Psychedelics and Invasive Brain–Computer Interfaces as Two “Bypass” Paradigms for Restoring Function: Convergent Learning Principles, Fundamental Differences, and a Roadmap for Hybrid Neurorehabilitation

Abstract

Two rapidly advancing translational strategies, psychedelic-assisted therapy and invasive brain–computer interfaces (iBCIs), are being explored as conceptual strategies to restore, compensate for, or remodel impaired neurological and psychiatric functions, but through fundamentally different implementations of a bypass principle. iBCIs create an extrinsic bypass, engineering an external control loop that directly links neural activity to an effector. In contrast, psychedelics act as intrinsic bypass enablers, transiently reopening plasticity-permissive, critical-period-like windows that allow experience to sculpt new internal pathways. This perspective proposes a conceptual learning framework uniting these approaches: beneath their technological and pharmacological distinctions, both paradigms operate on plasticity-gated learning under constraints. We review preclinical evidence suggesting that psychedelics promote structural plasticity and may modulate critical-period-like windows through mechanisms involving extracellular matrix regulation, alongside iBCI milestones enabling high-performance motor and communication restoration. While the approaches diverge in their precision, immediacy, scalability, and dependency profiles, their shared focus on constrained learning provides a foundation for a hybrid roadmap. We propose testable principles for integrating these strategies, in which pharmacologically opened plasticity windows are paired with structured iBCI training to accelerate skill acquisition, reshape maladaptive priors, and consolidate durable internal control policies, all within essential ethical and safety safeguards.

Graphical abstract

Keywords

psychedelics / critical period reopening / metaplasticity / extracellular matrix / invasive brain–computer interface / neural manifolds

Introduction

When clinicians, neuroscientists, and engineers discuss restoring function after stroke[1,2], spinal cord injury[3], neurodegenerative disease[4,5], or severe paralysis, “bypass” typically refers to circumventing a broken pathway[6]. For instance, stroke-induced hemiparesis is increasingly understood not only as a structural lesion but as a disorder of large-scale network dynamics, where compensatory contralesional hemispheric activation and interhemispheric inhibition can paradoxically suppress residual function in the ipsilesional cortex[7]. Similarly, in major depressive disorder, the therapeutic challenge involves not only localized neurochemical imbalances but also dysfunctional brain–body interactions, where maladaptive somatic feedback loops and chronic low-grade inflammation perpetuate rigid internal models of self[8]. Historically, bypass often implied a mechanical reroute, such as assistive devices or stimulation that substitutes for a failed muscle group.

Invasive brain–computer interfaces (iBCIs) exemplify bypass as hardware[9]. Signals recorded from cortex can be translated into commands for external effectors, effectively creating an engineered bridge from neural activity to output (Fig. 1A)[9,10]. Seminal demonstrations showed reach-and-grasp control of a robotic limb in tetraplegia, turning bypass into a practical clinical concept rather than a metaphor[11]. In parallel, brain–computer interface–functional electrical stimulation (BCI–FES) systems closed the loop from recorded intent to reanimated biological movement, illustrating a direct engineered substitute for interrupted descending pathways[12,13].

Psychedelics propose a qualitatively different bypass. Instead of inserting an external bridge, they may transiently reopen conditions for large-scale learning and reorganization, allowing the nervous system to build new internal routes around damage through experience-dependent plasticity[14,15] (Fig. 1B). Here, we use “intrinsic bypass” as a conceptual rather than anatomical term, referring to experience-dependent functional reorganization that may circumvent maladaptive network constraints without implying the formation of a literal anatomical pathway around a lesion. One influential preclinical study demonstrated reopening of a social-reward critical period in adult mice. Although highly informative mechanistically, its relevance to other behavioral domains, including motor rehabilitation, remains to be established[16].

This perspective proposes that these are not competing stories, but two complementary modalities operating along a shared axis: extrinsic bypass engineering versus intrinsic bypass induction. Their common operational engine is feedback-driven learning under structural constraints. Their core difference lies in where the new functional connection is instantiated (externally in hardware versus internally in neural circuits) and how it is stabilized (decoder convergence and device calibration versus synaptic consolidation and endogenous circuit remodeling).

A unifying language: plasticity-gated learning under constraints

The central scientific question bridging psychedelics and iBCIs is not whether plasticity occurs, but how pharmacological intervention alters the population dynamics of learning. When a user learns an iBCI mapping, performance is fundamentally constrained by whether the required population activity lies within the brain’s existing low-dimensional neural manifold. We formulate two dissociable theoretical mechanisms: either psychedelics induce true manifold expansion (increasing intrinsic dimensionality or geometric flexibility to enable access to previously unavailable activity patterns) or they facilitate subspace traversal (altering excitation to-inhibition [E/I] balance and lowering energetic barriers to accelerate exploration within pre-existing, normally suppressed subspaces). Because both mechanisms could operate concurrently across distinct microcircuits or post-stroke recovery timescales, forcing a statistical adjudication requires evaluating cross-validated shared dimensionality via Factor Analysis or Gaussian Process Factor Analysis (GPFA) and the effective rank of the rate-matched population covariance matrix before versus after drug exposure. Quantifying these metrics under condition-matched controls constitutes the primary theoretical engine of this Perspective.

