Testing the Entropic Brain Hypothesis Across Diverse States and Conditions

The Entropic Brain Hypothesis proposes that conscious states correspond to the entropy (diversity and unpredictability) of spontaneous brain activity. A recent cross-condition fMRI study found an entropic gradient: propofol anaesthesia reduced entropy, while LSD, DMT, modafinil, and schizophrenia increased it, with schizophrenia showing the largest increase. The present project aims to test this hypothesis more broadly by examining fMRI datasets from LSD, psilocybin, modafinil, propofol anaesthesia, natural sleep, schizophrenia, Mild Cognitive Impairment, Alzheimer's disease, and Frontotemporal Dementia. It hypothesizes that diminished states (sleep, anaesthesia) will show low entropy, psychedelic and stimulant states high entropy, and neurodegenerative conditions low entropy. A second objective is to compare a topology-derived entropy measure with a brain-state-derived Shannon entropy measure to assess robustness across operationalizations.

Study at a glance

Abstract

Understanding how different states of consciousness arise from brain activity remains a central challenge in consciousness science. Theoretical and empirical work has increasingly linked conscious states to the large-scale organization and dynamical complexity of brain activity, using measures ranging from network integration and graph-theoretical properties to signal diversity, perturbational complexity, and dynamic functional connectivity (Barttfeld et al., 2015; Casali et al., 2013; Demertzi et al., 2019; Schartner et al., 2015, 2017; Tagliazucchi et al., 2014). However, these measures have been operationalized differently across methodological traditions and experimental paradigms, making it difficult to determine whether diverse alterations of consciousness can be explained by a common underlying dynamical principle using the same analytical pipeline (Galván Rial et al., 2026). One influential framework addressing this question is the Entropic Brain Hypothesis, which proposes that the quality of conscious states is related to the entropy of spontaneous brain activity (Carhart-Harris et al., 2014). In information-theoretical terms, entropy quantifies uncertainty about the state of a system and is related to its degree of randomness or unpredictability (Ben-Naim, 2012; Carhart-Harris et al., 2014). Applied to brain activity, this framework proposes that differences in conscious experience correspond to differences in the diversity and unpredictability of spontaneous brain dynamics. The psychedelic state was initially proposed as a prototypical high-entropy state, characterized by less constrained brain dynamics and an expanded repertoire of transient functional connectivity configurations relative to ordinary waking consciousness (Carhart-Harris et al., 2014; Tagliazucchi et al., 2014). Conversely, states of reduced consciousness are expected to exhibit more restricted and stereotyped dynamics, consistent with findings of reduced neural complexity during non-rapid eye movement sleep and anaesthesia (Schartner et al., 2015; Varley et al., 2020). This framework therefore generates the broader prediction that heterogeneous states of consciousness may be systematically differentiated according to the entropy or complexity of their underlying brain dynamics. Recent work has provided an important cross-condition empirical test of this prediction. Galván Rial et al. (2026) applied a common analytical pipeline to five fMRI datasets spanning propofol anaesthesia, LSD, DMT, modafinil, and schizophrenia. The study quantified the temporal irregularity of large-scale network topology by applying sample entropy to time-resolved trajectories of dynamic small-worldness (SE dSW). Propofol anaesthesia was associated with progressively reduced dynamic-network entropy as sedation deepened, whereas LSD increased entropy, DMT and modafinil showed increases in the same direction, and schizophrenia exhibited the largest elevation relative to matched controls (Galván Rial et al., 2026). When placed on a common standardized scale, these conditions formed an entropic gradient extending from reduced network entropy under anaesthesia to increased entropy under psychedelic, stimulant, and clinical conditions. These findings provide initial evidence that the temporal diversity of large-scale network reconfiguration may constitute a shared dimension along which heterogeneous brain states can be organized (Galván Rial et al., 2026). However, empirical tests spanning a broad range of altered states within a common analytical framework remain limited, and the generalizability of the proposed entropic axis to additional conditions has yet to be established. Furthermore, the entropy measure used by Galván Rial et al. (2026) represents a topology-derived operationalization of entropy: sample entropy is applied to temporal fluctuations in dynamic small-worldness, a graph-theoretical property derived