The profound effects of psychedelics reshape subjective responses to internal and external sensations and are frequently reported as among the most meaningful experiences in life⁷. These states can manifest sustained therapeutic benefits, including reductions in depression, anxiety and addiction, alongside an increase in social connectedness and overall well-being^{3,4,9,10,11,12,13,14}.

The brain constructs perception and selfhood by integrating external sensory inputs with internal models of the environment¹⁵. These interactions between sensory and associative brain regions are enabled by structural and functional connectivity¹⁶. Psychedelics such as psilocybin disrupt these interactions by acting at the serotonin 5-HT_2A receptor, inducing structural and functional plasticity in preclinical models—rapid medial prefrontal cortex spinogenesis persisting for weeks, mechanistically linked to enduring behavioural effects^17,18—that can reshape macro-level connectivity^19,20,21,22. Studies that examine effective connectivity further suggest that associative network communication becomes reconfigured^{23,24,25,26,27}. Reflecting this reorganization, individuals frequently report an intensified sense of immersion, in which the context of space, time and selfhood feels deconstructed and interconnected^28,29. Although the exact mechanisms linking brain network-level shifts to subjective experience remain unknown, these dynamics have typically been characterized as entropic and desynchronized^5,6,30,31. A central mediator of these effects is the default mode network (DMN), which supports the integration of information from diverse associative regions spanning spatial, temporal and self-referential contexts^{32,33,34,35,36,37}. Under psilocybin, connectivity patterns that constrain networks relax, permitting novel interactions between sensory and associative regions to emerge^38,39. This reconfiguration alters connectivity patterns that underlie emotion, cognition and perception⁴⁰. Of note, in healthy adults, psilocybin produces a persistent reduction in anterior hippocampus (aHip)–DMN connectivity, and in rodents, preclinical studies have reported acute DMN hypoconnectivity^6,41. Although observed in non-clinical samples, this pattern has been proposed as mechanistically relevant to therapeutic effects. However, despite mounting evidence for these shifts in connectivity, how they give rise to meaningful and therapeutically relevant experiences remains unclear.

Psychedelic effects are also context-sensitive, shaped by mindset and setting⁸ and recognized in clinical guidelines⁴². Understanding how context influences therapeutic change, and the underlying brain connectivity that can be harnessed clinically, requires a comprehensive examination of how psychedelics reshape functional integration of the brain.

Existing studies often rely on small sample sizes, participants with psychedelic experience, limited imaging modalities or single-context designs, making it difficult to generalize findings. Structured cognitive tasks imposed during imaging also risk conflating task demands with the psychedelic state itself, rather than capturing how the state unfolds naturalistically. Some recent work has partly addressed these limitations^2,43,44, but no study to date has combined large-scale sampling, multimodal imaging and diverse naturalistic contextual manipulations within a single acute session and computational framework.

We acquired the largest single-site acute-phase psychedelic neuroimaging dataset to date, integrating multimodal fMRI and EEG with controlled contextual manipulations across eyes-open and eyes-closed states in a cohort with no lifetime psychedelic experience (n = 62). Experimental conditions of rest, meditation, music and movie were systematically varied, and extensive behavioural measures were collected (Extended Data Fig. 1).

Our analyses show that psilocybin redistributes integration, increasing global functional connectivity in associative regions while reducing it in sensory areas, altering the balance of internally and externally directed processing. Across fMRI and EEG, signals recorded during eyes-closed conditions became similar to those of an eyes-open condition. Critically, using machine learning, we uncovered context-aligned organization in brain activity that was missed by conventional analyses. This reorganization was experientially graded, emerging under positively felt self-dissolving and boundary-dissolving experiences, and not during negatively felt states or cognitive impairment.

These results identify a state-dependent neural signature that emerges when network organization becomes flexibly aligned with context, coinciding with the subjective state that we refer to as embeddedness. We find this state to be consequential for subsequent change and contextually modifiable. The finding that the organization of brain activity tracks the transformative quality of subjective experience helps to elucidate how the psychedelic state translates into psychological change.

Psilocybin redistributes cortical GFC

Psilocybin induced distinct shifts in global integration of functional connections across brain regions. This was quantified using global functional connectivity (GFC), a measure of the average correlation of blood oxygen level-dependent (BOLD) signals between each vertex and every other vertex in a high-resolution cortical surface map⁴⁵. During eyes-closed conditions (rest, meditation and music), psilocybin reduced the GFC of sensory regions (Fig. 1a), and these reductions formed statistically significant clusters (Extended Data Fig. 2a). Occipital areas showed the strongest overall reductions (percentage changes up to −39% and Cohen’s d = −0.55), with midline ventral parietal decreases prominent during rest, meditation and music. Conversely, GFC increased in associative areas during these conditions (percentage changes up to +72%, Cohen’s d = 0.53). This evidence indicates that, during eyes-closed conditions, psilocybin suppressed the integration of sensory regions with the broader cortical network and enhanced the integration of associative regions. By contrast, during the eyes-open movie condition, GFC increased in both sensory and associative areas (Fig. 1a), enhancing global synchrony, independent of head motion (Extended Data Fig. 2b). These eyes-closed shifts in GFC are directionally consistent with previous global connectivity studies under psychedelics during eyes-closed rest, which commonly report increases in associative or frontoparietal systems^46,47 and, in degree centrality analyses, decreases in sensory or visual areas⁴⁸. However, differences in metrics (such as global correlation and functional connectivity density), spatial domains (volumetric versus surface) and subcortical coverage limit direct comparisons.

