DMT shifts brain oscillatory dynamics away from criticality—towards subcritical regimes—in alpha and adjacent (theta) bands, increasing entropy and reducing complexity. These shifts, quantified via a functional excitatory–inhibitory ratio, correlate with the intensity of subjective self‑dissolution.
Psychedelics profoundly alter subjective experience and brain dynamics. Brain oscillations express signatures of near-critical dynamics, relevant for healthy function. Alterations in the proximity to criticality have been suggested to underlie the experiential and neurological effects of psychedelics. Here, we investigate the effects of a psychedelic substance (DMT) on the criticality of brain oscillations, and in relation to subjective experience, in humans of either sex. We find that DMT shifts the dynamics of brain oscillations away from criticality in alpha and adjacent frequency bands. In this context, entropy is increased while complexity is reduced. We find that the criticality-shifts observed in alpha and theta bands correlate with the intensity ratings of self-dissolution, a hallmark of psychedelic experience. Finally, using a recently developed metric, the functional excitatory-inhibitory ratio, we find that the DMT-induced criticality-shift in brain oscillations is toward subcritical regimes. These findings have major implications for the neuronal understanding of the self and psychedelics, as well as for the neurological basis of altered states of consciousness.
The authors frame their study within recent interest in how classic psychedelics alter subjective experience and brain dynamics. They explain the concept of criticality — a dynamical regime in which correlations persist over long distances and timescales, producing scale-invariant, complex fluctuations — and note that the human brain is often thought to operate near criticality. Long-range temporal correlations (LRTCs) in oscillatory amplitude envelopes are presented as an empirical signature of proximity to criticality, and detrended fluctuation analysis (DFA) is introduced as a method for quantifying LRTCs. The introduction also notes that DFA alone cannot distinguish shifts toward inhibition-dominated (subcritical) versus excitation-dominated (supercritical) regimes, and that a recently developed functional excitatory/inhibitory (fE/I) ratio can provide directional information. This study set out to quantify markers of criticality in human EEG recordings under the influence of the short-acting psychedelic DMT, and to relate those markers to subjective reports of self-dissolution (measured with a specific visual analogue scale item). M. and colleagues combined two placebo-controlled, single-blind, within-participant EEG datasets to test whether DMT alters DFA-estimated LRTCs and fE/I, whether those changes indicate sub- or supercritical shifts, and whether such changes correlate with disruption of self-related processing during the psychedelic state. The authors emphasise that linking criticality metrics with subjective self-dissolution could provide novel neural correlates of self-related processing in altered states of consciousness.
Papers cited by this study that are also in Blossom
Aqil, M., Roseman, L. · Neuropharmacology (2022)
Carhart-Harris, R. L. · Neuropharmacology (2018)
Carhart-Harris, R. L., Friston, K. J. · Pharmacological Reviews (2019)
Carhart-Harris, R. L., Leech, R., Shanahan, M. et al. · Frontiers in Human Neuroscience (2014)
van Elk, M., Yaden, D. B. · Neuroscience and Biobehavioral Reviews (2022)
M. and colleagues combined EEG data from two within-participant, single-blind, placebo-controlled studies to produce a final sample of 27 healthy adults (12 female; mean age 34.1, SD 8.7) after exclusion of recordings with excessive artefacts. Participants were screened for physical and mental health and excluded for underage status, MR contraindications, lack of prior psychedelic experience, prior adverse psychedelic reactions, psychiatric or physical illness rendering them unsuitable, family history of psychosis, or excessive substance use. Ethics approval and regulatory oversight were obtained and all participants gave written consent. Study 1 contributed 12 participants and was a dose-finding design in which participants received placebo on a first visit and DMT on a second visit one week later; administered DMT fumarate doses varied across participants (reported doses: 7 mg, 14 mg, 18 mg, 20 mg). Study 2 contributed 15 participants who received 20 mg DMT fumarate and placebo in counterbalanced order with visits two weeks apart. When participants took part in both studies only recordings from the second study participation were used to improve dose homogeneity. EEG was recorded at baseline and for 20 minutes after intravenous DMT administration (the bolus given over 30 seconds followed by a 15-second saline flush). After sessions, participants completed questionnaires and visual analogue