from time-resolved functional connectivity networks. This raises the question of whether the proposed entropic ordering is robust to alternative, more direct operationalizations of entropy. Different measures of neural complexity may capture distinct properties of brain dynamics, as demonstrated by the dissociation between topological entropy and connectivity-magnitude entropy observed under psychedelic conditions (Galván Rial et al., 2026). The primary aim of the present project is therefore to provide a broader empirical test of the Entropic Brain Hypothesis by investigating whether brain entropy can serve as a common dimension for ordering diverse states of consciousness. Using fMRI data, we will examine datasets representing LSD, psilocybin, modafinil, propofol anaesthesia, natural sleep, schizophrenia, Mild Cognitive Impairment, Alzheimer's disease, and Frontotemporal Dementia, together with the relevant healthy, placebo, or waking baseline conditions. This selection extends previous cross-condition analyses (Galván Rial et al., 2026) by incorporating the psychedelic compound psilocybin, natural sleep, as well as neurodegenerative disorders as additional datasets to examine, allowing us to investigate a broader range of pharmacological, physiological, psychiatric, and neurodegenerative brain states. We hypothesize that these conditions will be distributed along an entropic gradient, with diminished states of consciousness, including sleep and propofol anaesthesia, occupying the lower end; psychedelic states and central nervous system stimulant conditions occupying the higher end; and healthy waking consciousness occupying an intermediate position. The prediction of reduced entropy under diminished states and increased entropy under psychedelic states is motivated by previous theoretical and empirical work demonstrating restricted dynamics during diminished consciousness and expanded or more diverse dynamics during psychedelic states (Carhart-Harris et al., 2014; Galván Rial et al., 2026; Schartner et al., 2015, 2017; Tagliazucchi et al., 2014; Varley et al., 2020). We additionally hypothesize that the neurodegenerative conditions examined here—Mild Cognitive Impairment, Alzheimer's disease, and Frontotemporal Dementia—will occupy the lower portion of the entropic gradient. This represents a novel extension of the framework and a central empirical prediction of the present project, rather than a prediction directly established by the primary papers motivating this preregistration. A second major objective is to determine whether the proposed entropic ordering depends on how entropy is operationalized. We will compare the topology-derived entropy measure with a brain-state-derived measure based on Shannon entropy. Shannon entropy provides a direct measure of uncertainty associated with a probability distribution and has previously been used to characterize changes in the diversity of neural dynamics under psychedelics (Carhart-Harris et al., 2014). We hypothesize that our brain-state-derived Shannon entropy measure will distinguish among the examined states of consciousness at least as well as, and potentially more accurately than, the topology-derived entropy metric. Applying both approaches to the same datasets will allow us to assess whether the proposed entropic gradient is robust across alternative operational definitions of entropy and whether a more direct, less assumption-dependent entropy measure provides greater sensitivity for differentiating brain states. If our hypotheses are supported, the findings would strengthen the proposal that entropy captures a general property of brain dynamics that varies systematically across markedly different alterations of consciousness (Carhart-Harris et al., 2014; Galván Rial et al., 2026). Demonstrating a common entropic organization across pharmacological, physiological, psychiatric, and neurodegenerative conditions could support the use of entropy as a sensitive quantitative marker for distinguishing diverse brain states. Conversely, failure to identify the predicted gradient would provide evidence against a simple unidimensional interpretation of the Entropic Brain Hypothesis, at least for the datasets and entropy definitions examined here. Differences between the topology-derived and brain-state-derived measures would also be theoretically informative, potentially indicating that different entropy metrics capture distinct and complementary dimensions of neural dynamics rather than a single underlying property (Galván Rial et al., 2026). Regardless of the outcome, systematically applying common entropy metrics across heterogeneous datasets will help clarify the explanatory scope and limitations of the Entropic Brain Hypothesis. Planned outputs include a manuscript, a public code repository, and a web application intended to support the collaborative mapping of different states of consciousness using entropy and potentially other neurobiological metrics.