How do psychedelics work?

Psychedelics such as LSD, psilocybin and DMT have re-attracted attention across psychiatry, cognitive neuroscience and other disciplines, driven by promising small clinical trials, human brain imaging studies and renewed public interest. Carhart-Harris frames these compounds as unusual brain drugs that historically influenced culture and psychiatry and poses a central scientific question: by what mechanisms do psychedelics exert their characteristic experiential and therapeutic effects? This brief opinion piece aims to offer a multi-scale perspective linking known molecular actions to anatomical, physiological, network-level and experiential phenomena. The paper proposes that psychedelics trigger a cascade—from serotonin 2A receptor agonism through heightened plasticity and increased neural entropy to network disintegration and, ultimately, a relaxation of high‑level beliefs—and that psychedelic therapy seeks to use this transient belief‑relaxation to enable adaptive revision of pathological beliefs.

This article is an opinion/theoretical synthesis rather than an empirical study or systematic review. Carhart-Harris integrates findings from pharmacology, neurophysiology, human neuroimaging and clinical reports to construct a mechanistic account. The text draws on antagonist pretreatment studies in humans, receptor expression data, single-unit and population electrophysiology, resting-state network imaging, entropy-related metrics of brain activity, and clinical observations from small trials. No formal methods section or systematic search strategy is reported in the extracted text, so the piece should be read as a conceptual synthesis built on selected empirical studies rather than a comprehensive, reproducible review. The argument proceeds by examining evidence at successive scales—molecular (5-HT2A receptor action), cellular/anatomical (layer 5 pyramidal neurons and cortical localisation), neurophysiological effects (glutamate dynamics and oscillatory changes), whole-brain network alterations (disintegration and desegregation of intrinsic networks), and information‑theoretic measures (entropy)—before linking these to predictive processing frameworks and proposing mnemonic models (TIBER and REBUS). Images and figures are referenced to illustrate network changes, but the extracted text does not include methodological details for those visualisations.

Carhart-Harris synthesises empirical findings to support several interrelated propositions rather than reporting new experimental results. At the molecular and anatomical level, classical psychedelics share agonism at the serotonin 2A receptor (5-HT2AR); antagonist pretreatment studies in humans indicate that blocking this receptor abolishes the hallmark effects. These receptors are densely expressed in high‑level association cortex and particularly on layer 5 pyramidal neurons, suggesting a high‑hierarchical locus of action. Neurophysiological evidence cited indicates that 5-HT2AR agonism induces asynchronous glutamate release and spike-to-field decoherence at the single‑unit level, which is associated with decreased oscillatory power across several frequency bands (including alpha and beta). At the systems level, resting-state or intrinsic brain networks show marked disruption under psychedelics: functional integrity and segregation of canonical networks are reduced, while global functional integration increases, a pattern the author terms disintegration and desegregation. These network changes correlate with high-level psychological phenomena such as ego-dissolution and the unitive or mystical experience, which in turn predict better mental health outcomes in some studies. The entropic brain hypothesis is presented as a unifying principle: entropy, a measure of unpredictability, is proposed to index the richness or diversity of subjective experience. Entropy-related metrics (for example, Lempel-Ziv complexity) reportedly track subjective intensity of psychedelic effects across compounds and predict longer-term changes such as increased openness. Within a predictive processing framework, the paper argues that psychedelics relax or decompose high‑level priors (beliefs) — because 5-HT2ARs are located on neurons that instantiate high-level prediction units and because psychedelics suppress rhythms thought to carry priors — allowing prediction errors to percolate to lower levels and thereby producing the characteristic perceptual and cognitive changes. Clinically, the synthesis notes that psychedelics have shown promise in small trials for conditions associated with rigid or pathological beliefs, including depression (with reports of correction of overly pessimistic expectations in treatment‑resistant depression following psilocybin therapy), obsessive–compulsive disorder, end‑of‑life existential distress, and addictions; anecdotal reports of ayahuasca use in eating disorders are also mentioned. Two conceptual models are proposed: TIBER (Tightened Beliefs in Response to Uncertainty), which casts trauma/uncertainty as leading to tightened, maladaptive implicit beliefs, and REBUS (RElaxed Beliefs Under pSychedelics), which summarises the hypothesised acute action and downstream potential for belief revision. The author emphasises that belief revision is contingent on contextual factors such as psychotherapeutic support, music, priming and empathic listening.

Carhart-Harris interprets the assembled evidence as supporting a coherent mechanistic account in which 5-HT2AR agonism precipitates heightened plasticity and increased brain entropy, culminating in a relaxation of high‑level priors or beliefs. This relaxation, when combined with appropriate contextual support, may permit longer‑term revision of pathological beliefs and thereby explain rapid and enduring therapeutic effects observed in some studies. The paper situates this account within prior theoretical work, in particular the entropic brain hypothesis and the predictive processing framework, arguing that these perspectives complement one another and together offer explanatory reach from molecules to experience. Important caveats are acknowledged: the TIBER and REBUS models are largely conceptual at present and require substantial empirical testing to clarify how receptor-level, physiological, network and psychological components interrelate dynamically and biologically. The author stresses that belief‑relaxation is an acute pharmacological action, whereas beneficial belief revision is more contingent on context and psychological integration. Implications highlighted by the author include the potential for psychedelic therapy to remediate disorders marked by inflexible beliefs, and the importance of set and setting (for example, music and therapeutic support) in guiding adaptive outcomes. At the same time, the discussion cautions that further research is necessary to determine how useful these models will be for prediction and clinical translation.

The paper concludes by reiterating that 5-HT2A receptor agonism, heightened plasticity and increased brain entropy—mirrored by the relaxation of high‑level priors—constitute a parsimonious mechanistic account of psychedelic action. When this transient decomposition of beliefs is mediated appropriately, it may allow corrective belief‑revision that underlies rapid and sustained improvements in mental health reported in emerging studies. Carhart-Harris presents REBUS as a mnemonic for this framework while emphasising that these proposals remain conceptual and that significant empirical work is needed to validate and refine them.