a, GFC percentage change (psilocybin versus no psilocybin) averaged across all participants for each condition (row). The significant clusters after threshold-free cluster enhancement are shown in Extended Data Fig. 2a. b, Average functional connectivity difference (psilocybin minus no psilocybin) across participants. Psilocybin reduced functional connectivity within networks during eyes-closed conditions. The network names are: subcortex (sub), visual (vis), somatomotor (sommot), dorsal–ventral attention (dorsventattn), limbic, salience–ventral attention (salventattn), default mode (default) and control (cont). A larger version of these functional connectivity matrices is provided in Extended Data Fig. 4b. c, s.d. percentage change (psilocybin minus no psilocybin) computed for each participant and then averaged across participants, for each condition. Only cortical vertices where s.d. changes were not significantly associated with framewise displacement are shown. d, Histograms of GFC values for seven resting-state networks (columns) and four conditions (colour-coded). The gap between eyes-closed and eyes-open conditions at baseline (no psilocybin; top row) vastly reduces after psilocybin (bottom row). This reduction was most prominent for the visual, somatomotor and dorsal attention networks, but was larger than −66% in all networks and statistically significant (Extended Data Fig. 4a). e, Histograms of GFC values combining all networks. f, Psilocybin reduces functional connectivity (FC) modularity across conditions (Cohen’s d ≤ −0.6) for all comparisons. Each point corresponds to one participant (rest n = 60, meditation n = 59, music n = 56 and movie n = 57), the boxes indicate the median and quartiles, and the whisker length is 1.5 times the interquartile range. The asterisks indicate a significant decrease using the Mann–Whitney U-test (right-tailed, all P < 10⁻³).

Group-level findings in eyes-closed conditions were robust to z-scoring and global signal regression (Extended Data Fig. 3; see Methods for the mathematical rationale for omitting these steps). Substantial individual variability in GFC changes was observed across participants (Supplementary Fig. 1).

Context reshapes spatial BOLD variance

Mapping the standard deviation (s.d.) of the BOLD signal within participants across the cortex revealed that signal variability was spatially reorganized and condition dependent under psilocybin.

Compared with baseline (no psilocybin), the s.d. increased during the eyes-open movie condition in the orbitofrontal cortex, occipitotemporal gyri, inferior temporal lobes (particularly anterior regions) and the right hemisphere primary somatosensory cortex (percentage changes up to +19%, Cohen’s d = 0.54).

By contrast, s.d. decreased in early visual areas across all eyes-closed conditions, most strongly during music (percentage changes up to −10%, Cohen’s d = −0.38). During the eyes-open movie, we also detected spatially constrained s.d. decreases in early visual regions, as well as in the posterior cingulate cortex and precuneus (Fig. 1c), which otherwise showed increased s.d. as part of widespread cortical increases observed during eyes-closed conditions.

Increases in s.d. under psilocybin were partially lateralized to the right hemisphere during eyes-closed rest, meditation and music, and s.d. decreases in sensory regions were observed in the occipital lobe, particularly during music (Fig. 1c). Finally, the maps of s.d. change across the brain in all states were broadly consistent with the maps of GFC change (Fig. 1a), and this redistribution of signal variability is consistent with increased Shannon entropy (see Methods).

Eyes-closed and eyes-open GFC converge

Complementing the spatial maps of GFC changes, we examined the overall distribution of GFC values across all cortical vertices, independent of their spatial organization. Under psilocybin, the histograms of GFC values in eyes-open and eyes-closed states—which were distinct at baseline (no psilocybin) imaging—largely overlapped (Fig. 1e).

This convergence was consistent and statistically significant across sensory, limbic and associative resting-state networks (Fig. 1d and Extended Data Fig. 4a) but particularly pronounced in the visual network, where the gap between eyes-closed and eyes-open GFC values decreased by 85% after psilocybin (Cohen’s d = −3.23, P < 10⁻³⁰⁷).

Decreased functional modularity

Functional modularity—the degree to which brain activity is organized into distinct, segregated networks—decreased under psilocybin across all conditions (Fig. 1f; Cohen’s d = −0.63, P = 6.5 × 10⁻⁵ for rest; d = −0.60, P = 1.0 × 10⁻⁴ for meditation; d = −0.70, P = 8.8 × 10⁻⁵ for music; d = −0.91, P = 2.8 × 10⁻⁶ for movie; Mann–Whitney U-test, right-tailed). Comparable decreases in modularity have also been reported in clinical samples, where lower post-treatment modularity was associated with longer-term symptom improvement following psilocybin therapy⁴⁹. Previous studies have consistently reported disruptions in resting-state network integration and segregation under psychedelics^46,50, and have primarily analysed DMN connectivity^1,51,52.

Analysis of our larger dataset revealed that psilocybin-induced reorganization of brain network architecture was primarily attributable to decreased within-network connectivity (Fig. 1b), with significant reductions observed in all networks across eyes-closed conditions (rest, meditation and music), and only in the DMN and dorsal attention network during the eyes-open movie (Supplementary Figs. 2 and 3).

Context-aligned trajectories emerge

Dimensionality reduction methods have been adopted in neuroscience for uncovering meaningful low-dimensional structures in neural data⁵³. In machine learning, an embedding is a learned mapping from high-dimensional data to a low-dimensional representation, expressed as vector coordinates in an embedding space that preserve latent structure^53,54,55. Here we used contrastive embeddings for behavioural and neural analysis (CEBRA) to generate low-dimensional embeddings of the preprocessed, frame-by-frame regional BOLD time series and used support vector machine (SVM) classification as a descriptive readout of how well session-specific embeddings distinguished between rest, meditation, music and movie conditions⁵⁶ (see Methods). We then examined the relationship between classification accuracy and subjective experiences assessed via the Mystical Experience Questionnaire (MEQ30)^29,57, administered at the end of the session.