In an iBCI system, learning is a bidirectional co-adaptation. The user learns to volitionally modulate population neural activity to achieve goals through a decoder, while the algorithm itself adapts to the user’s neural statistics and drift[17,18]. In psychedelic-enabled rehabilitation, learning occurs during a transient, pharmacologically-induced state of heightened plasticity[19]. The substance does not directly encode information. Instead, it shifts the brain’s metaplastic rules, temporarily relaxing molecular and network-level brakes on change[16,20,21]. This creates a permissive window during which therapeutic experience, such as guided therapy or physical practice, can reshape circuits that are typically resistant to revision[16]. In both cases, the “bypass” of neurological limitation is ultimately mediated by the brain’s capacity to learn under a new set of rules.

A powerful framework for understanding these learnable mappings comes from modern neuroscience and the concept of neural manifolds[22,23]. Research in iBCIs reveals that neural population activity is not arbitrary. It is constrained to a low-dimensional subspace, or manifold, shaped by the brain’s inherent connectivity and dynamics[24–27]. Critical studies show that learning is efficient when a task demands neural patterns that lie within this intrinsic manifold, but becomes profoundly difficult when targets fall outside it. This demonstrates that the existing architecture of neural circuits imposes a strict “geometry of learnability” on behaviorally relevant timescales[27,28].

This manifold perspective translates naturally to the psychedelic state[29]. If psychedelics alter excitability, synaptic gain, and network stability, they may effectively relax the constraints of the intrinsic manifold[30]. The central, testable question then becomes: does this pharmacological intervention expand the geometry of learning itself? In other words, do psychedelics increase the dimensionality or flexibility of task-relevant neural manifolds, allowing the exploration of entirely new patterns, or do they primarily accelerate learning within the brain’s existing subspaces? Answering this question with rigorous, measurable neural metrics is a primary objective of the hybrid experimental roadmap proposed later in this paper, providing a common language to quantify how both strategies reshape the brain’s capacity for change[24].

It is also crucial to acknowledge that in many clinical conditions targeted by hybrid rehabilitation, the very plasticity mechanisms upon which both paradigms depend are themselves compromised by disease-specific molecular constraints. For example, in Alzheimer’s disease and related tauopathies, dysregulated alternative splicing of synaptic genes directly impairs the structural plasticity required for experience-dependent learning[31]. Similarly, in major psychiatric disorders, early neurodevelopmental risk factors, including immune-related morphological changes and genomic variations, shape a lifelong vulnerability to maladaptive circuit stabilization, effectively narrowing the dynamic range of the intrinsic manifold even before symptom onset[32]. We hypothesize that adult network remapping is constrained by a “baseline plasticity barrier” maintained by structural extracellular matrix brakes and inhibitory circuit consolidation. Recognizing these barriers is essential for setting realistic expectations for hybrid protocols: in some patients, the primary therapeutic goal may be to restore the manifold’s capacity for local learning, while in others, it may be to compensate for an irreversibly reduced dimensionality through an external BCI scaffold.

To operationalize this manifold expansion, we must identify the biological correlates that permit such geometric flexibility[24]. It is likely that the relaxation of manifold constraints is mediated by a shift in the global E/I balance and a reduction in lateral inhibition[33,34]. By temporarily lowering the threshold for synaptic firing and reducing the “winner-take-all” dynamics of stabilized adult networks, psychedelics may allow the neural population to escape low-dimensional attractors and traverse previously inaccessible regions of state space[30,35,36]. This biological loosening provides the substrate for the increased “dimensionality” required to learn complex, non-native iBCI mappings[28,37]. This interpretation aligns with the Relaxed Beliefs Under Psychedelics (REBUS) framework, where relaxed high-level priors allow bottom-up sensory evidence to dominate, effectively unmasking latent circuits that are ordinarily inhibited by top-down predictions. Operationally, because raw principal component analysis (PCA) and GPFA dimensionality estimates are easily confounded by drug-induced alterations in baseline firing rate, global arousal, and unit yield, isolating true manifold reorganization from non-specific neural noise requires three strict analytical safeguards: (1) rate-matched control sub-sampling (matching trial-averaged firing rate distributions between drug and vehicle conditions); (2) cross-validated dimensionality estimation (e.g., cross-validated Factor Analysis or PCA on held-out trials to prevent overfitting noise); and (3) within-session, condition-matched comparisons tracking task-aligned covariance structure rather than unconstrained variance inflation.

