Models of psychedelic drug action: modulation of cortical-subcortical circuits

Psychedelic drugs that act as agonists or partial agonists at the serotonin 2A (5-HT2A) receptor—examples include psilocybin and lysergic acid diethylamide (LSD)—produce profound alterations in perception and cognition and have renewed interest because of possible therapeutic benefits across mood, substance use and headache disorders. Prior animal and human work implicates cortical 5-HT2A signalling, especially in layer V pyramidal cells, in initiating cascades that alter cortical and subcortical function; complementary findings include increased cortical glutamate, changes in interneuron activity, and the requirement of 5-HT2A signalling for many subjective and behavioural effects of classic psychedelics. Despite this progress, there is no single accepted circuit-level model that accounts for the range of neural and behavioural effects reported, and multiple, sometimes conflicting, models have emerged focusing on different subcortical hubs and mechanisms of network disruption. Doss and colleagues set out to review and compare three circuit-level models of psychedelic action: the cortico‑striatal‑thalamo‑cortical (CSTC) model, the relaxed beliefs under psychedelics (REBUS) model, and a newly proposed cortico‑claustro‑cortical (CCC) model centred on the claustrum. They describe the anatomical and physiological bases of each model, review behavioural and neuroimaging evidence bearing on them, identify inconsistencies and gaps in the literature, and suggest how these frameworks might be revised or integrated to guide future research into both serotonergic and non‑serotonergic hallucinogens.

CSTC model: The CSTC model conceptualises psychedelics as activating medial prefrontal layer V 5-HT2A‑expressing pyramidal neurons that influence ventral striatal GABAergic neurons and downstream pallido‑thalamic inhibition, leading to thalamic disinhibition and an influx of sensory information to cortex. Human neuroimaging studies commonly report that LSD increases functional connectivity between the thalamus and widespread cortical regions under task‑free conditions; directional analyses have shown increased thalamic influence on posterior cingulate cortex that depends on 5-HT2A signalling, while changes in striatal input to the thalamus were reported to be 5-HT2A‑independent. Behavioural assays linked to thalamocortical gating—pre‑pulse inhibition (PPI; a reduction in startle when a weak stimulus precedes a startling stimulus) and inhibition of return (IOR; suppressed processing of a recently attended stimulus)—show mixed effects: animal studies display reductions in PPI with LSD/DOI, but human PPI findings vary with interstimulus interval (reductions at very short intervals, variable or null effects at longer intervals), and DMT studies have failed to show consistent PPI changes. Similarly, DMT attenuates IOR but also blunts primary sensory cortical responses during sustained attention tasks, an observation that could reflect increased neural noise rather than increased sensory throughput. Overall, reports of increases or decreases in prefrontal and thalamic activity under psychedelics are inconsistent, and the authors note that task‑free paradigms and coarse thalamic parcellation may contribute to these discrepancies. The review also emphasises bidirectional corticothalamic projections and the concept of ‘efference copies’—motor-related predictions sent from cortex to thalamus and subcortical motor centres—as additional mechanisms by which psychedelics might generate aberrant predictions and sensory distortions. REBUS model: The REBUS account frames psychedelics as weakening hierarchical top‑down constraints so that bottom‑up prediction errors exert greater influence on higher cortical levels, thereby increasing cortical entropy (in the information‑theoretic sense) and permitting a broader repertoire of brain states. The default mode network (DMN) is posited near the top of this hierarchy and implicated in self‑referential processing; the model links DMN disruption to phenomena such as ‘ego‑dissolution’. Empirical findings partially support REBUS: psychedelics reliably decrease within‑DMN functional connectivity and increase measures of signal complexity or entropy across the brain, and scalp electrophysiology shows desynchronisation of low‑frequency oscillations. However, these effects are often as strong or stronger in other networks (salience, frontoparietal, sensory) than in the DMN, and rodent studies under anaesthesia have not consistently replicated DMN‑selective changes. Contrary to a core REBUS prediction, psilocybin and LSD have been reported to decrease hippocampal and parahippocampal connectivity with the DMN, and some directed‑connectivity analyses have found increased top‑down flow from parahippocampal regions to visual cortex. The authors highlight that treating hippocampal and parahippocampal structures as unitary masks may obscure subregion‑specific effects; they recommend finer parcellation (e.g. entorhinal, perirhinal, parahippocampal cortices; anterior vs posterior hippocampus) to test bottom‑up modulation hypotheses. Behavioural evidence relevant to REBUS is mixed. The model predicts greater impairment of higher‑level cognition than of low‑level perception, and some studies show selective effects (e.g. impaired motion perception but preserved contrast sensitivity). Psychedelics can both impair and enhance aspects of memory and attention depending on the task and subcomponent: episodic recollection (hippocampal‑dependent) is often impaired, whereas familiarity (perirhinal‑dependent) may be preserved or enhanced. Animal data show heterogeneous effects on extinction learning and spatial learning, and measures of cognitive flexibility yield inconsistent results across species and compounds. The authors caution that many findings derive from task‑free paradigms where unconstrained cognitive activity under drugs may drive apparent increases in state diversity. CCC model (claustrum‑centred): The CCC model proposes the claustrum—a thin, ribbon‑like telencephalic nucleus with dense 5-HT2A expression and extensive bidirectional connectivity with cortex—as a key coordinator of cortical network states. In rodents, the claustrum receives prominent inputs from association cortices (prefrontal, cingulate) and provides extensive ipsilateral output to prefrontal, sensory and parietal association areas; optogenetic activation produces transient cortical suppression followed by rebound and widespread synchronisation. The model argues that prefrontal → claustrum → cortex loops enable prefrontal regions to entrain canonical network states (for example DMN or frontoparietal network activity) necessary for goal‑directed cognition, and that psychedelic activation of 5-HT2A receptors in claustrum and prefrontal cortex decouples this entrainment, destabilising network states. Supporting human evidence is limited but suggestive: psilocybin reduces resting‑state claustrum activity and decreases functional connectivity between the claustrum and cortical networks, with lateralised patterns (right claustrum decreases with DMN and auditory networks; left claustrum decreases with frontoparietal network, though an unexpected increase in right claustrum–FPN connectivity was also observed). Reduced claustrum activity correlated with subjective ineffability. The claustrum is functionally coupled with the thalamus, striatum and parahippocampal gyrus, positioning it as a candidate hub intersecting features of the CSTC and REBUS models. The authors note the current human claustrum data are sparse and call for task‑based studies to clarify lateralised findings. Opioid‑receptor hallucinogens: The claustrum also expresses κ‑opioid receptors, the target of atypical hallucinogens such as salvinorin A. Single small fMRI and EEG studies indicate that salvinorin A produces decreases in DMN connectivity, reductions in EEG power across bands, and increased brain entropy—effects that superficially resemble those of classic 5‑HT2A psychedelics. However, the magnitude and spatial pattern of network changes under salvinorin A appear less extensive than with classic psychedelics, and the evidence is currently limited to very small samples. The authors suggest that claustrum disruption could be a common pathway for some ‘psychedelic‑like’ effects across pharmacological classes, but emphasise the preliminary nature of these findings.