Under psilocybin, structured organization emerged in the temporal dynamics of brain activity. Of note, stronger subjective effects were associated with tighter clustering of time point embeddings within the same condition, making them more easily separable from embeddings of other conditions (Fig. 2a). This condition-specific clustering was moderated by the timing of subjective effects, as demonstrated by a participant who reported a late onset of substantial effects but minimal subjective experience during the imaging session (Fig. 2b). Their embeddings resembled baseline (no psilocybin) scans, indicating that the observed neural reorganization covaried with the occurrence and intensity of subjective effects within the imaging window (Fig. 2b).

a, Using CEBRA, for each participant, we mapped the fMRI time series of all 332 brain parcels at each time point into a three-dimensional space, producing a trajectory spanning the four conditions: rest (pink), meditation (blue), music (green) and movie (yellow). b, Network embeddings (CEBRA-derived trajectories) for participant PC201, who reported no subjective effect onset during the MRI session on the day of psilocybin administration (late-onset post-MRI); their embeddings resembled the baseline (no psilocybin) scan. c, For the CEBRA trajectory of each participant, a SVM classified the condition label at each time point. Pearson correlation between classification accuracy and acute psilocybin subjective effects (11D-ASC and MEQ30) is shown (left). Positively felt (euphoric) self-dissolving and boundary-dissolving effects (blue), sensory–hallucinogenic effects (orange), negative effects (red) and other (combination) effects (purple) are displayed. The asterisk indicates statistical significance (*P < 0.05; Fig. 4 and Extended Data Fig. 6; Methods). d, Pearson correlations between per-participant functional modularity (psilocybin) and two behavioural outcomes (MEQ30 mean, n = 54; next-day mindset change, n = 53) across the four conditions. Modularity did not predict either outcome (all ∣r∣ < 0.15, P > 0.28). The dashed lines mark the corresponding CEBRA classification accuracy correlations in the same participants (MEQ30: r = 0.64, P < 10⁻⁶; mindset: r = 0.40, *P < 0.01). NS, not significant. e, A TAVRNN mapped the 332 ROIs into a two-dimensional space. ROIs were projected into a common two-dimensional t-distributed stochastic neighbour embedding space and visualized separately for each condition. Average two-dimensional embedding of participants with the highest MEQ scores (80–100; top row), and the lowest MEQ scores (0–20; bottom row) are shown. Each node corresponds to one ROI, colour-coded by brain network. The axes represent abstract latent dimensions in arbitrary units (a.u.) learned by the model and do not correspond to physical coordinates.

Further analysis of 12 representative participants, 6 with high MEQ scores and 6 with low MEQ scores, reinforced these findings, revealing a gradient in brain embeddings that scaled with strength of subjective effects (Fig. 3a–d). This gradient reflected a reorganization of neural trajectories under psilocybin that was both context-specific and effect-dependent, suggesting that as subjective effects intensified, context increasingly differentiated neural trajectories into distinct, cohesive patterns.

a,b, CEBRA embeddings of n = 6 participants with low (a) and high (b) MEQ scores, ordered by classification accuracy. Psilocybin (top) and no psilocybin (bottom) are shown. c,d, Silhouette scores quantifying within-condition clustering for the same 12 participants at baseline (no psilocybin; c) and under psilocybin (d). The boxes indicate the median and quartiles, the whisker length is 1.5 times the interquartile range, and the open circles denote outliers beyond the whiskers (n = 6 participants per box). Groups did not differ at baseline (all P > 0.19) but differed significantly across all conditions under psilocybin (rest: Δs = 0.51, P = 0.0002; meditation: Δs = 0.71, P = 0.0023; music: Δs = 0.52, P < 0.0001; movie: Δs = 0.63, P = 0.0007; Welch’s t-test), establishing the gradient as psilocybin-specific. CEBRA training was repeated 10 times per participant. *P < 0.05. e,f, Run-to-run variability of silhouette scores per participant: mean s.d. (e) and median absolute deviation (MAD; f) across conditions. Both remain small (s.d. ≈ 0.05–0.25; MAD ≈ 0.03–0.14), confirming clustering stability. P1–12, participant 1–12. g–i, Network-attributed decomposition from a perturbation analysis (n = 6; see Methods): for each participant, the psilocybin time series of one network was replaced with their no-psilocybin dynamics; the drop in classification accuracy was attributed to that network. Radar plot (g) of averaged per-network contributions. Waterfall plot (h) decomposing the cumulative accuracy gain from baseline (~62%) to psilocybin (~98%); DMN (+22.6%) and visual network (+23.1%) together account for nearly half of the gain, indicating that the gradient end points jointly enable context alignment; dashed lines mark group mean SVM classification accuracy at baseline (lower) and under psilocybin (upper). Per-participant stacked bars (i) showing the contribution of each network, demonstrating consistency across individually trained models. See Methods for details.

To assess the robustness of these effects, we compared CEBRA embeddings with three common dimensionality reduction methods (principal component analysis, t-distributed stochastic neighbour embedding and Isomap) in participants with high MEQ. Despite methodological differences, each approach recovered the context-dependent embedding structure (Extended Data Fig. 5), confirming that the observed organization was not specific to CEBRA.