scales (VAS) about their subjective experience. EEG acquisition used 31 scalp electrodes (10-20 system) with an MR-compatible BrainAmp MR amplifier; FCz served as reference and AFz as ground, and an ECG channel recorded heart rate. Preprocessing included visual inspection, removal of segments contaminated by muscle and movement artefacts, and cleaning with Independent Component Analysis. For LRTC estimation the authors bandpass-filtered signals, extracted amplitude envelopes (Hilbert transform), then applied detrended fluctuation analysis (DFA). DFA estimates the slope of root-mean-square fluctuations of the integrated, detrended amplitude envelope across window sizes; a DFA exponent near 0.5 indicates uncorrelated (white-noise-like) dynamics, while values approaching 1 indicate long-range temporal correlations consistent with proximity to criticality. DFA was calculated in specified time-scale ranges (e.g. 2–30 s for alpha). To infer the direction of criticality shifts, the authors used the fE/I algorithm, which estimates a functional excitatory/inhibitory ratio from the windowed covariation between amplitude and short-window estimates of temporal autocorrelation structure (a normalized fluctuation function proxy for DFA). The fE/I computation pipeline included bandpass filtering, Hilbert-derived amplitude envelopes, segmentation into overlapping 5-second windows with 80% overlap, normalization and detrending, computation of windowed normalized fluctuation functions, and finally deriving fE/I as 1 minus the Pearson correlation between windowed normalized fluctuations and windowed amplitudes. The fE/I value was only computed when the DFA exponent exceeded a threshold of 0.6, since far-from-critical signals lack the covariation necessary for reliable fE/I estimation. Statistical analyses were performed per channel with nondirectional paired t-tests (alpha p < 0.05) comparing placebo and DMT conditions; average DFA across statistically significant electrodes was reported. Changes in DFA and fE/I were calculated as differences between placebo and DMT to avoid order effects. Correlations between criticality metrics and subjective experience used change scores from DMT baseline to post-injection DMT and Pearson's correlation. Multiple comparisons across electrodes within a frequency band were controlled using the Benjamini–Hochberg false discovery rate with adjusted q-values < 0.10.
DFA (LRTC) changes: Paired-sample comparisons between placebo and DMT showed significant reductions in DFA exponents in theta, alpha and beta frequency bands. Reported group mean differences were ΔDFA = -0.06 for theta (p < .0001), ΔDFA = -0.09 for alpha (p < .0001), and ΔDFA = -0.06 for beta (p = .0004). These reductions were described as widespread across the scalp and robust across subjects according to the authors' electrode-wise tests and subject-level boxplots. The authors interpret reduced DFA exponents in the 0.5–1 range as reflecting a shift from more complex, less entropic (pink-noise-like) dynamics toward less complex, more entropic (white-noise-like) dynamics in alpha and adjacent bands under DMT. fE/I (directionality) changes: To determine the direction of the criticality shifts, the authors computed fE/I where DFA exceeded the threshold for reliable estimation. Compared with placebo, DMT produced a significant decrease in fE/I in the alpha band (ΔfE/I = -0.18; p = .0003), with effects most pronounced over parietal and occipital electrodes. In the beta band a significant reduction in fE/I was reported only in the left occipital region (ΔfE/I = -0.14; p = .0005). The authors summarise these results as evidence that the observed decreases in DFA in alpha and neighbouring bands reflect shifts toward subcritical (inhibition-dominated) regimes. Correlations with subjective experience: The authors tested correlations between criticality shifts and selected VAS items related to disruptions of self-related processing. They found statistically significant negative correlations between DFA reductions and ratings of self-dissolution (VAS item: "I experienced a disintegration of my sense of self or ego"): theta band DFA change correlated with self-dissolution r(25) = -0.61, p = .001, and alpha band DFA change correlated with self-dissolution r(25) = -0.56, p = .005. These correlations were reported as significant across many electrodes. The authors therefore report that the magnitude of the DMT-induced reduction in LRTCs in theta and alpha bands relates to the intensity of subjective self-dissolution. Additional notes reported by the authors: The authors emphasise that DFA reductions are distinct from reductions in oscillatory power and represent novel evidence of altered temporal structure (entropy and complexity) rather than merely amplitude changes. They also note that fE/I reductions and DFA changes were localised mainly to posterior regions for alpha and left occipital for beta where reported.
