This re-analysis of a brain imaging study (n=17) used magnetoencephalography (MEG) in people given LSD or placebo and found that LSD loosened the usual link between brain structure and low-frequency activity, while gamma activity changed in a more selective way. Changes in default-mode network regions were linked to stronger feelings of ego dissolution.
Psychedelics profoundly alter conscious experience, yet how they reshape the relationship between brain anatomy and electrophysiological dynamics remains unclear. Here we use source-localized magnetoencephalography mapped onto connectome harmonics to quantify structure–function coupling in humans under lysergic acid diethylamide (LSD) and placebo. LSD induces a robust decoupling of low-frequency (theta, alpha and beta) activity from anatomical constraints, indicating a global loosening of structure-aligned large-scale dynamics. High-frequency gamma activity shows selective reorganization rather than uniform disruption. Decoupling within core default-mode network regions predicts ego dissolution intensity across individuals, linking frequency-selective DMN reorganization to subjective loss of self. Functional decoding further reveals system-specific rebalancing: visual and attentional systems preferentially decouple while auditory networks exhibit strengthened coupling. Together, these findings provide electrophysiological evidence that psychedelic states emerge from a frequency-dependent relaxation of structural constraints on brain activity and identify default-mode reorganization as a neural correlate of ego dissolution. These results offer a mechanistic framework for understanding how LSD may exert therapeutic effects by transiently relaxing rigid structural constraints and enhancing dynamical flexibility within networks involved in self-related processing.
Subramani and colleagues frame the study around a central question in psychedelic neuroscience: how LSD alters the relationship between brain anatomy and neural dynamics. Earlier work had shown that structure-function coupling can be studied by comparing structural connectomes with functional signals, and fMRI-based connectome harmonic studies had suggested that psychedelics shift brain activity towards more distributed, less structure-bound patterns. However, those findings relied on hemodynamic signals, which may not directly reflect neural activity and can be influenced by vascular effects. The authors therefore note that it remained unclear how LSD reshapes structure-function coupling in direct electrophysiological signals, and whether this varies across frequency bands. The study set out to test whether LSD produces a frequency-dependent reconfiguration of structure-function coupling in human cortex. Specifically, the researchers aimed to map source-localised MEG activity onto connectome harmonics to quantify how closely neural dynamics align with the structural connectome under LSD versus placebo. They hypothesised that LSD would selectively relax anatomical constraints on slower, integrative rhythms while reshaping faster, locally expressed activity, and that these changes would relate to subjective psychedelic experiences such as ego dissolution and complex imagery.
Papers cited by this study that are also in Blossom
Carhart-Harris, R. L., Leech, R., Shanahan, M. et al. · Frontiers in Human Neuroscience (2014)
The researchers analysed MEG data from a previously collected single-blind, within-subject, placebo-controlled study. Seventeen participants received intravenous LSD (75 µg) or placebo in counterbalanced order across two sessions separated by 14 days. MEG recordings were obtained about 4 h after administration during eyes-closed rest, both with music and without music, in 7-minute blocks. The paper states that the original study design was used and that further details were reported elsewhere. MEG preprocessing followed the original pipeline and was expanded with automated artefact detection using Autoreject. The continuous recordings were segmented into non-overlapping 2-second epochs and cleaned with a combination of automated epoch rejection and independent component analysis to reduce physiological and non-physiological artefacts. The authors report that, on average, about 200 clean epochs were retained per recording, with similar retention in LSD and placebo sessions. For source analysis, individual T1-weighted MRI scans were processed with FreeSurfer, and source estimates were computed using dynamic statistical parametric mapping (dSPM) with boundary element forward models. Source time series were morphed to fsaverage and parcellated into 360 cortical regions using the HCP-MMP atlas. Power spectral density was estimated at source level with a multitaper approach, and band-limited analyses covered theta, alpha, beta, low-gamma, mid-gamma and high-gamma ranges. Because no diffusion MRI was available for the MEG dataset, the structural connectome was derived from high-quality diffusion MRI data from the Human Connectome Project S1200 sample (N = 1,063) and averaged into a consensus connectome. Structural weights were based on streamline density between HCP-MMP regions. The authors then applied connectome harmonic decomposition, using the graph Laplacian of the structural network to define spatial eigenmodes ranging from smooth, structure-aligned patterns to more localised, structure-decoupled patterns. MEG activity was projected onto this basis, and the resulting graph power spectral density (gPSD) and Structural Decoupling Index (SDI) were used to quantify structure-function coupling and decoupling. To relate these spatial effects to cognition and experience, the researchers used NiMARE with the Neurosynth meta-analytic database to decode SDI maps into broad cognitive and affective domains. They also examined associations between SDI changes and phenomenological ratings using Spearman correlations in a priori regions and networks, correcting for multiple comparisons with Bonferroni adjustment. Statistical significance for graph power and SDI changes was assessed with cluster-based permutation testing, and the paper distinguished main drug effects from drug-by-music interactions.