To quantify the degree of this organization, we classified the condition label at each time point for each participant with a SVM classifier on the CEBRA-derived trajectories. We examined the correlation between classification accuracy and each of the 11-Dimension Altered States of Consciousness (11D-ASC)⁵⁸ and MEQ30 scores, after psilocybin administration. Classification accuracy was most strongly associated with positively felt self-dissolving and boundary-dissolving effects, followed by weaker correlations with sensory-hallucinogenic effects, and low or negative correlations with anxiety and impaired control and cognition (Fig. 2c and Extended Data Fig. 6). We also found a correlation between classification accuracy and mindset change 1 day after psilocybin (Extended Data Fig. 6, Supplementary Table 1 and Supplementary Fig. 4). Participants with higher MEQ30 mean scores generally exhibited higher classification accuracy, indicating that the context-specific organization of neural dynamics scales with the intensity of subjective experience (Extended Data Fig. 7a). We refer to this structured alignment of neural dynamics with experiential context as context alignment. Of note, these brain–behaviour associations depended on preserving temporal dynamics: per-participant functional modularity—a time-averaged measure of network segregation that, despite robust group-level reductions under psilocybin (Fig. 1f; Cohen’s d = −0.60 to −0.91)—was not significantly associated with MEQ30 mean or next-day mindset change in any condition (Fig. 2d).

A network perturbation analysis (see Methods) revealed that this context-aligned organization is globally distributed but depends most strongly on altered dynamics in the DMN and the visual network (approximately 6–8 percentage point accuracy reduction each when their psilocybin-induced dynamics are replaced with baseline; Fig. 3g,h)—the two systems that anchor opposite poles of the principal cortical gradient from internal to external processing³⁴. This dependence was uniform across all four conditions rather than condition-specific (Extended Data Fig. 7b), indicating that the joint alteration of the systems at both gradient end points sets the conditions for brain activity to align with context. This convergence held across individually trained models (Fig. 3i), providing anatomical grounding for how psilocybin reorganizes neural trajectories into structured patterns that reflect context. Crucially, if the observed effect merely increased random noise, trajectories would become diffuse and overlapping, degrading classification accuracy. The opposite was observed. Participants reporting stronger effects showed tighter clustering (higher silhouette scores; Fig. 3d) and higher accuracy, indicating that this reorganization was structured rather than random. No structured embedding emerged at baseline or with low MEQ, and clustering was already strong during rest (the first condition acquired), before any sequence could accrue, indicating that the low-dimensional organization is unlikely to be driven by the fixed order of conditions (see Supplementary Information for condition order considerations).

These findings were independently supported using another machine learning-based technique called temporal attention-enhanced variational graph recurrent neural network (TAVRNN)⁵⁹, which captured a lower-dimensional representation of individual network connectivity during each condition. These two-dimensional maps represent abstract latent dimensions learned from brain functional connectivity patterns, not physical brain coordinates, with each point corresponding to one region of interest (ROI) placed closer to others when their connectivity evolves similarly over time. TAVRNN captures temporal changes in network structure by modelling sequential snapshots of brain connectivity, enabling the identification of key connectivity patterns and shifts in communicability (see Methods for TAVRNN details).

TAVRNN revealed two distinct patterns under psilocybin. Nodes within individual networks exhibited tighter clustering, reflecting more cohesive within-network dynamics (Extended Data Fig. 8 and Supplementary Information). Simultaneously, embeddings across all networks showed tighter clustering, indicating a shift towards more integrated brain-wide organization and increased global cohesion across contexts. This dual pattern of local within-network cohesion alongside global reorganization scaled with subjective effects, suggesting that the increased within-network and between-network cohesion observed under psilocybin depended on the strength of subjective experience and was broadly preserved across contexts (Fig. 2e).

Structure of psychedelic phenomenology

Participants enrolled in the study had no previous psychedelic experience (see Methods). Reports 1 day after psilocybin confirmed that their experiences during imaging were profoundly altered and meaningful. Half of the participants retrospectively ranked it among the most meaningful experiences of their lives (Fig. 4a), and most reported it as substantially intense (at least 9 out of 10; Fig. 4b). Semi-structured reports included accounts that described the experience as ‘one of the most peaceful and profound things I could ever experience’, ‘…meld[ing] with the MRI machine, floor, walls, air’, ‘los[ing] the plot of who I was, where I was, if I was even here, what was happening’ and ‘los[ing] all sense of self… becom[ing] at one with all my surroundings…as part of a bigger network of things’ (see Supplementary Information for tables of qualitative excerpts).

a, Distribution of meaningfulness ratings (see Methods). The first two bars represent participants who did (second) and did not (first) rate the experience as personally or spiritually meaningful; the next three bars indicate the number rating it in the top 50, top 10 and top 5 most meaningful life experiences, respectively. See Methods for expanded range. b, Histogram of intensity scores from ‘how intense would you rate the psilocybin experience?’ reported the day after psilocybin. c, Histograms of 11D-ASC scores. The colours indicate theoretical distinctions: positively felt associative effects (blue); sensory–hallucinogenic effects (orange); negative effects (anxiety and impaired control and cognition; red); and other (combination) effects (purple). d, Radar plot of 11D-ASC: mean scores across participants after psilocybin (solid red) and at baseline (solid blue), with the minimum–maximum range shaded. e, Histograms of MEQ30 scores, composed of three positively felt associative subdimensions and one valence-neutral (ineffability) subdimension. f, Radar plot of MEQ30 mean scores with the minimum–maximum range shaded. g, Associations between 11D-ASC effects show strong correlations among positively felt (blue) and sensory–hallucinogenic (orange) and negative (red) groupings. Grey marks paired subscales from different groupings, shown to indicate the absence of association between them. Impaired control and cognition showed no correlation with spiritual experience. The asterisks denote correlations significant after Bonferroni correction (*P < 0.001) across the 105 unique pairs (α = 4.76 × 10⁻⁴; two-sided Pearson). See Extended Data Fig. 9 for the complete 11D-ASC–MEQ pairwise correlation matrix and respective n values. h, Correlations between averaged next-day mindset change and acute psilocybin subjective effects (11D-ASC and MEQ30). *P < 0.05, two-sided Pearson, uncorrected; exact P values and per-subscale n are given in Methods. See Supplementary Information for mindset score dimensions. i, Music experience scores (see Methods) during EEG and MRI at baseline and under psilocybin, and AES-M scores comparing aesthetic music experience at baseline and under psilocybin. The asterisks (*P < 0.001) show Mann–Whitney U-test, two-tailed (n = 55 for all boxplots). The boxes show the median and quartiles, and the whiskers are 1.5 times the interquartile range.