M. and colleagues interpret their findings as showing that DMT reduces long-range temporal correlations and shifts brain oscillations in alpha and neighbouring frequency bands toward a more entropic, less complex, and subcritical regime. They highlight three core observations: DMT caused statistically significant reductions in DFA in theta, alpha and beta bands; fE/I results indicate these shifts are in the subcritical direction particularly for posterior alpha; and the magnitude of DFA reductions in theta and alpha correlated with subjective ratings of self-dissolution. The authors propose that weakened LRTCs in alpha and adjacent bands may index disruption of the temporal continuity of self-referential processing, linking a measurable neural dynamic to the phenomenology of self-dissolution. The discussion situates the results relative to prior literature. The authors note that prior psychedelic studies have reported increases in entropy and reductions in alpha power and default-mode network connectivity; their findings are consistent with entropy increases but differ from some formulations of the entropic brain hypothesis that predict movement toward criticality. They suggest this discrepancy can reflect system-specific relationships between entropy and criticality: in their data reduced DFA (closer to white-noise-like dynamics) accompanies increased entropy and reduced complexity. The authors also compare their findings to reports of reduced LRTCs in anaesthesia and deep meditation, proposing that despite phenomenological differences (e.g. richness of experience in psychedelics versus paucity in anaesthesia), these states may share disruption of coherent, temporally extended self-referential processing, which could be captured by LRTC metrics. Mechanistic and frequency-specific interpretations are discussed. The authors highlight alpha oscillations' putative role in top-down predictive processes and note that alpha suppression and DMN dysregulation under psychedelics have been linked to experiential effects. They acknowledge that alpha activity also reflects low-level sensory processes and that intense visual content during DMT may interact with posterior alpha dynamics. The authors speculate that richness of phenomenal content under psychedelics may be indexed by higher-frequency activity (high gamma > 40 Hz), which was not accessible in their dataset, and suggest that different frequency bands may show opposing criticality shifts, motivating future time-resolved, frequency-specific analyses that align subjective reports with neural moments. Methodological caveats and limitations acknowledged by the authors include the inherent difficulty of proving that a real-world system is strictly at criticality (their measures test for consistent signatures), constraints of EEG in accessing high-frequency components, and the potential ambiguity in questionnaire items since participants were not asked to clarify how they understood particular items. They defend the use of DFA as robust to nonstationarities and appropriate for assessing LRTCs. The authors conclude that their results provide novel, frequency-dependent evidence on how DMT alters criticality, entropy and complexity of brain oscillations and that weak, subcritical LRTCs in alpha and adjacent bands may be a candidate neural correlate of disrupted self-related processing across several altered states of consciousness.