At the level of temporal dynamics, LSD produced robust reductions in low-frequency power spanning theta through beta bands, alongside increases in gamma power. The spatial extent of these effects varied by band: theta reductions were broadest and included posterior cortex and default mode network regions, alpha reductions were prominent in visual cortex, posterior cingulate cortex and temporal areas, and beta reductions were more focal around anterior and posterior cingulate cortices. In contrast, low-gamma effects were focal, whereas mid- and high-gamma power increases were widespread and highly similar to each other. The authors report that no frequency band showed a statistically significant drug-by-music interaction for total graph power, although effect-size estimates suggested a possible attenuation of LSD-related changes in the presence of music. When cortical activity was re-expressed in connectome harmonic space, LSD showed significant changes in graph power across all bands except low-gamma. Low-frequency bands showed weakened graph power across a broad range of eigenmodes, while mid- and high-gamma bands showed increased graph power concentrated in low-to-mid eigenmodes. Overall, LSD produced a broadband reduction in total graph power, with the strongest effect in the beta band. The Structural Decoupling Index showed a clear frequency-dependent pattern. In theta, alpha and beta bands, LSD induced decoupling relative to placebo, and these low-frequency maps were highly similar to one another (Spearman's ρ > 0.82, p < 10^-15). In higher frequencies, the pattern was mixed: some regions showed decoupling, while others, especially in temporal cortex, showed stronger coupling. Mid- and high-gamma maps were almost identical (ρ = 0.98, p < 10^-15), indicating a distinct high-frequency pattern. Functional decoding suggested system-specific reorganisation rather than a global shift. Under LSD, visual, executive control, action and attention-related domains tended towards decoupling, whereas auditory, language and emotion-related domains showed stronger coupling, particularly in alpha and high-gamma bands. The auditory system showed especially strong coupling in high-gamma. Exploratory analyses of the drug-by-music interaction suggested reversals in several systems, but these were not robust at the group level. Music alone was associated with decoupling in auditory, emotional and visual cortices. Phenomenology analyses showed several significant associations. Complex imagery was positively correlated with drug-by-music-related decoupling in ventral visual cortex in the low-gamma band (ρ = 0.61, Bonferroni-corrected 95% CI [0.09, 0.86]). Ego dissolution was associated with stronger decoupling in default mode network subnetwork 2, including posterior cingulate cortex and medial prefrontal cortex, in mid- and high-gamma bands (ρ = 0.63 and 0.68, both p < 0.05 after Bonferroni correction). Emotional arousal correlated with stronger decoupling in medial prefrontal cortex in theta and mid-gamma bands (ρ = 0.65 and 0.64), whereas higher positive mood correlated with stronger coupling in auditory association cortex in mid- and high-gamma bands (ρ = -0.65 and -0.70, with the negative sign indicating that greater coupling corresponded to higher mood under the study’s coding).
Subramani and colleagues interpret the findings as evidence that LSD loosens the structural constraints on brain dynamics in a frequency-dependent way. They argue that low-frequency desynchronisation in theta through beta bands is reflected in reduced graph power when MEG signals are projected onto the structural connectome, and that SDI reveals a more specific effect: low-frequency activity becomes decoupled from anatomical architecture, while higher-frequency gamma activity is reorganised more heterogeneously, with both coupling and decoupling depending on region. The authors place these results in the context of earlier connectome harmonic work, which had mainly used fMRI and reported shifts from smooth to more distributed spatial modes under psychedelics. They emphasise that MEG offers direct neural measurement and avoids ambiguity from vascular or metabolic effects. On that basis, they suggest that LSD-induced decoupling of slow rhythms may reflect weakened large-scale coordination and an expanded repertoire of accessible brain states. For gamma-band effects, they link the results to hierarchical models of cortical processing and the REBUS framework, proposing that psychedelics relax top-down constraints while relatively preserving, or in some regions enhancing, structure-aligned faster dynamics that may support bottom-up sensory processing. The authors also interpret the functional decoding as evidence against a uniform account of psychedelic “disorganisation”. Instead, they describe a system-specific rebalancing in which visual and attention/executive systems tend towards decoupling, whereas auditory and some affective/language-related systems show stronger coupling. They note that this pattern is consistent with prior work linking psychedelic visual phenomena to reduced occipital alpha power and altered visual connectivity. Regarding subjective experience, the authors argue that the correlations with phenomenology support the behavioural relevance of the SFC changes. They highlight the association between default mode network decoupling and ego dissolution, the link between ventral visual decoupling and complex imagery, and the relationships between prefrontal decoupling or auditory coupling and emotional arousal or positive mood. They caution, however, that the correlations were based on targeted regions and networks, with wide bootstrap confidence intervals, so they should be interpreted carefully. The main limitations they acknowledge are that the SDI analysis combined periodic and aperiodic components of the MEG signal, making it impossible to attribute the effects exclusively to oscillatory dynamics; that the sample size of 17 limits precision and generalisability; and that retrospective phenomenology ratings, which were not fully time-aligned with the neural recordings, likely reduced sensitivity. They suggest that future work should separate periodic and aperiodic spectral components, replicate the findings in larger samples, and collect denser, temporally aligned subjective reports.
The authors conclude that LSD alters structure-function organisation in the human brain in a frequency- and system-specific manner. In their view, low-frequency activity becomes broadly decoupled from the structural connectome, whereas higher-frequency dynamics are selectively reorganised rather than uniformly disrupted. They further conclude that these neural changes are related to subjective psychedelic experiences, including ego dissolution, imagery and affect, and that connectome harmonic analysis of MEG offers a useful way to study how psychedelics reshape the neural basis of consciousness.
We analyzed MEG data collected by Carhart-Harris and co.. The original study followed a single-blind, within-subject, placebo-controlled setup in which 17 participants received intravenous LSD (75 µg) or placebo in counterbalanced order across two sessions separated by 14 days. MEG was acquired 4 h post-administration during eyes-closed rest with (Music) or without music (NoMusic; 7 min per condition). Further details are reported elsewhere.
MEG preprocessing followed the pipeline of the original dataset, augmented with automated artefact detection using Autoreject(Supp. Fig.). Continuous recordings were epoched into non-overlapping 2-s segments and cleaned using a combination of Autoreject and independent component analysis to attenuate physiological and non-physiological artefacts (See Supplementary Material for details).
Source localization was performed using individual T1-weighted MRI scans processed with FreeSurfer. Forward models were computed using boundary element method, and source estimates were obtained using dynamic statistical parametric mapping (dSPM). Individual source estimates were morphed to a common template (fsaverage) and parcellated into 360 cortical regions using the HCP-MMP atlas. Source-level power spectral density was estimated using a multitaper approach and served as the basis for the spectral analysis. Structure-function coupling was quantified for the bandlimited activity (butterworth filter, order of 4) for theta (4-8 Hz) through high-gamma (90-120 Hz) range. Full source reconstruction parameters are reported in the Supplementary Methods.