Our large sample enabled a view of within-group subjective reports given a standard 19 mg dose of psilocybin. 11D-ASC and MEQ30 scores illustrate substantial group-level intensity across key scales alongside wide individual variability (Fig. 4c–f). Subscales were grouped to reflect key conceptual distinctions in psychedelic effects (colour-coded in Fig. 4c). Positively felt self-dissolving and boundary-dissolving effects such as bliss and unity received higher ratings across our sample than cognitive insights or spiritual experiences. Negative effects, particularly anxiety, were infrequent or rated as low.

These subscales were analysed post hoc to assess patterns of covariation⁶⁰, identifying groups of subscales that showed high intercorrelations in our sample (Fig. 4g and Extended Data Fig. 9).

The Life Attitudes Profile Revised⁶¹, used to assess death acceptance and changes in personal meaning, and Nature Relatedness Scale⁶² were measured before and 1 month after psilocybin administration. Results showed group-level improvements in death acceptance (t(54) = −3.66, P = 0.0006, d = 0.49), personal meaning (t(54) = −3.77, P = 0.0004, d = 0.51) and nature relatedness (t(55) = −3.37, P = 0.0014, d = 0.26), alongside individual variability (Supplementary Figs. 5 and 6).

Intensity scores on blissful state and unity matched those of higher-dose studies (300–400 μg kg⁻¹)⁶³, suggesting that our supportive context (space design, minimized interruptions, immersive music and inward focus) facilitated the depth of the subjective experience.

Acute experience tracks mindset change

We examined the relationship between the post-psilocybin mindset change score (measured 1 day after administration; see Supplementary Information) for each participant and their subjective experience scores from the 11D-ASC scale and the MEQ30. Mindset change, assessed across dimensions such as connection to self, others and nature, as well as patience, harmony and inner peace, correlated moderately with associative dimensions of the psychedelic experience. Insightfulness showed the strongest correlation (r = 0.65, P = 9.9 × 10⁻⁸), followed by mystical, positive, blissful and spiritual sub-dimensions (r = 0.52–0.60, all P < 10⁻⁴). Sensory–hallucinogenic effects were less correlated (r = 0.18–0.46), with four of five subscales reaching significance (P < 0.05). Negative experiences showed minimal associations (r < 0.1, not statistically significant; Fig. 4h). These findings suggest that positively felt associative dimensions (self-dissolving and boundary-dissolving effects) are the primary drivers of psychological shifts, with cognitive insights and transcendent states, such as unitive, blissful, mystical and spiritual experiences, contributing more to mindset change than sensory–hallucinogenic or negative effects (anxiety and impaired control and cognition). The observed relationship between subjective experience and psychological shifts provides empirical context for ongoing debates about the necessity of the psychedelic experience for clinical efficacy^64,65,66.

Context directs hippocampus–DMN coupling

Dynamic causal modelling (DCM)^67,68,69 estimated effective connectivity between the aHip and core DMN nodes across rest, meditation, music and eyes-open movie, contrasting psilocybin and baseline (no psilocybin) conditions (Fig. 5). Group parametric empirical Bayes contrasts revealed condition-specific changes in effective connectivity. Eyes-closed conditions showed modest changes, whereas the movie condition exhibited the largest number and magnitude of changes. The total strength of between-region parameter changes (sum of absolute changes with posterior probability ≥ 0.99) was: rest ≈ 0.75 Hz, meditation ≈ 0.66 Hz, music ≈ 0.73 Hz and movie ≈ 1.48 Hz (largest). The greater magnitude of effective connectivity changes during movie indicates context-dependent reorganization within the DMN. See Supplementary Tables 2–5 for parameter estimates and credible intervals, and Supplementary Information for condition-level interpretation.

The brain regions included: left aHip (laHip), right aHip (raHip), left inferior parietal cortex (lIPC), right inferior parietal cortex (rIPC), medial prefrontal cortex (mPFC) and posterior cingulate cortex (PCC). Scan sequences (from top to bottom): rest, meditation, music and movie watching. The model was specified as fully connected, allowing all possible causal interactions between regions. a, Spatial mapping of ROIs with estimated changes in effective connectivity. Red shades denote increases and blue shades denote decreases (Hz), based on posterior means averaged across participants using Bayesian model averaging. b, Mean effective connectivity at baseline. For mean effective connectivity, the warm colours represent excitation and the cool colours represent inhibition. c, Matrix of effective connectivity changes from baseline to psilocybin, corresponding to panel a. d, Mean effective connectivity under psilocybin. Connection strengths (posterior expectations) are reported in Hz. Self-connections are not log-scaled. See Supplementary Information for tables of posterior expectations and credible intervals, and further details. Displayed connections have a posterior probability > 0.99, indicating very strong evidence. Illustration of brain in panel a by Kristina Bulgakova/iStock by Getty Images.