For this study the EEG data from two placebo-controlled, single-blind, within-participant studieswere combined resulting in 27 healthy participants (12 female, mean age= 34.1 SD=8.7) after removing participants with unusable EEG data due to muscle and movement artifacts. An initial screening visit to the Imperial College Clinical Research Facility (CRF) assessed physical and mental health to ensure suitability of participants. Exclusion criteria included: being under 18 years old, MR contraindications, no previous psychedelic experiences, an adverse reaction to a psychedelic, a history of psychiatric or physical illness rendering participants unsuitable for participation, a family history of psychotic disorder, or excessive use of alcohol or drugs of abuse. All participants provided written informed consent for participation in the study. Both studies were approved by the National Research Ethics Committee London-Brent and the Health Research Authority and was conducted under the guidelines of the revised Declaration of, the International Committee on Harmonization Good Clinical Practices guidelines, and the National Health Service Research Governance Framework. Imperial College London sponsored the research conducted under a Home Office license for research with Schedule 1 drugs. Study 1 (12 participants from the final sample) was a dose-finding study where participants always received placebo (saline) on a first visit and DMT on a second visit one week after, while EEG recordings took place. The doses administered were as follows: 2 participants received 7 mg, 3 participants received 14 mg, one participant received 18 mg, and 4 participants received 20 mg of DMT fumarate -for full recruitment procedure and protocol see. In study 2 (15 participants from the final sample), participants received 20 mg of DMT fumarate and placebo (saline) in a counterbalanced order in visits separated 2 weeks apart -for full recruitment procedure and protocol see:. For both studies, EEG recordings occurred at baseline (prior to administration) and for 20 minutes after intravenous administration of DMT fumarate, which was performed during 30 seconds and followed by a 15-second flush of saline. If participants partook in both studies, we only used the recordings corresponding to their participation in the second study to increase homogeneity in the doses. Participants attended the experimental sessions at the National Institute of Health Research (NIHR) Imperial Clinical Research Facility (CRF) for Study 1, and the Clinical Imaging Facility (CIF) at Imperial College London for Study 2. For both studies, following EEG recordings and once subjective effects subsided, participants completed questionnaires and visual analogue scales designed to assess the subjective effects experienced during the administration of DMT (seefor full procedures). Of these metrics, we selected 5 visual analogue scales for correlations with our EEG results, as these reflect central features of subjective experience we hypothesised are susceptible to changes in LRTC and changes in criticality: 1) the sense of self ('I experienced a disintegration of my usual sense of 'self' or 'ego'), 2) the sense of time ('my sense of time was altered'), 3) the sense of space ('my sense of size or space was altered'), 4) cognition ('my thoughts wandered freely'). Additionally, we assessed how LRTC related to generic subjective effects induced by DMT ('how intense was the drug experience').
The same EEG setup was used for both studies. Data were collected from 31 scalp locations in accordance with the 10-20 system using an MR-compatible BrainAmp MR amplifier (BrainProducts, Munich, Germany) and a compatible cap (BrainCap MR; BrainProducts GmbH, Munich, Germany). The system used FCz as the reference for all electrodes, and AFz was the ground electrode. Additionally, an ECG channel was used to capture heart rate recordings. For both recordings data were visually inspected, and segments of data containing muscle artifacts, head motion, and other gross artifact were removed prior to cleaning using Independent Component Analysis (seefor a full list of preprocessing details).
Detrended Fluctuation Analysis (DFA) measures the scale-invariance in temporal correlations, which is reflected by power-laws in fluctuation sizes, a marker of criticality. DFA proceeds from an analogy between the signal (e.g., alpha-band amplitude envelopes) and fractional Gaussian noise; validated by the linearity of the log-log fit in fluctuation scalings over a range of timescales. A DFA exponent (slope of the log-log linear fit) of 0.5 (analogous to white noise) indicates no long-range temporal correlations, indicative the underlying system is far from criticality; a DFA exponent near 1 (analogous to pink noise) is consistent with the system being near criticality, and between 0.5 and 1 indicates a degree of LRTCs increasingly proximal to a potential critical state. The relationship between entropy, complexity, and criticality is not trivial. In the present context, entropy refers to the unpredictability or randomness of a signal. A whitenoise signal exhibits maximal entropy, due to its unpredictability and lack of temporal correlations (no LRTCs; DFA exponent near 0.5), thus reflecting minimal statistical complexity, as it lacks any temporal structure. Colored noise has lower entropy, as its structured temporal correlations make it more predictable (LRTCs, DFA exponent > 0.5), and is statistically more complex, due to the rich variety of patterns elicited by temporal correlations over a wide range of timescales. Hence, in the context of DFA, entropy and complexity are by definition inversely related. Higher DFA exponents indicate signals with temporal structure over long timescales, higher complexity, lower entropy, potentially proximal to criticality; lower DFA exponents indicate signals with little or no temporal structure, lower complexity, higher entropy, farther away from criticality. In statistical descriptions, as is the case here, entropy can increase away from criticality in both directions, as dynamics become increasingly unstructured and white-noise-like. As complexity is generally maximal at criticality due to the range of potential spatiotemporal patterns, these distinctions imply that entropycomplexity-criticality relationships depend on the theoretical framework, and on assumptions on the underlying system. Here we assessed LRTC and E/I with the Neurophysiological Biomarker Toolbox (NBT) written in Python, github links available at. Several measures of E/I balance have been proposed, including LRTC, which is a robust empirical feature of oscillations associated with E/I balance in computational models. Detrended fluctuation analysis (DFA) is used to assess LRTC in the signal in the time scales of interest via a DFA exponent. A DFA exponent of 0.5 indicates an uncorrelated random signal (i.e., absence of LRTC), whereas the value > 0.5 indicates the presence of positive auto-correlations and their strength. To quantify the strength of long-range temporal correlations (LRTC) in the amplitude modulation of the EEG oscillations, we first extracted the amplitude envelope using band-pass filters (FIRfilter, Blackman window with transition bandwidth of 1 Hz, (Figure) and the Hilbert transform (Figure). Next, the root-mean-square fluctuation of the integrated and linearly detrended signals, F(t), was calculated as a function of time window size, t (with an overlap of 50% between windows) and plotted in double-logarithmic coordinates (Figure). The DFA exponent is the slope of the fluctuation function F(t) in a given interval, which was set to 2 to 30 seconds for the alpha band.