Because diffusion-weighted imaging was not available for the MEG dataset, we derived the structural connectome from high-quality diffusion MRI data provided by the Human Connectome Project (HCP S1200 release; N = 1,063). Individual connectivity matrices were reconstructed and averaged to obtain a consensus connectome. Edge weights were defined as streamline density between pairs of cortical regions parcellated according to the HCP-MMP atlas, matching the atlas used for source-localized MEG activity. To obtain a group-level connectome suitable for our analyses, individual connectivity matrices were averaged across participants to form a consensus structural connectome (N = 1,063) (Fig.). Full tractography and reconstruction details are provided in the Supplementary Methods.
To characterize how electrophysiological activity aligns with anatomical structure, we combined source-localized MEG with connectome harmonic analysis (Fig.). A consensus structural connectome derived from diffusion MRI defines an intrinsic set of spatial harmonics spanning smooth, structure-aligned patterns to increasingly localized patterns that deviate from large-scale anatomical constraints (Fig.). Source-level MEG activity was projected onto this harmonic basis, allowing cortical dynamics to be expressed as a weighted combination of structure-coupled and structure-decoupled components (Fig.). We quantified the relative dominance of these components using the Structural Decoupling Index (SDI), which captures the ratio between activity aligned with the structural connectome and activity expressed in higher-order, spatially localized modes (Fig 1D ). Following, binary log is subsequently applied, resulting in the negative and positive tail corresponding to coupling and decoupling respectively. This framework provides an anatomically grounded measure of structure-function coupling that is directly applicable to fast electrophysiological signals. Full mathematical definitions and implementation details are provided in the Supplementary Methods. Consistent with prior studies under normal wakeful state, we computed SDI on broadband cortical signals, without separating periodic oscillatory activity from the aperiodic (1/f -like) component. This allows for a direct comparison of electrophysiological SFC across normal wakeful state and psychedelics-induced altered states.
To interpret LSD-induced changes in structure-function coupling at the systems level, we investigated the functional correlates of SDI contrast maps using NiMARE, following established approaches in prior work. Unthresholded SDI maps were compared against the Neurosynth meta-analytic database to identify large-scale cognitive and affective domains associated with regions exhibiting the strongest coupling and decoupling effects under LSD. This approach enables a systems-level interpretation of structure-function reorganization by relating spatial patterns of SDI change to distributed functional domains (e.g., perception, attention, emotion, language), rather than to individual tasks or regions. Full decoding procedures and statistical thresholds are detailed in the Supplementary Methods.
We examined the correspondence between changes in SDI and phenomenological scores using Spearman's correlation coefficient ρ. To this end, the regression analysis were performed within a priori regions or functional networks, selected based on previous studies. Multiple comparisons were controlled using Bonferroni correction across number of spatial tests. The 95% CI was established using bias-corrected bootstrappingtechnique (N = 10,000) implemented in scipy.stats.bootstrap.
We employed cluster-based permutation testing (mne.stats.permutation_cluster_test) to assess the statistical significance of changes in graph power and SDI. For gPSD, clusters were defined along the one-dimensional eigenmode axis, reflecting adjacency in graph-frequency space. For SDI, clusters were defined based on spatial adjacency among HCP-MMP parcels. Cluster-level significance was evaluated at α = 0.05. The Main Effect of Drug and Music assessed the impact of each factor independently of the other. The Interaction effect (Drug × Music) tested whether the effect of LSD relative to placebo differed as a function of music presence.
Using connectome harmonics as the basis of brain activity, we characterize how LSD reorganizes the structural constraints on brain activity, mapping these shifts to large-scale functional systems and subjective phenomenology. A Anatomical MRI and diffusion-weighted imaging are used to segment white and gray matter and to reconstruct an anatomically constrained tractography. Streamline densities between cortical regions defined by the HCP-MMP atlas are used to build a group-level structural connectome. Source-localized MEG provides time-resolved cortical activity S t . B Its harmonic modes ψ k are obtained via eigendecomposition of the graph Laplacian (See Supplementary). Low-order eigenmodes capture spatially smooth, global patterns constrained by anatomy, whereas high-order modes capture increasingly localized patterns weakly constrained by structure. C Cortical MEG activity is decomposed at each time point as a weighted linear combination of connectome harmonic modes, yielding a representation of functional activity in the graph spectral domain. The resulting graph power spectral density (gPSD) quantifies the contribution of each harmonic mode over time. Graph-domain filtering (median split of the gPSD) separates structure-coupled and structure-decoupled components of the signal. D The relative balance between decoupled and coupled components defines the Structural-Decoupling Index (SDI), which is computed across experimental conditions (LSD vs placebo; music vs no-music) to quantify changes in structure-function coupling (See Supplementary). MEG system image credits to CTF MEG Neuro Innovations, Inc.
LSD-induced modulations (i.e. Main Effect Drug) in the spatial organization of cortical activity were revealed by mapping the source-localized cortical activity onto the structural connectome (Fig.). The resulting graph power (gPSD) describes the distribution of the weights of each mode along the spatial scales ranging from smooth, large-scale patterns (low eigenvalues) to more spatially localized patterns (high eigenvalues). Significant changes (p < 0.05, cluster-corrected, N = 10,000 permutations) in graph power were observed across all frequency bands except low-gamma (Fig.). Significantly weakened graph power was observed in theta through beta range across a broad range of eigenmodes (Figfor the spatial topography). In contrast, mid-and high-gamma bands showed increased graph power, with effects concentrated in low-to mid-range eigenmodes, corresponding to large-scale and intermediate spatial patterns along the structural connectome. To summarize these effects, we computed total graph power by summing gPSD across all eigenmodes. Total graph power was examined for the main effect of drug and the drug × music interaction using paired t-tests, with effect sizes visualized using Cohen's d (Fig.). The main drug effect condition revealed a broadband modulation of total graph power, with the strongest effect observed in the beta band. No frequency band exhibited a statistically significant drug × music interaction. Nevertheless, effect-size estimates showed a consistent reversal in the sign of Cohen's d across several frequency bands, suggesting a systematic attenuation of LSD-induced power changes in the presence of music, although this trend did not reach statistical significance.