Meditation training and music context

We assigned half of the participants to an 8-week mindfulness-based cognitive therapy programme (Extended Data Fig. 1; see Methods). No statistically significant differences emerged between meditators and non-meditators in acute psilocybin effects (11D-ASC and MEQ30) or connectivity metrics under psilocybin (for example, functional modularity; Extended Data Fig. 10). Previous reports of meditation–psilocybin synergy have typically involved experienced practitioners in intensive retreat-based practice^70,71, whereas brief structured training did not produce detectable group-level effects in analyses of our sample.

Music, a contextual element known to shape subjective experiences under psychedelics⁷², had a central role in the study design (see Supplementary Information). Participants rated the ethereal music as significantly more emotionally resonant and engaging under psilocybin during MRI and EEG sequences (fMRI: +16%, Cohen’s d = 0.67, P = 6.3 × 10⁻⁴; EEG: +21%, Cohen’s d = 1.08, P = 3.2 × 10⁻⁷). This was independently confirmed across the administration day on the Aesthetic Experience Scale–Music (AES-M; +25%, Cohen’s d = 0.75, P = 2.4 × 10⁻⁴; Fig. 4i; see Methods for details). Moreover, TAVRNN embeddings showed the greatest global network organization during music (Fig. 2e).

Context modulates power and complexity

Context-sensitive psilocybin-induced brain dynamics observed using MRI were confirmed using 64-channel wet EEG, in both the power spectrum and the signal complexity. Our findings extend preclinical and human observations of desynchronized local brain activity under psilocybin^5,73 by demonstrating that these alterations are modulated by sensory context. We found that psilocybin expanded the power spectrum, with decreases in theta, alpha and, to a lesser extent, beta power, accompanied by modest gamma increases during eyes-closed conditions, concentrated in frontal regions. These became attenuated during movie, with increases localized to early visual areas, indicating condition-dependent topography (Fig. 6a). This attenuation in power and complexity during movie, relative to eyes-closed conditions, is consistent with reduced network-level disruption during externally directed attention, as previously shown using fMRI functional connectivity⁶. Meditation and music recordings aligned in EEG, showing similar power profiles across frequencies at baseline and under psilocybin, converging despite distinct stimulus properties, whereas rest modestly differed (Fig. 6b). Psilocybin also reduced the difference between eyes-open and eyes-closed alpha power by 48% (Cohen’s d = −0.79, Mann–Whitney U-test right-tailed P < 10⁻¹⁹), suggesting an integration of internally and externally focused neural processing (Fig. 6b,c and Supplementary Fig. 7).

a, Spatial difference in power (psilocybin minus baseline; dB units) for each frequency band. b, Group-averaged power spectra for each condition at baseline (top) and under psilocybin (bottom). Under psilocybin, all eyes-closed conditions (rest, meditation and music) showed broad-band reductions in theta, alpha and beta power (meditation and music overlap; rest modestly differed), whereas the eyes-open movie spectrum was relatively unchanged. c, Group-averaged power spectra plotted separately for each condition, to compare psilocybin versus no psilocybin. d, Lempel–Ziv complexity (psilocybin minus baseline) showed similar values across eyes-closed conditions and reduced complexity during the eyes-open movie.

Signal diversity, quantified via Lempel–Ziv complexity, was highest in eyes-closed conditions, consistent with recent findings⁴⁴, suggesting that internally generated perceptions are associated with increased signal complexity (Fig. 6d). Alpha-band activity (8–12 Hz), typically associated with suppressing visual input during eyes-closed conditions^74,75, decreased under psilocybin, particularly in visual regions. This reduced sensory filtering, combined with changes in MRI functional connectivity patterns, suggests diminished distinctions between eyes-open and eyes-closed conditions. This convergence, observed in both fMRI and EEG, is unlikely to reflect lower vigilance during the eyes-closed conditions, given robust baseline eyes-closed alpha and elevated eyes-closed signal complexity under psilocybin (see Supplementary Information for vigilance considerations).

Comparable decreases in alpha power and increases in signal diversity have also been observed under intravenous administration of the serotonergic psychedelic N,N-dimethyltryptamine⁴³, suggesting that these may be conserved markers of serotonergic psychedelic action, despite differences in compound, route of administration, timing and EEG methodology. The convergence of the MRI and EEG findings confirms that this effect cannot be explained solely by vascular changes caused by psychedelics¹. Together, these results establish context alignment as a property of brain dynamics under psilocybin that is evident across imaging modalities, timescales and analytical approaches.

Embeddedness as brain–mind continuity

Machine learning embeddings revealed an association between the positively felt experience of self-dissolution and boundary dissolution and the reorganization of brain networks into context-aligned clusters (Fig. 2b,e). The relative activation of these networks regulates the balance between internally (for example, DMN) and externally directed processing, a dynamic that shapes perception and cognition^33,76. Their increased integration under psilocybin suggests a shift towards more flexible connectivity and altered functional interactions (Fig. 2e).