The fE/I algorithm was developed based on an extended version of the Critical Oscillations (CROS) computational model of neuronal oscillationswhich mimics the signals observed in human M/EEG recordings. In fE/I, the E/I ratio is estimated from the windowed covariation of the average amplitude and amplitude modulation (the temporal auto-correlation structure) of frequency-specific activity, in the presence of significant LRTC in the signal (Figure,I), that have been associated with E/I balance in computational models. It estimates the temporal structure of oscillation amplitude and reflects the level of criticality in the network. Detrended fluctuation analysis (DFA) is used to assess LRTC in the signal in the time scales of interest via a DFA exponent (Figure). The DFA value of 0.5 indicates an uncorrelated random signal (i.e., absence of LRTC), whereas the value > 0.5 indicates the presence of positive auto-correlations and their strength. Taken alone, however, the DFA exponent cannot distinguish sub-from super-critical activity (Figure). Neither can the amplitude of oscillations, which changes monotonously with E/I balance, tell where the critical point is. The combination of the two, on the other hand, can be used to tell apart sub-, critical, and super-critical dynamics (Figure). The correlation of amplitude and DFA is positive for a network operation in a slightly sub-critical state (FigureF left), zero in a critical state (Figure, middle), and negative in a slightly super-critical state (Figure, right). Given that networks operating in these regimes exhibit co-variation in amplitude and temporal structure, we can use a slidingwindow approach to quantify this co-variation and, thus, infer the E/I balance of the underlying networks. Of note, when networks are far from criticality as reflected in DFA exponents < 0.6, there is no correlation between the covariation of amplitude and temporal structure and, therefore, the DFA exponent of 0.6 is used as a threshold to compute the fE/I ratio. Given the presence of LRTC in the signal reflected by the DFA exponent greater than 0.6, fE/I is then computed by correlating the amplitude and LRTC in short windows. Windowed LRTC, in this case, is estimated through the normalized fluctuation function which serves as a reliable proxy of the DFA exponent on short time scales. Overall, the main steps of the fE/I algorithm are as follows: J N e u r o s c i A c c e p t e d M a n u s c r i p t (1) The signal is bandpass-filtered in the desired frequency range using a finite-impulse-response (FIR) filter. (2) The amplitude envelope of the filtered signal is extracted using Hilbert Transform. (3) The signal profile is calculated and segmented into 80%overlapping 5-second windows. (4) The windows are normalized using the mean of the amplitude envelope calculated per window. (5) Subsequently, the normalized windows are detrended. (6) The normalized fluctuation function for each window is computed as the root-mean square fluctuation of the detrended amplitude-normalized signal profile. (7) Finally, the fE/I value is obtained by subtracting the Pearson correlation between the windowed normalized fluctuations and windowed amplitudes from 1. The fE/I is set to NaN (i.e., missing) if DFA exponent does not exceed the DFA threshold of 0.6.