With the global reduction in the graph power observed, characterizing the relative proportion of the graph power between localized eigenmodes and smooth eigenmodes revealed the SFC organization. Group-level maps (p < 0.05, cluster-corrected, N = 10, 000 permutations) for the main effect of drug are shown in Fig.. Positive values (red) indicate LSD-induced decoupling of functional activity from the underlying structural architecture relative to PLA, whereas negative values (blue) indicate stronger structure-function coupling under LSD. In the low-frequency range (theta-beta) LSD induced decoupling (relative to PLA), with effects that were spatially focal yet consistent across bands. In contrast, higher frequencies exhibit a more heterogeneous pattern, with both significant coupling and decoupling observed depending on cortical region. Quantifying the spatial similarity across frequency bands using Spearman's ρ revealed a strong correspondence among theta, alpha, beta maps (ρ > 0.82, p < 10 -15 ; Supplementary figure), shedding light on a coherent low-frequency mode of LSD-induced decoupling. In the high-frequency range, alongside regions of decoupling, we observed increases in structure-function coupling, most prominently localized to temporal cortical areas. Spatial similarity analysis revealed mid-and high-gamma were near identical (ρ = 0.98, p < 10 -15 ), highlighting a distinct and consistent high-frequency mode of structural reorganization. Low-gamma occupied an intermediate position, showing moderate similarity to both low-
To assess the functional relevance of LSD-induced reorganization in SFC, we mapped the SFC changes onto large-scale cognitive and affective domains using NiMARE (Fig.). This analysis focused on the main effect of drug (Fig.). In an exploratory analysis, we additionally examined the drug × music interaction (Supp. Fig.), and the main effect of music (Supp. Fig.), enabling assessment of how distinct experimental factors modulate SFC organization and its functional relevance. Functional domains associated with LSD-induced SDI changes are summarized in Fig.. The horizontal axis represents a gradient from stronger structure-function coupling (left) to stronger decoupling (right) under LSD relative to PLA. Across alpha and high-gamma bands, the spatial topography of effects was consistent in some functional systems but diverged in others. Notably, the auditory system exhibited stronger coupling most prominently in the high-gamma band. Furthermore, LSD-induced strengthening of coupling within auditory, language, and emotion domains was more pronounced in alpha band than in high-gamma band. In contrast, across both frequency bands, domains associated with visual processing, executive control, action, and attention are preferentially positioned toward the decoupling end of the gradient. Exploratory analyses of the drug × music interaction revealed a marked reversal in several functional systems (Supplementary Fig.). Functional domains that exhibited LSD-induced structure-function coupling in the main drug effect shifted toward decoupling in the presence of music, with the strongest reversals observed in auditory and emotion-related systems. Conversely, functional systems that exhibited prominent LSD-induced decoupling, such as visual and attention-related domains, shifted toward stronger coupling when LSD was administered in conjunction with music. Finally, assessing the functional correlates of main effect of music independent of drug state (Music-NoMusic SDI contrasts; Supplementary Fig.) revealed that music was associated with structure-function decoupling in auditory, emotional, and visual cortices, indicating reduced alignment with underlying structural constraints during music listening irrespective of pharmacological condition.
The relationship between reorganization in SFC and subjective experience was assessed by computing Spearman's correlation coefficient (ρ) between SDI values and phenomenological dimensions. Four phenomenological dimensions were examined : Complex Imagery, Ego Dissolution, Emotional Arousal and Positive Mood. These associations were evaluated separately for the main effect of drug, and the interaction between drug and music. Figurepresents the significant correlation between changes in SFC and phenomenology (p < 0.05, Bonferroni-corrected for number of spatial tests). Regression analyses were performed within a subset of a priori regions selected based on prior literature. Heatmaps display the For Complex Imagery (Fig.), analyses focused on primary visual cortex (V1), the intraparietal sulcus (IPS), the fusiform gyrus (FFG), and ventral visual cortex (VVC). SDI changes within these regions were examined for correspondence with the reported complex imagery across frequency bands and statistical contrasts. Notably, a strong positive correlation (ρ = 0.61, bonferroni-corrected; 95% CI : [0.09, 0.86]) was observed for the drug and music interaction in VVC within the low-gamma band, indicating that LSD-induced, music-modulated decoupling in ventral visual regions were associated with enhanced complex visual imagery. For Ego Dissolution (Figure), analyses targeted large-scale functional networks defined by the 17-network Yeo-Krienen parcellation, with a focus on the Default Mode Network (DMN) (three subnetworks) and the salience/ventral attention network (two subnetworks). Stronger decoupling within the DMN, subnetwork 2 -encompassing the Posterior Cingulate Cortex (PCC) and medial prefrontal cortex (mPFC) -showed strong correlations with ego dissolution in the mid-and high-gamma bands under the main effect of drug (ρ = 0.63, 95% CI : [0.16, 0.87] and 0.68, 95% CI : [0.26, 0.88], p < 0.05, bonferroni-corrected). These associations indicate that LSD-induced decoupling within core DMN regions is robustly associated with the intensity of ego dissolution. For Emotional arousal and Positive Mood (Figure-D), analyses targeted auditory and prefrontal regions, including primary auditory cortex (AudCortex), auditory complexes 4 and 5 (AudCmplx 4/5), superior temporal gyrus (STG), medial prefrontal cortex (mPFC), and orbitofrontal cortex (OFC). LSD-induced changes in SDI were associated with both emotional arousal and positive mood. Specifically, stronger decoupling in the mPFC was associated with increased emotional arousal across theta (ρ = 0.65, 95% CI : [0.14, 0.90], p < 0.05, bonferroni-corrected) and mid-gamma bands (ρ = 0.64, 95% CI : [0.23, 0.89], p < 0.05, bonferroni-corrected). In contrast, increased coupling within AudCmplx 4 correlated with higher positive mood in the mid-(ρ = -0.65, 95% CI : [-0.86, -0.25], p < 0.05, bonferroni-corrected) and high-gamma (ρ = -0.70, 95% CI : [-0.90, -0.32], p < 0.05, bonferroni-corrected) ranges.