The correlation between self-dissolving and boundary-dissolving effects and classification accuracy based on machine learning embeddings (Fig. 2c) indicates that these embeddings capture a connectivity state in which internal and external processes become less distinct. This is consistent with the reports of participants of feeling integrally part of a broader physical and psychological relational context. We refer to this state as embeddedness, the phenomenological correlate of context alignment. Unlike connectedness, which implies links between separate entities, embeddedness is the subjective experience of being continuous with, rather than separate from, the environment, which emerges when brain dynamics become aligned with context in proportion to the depth of self-dissolving and boundary-dissolving effects. Psychedelics appear to facilitate this state—a perceptual and cognitive realization of being part of a unified whole—by reorganizing brain network dynamics to increase integration within and between networks (Fig. 2e). The context sensitivity of this reorganization—its dependence on what the participant is doing and experiencing during psilocybin administration—suggests that the underlying synaptic changes may themselves be shaped by ongoing brain activity.

Although embeddedness refers to an experiential state, CEBRA and TAVRNN are machine learning methods that learn embedding spaces from imaging data. Although the experiential construct and the computational representations are conceptually distinct, the context-organized clustering of low-dimensional coordinate vectors within the learned embedding spaces covaried with the intensity of subjective effects (Figs. 2b,e and 3a,b), and classification accuracy covaried specifically with positively felt, self-dissolving and boundary-dissolving effects (for example, spiritual, blissful and unitive; Fig. 2c), with weaker associations for sensory–hallucinogenic effects and none for negative effects (for example, impaired control and cognition, and anxiety).

This pattern supports interpreting embeddedness as a state central to the therapeutic relevance of psychedelics, potentially by addressing distress arising from separation. We used separation to mean the pervasive, felt disconnection between self and world (including self–other boundaries and rigid self-referential attachment). Interpreted this way, embeddedness marks a transient reduction of that separation, an existentially salient state that may help to explain reported reductions in death-related anxiety and existential distress following psychedelic treatment^3,4,77, and, in our non-clinical cohort, is consistent with self-dissolving and boundary-dissolving experiences (Fig. 4c–g and Extended Data Fig. 9), increased personal meaning and death acceptance, and greater nature-relatedness (see Supplementary Information).

Furthermore, the link between positively felt associative effects and improved mindset (Fig. 4h) was replicated using CEBRA embeddings, where classification accuracy was associated with next-day mindset change (r = 0.4, P < 0.01; Extended Data Fig. 6). Although this association is probably mediated by subjective experience, it supports the view that brain embeddings capture meaningful experiential features of the psychedelic state that relate to subsequent psychological change. This suggests that brain-derived embeddings may offer a complementary approach for tracking participant-specific correlates of change where self-report is unavailable or unreliable.

The blissfully felt state of boundary-dissolving embeddedness may reflect a general mechanism of psychological adaptation, one that opens a sense of boundlessness and remediates separateness. Under supportive conditions, this can enable personally meaningful acute experiences that translate into persisting psychological benefits⁷⁸. This interpretation is consistent with therapeutic efficacy observed in depression⁹, and with reports from participants of communitas, empathy and nature-relatedness, reflecting an attenuation of self–world boundaries characteristic of embeddedness^12,13,14,79. Embeddedness thus offers a construct that may inform clinical approaches to mental health, by characterizing the extent to which the brain and mind of a person are aligned with context under pharmacologically altered experience.

The learned embedding spaces point to a deeper principle. Ordinarily, segregation between internally and externally directed systems buffers brain dynamics from direct environmental coupling, maintaining the separation between internal models and sensory context on which predictive processing depends⁸⁰. Under psilocybin, this boundary dissolves. Rather than maintaining separation from context, brain activity differentiates more coherently across contexts and becomes temporally coherent within each context (Fig. 2a,e).

These results address a key gap in systems neuroscience: how large-scale functional brain embeddings can shift under multiple contextual demands in a pharmacologically altered state. Network perturbation analysis grounds this organization anatomically, identifying the DMN and visual network as its primary contributors (Fig. 3g–i). This anatomical anchoring is at the level of large-scale functional systems, not lower-level neurophysiology, and does not assign unique network generators to individual contexts (Extended Data Fig. 7b). Context alignment is a distributed, temporal property of whole-brain dynamics, and the perturbation analysis identifies which systems’ dynamics are necessary for that organization to emerge, providing a neurobiological account at the grain that whole-brain fMRI can rigorously address, and a concrete entry point for finer-grained mechanistic investigations to follow. These results also position embeddedness as a data-driven construct that bridges measurable organization of brain activity with both acute and enduring effects, providing an alternative to accounts of psychedelic experience that are difficult to operationalize, such as ego dissolution⁸¹.

Crucially, machine learning embedding techniques show that the prevailing interpretation of psychedelic resting-state network integration–segregation dynamics, commonly summarized by modularity changes, is incomplete. Reductions in within-network functional connectivity, often interpreted as a loss of within-network organization, coexisted with tighter clustering of regional dynamics in the TAVRNN embedding (Fig. 2e), and modularity did not predict the individual differences in subjective experience or mindset change that CEBRA classification accuracy predicted, indicating that experientially meaningful organization resides within the temporal dynamics that time-averaged summaries do not preserve.

By revealing structured organization aligned with context where previous studies found desynchrony^5,6, these embeddings recast the psychedelic state: apparent disorder in time-averaged measures masks dynamic organization that emerges in proportion to subjective experience. This view is now supported at the circuit level by preclinical evidence that psilocybin selectively weakens corticocortical recurrent pathways while strengthening specific feedforward routes from perceptual and medial regions (the latter a rodent DMN homologue) in an activity-dependent manner⁸².

Although further research is needed to determine whether machine learning embedding signatures generalize across different populations, CEBRA demonstrates how dynamic functional trajectories can uncover structured organization missed by static connectivity approaches, with broad applications in consciousness research and psychiatry.

