters p < 0.025) but not relative to PLA (p > 0.025). Notably, HAR (vs PLA) increased neural responses to familiar faces at 265–480 ms at right central-parietal-temporal sites (p = 0.01). ERPs to unknown faces: MANOVA revealed two positive clusters at 125–205 ms and 215–600 ms (both p < 0.001). DMT/HAR enhanced posterior-occipital responses to unknown faces in the 130–205 ms window relative to PLA (p = 0.02) but not relative to HAR (p > 0.025). At later stages (300–600 ms), DMT/HAR decreased responses at right central-posterior-temporal electrodes compared with PLA (p < 0.001) and HAR (p = 0.004). Interim and interaction findings: Across face categories, DMT/HAR consistently modulated early ERPs (<200 ms) with increased parieto-occipital activity (P1) and attenuated later face-encoding markers (N170). Crucially, interaction analyses testing whether DMT/HAR reduced neural differentiation between self and other faces confirmed that differential self-minus-other ERP responses were significantly reduced under DMT/HAR. A negative cluster approximately between 320 and 550 ms showed smaller Self−Unknown and Self−Familiar differences for DMT/HAR versus PLA (Self−Unknown p = 0.024; Self−Familiar p = 0.004) and versus HAR (Self−Familiar p = 0.002). Post-hoc analysis at central-posterior electrodes (CPz, POz, Pz) in the 300–500 ms P300 window showed that the typical P300 elevation for self faces observed in PLA and HAR was substantially reduced in DMT/HAR. Correlations with subjective experience: Spearman correlations using DMT−PLA difference scores revealed that reductions in N170 amplitude correlated positively with several 11D-ASC scales: Spiritual Experience (r = 0.370, p = 0.044), Impaired Cognition and Control (r = 0.394, p = 0.031), global ASC (r = 0.384, p = 0.036), Anxious Ego Dissolution (r = 0.422, p = 0.020), and Complex Imagery (r = 0.370, p = 0.044). Reductions in P300 amplitude correlated with Anxious Ego Dissolution (r = 0.384, p = 0.037) and Elementary Imagery (r = 0.405, p = 0.027). No significant correlations were found for P1 or other subjective measures (all p > 0.05).

Suay and colleagues interpret these results as evidence that the DMT/HAR formulation differentially affects stages of visual and self-referential processing. At early stages (<200 ms) DMT/HAR increased parieto-occipital responsiveness, reflected in larger P1 amplitudes, while attenuating the N170 at lateral occipital-temporal sites. The authors characterise this pattern as enhanced sensory gain paired with impaired structural encoding of faces, consistent with prior reports that psychedelics amplify low-level visual signals but disrupt higher-level perceptual integration. At later stages, analyses showed that DMT/HAR reduced P300-based differentiation between self and other faces at central-posterior electrodes, which the investigators link to a weakening of attentional prioritisation for self-relevant stimuli. The P300 is discussed as a marker of attentional resource allocation and has ties to networks implicated in self-referential cognition; thus the reduction in self-specific P300 under DMT/HAR is interpreted as a neural correlate of blurred self-other boundaries akin to ego dissolution. Importantly, the authors note a spatial dissociation: despite posterior attenuation of self–other differentiation, frontal electrodes showed increased responses to self-faces under DMT/HAR, while responses to unknown faces were decreased at right central-parietal sites. They propose this as a reorganisation rather than a uniform loss of self-salience, where higher-order frontal processes may partially preserve or compensate for self-representation while lower-level perceptual distinctions become less pronounced. The discussion links ERP changes to subjective reports: reductions in N170 and P300 correlated with measures related to anxious ego dissolution and imagery intensification, suggesting that early perceptual instability may cascade into altered late-stage self-referential processing. The authors situate these findings within hierarchical predictive-coding frameworks, proposing that psychedelics disrupt lower-level perceptual priors and thereby reweight higher-order representations of social and self-related information. Limitations acknowledged by the study team include the specificity of the observed effects to the particular DMT/HAR dose and administration route, the limited spatial precision of EEG (suggesting combined fMRI or MEG for future localisation), and uncertainty whether these are transient state effects or relate to long-term changes. The authors also note the all-male healthy sample and their choice of celebrity faces as "familiar" stimuli, which may not capture personally familiar, emotionally salient faces. They recommend future work to test other compounds and doses, integrate multimodal imaging, explore longitudinal effects, and assess individual differences that may modulate whether ego-dissolution experiences are experienced as therapeutic or distressing.

The investigators conclude that the ayahuasca-inspired DMT/HAR formulation substantially alters both early and late neural processing of faces. Enhanced P1 amplitudes indicate heightened early sensory processing, while attenuated N170 amplitudes point to disrupted structural face encoding. Critically, reduced P300 differentiation for self versus other faces at posterior sites suggests a neural blurring of self–other boundaries consistent with ego-dissolution experiences, though frontal increases in self-related responses indicate a reorganised, not entirely lost, self-salience. The authors propose that these reorganisations of self-referential processing may help explain aspects of psychedelic-induced changes in social cognition and might be relevant to therapeutic mechanisms for conditions characterised by rigid self-representation.