We utilized connectome harmonics to show that LSD induces robust decoupling of low-frequency activity from anatomical structure, alongside frequency-specific increases in coupling within temporal cortices at higher frequencies. Functional correlates further reveals that these effects are organized in a cognitive-and affective system-specific manner. Finally, inter-individual variability in SFC reorganization relates to subjective experience, including visual imagery, ego dissolution, emotional arousal, and positive mood.
Connectome harmonicsenable direct characterization of structure-function organization across spatial scales. By incorporating indirect, polysynaptic pathways, this approach captures dominant patterns of activity propagation beyond direct anatomical connections. As in most SFC studies, however, the topology considered is restricted to cortico-cortical connectivity. Many studies have utilized structural connectome and its harmonics to investigate SFC primarily in the normal wakeful state. Applications of this framework to psychedelic states have relied largely on fMRI, reporting redistribution of BOLD power from spatially smooth to more distributed eigenmodes. While often interpreted as reflecting increased dynamical flexibility, such findings are difficult to disentangle from vascular and metabolic influences. Here we characterize for the first time how neural activity measured with source-localized MEG reorganizes its alignment with the underlying structural connectome under LSD. Consistent with prior electrophysiological work, LSD produced broadband low-frequency attenuation (theta-beta) and increased gamma power, alongside elevated temporal signal diversity. When projected onto connectome harmonics, this temporal desynchronization manifests as a global reduction in graph power across spatial modes (Fig.). Because the graph Fourier transform preserves signal energy, reduced oscillatory amplitude necessarily entails reduced total graph power. Thus, gPSD attenuation represents the structural-domain expression of classical low-frequency desynchronization. These effects arise from fast electrophysiological dynamics and should not be directly equated with graph-spectral findings derived from BOLD signals. Exploratory analyses suggested that music attenuated LSD-related graph power changes, although the interaction was not statistical significant, mirroring previously reported music-related modulations of spectral and temporal features in the same dataset.
Given that LSD induces robust low-frequency desynchronization (Fig.), a concomitant weakening of graph power (Fig.) is expected. The more informative question is whether LSD merely attenuates the contributions of connectome harmonic modes globally, or selectively reweights how neural activity is distributed across the structural connectome. To address this, we used SDI, which quantifies the relative contribution of localized (higher-order) harmonics over spatially smooth (lower-order) harmonics. Under normal wakeful conditions, fMRI-based SFC follows a unimodal-transmodal hierarchy, paralleling large-scale functional organization. However, recent electrophysiology-derived SFC revealed modality-specific differences: while unimodal cortices show consistent coupling across modalities, electrophysiology-derived SFC show a coupling-decoupling gradient rather than SFC of BOLD typically occupying the decoupling gradient. These distinctions are reinforced by direct comparisons of MEG-and fMRI-derived SFCand are consistent with evidence that BOLD responses may dissociate from underlying oxygen metabolism. Against this backdrop, examining SFC with direct neural measures provides a less ambiguous characterization and access to a broader frequency range. Within this framework, we find that LSD robustly increases decoupling of low-frequency activity, particularly in the theta-beta range. Low-frequency rhythms reflect large-scale synchronous population dynamicsand are thought to support long-range communication, rendering them typically well aligned with anatomical pathways. Disruption of large-scale coordination under LSD would therefore be expected to preferentially affect these rhythms. The observed low-frequency decoupling is consistent with this account and suggests selective weakening of structure-aligned integration. More broadly, theta-beta decoupling reflects altered alignment between functional dynamics and anatomical constraints. Such loosening of structural constraints may contribute to the emergence of atypical brain states characteristic of psychedelic experience. This interpretation aligns with the notion of an expanded repertoire of accessible brain states under psychedelics.
Previous work has demonstrated a largely consistent SFC topography across MEG frequency bands during the normal wakeful state. Under LSD, however, SFC reorganizes in a frequency-selective manner. Low-frequency bands (theta-beta) predominantly exhibit decoupling, whereas high-frequency bands (gamma) display a mixture of coupling and decoupling effects (Fig.), with coupling most prominently localized to temporal cortices. Across-frequency similarity analyses support this clustering, with low-gamma occupying an intermediate, transitional position (Supp. Fig.). This dissociation is broadly consistent with hierarchical models of cortical communication, in which alpha-beta rhythms are associated with top-down predictive control, while gamma activity is associated with bottom-up, stimulus-driven processing via cross-frequency interactions. From this perspective, the REBUS frameworkoffers a useful interpretative context: psychedelics relax high-level priors and top-down constraints, thereby overweighting bottom-up sensory and perceptual processes. Consistent with this account, LSD appears to weaken structure-imposed constraints on slow, integrative coordination mediated by theta-beta rhythms, while preserving-or in selected regions enhancing-structure-aligned coordination in faster gamma dynamics. Although gamma oscillations are often locally generated, evidence suggests they can contribute to long-range communication under specific conditions, rendering such coupling effects physiologically plausible.
The functional relevance of LSD-induced SFC changes was assessed using NiMARE. Analyses focused on the main effect of drug (Fig.; see Supplementary Fig.for the Drug × Music interaction and main effect of music). Robust spatial effects were observed only for the main drug effect (Fig., cluster-corrected), whereas interaction and music effects were nominal (p < 0.05, uncorrected). Because NiMARE operates on unthresholded statistical maps, results for the latter conditions are interpreted cautiously. Functional correlates of the SDI patterns were assessed in the alpha and high-gamma bands, which capture the two dominant frequency regimes identified in the SFC analyses (Supplementary Fig.). Rather than inducing a global shift toward decoupling, LSD produces a selective rebalancing of SFC across cognitive and affective systems, arguing against a unitary account of psychedelics-induced "disorganization" often inferred from BOLD-based studies. Specifically, LSD strengthens SFC within auditory systems while inducing pronounced decoupling within visual systems across both alpha and high-gamma bands. Although this pattern is shared across frequencies, auditory coupling is especially prominent in the high-gamma bands, where effects localize to temporal cortices. In contrast, visual systems consistently occupy the decoupling end of the SDI gradient. This interpretation aligns with prior findings linking LSD-induced visual phenomena to reduced occipital alpha power and increased functional connectivity of primary visual cortex, consistent with disinhibition and expanded influence rather than functional breakdown. In this context, visual decoupling may reflect a relaxation of structural constraints that facilitates internally generated imagery. Supporting this view, mental imagery-related topics are preferentially associated with decoupled regions in the alpha band, consistent with reports linking alpha reductions to both simple and complex imagery under LSD.
