In the past decade, two theoretical frameworks developed primarily by Robin Carhart-Harris and colleagues have substantially influenced how the neuroscience of psychedelics is discussed. The first is the entropic brain hypothesis. The second is the REBUS model — Relaxed Beliefs Under Psychedelics. Both frameworks attempt to explain what psychedelics do to information processing in the cortex and why these effects might produce the kinds of subjective experiences and clinical outcomes they do.
This article explains the frameworks in non-technical terms, examines what evidence supports them, and discusses where they remain provisional.
The Brain as a Prediction Machine
The starting point for both frameworks is a view of the brain that has become standard in cognitive neuroscience: the brain as a prediction machine. On this view, the brain is constantly generating expectations about what it will encounter — visually, auditorily, conceptually, socially — and updating those expectations based on incoming sensory data.
When predictions match incoming data, processing is efficient and the prediction is reinforced. When predictions are violated, the brain registers a “prediction error” and updates its model. Most of what counts as perception, on this view, is the brain’s predictions about what is there, with sensory input serving primarily to correct or refine those predictions rather than to construct perception from scratch.
This framework is sometimes called “predictive coding” or “active inference.” It has become widely adopted across cognitive neuroscience because it accounts for many features of perception, attention, learning, and other cognitive processes in a unified way.
The Entropic Brain Hypothesis
The entropic brain hypothesis, proposed in a 2014 paper by Carhart-Harris and colleagues, applies these ideas to psychedelics. The central claim is that psychedelics increase the entropy — the disorder or unpredictability — of brain activity, particularly in higher-order cortical regions.
Under normal conditions, brain activity is relatively constrained. Patterns of activity within and between regions are reasonably predictable; the system operates within a particular “regime” that is efficient for ordinary perception and cognition. Psychedelics, on this account, expand the range of brain states the system can occupy. Activity becomes less predictable and more varied.
The hypothesis predicts measurable increases in various entropy-related metrics during psychedelic states, and these have generally been observed in the studies that have measured them. Brain activity under psilocybin shows increased variability, less coordination of the kind characteristic of resting-state networks, and patterns that are statistically distinguishable from baseline.
The REBUS Model
The REBUS model, proposed in a 2019 paper by Carhart-Harris and Friston, builds on the entropic brain hypothesis but develops a more specific account of what the increased entropy means functionally.
In standard predictive-coding models, top-down predictions and bottom-up sensory data are constantly balanced against each other. The strength of top-down predictions — the “precision” with which the brain trusts its own model — varies across regions and conditions.
REBUS proposes that psychedelics specifically reduce the precision of high-level beliefs — the most abstract, most established models the brain has built about itself, the world, and what to expect. When these high-level priors are weakened, lower-level signals have more relative influence on processing. Sensory input is less constrained by expectations; conceptual associations that would normally be filtered out can come through.
This account does several explanatory things. It frames the visual phenomena of psychedelic experience as arising from reduced top-down filtering of perceptual data. It explains why psychedelics tend to produce ego-dissolution by suggesting that the high-level model of self is among the priors that weakens. It accounts for the often-noted therapeutic potential by framing the experience as an opportunity to revise long-held beliefs that have become rigid.
What the Evidence Shows
Several lines of evidence are consistent with the broad outlines of REBUS.
Brain imaging studies have shown the predicted patterns of reduced default-mode network coordination, increased between-network communication, and increased measures of brain activity entropy. These findings have been replicated across multiple research groups and across different psychedelic substances.
Studies of how psychedelics affect specific perceptual phenomena have shown effects consistent with reduced top-down constraints — for example, weakened binding of features into integrated objects, altered sensitivity to expectation effects in visual processing.
Clinical evidence on the relationship between certain features of psychedelic experiences (particularly the “mystical-type” experiences associated with reduced sense of self) and therapeutic outcomes is consistent with the prediction that revising high-level beliefs about self can produce lasting change.
Limitations and Open Questions
Both the entropic brain hypothesis and REBUS are theoretical frameworks rather than fully validated models. Several limitations and open questions deserve attention.
The relationship between population-level brain activity measures (the kind of thing fMRI captures) and the more granular cellular processes that REBUS describes is not direct. The metrics that have been used to support the framework are reasonable proxies but are not direct measures of “precision” in the technical sense the model uses.
The framework predicts certain things about how psychedelics interact with prior beliefs that have not been directly tested in many studies. The clinical relevance of “weakened priors” — whether weakened priors are actually what produces lasting clinical change, as opposed to other mechanisms — is plausible but not definitively shown.
Alternative theoretical frameworks have been developed. The “claustrum hypothesis” focuses on a specific brain region that may have a particular role in coordinating cortical activity. Various network-level frameworks emphasize different aspects of how psychedelics reshape brain dynamics. REBUS is the most prominent of these but is not the only viable account.
Why These Frameworks Matter
Whether or not the specific predictions of REBUS hold up to further testing, the framework has been useful for organizing thinking about psychedelics in neuroscience. It provides a conceptual structure that connects molecular pharmacology (5-HT2A activation), cellular effects (changes in cortical excitability), network-level dynamics (reduced DMN coordination), subjective experience (perceptual changes, ego dissolution), and clinical outcome (revised beliefs about self) into something approaching a coherent story.
The honest assessment is that the story is partial and provisional. Each step of the chain has supporting evidence, and the connections between steps are plausible, but the whole framework remains an active research program rather than a settled account.
For non-specialists trying to make sense of news about psychedelic neuroscience, the entropic brain and REBUS are useful concepts to know. They are the frameworks within which much of the current research is conducted and discussed. Understanding what they propose — and what they leave open — is part of being able to read research on this area as more than a series of disconnected findings.
The next decade of work will substantially refine these accounts. Some elements will likely be retained; others will need revision. This is normal for theoretical frameworks at this stage of development. The fact that we have accounts at this level of specificity at all, where a decade ago we did not, is a substantial advance.