The EEG analysis was performed per channel with a nondirectional paired t-test (level: p < 0.05), with average DFA across statistically significant electrodes reported. Changes in DFA and fE/I were calculated between the placebo and DMT condition to avoid order effects, while the correlations with changes in subjective experience was performed with changes from DMT baseline to DMT after injection, to minimize noise. Due to the continuous nature of subjective experience, the parametric Pearson's correlation coefficient was used to test for correlations with experience. To prevent chance-level effects, we used the Benjamini Hochberg false discovery rate (FDR) multiple comparison correction methodas post hoc test with FDR adjusted q-values of < 0.10 within a specific frequency band.
In order to investigate the impact of DMT on the criticality of brain oscillations, we quantify the presence of LRTC following DMT and placebo administration in the same participants. We computed the DFA exponent as an index of LRTC. Paired-samples t-tests were conducted to compare the difference in DFA between placebo and DMT conditions for the theta, alpha and beta band. Compared to placebo, DMT induced a significant decrease of DFA in theta (ΔDFA = -0.06; p = <.0001), alpha (ΔDFA = -0.09; p = <.0001) and beta (ΔDFA = -0.06; p = .0004). These observed reductions in DFA exponents were widespread throughout the scalp (Figure, left column), and are robust across subjects (Figure, left column boxplots). In sum, we find that DMT shifts the dynamics of brain oscillations in alpha and adjacent frequency bands away from criticality and in the direction of a more entropic, sub-critical regime. As implied above, DFA exponents also provide information on the entropy and complexity of brain signals. A DFA exponent of 1.0 represents a pink-noise-like signal (1/f spectrum), while a DFA exponent of 0.5 represents a white-noise-like signal (flat spectrum). Exponents with intermediate values represent the spectrum of fractional Gaussian noises between these two extrema. In this context, entropy and complexity are inversely related. A white-noise-like signal has higher entropy but lower complexity, while a pink-noiselike signal has lower entropy but higher complexity. As such, a reduction in DFA exponent estimates, in the range between 0.5 and 1, such as that we observe here, indicates a shift away from a more complex, less entropic, pink-noise-like signal towards a less complex, more entropic, white-noiselike signal. In sum, our findings imply that brain oscillations in alpha and adjacent frequency bands under DMT become less complex, but more entropic, throughout the brain.
DFA does not provide information on the direction of the criticality shifts it quantifies. Starting from a near-critical point, both shifts towards subcritical and supercritical dynamics are characterized by reductions in DFA exponents, such as those we observe here. To distinguish the directionality of the shift we observed, we computed a metric recently introduced for this purpose, the functional Excitatory/Inhibitory ratio. Paired-samples t-tests were conducted to compare the difference in fE/I between the placebo and DMT condition. Compared to placebo, DMT induced a statistically significant decrease of fE/I for the alpha band (ΔfE/I = -0.18; p = .0003), which were especially pronounced in parietal and occipital electrodes. In the beta band the reduction only reached significance in the left occipital region (ΔfE/I = -0.14; p = .0005) (Figure, right column). In sum, we find evidence that the observed criticality shifts in alpha and theta bands are in the subcritical direction.
To test whether shifts in criticality of neuronal oscillations measured with DFA and the fE/I ratio were associated with DMT-induced alterations in self-related processing, we performed correlations with 4 items of the Visual Analogue Scales (VAS) related to 3 elements related to disruptions in self-related processing as well as a generic item broadly assessing disruptions in the sense of self (see methods). We found statistically significant correlations between criticality shifts (measured with DFA estimates) and the generic item of self-disruption (measured with the item: "I experienced a disintegration of my sense of self or ego") in the theta (r(25)=-0.61, p=.001) and alpha bands (r(25)=-0.56., p=.005). The correlation were statistically significant for most electrodes across the scalp (Figure). In sum, we find that the magnitude of the shift away from criticality elicited J N e u r o s c i A c c e p t e d M a n u s c r i p t by DMT in theta and alpha bands relates to the disruptions in self-related processing.