Brain activity aligns with context

Our analyses converge on a redistribution of integration under psilocybin, which provides the systems-level basis for how context shapes brain dynamics during the psychedelic state. Time-averaged connectivity (fMRI) and spectral power (EEG) confirmed this redistribution across modalities, whereas temporally resolved trajectory analysis revealed that context increasingly differentiated brain dynamics as subjective effects intensified (Fig. 3a–d), indicating that context alignment is graded by experiential depth rather than imposed by external stimulation alone.

The GFC analysis demonstrated that psilocybin rebalanced sensory and associative connectivity. During eyes-closed conditions, sensory integration decreased, whereas associative integration increased, consistent with a redistribution across the cortical hierarchy.

Our larger sample helps to clarify discrepancies in earlier reports of GFC patterns from serotonergic psychedelic studies, in which spatial patterns varied across datasets and global signal regression^43,83,84. Further analysis using histograms of GFC values revealed that psilocybin reduced the separation between eyes-open and eyes-closed states, both globally and across individual networks, and was particularly evident in the visual network (Fig. 1d). Similar sensory region degree centrality reductions have been reported under lysergic acid diethylamide (LSD), 3,4-methylenedioxymethamphetamine (MDMA) and d-amphetamine⁴⁸, indicating that sensory global connectivity decreases can occur outside the class of serotonergic psychedelics.

Psilocybin also redistributed BOLD signal variability across sensory and associative regions. During eyes-closed states, variance increased in ventral, temporal and somatosensory areas and decreased in the early visual cortex, consistent with the redistribution of integration between sensory and associative regions observed in our GFC findings. These changes align with previous evidence of increased entropy under psychedelics, measured as magnetoencephalography–EEG signal diversity, fMRI sample entropy and model-derived neuronal firing rate entropy^{31,85,86,87,88}, demonstrating how psilocybin reshapes spatial signal dynamics in a context-sensitive manner.

The 64-channel wet EEG provided a separate modality that confirmed the pattern of context-sensitive changes that we identified in fMRI. The reduced alpha-band inhibition observed under psilocybin during eyes-closed states suggests a mechanism for the spontaneous production of internally generated visual effects⁷⁴. Lempel–Ziv complexity increased brain-wide during eyes-closed conditions, extending previous evidence of these effects in occipital–parietal regions⁸⁵. Meditation and music show visually similar group-averaged power spectra at baseline and under psilocybin (Fig. 6b); owing to our fixed-order design, we have reported these findings descriptively (see Methods). Similar reductions in signal diversity (Lempel–Ziv complexity) under external stimulation have also been reported under psychedelics⁴⁴.

Therefore, although the spatial organization of connectivity (fMRI surface maps) and brain dynamics (EEG scalp topographies) remains distinct from baseline (no psilocybin), the overall distributions of connectivity strength (GFC values; Fig. 1d,e) and spectral power (EEG power; Fig. 6b) converge under psilocybin, indicating that eyes-closed states shift towards a more externally engaged profile while preserving condition-specific topology. This pattern reflects a redistribution of integration in which associative networks participate more broadly during eyes-closed states while sensory networks become less constrained by baseline organization. The convergence of sensory state boundaries in time-averaged functional connectivity (Fig. 1b,d) is not a loss of structure but a reorganization compatible with the context-aligned organization captured by low-dimensional trajectory analysis (Fig. 2a). As rigid distinctions between eyes-open and eyes-closed states relax, temporally ordered dynamics become more distinctly locked to each context.

DCM analysis also linked these findings to emerging preclinical and human evidence of aHip–DMN neuroplasticity following psilocybin⁶. Using DCM, we demonstrated that eyes-open versus eyes-closed context differentially tunes the effective connectivity of these brain circuits during the acute effects, with the eyes-open movie showing the greatest magnitude of directed change, in line with distinctions in our fMRI and EEG findings.

Machine learning embeddings distinguished patterns of psychedelic brain activity across experimental contexts (rest, meditation, music and movie), with classification accuracy scaling with the intensity of acute subjective experience—particularly positively felt, immersive self-dissolving and boundary-dissolving dimensions (for example, mystical, blissful and unitive). Together, increased integration of cortical–subcortical connectivity—including interactions between systems that ordinarily segregate internal (self-referential) and external (environment-focused) processing—modelled with TAVRNN, and the context separability uncovered by CEBRA, converge to support the construct of embeddedness: a state in which individuals feel fundamentally continuous with their environment, marked by diminished self–world separation. These signatures were detectable at the individual level and associated with next-day mindset improvements, providing a brain-derived marker of large-scale brain reorganization that connects the quality of subjective experience to the change that follows.

Our findings extend prevailing accounts that characterize the psychedelic state as desynchronized or entropically disordered by revealing structured, context-aligned organization in BOLD signal dynamics—a latent order harboured within temporal dynamics that time-averaging obscures—which emerges in proportion to the depth of self-dissolving and boundary-dissolving experience. This organization is distributed across networks but depends on the joint alteration of DMN and visual systems: the gradient end points that ordinarily segregate internal from external processing. When both are altered, that functional boundary dissolves, setting the conditions for brain activity to align with context, an integration that provides a neurobiological rationale for how structured settings can shape outcomes under psychedelics. That the felt boundary between self and world reorganizes when these dynamics shift implies that the boundary is not a fixed datum of conscious experience but a construction actively maintained by neural activity. Embeddedness—the continuity that emerges when this construction relaxes—may thus reveal less about what psychedelics add to consciousness than about what ordinary neural dynamics keep apart.