To identify neural correlates of psychedelic experience, we examined whether LSD-induced alterations in SFC relate to phenomenological reports. Rather than adopting a whole-brain exploratory approach, we focused on a priori regions and networks consistently implicated in psychedelic and music neuroscience. This targeted strategy reduces multiple-comparisons burden and facilitates interpretation within an established neurobiological framework. However, a wide CI range revealed by bootstrapping analyses warrant a careful interpretation of these findings. Visual imagery under psychedelics has repeatedly been linked to early and ventral visual cortices. Carhart-Harris and colleaguesreported that increased blood flow in primary visual cortex (V1) predicted complex imagery under LSD, while more recent effective-connectivity work under psilocybin implicates a distributed circuit involving early visual cortex, ventral visual regions (including fusiform gyrus), IPS, and inferior frontal gyrus, with imagery related to altered top-down influences. Guided by this literature, we examined whether changes in SFC within these regions tracked imagery ratings. We observed that drug × music-related changes in SDI within VVC were positively associated with complex imagery. Notably, this relationship emerged despite the absence of a statistically significant group-level interaction effect on SDI, indicating that individual differences in music-modulated decoupling scale with imagery vividness even when the interaction is not robust at the group level. Disruptions in the sense of self under psychedelics have been consistently associated with changes in the DMN and salience network. Prior electrophysiological studies further suggest that reductions in low-frequency oscillatory power (delta and alpha) within DMN regions predict ego dissolution. Extending this literature, we show that SDI derived from electrophysiological signals within core DMN regions robustly predicts ego dissolution in the mid-and high-gamma bands. This finding suggests that a reorganization of how neural activity decouples from the underlying structural scaffold of transmodal networks could potentially manifest a neural correlate predictive of ego dissolution. Emotional responses to music and psychedelics have been linked to temporal association cortices, mPFC, and ventromedial prefrontal regions, with emotional arousal and positive mood showing substantial covariance. Anchoring our analyses in this framework, we related SDI changes in auditory and prefrontal regions to emotional arousal and positive mood. We found that stronger decoupling of mPFC tracked increased emotional arousal across theta and mid-gamma bands, whereas stronger coupling within auditory association cortex (Auditory Complex 4) in mid-and high-gamma bands predicted higher positive mood. These dissociable relationships suggest that emotional intensity and affective valence may rely on distinct structure-function regimes, with emotional arousal linked to reduced structural constraint in prefrontal regions and positive mood linked to strengthened coupling in auditory cortices.
We report the limitations here as it may guide future research. First, this study tackles structure-function coupling of cortical activity that includes both rhythmic (periodic) and arrhythmic (aperiodic) activity. LSD induces widespread alterations in the aperiodic component of the power spectrum. Consequently, the observed SFC changes cannot be attributed exclusively to oscillatory dynamics. This distinction is non-trivial: synchronous rhythmic activity and asynchronous aperiodic activity may exhibit fundamentally different relationships to the anatomical scaffold. Disentangling the structure-function coupling of periodic and aperiodic components -potentially by computing SFC separately on parametrized spectral components -represents an important avenue for future work. Such analyses would clarify whether the observed decoupling primarily reflects disrupted large-scale synchrony, altered excitation-inhibition balance, or a combination of both. Second, the sample size of N = 17, while comparable to other controlled psychedelic neuroimaging studies, limits the precision of effect size estimates. Brain-phenomenology correlations in particular should be interpreted as preliminary, as reflected in the wide bootstrap confidence intervals reported. Replication in larger samples remains necessary to establish generalizability and more precisely characterize the magnitude of LSD-induced structure-function reorganization. Finally, the subjective ratings were collected retrospectively and referenced to participants' perceived peak drug effects, whereas in-scanner ratings were acquired approximately 150 minutes after the peak. This mismatch likely introduces the noise and reduces the sensitivity of brain-phenomenology correlation. Future studies would benefit from a denser, temporally aligned sampling of phenomenology -ideally collected concurrently with neural recordings -to better capture dynamic brain-experience coupling and improve statistical power.
By mapping source-localized MEG activity onto connectome harmonics, this study provides evidence for how LSD reshapes structure-function relationships in the human brain. At the level of temporal dynamics, we replicated the canonical signature of the psychedelic state characterized by widespread attenuation in low-frequency bands (theta-beta) alongside increases in gamma power. Re-expressing these temporal signals in the spatial domain using harmonics of the structural connectome revealed a corresponding reduction in graph power across theta-beta and a reversal in mid-and high-gamma. This finding demonstrates how classical desynchronization of temporal dynamics organize in the spatial domain along the anatomical scaffold. Crucially, beyond global power changes, LSD induced a frequency-selective reorganization of structure-function coupling. Low-frequency activity exhibited robust decoupling from the structural connectome, whereas higher frequencies showed a more heterogeneous pattern, including strengthened coupling within temporal cortices alongside focal decoupling elsewhere. Functional correlates of these effects identified using NiMARE revealed a system-level reconfiguration rather than uniform disorganization: LSD preferentially induced decoupling in visual and attention/executive systems while strengthening coupling in auditory and affective/language-related domains. Importantly, these alterations in structure-function organization were meaningfully related to subjective experience. Default-mode decoupling predicted ego dissolution and sensory/prefrontal SDI changes predicted imagery and affective reports. Together, these findings demonstrate that LSD induces decoupling and coupling of structure-function alignment in a system-, frequency-specific manner, and predicts subjective experience. Cluster grid thanks to the BC DRI Group, Calcul québec, and the Digital Research Alliance of Canada, and we really appreciate their continued support. K.J. is supported by funding from the Canada Research Chairs (950-232368) program and a Discovery Grant from the Natural Sciences and Engineering Research Council of Canada (2021-03426). We utilized a subset of the data collected by Carhart-Harris and colleagues. Data collection of this study was supported by Safra Foundation and the Beckley Foundation as part of the Beckley-Imperial research programme, and by supporters of the Walacea.com crowdfunding campaign.