Here, we investigated the effects of DMT, a classic psychedelic, on markers of criticality, and their relation to subjective experience. We found statistically significant shifts away from criticality for oscillations in theta, alpha, and beta bands, implying a reduction in complexity and an increase in entropy. We found statistically significant shifts towards subcritical regimes, in occipital and parietal alpha, and left-occipital beta bands. Finally, we found statistically significant correlations between DFA reductions in alpha and theta bands and ratings of self-dissolution. Our findings provide novel information on the effects of DMT (shifts away from criticality, and toward a more entropic, less complex subcritical regime), in alpha and adjacent frequency bands. Furthermore, our findings demonstrate a relationship between criticality shifts and the disruption of the self, providing novel candidates of neural correlates of self-related processing in the human brain. Classic psychedelics act through the (excitatory) 5-HT2A receptor. Previous studies have reported increases in measures of entropy under psychedelics. Shifts away from criticality have also been observed in states of anesthetic unconsciousness and meditation, which certainly have profound differences with psychedelic states. On one hand, psychedelics generally elicit states of high phenomenal richness; on the other, anesthetics generally elicit states of little or no phenomenal experience. If the effects of anesthetics and psychedelics are taken to be completely distinct, our observation of weaker LRTCs and shifts towards subcritical dynamics in alpha and adjacent frequency bands may appear puzzling. However, high-dose psychedelic and anesthetic states may also share notable features. In particular, the loss or severe disruption of the coherent, self-referential stream of thought, unfolding over medium-long timescales, which is a cornerstone of daily waking experience. Our findings suggest that reduction in DFA and fE/I metrics in alpha and adjacent frequency under DMT might specifically track the disruption of this process shared by high-dose psychedelic states as well as anesthetic or deep-meditation states. Indeed, we believe that the measures employed in our study selectively capture the shared nature of the latter similarity between otherwise distinct states of consciousness. As such, our study represents a step forward towards understanding the neural correlates of multidimensional consciousness. Alpha frequencies have been implied as top-down carriers of predictive models instantiated by high-level regions and networks. Theoretical frameworks and empirical evidence have indeed implicated reductions of alpha oscillations and default-mode network (DMN) connectivity (a system known to relate to alpha oscillations), as key players in psychedelic effects. In fact, both systems have been found to be significantly dysregulated under DMT in an interrelated fashion. However, alpha oscillations may also relate to activity in low-level sensory regions, which coexist in a dialogical interplay with high-level ones. Alpha oscillations and DMN activity are strongly suppressed by visual content, which is intensely present during DMT states. In fact, DMT-induced dysregulation of alpha oscillations has been found to significantly correlate with the intensity of visual experience. Further work is needed to clarify the role of alpha oscillations in the interplay between low and high-level cognition. We speculate that supercritical regimes under DMT and other classic psychedelics (and the corresponding richness of phenomenal content) may be apparent in high frequency oscillations (high gamma), which are not accessible here because of experimental limitations. Changes of this nature have been observed in animal models of psychedelics. Alternatively, it is possible that theta oscillations may reflect supercritical regimes under DMT, as increases in theta have been found to be correlated with the visual experience induced by DMT. Future studies should narrow the EEG analysis to moments of experience in which increases in content richness are present to advance these questions. This could be achieved by employing methodological paradigms that carefully match subjective experience with specific moments of the psychedelic state. Aside from mathematically-defined theoretical systems, it is generally not possible to strictly confirm that a realworld system is at criticality. Rather, it is possible to test for signatures that are consistent with critical dynamics. One such signature is the presence of long-range temporal correlations (LRTCs) quantified by the DFA exponent. The DFA exponent is a direct generalization of the Hurst exponent. However, DFA is considered more robust to nonstationarities and trends in the data and a wealth of results implicates its ability to measure functionally relevant brain properties. Thus, we chose DFA over other alternative measures such as the Hurst exponent. A reduction in alpha oscillatory power does not automatically imply a shift