Doss and colleagues interpret the literature as indicating that no single circuit model fully explains the multifaceted neural and behavioural effects of psychedelic drugs; instead, CSTC, REBUS and CCC frameworks each capture complementary aspects of psychedelic action. They argue that the CSTC model highlights thalamic gating and disinhibition as a plausible route to increased cortical sensory input, but it underestimates bidirectional corticothalamic signalling and the role of efference‑copy motor predictions that could produce aberrant sensory expectations. Likewise, the REBUS model usefully frames psychedelics as relaxing top‑down priors and increasing cortical entropy, yet it needs clearer anatomical specification of what constitutes higher versus lower hierarchical levels and finer parcellation of medial temporal and thalamic subregions to test bottom‑up signalling claims. The authors advance the CCC model as a third perspective in which the claustrum coordinates prefrontal entrainment of canonical cortical network states; psychedelic‑mediated decoupling of prefrontal–claustral interactions could therefore destabilise network integrity and cognitive control. They note that the claustrum’s dense expression of 5-HT2A and κ‑opioid receptors makes it a plausible locus for both serotonergic and some atypical dissociative hallucinogen effects, but also stress that empirical human data on claustrum function under psychedelics are currently sparse. Methodological limitations recur throughout the literature: many studies rely on task‑free imaging where uncontrolled mental activity may confound interpretations, structures such as the thalamus and medial temporal lobes are often treated as unitary despite functional subregions, sample sizes are small for several key pharmacological fMRI/EEG studies, and causal inferences are limited by correlational imaging methods. The authors therefore recommend more targeted, task‑based paradigms, finer anatomical parcellation, larger samples, and the use of newer methods that can probe the claustrum and subnuclei of relevant structures. Regarding implications, the paper suggests that different forms of network‑state disruption—(1) loss of sensory gating with increased sensory flow, (2) relaxation of predictive codes from past experience, and (3) disengagement of executive control over network states—could each contribute to how psychedelics enable new interpretations of familiar stimuli and potentially facilitate therapeutic change. The authors propose that modulation of maladaptive, inelastic network states observed in clinical conditions such as addiction, depression and obsessive‑compulsive disorder might underlie psychedelic therapeutic effects, but they emphasise that this mechanistic possibility remains to be tested empirically. Finally, the review notes that differential modulation of these circuits may account for both shared and distinct effects across pharmacological classes of hallucinogens, and that examination of non‑serotonergic targets (e.g. NMDA and κ‑opioid receptors) in the claustrum and cortex represents a promising avenue for future work.

The authors conclude that integrating three circuit perspectives—the CSTC, REBUS and CCC models—provides a more complete framework for understanding psychedelic drug action than any single model alone. They recommend revising each model to incorporate bidirectional corticothalamic projections, more precise definitions of hierarchical levels and anatomical subregions, and explicit consideration of the claustrum as a putative network‑state hub. The review identifies three candidate modes by which psychedelics could disrupt network states (gating loss, relaxed predictive codes, and executive disengagement) and suggests these mechanisms may help explain both subjective phenomena and potential therapeutic benefits. Finally, the authors call for focused empirical work using task‑based paradigms, finer anatomical parcellation, larger samples, and emerging methods to directly probe claustrum function to validate and refine these models.