We used OpenAI's ChatGPT 5.2 to improve writing style, fluidity, and brevity. All AI-suggested content was reviewed and integrated into the manuscript when appropriate. The authors take full responsibility for the manuscript content.
Below contain details about the methodology of preprocessing MEG signal, and Structural Connectome reconstruction. Furthermore, a full mathematical description to quantify structure-function coupling is provided. Supporting results guide our interpretation, alongside additional exploratory analysis of main effect music and interaction effect drug x music.
MEG preprocessing followed the same general pipeline as in the original dataset, with the addition of automated artefact detection procedures implemented using the Autoreject package. A schematic overview of the preprocessing workflow is provided in Fig.. In brief, continuous recordings were filtered, visually inspected, and segmented into non-overlapping 2-s epochs. Artefacts were then attenuated through a two-stage procedure combining automated epoch rejection and independent component analysis (ICA). Specifically, an initial pass of Autoreject was used to identify and mask artifactual epochs using data-driven, sensor-specific thresholds prior to ICA decomposition. ICA (Picard algorithm) was then applied to remove components associated with physiological noise, after which cleaned signals were reconstructed by back-projecting the retained components. A second pass of Autoreject was subsequently applied to the reconstructed data to further attenuate residual artefacts, yielding the final cleaned epochs used for all subsequent analyses. Following this pipeline, approximately 200 clean epochs per recording on average (SD = 20) were available for subsequent analyses, with comparable data retention across LSD and PLA conditions (t = -1.69, p < 0.12).
High-resolution T1-weighted anatomical MRI scans were available for all participants and sessions. Structural images were processed using FreeSurfer's standard reconstruction pipeline (recon-all, v7.4.1), which includes motion correction, intensity normalization, skull stripping, and white-gray matter surface segmentation. All reconstructions were visually inspected for accuracy prior to source analysis. Subject-specific cortical surfaces were used to constrain MEG source estimation. Forward models were computed using Boundary Element Model (BEM) surfaces generated with FreeSurfer's watershed algorithm as implemented in MNE-Python. A standard single-layer (inner-skull) BEM was used with conductivity parameter set to 0.3. Individual source spaces were defined using an ico5 tessellation, yielding 10242 vertices per hemisphere. MEG sensor positions were co-registered to individual anatomical images using three fiducial points (nasion, left and right preauricular points), resulting in an affine transformation mapping MEG sensor coordinates to MRI space. Inverse solutions were computed using dynamic Statistical Parametric Mapping (dSPM). Because empty-room noise recordings were unavailable, noise covariance was modeled using an identity matrix, assuming uniform sensor noise. Source orientation constraints were set to be mostly perpendicular to the cortical surface, adopting a loose orientation model (loose = 0.2). Depth weighting was applied to the forward model with a depth prior of 0.8. This procedure yielded source-level time series at each vertex of the cortex. For group-level analyses, individual source estimates were morphed to the FreeSurfer fsaverage (v5) template. Morphed source time series were subsequently parcellated onto the HCP-MMP atlasusing MNE's extract_label_time_course function. Regional activity was obtained by averaging across vertices within each parcel, resulting in 360 cortical region time series per participant and condition. Power Spectral Density (PSD) estimates were computed at source level using multitaper method (implemented in MNE's compute_psd_epochs), followed by performing source space morphing to extract the PSDs in HCP-MMP ROIs. Source inversion was performed on epochs basis (Clean Signal; Fig.). Because these epochs are 2-s windows, performing bandpass filtering risks edge artefacts. We therefore padded zeros of duration 1 second on both side creating 4-s epochs to apply filters (butterworth, order of 4): theta (4 -8Hz), alpha (8 -13 Hz), beta (15 -30 Hz), low-gamma (30 -60 HZ), mid-gamma (60 -90 Hz) and high-gamma (90 -120 Hz).
We derived the structural connectome from high-quality diffusion MRI data provided by the Human Connectome Project (HCP), since diffusion-weighted imaging (DWI) was not available for our dataset. We used the S1200 data release, comprising 1,063 healthy participants with preprocessed diffusion and T1-weighted anatomical images. Diffusion data were processed using a state-of-the-art tractography pipeline implemented in QSIRecon(automated report below):
Reconstruction was performed using QSIRecon 1.1.1.dev0+gaf43da9.d20250414, which is based on Nipype 1.9.1 ([50];; RRID:SCR_002502). A hybrid surface/volume segmentation was created. FreeSurfer outputs were registered to the QSIRecon outputs. Anatomical data for DWI reconstruction T1w-based spatial normalization calculated during preprocessing was used to map atlases from template space into alignment with DWIs. Brain masks from antsBrainExtraction were used in all subsequent reconstruction steps. The following atlases were used in the workflow: the glasser atlas. Cortical parcellations were mapped from template space to DWIs using the T1w-based spatial normalization. MRtrix3 Reconstruction Multi-tissue fiber response functions were estimated using the dhollander algorithm. FODs were estimated via constrained spherical deconvolutionusing an unsupervised multi-tissue method. Reconstruction was done using MRtrix3. FODs were intensity-normalized using mtnormalize. Many internal operations of QSIRecon use Nilearn 0.10.1and Dipy 1.8.0. For more details of the pipeline, see the section corresponding to workflows in QSIRecon's documentation. In line with previous SFC studies, we defined the adjacency matrix of the structural connectome as the density of fibers connecting pairs of HCP-MMP ROIs, same atlas in which sources of the MEG activity are estimated.