away from criticality. The reduction in DFA and fE/I measures we observe therefore represent a novel finding, distinct from the previously known reduction in power, and providing evidence of a shift towards subcritical regimes in alpha and neighboring frequency bands elicited by DMT. Self-referential processing is thought to encompass functional connections between widespread brain regions. Longrange amplitude coupling is optimized for brain networks near criticality. Shifts in DFA have previously been found in deep meditation and anesthetic states, which also present a reduced or absent self-referential J N e u r o s c i A c c e p t e d M a n u s c r i p t processing. Hence, we hypothesized that LRTCs might underlie the coherent, temporally extended stream of consciousness in our daily experience, disrupted in states such as high-dose psychedelics, anesthesia, or deep meditation (despite these states differing on other dimensions, such as richness of phenomenal content). As such, we carried out correlations only with selected questionnaire items that related to disruptions of the sense of self. A limitation of our study is that we did not ask participants to explain what they understood a particular questionnaire item to mean. This could be a useful consideration for future studies, in order to reduce potential ambiguity. The entropic brain hypothesissuggests that during psychedelic states 1) brain entropy increases and 2) brain dynamics move closer to criticality, starting from slightly subcritical states during normal waking consciousness. Here, we find that 1) brain entropy increases during psychedelic states, and 2) alpha oscillations and neighboring frequencies move in the subcritical direction. Hence, our results are consistent with the entropic brain hypothesis insofar as entropy increases; but not consistent insofar as we observe shifts away from criticality. The discrepancy can be explained by the different relationships between entropy and criticality in different systems. In the context of the original entropic brain hypothesis, entropy is assumed to increase monotonically from subcritical, to critical, to supercritical states, as indeed can be the case in some dynamical systems. However, in this context, a decrease in DFA exponent indicates a shift from a more complex, less entropic pink-noise-like signal to a more entropic, less complex white-noise-like signal. We suggest that subcritical shifts in alpha and neighboring frequencies might relate to the disruption of self-referential processing, as indeed they correlate with the intensity of self-dissolution. We speculate that high gamma oscillations (>40hz) might instead display shifts towards criticality or beyond, into supercritical regimes, potentially indexing the richness and complexity of phenomenal content during psychedelic states. This is consistent with recent findings showing how increases in complexity in psychedelic states are mostly attributed to high-frequency components of brain signals. Our findings suggest a nuanced, potentially frequency-dependent relationship between criticality of brain oscillations and different dimensions of psychedelic subjective experience, moving beyond global measures of brain entropy and a generic notion of 'psychedelic state'. In sum, we find that DMT alters the criticality signatures of brain oscillations, and that the observed shifts in criticality correlate with the subjective experience of self-dissolution. In particular, we find that DMT shifts oscillations in alpha and adjacent frequency bands away from criticality and towards subcritical regimes characterized by increased entropy but reduced complexity. Weak, subcritical LRTCs in alpha and adjacent bands may be a potential shared neural correlate of the disruption of conscious self-related processing, common both to high-dose psychedelic experiences, deep meditation states, and anesthesia. Overall, our findings provide novel information on the effects of psychedelics on criticality, entropy, and complexity of brain oscillations, and their relation with subjective experience.
Create a free account to open full-text PDFs.
Griffiths, R. R., Richards, W. A., Mccann, U. et al. · Journal of Psychopharmacology (2006)
Halberstadt, A. L. · Behavioural Brain Research (2014)
Inserra, A., De Gregorio, D., Gobbi, G. · Pharmacological Reviews (2021)
Kałużna, A., Schlosser, M., Gulliksen Craste, E. et al. · Journal of Psychedelic Studies (2022)
Kwan, A. C., Olson, D. E., Preller, K. H. et al. · Nature Medicine (2022)
Kyzar, E. J., Nichols, C. D., Gainetdinov, R. R. et al. · Trends in Pharmacological Sciences (2017)
Roseman, L., Nutt, D. J., Carhart-Harris, R. L. · Frontiers in Pharmacology (2018)
Schartner, M., Carhart-Harris, R. L., Barrett, A. B. et al. · Scientific Reports (2017)
Timmermann, C., Vollenweider, F. X. · Trends in Cognitive Sciences (2023)
Timmermann, C., Roseman, L., Haridas, S. et al. · PNAS (2023)
Timmermann, C., Roseman, L., Schartner, M. et al. · Scientific Reports (2019)
Vollenweider, F. X., Preller, K. H. · Nature Reviews Neuroscience (2020)