The harmonics of the structural connectome are derived using Eigendecomposition. Let us define an undirected weighted graph G =< V, E, W >, where V is a set of N elements called vertices (i.e., 360 HCP-MMP ROIs) and E ⊂ V × V the set of edges connecting unordered pairs of vertices with scalar weights w (i.e., fiber density). The degree matrix which each element is different from zero only if the corresponding edge exist. Following previous studies, the graph Laplacian is defined as L = I -D -1/2 AD -1/2 , with I the identity matrix of order N . Laplacian matrix L is a real symmetric matrix and can be diagonalized as where ψ represents the matrix of eigenvectors and Λ the diagonal matrix of real eigenvalues, sorted in increasing order. The eigenvectors are orthogonal to each other and form a complete basis set, which can be leveraged to re-express any spatiotemporal activity. The eigenvectors of the Laplacian of the structural connectome represent the brain's intrinsic eigenbasis -the fundamental spatial patterns of the connectome. These modes are orthogonal to each other and form a complete basis set, which can be leveraged to re-express any spatiotemporal activity. A subset of these eigenmodes is displayed in Fig.. Functional activity was mapped onto the structural connectome using Graph Fourier Transform (GFT;), which decomposes the MEG activity S t into connectome-derived spatial harmonics: The resulting coefficients Ŝt represent the weights of the contributions of the eigenmodes, describing how neural activity is distributed across connectome harmonics. Squaring these coefficients defines the graph power spectral density (gPSD). The original MEG activity can be reconstructed via the inverse GFT as Using the GFT framework, spatial graph filters were defined to decompose the MEG activity into low graph-frequency components (S coupled ), corresponding to spatially smooth activity patterns aligned with the structural connectome, and high graph-frequency components (S decoupled ), corresponding to more spatially localized activity patterns that deviate from large-scale anatomical constraints (seefor a detailed formulation). Following, these components were separated using a median split of the gPSD, yielding a critical graph frequency C that defines the boundary between low-and high-frequency harmonics. This critical frequency was computed separately for each subject and each temporal frequency band and varied across individuals and spectral bands from theta to gamma (see Supplementary Fig.). The Structural-Decoupling Index (SDI) quantifies the ratio of norm (l1-norm) of decoupled activity over coupled activity, and is defined as Binary log is applied subsequently, resulting in the positive tail corresponding to decoupling and the negative tail coupling.
We performed functional decoding of the SDI contrast maps using NiMARE, following established approaches in prior work. The goal of this analysis is to relate the LSD-induced SFC changes to broad cognitive and affective domains, thereby providing a systems-level interpretation of structure-function reorganization under LSD. Functional decoding was performed using the Neurosynth database, a large-scale automated meta-analytic resource comprising activation maps derived from more than 14,000 neuroimaging studies. Neurosynth organizes the neuroimaging literature into topics, which correspond to latent semantic components extracted from the co-occurrence of terms in article abstracts and their associated activation patterns. Each topic thus represents a broad cognitive or affective construct (e.g., perception, memory, emotion, language), rather than a specific task or anatomical region. For each SDI contrast (LSD vs. PLA), unthresholded regional SDI maps were rank-ordered and segmented into ten equally sized deciles (10% increments). Each decile was binarized and submitted to the ROIAssociationDecoder implemented in NiMARE, which quantifies the spatial correspondence between the input SDI masks and Neurosynth topic maps. The resulting correlation coefficients reflect the degree to which LSD-induced SDI changes preferentially overlap with brain regions associated with each cognitive or affective topic. Correlation values were converted to z-statistics, and only associations surviving a threshold of p < 0.001 were retained. Following previous work, topic terms were ranked based on the weighted average of their z-scores across deciles, highlighting the cognitive and affective domains most strongly associated with regions exhibiting the largest LSD-induced alterations in SFC.
LSD-induced modulations in temporal dynamics were revealed by the Fourier analysis across canonical frequency bands: theta (4 -8 Hz), alpha (8 -13 Hz), beta (13 -30 Hz), low-gamma (30 -60 Hz), mid-gamma (60 -90 Hz) and high-gamma (90 -120 Hz). Supplementary Figureshows statistically significant LSD-PLA contrasts (p < 0.05, permutation-corrected; 50,000 permutations), with blue indicating power reductions and red indicating power increases. LSD produced robust reductions in low-frequency power spanning theta through beta. Although this attenuation follows a consistent desynchronization, its spatial extent exhibits clear frequency-specific organization. Theta band power reductions were the most widespread, encompassing large portions of posterior cortex and core regions of default mode network (DMN). Alpha band reductions followed a similar but more spatially restricted pattern, remaining prominent in visual cortex, posterior cingulate cortex (PCC), and temporal regions. Beta band power reductions were more focal, primarily involving anterior and posterior cingulate cortices. In contrast, higher frequencies exhibited a reversal of this pattern. LSD was associated with significant increases in gamma band power, with low-gamma effects remaining focal, whereas mid-and high-gamma bands showed widespread power increases. The spatial topographies of mid-and high-gamma effects were highly similar, with high-gamma increases extending more prominently into frontal regions (both laterally and medially) and posterior cortical areas.
Create a free account to open full-text PDFs.
Carhart-Harris, R. L. · Neuropharmacology (2018)
Luppi, A. I., Vohryzek, J., Kringelbach, M. L. et al. · Communications Biology (2023)
Atasoy, S., Leor, R., Kaelen, M. et al. · Scientific Reports (2017)
Atasoy, S., Vohryzek, J., Deco, G. et al. · Progress in Brain Research (2018)
Carhart-Harris, R. L., Muthukumaraswamy, S., Roseman, L. et al. · PNAS (2016)
Subramani, V., Nest, T., Pascarella, A. et al. · Biorxiv (2026)
Muthukumaraswamy, S. D., Carhart-Harris, R. L., Moran, R. J. et al. · Journal of Neuroscience (2013)
Timmermann, C., Roseman, L., Schartner, M. et al. · Scientific Reports (2019)
Mediano, P. A. M., Rosas, F. E., Timmermann, C. et al. · ACS Chemical Neuroscience (2024)
Carhart-Harris, R. L., Friston, K. J. · Pharmacological Reviews (2019)
Stoliker, D., Preller, K. H., Novelli, L. et al. · Molecular Psychiatry (2024)
Lebedev, A. V., L€ Ovd En, M., Rosenthal, G. et al. · Human Brain Mapping (2015)