To ensure scientific rigor, this conceptual framework is explicitly falsifiable. The proposed synergy would be disconfirmed under any of the following: pharmacological plasticity induction alters global cortical gain or arousal without modifying task-relevant manifold geometry or accelerating out-of-manifold decoder learning; closed-loop iBCI feedback during the acute/sub-acute window fails to steer plastic reorganization, yielding equivalent outcomes to non-contingent or open-loop stimulation; or the rate and ceiling of iBCI co-adaptation are strictly bottlenecked by structural subcortical wiring regardless of cortical metaplastic state.

Psychedelics as intrinsic bypass enablers

A foundational insight for understanding psychedelics as an intrinsic bypass strategy is their capacity to function as “psychoplastogens”, agents that rapidly promote structural and functional neural plasticity[14,19]. Preclinical evidence demonstrates that certain serotonergic psychedelics, particularly lysergic acid diethylamide (LSD), psilocin, and N,N-dimethyltryptamine (DMT), can enhance neurite outgrowth, dendritic spinogenesis, and synaptic function in prefrontal and hippocampal regions. However, these effects are compound-specific, dose-dependent, and have not been uniformly observed across all brain areas or in all disease models[14]. This is not merely a biochemical detail but a core translational principle: if a brief pharmacological intervention can expand the brain’s inherent capacity to change, then the therapeutic outcome becomes critically dependent on the specific training and experiences—the “dataset”—that the brain encounters during and after this transient window of plasticity[29].

This concept gains support from studies that have moved beyond metaphor to demonstrate the reopening of defined critical-period-like states in specific adult brain circuits[16,38,39]. Critical periods represent developmental windows of heightened experience-dependent plasticity that gradually close through multiple stabilizing mechanisms, including maturation of inhibitory circuits and formation of extracellular structures such as perineuronal nets[40]. In mice, psychedelics were shown to reopen a specific social-reward learning critical period, with the duration of this effect varying among compounds. The relationship between the duration of this mouse plasticity window and the reported duration of subjective drug effects in humans remains an intriguing but unvalidated cross-species comparison[16]. At the molecular level, classic serotonergic psychedelics (e.g., psilocybin, LSD, and DMT), primarily through 5-HT2A receptor signaling, have been linked to extracellular matrix (ECM)-related remodeling, including alterations in perineuronal-net-associated pathways and matrix-modifying enzymes such as MMP-9[14,15,41]. In contrast, non-hallucinogenic psychoplastogens (e.g., tabernanthalog and ibogaine analogues) may promote structural plasticity through mechanisms involving pathways such as TrkB–mTOR signaling, potentially contributing to critical-period-like states[41]. Together, these findings support the possibility that pharmacological interventions can transiently enhance adult neural plasticity in a context-dependent manner, with subsequent experience required to shape and stabilize newly formed circuits. This suggests a two-phase conceptual model: a temporary pharmacological enhancement of learning capacity followed by experience-dependent consolidation into more durable circuit adaptations[28]. However, extrapolating findings from rodent social-reward learning to human motor rehabilitation remains a major translational challenge. Critical-period mechanisms in limbic and social-learning circuits may not directly correspond to the population dynamics governing primate motor learning or iBCI skill acquisition. Therefore, motor-specific preclinical validation will be essential before proposing psychedelic-assisted approaches for neurorehabilitation.

Translational history highlights that pharmacologically enhancing plasticity is insufficient on its own. While initial small-scale studies suggested motor gains with fluoxetine after stroke (FLAME)[42], subsequent large-scale phase III randomized controlled trials involving over 6,000 patients (FOCUS[43], AFFINITY[44], and EFFECTS[45]) showed no improvement in functional motor outcomes, mirrored by earlier null trials pairing amphetamines with physiotherapy[46,47]. These failures highlight a fundamental bottleneck: broad, systemic plasticity elevation without precise, high-frequency reinforcement signals leads to unguided synaptic updating and uncoordinated circuit remodeling. The proposed hybrid framework directly addresses this failure mode by replacing unstructured post-stroke care with dense, closed-loop iBCI training. Here, the iBCI supplies real-time, objective error signals and adaptive reinforcement contingencies that actively steer and consolidate the pharmacologically opened plasticity window toward functional target manifolds.

At a systems level, computational models provide a complementary explanation[48,49]. From a predictive processing perspective, learning is constrained not only by synaptic potential but by the strength of prior beliefs[50]. The REBUS model posits that psychedelics relax the precision weighting of these high-level priors, increasing the brain’s receptivity to new, bottom-up sensory evidence and prediction errors[29]. This is directly relevant to rehabilitation, where maladaptive priors, like learned non-use after stroke or catastrophic beliefs about pain, can functionally block learning and exploration, even when underlying biochemical plasticity is possible[51,52]. Thus, psychedelics may act as a computational bypass by broadening the brain’s search of the solution space and enabling the discovery of compensatory strategies that can then be stabilized[36].

A significant translational frontier involves efforts to decouple plasticity promotion from the acute hallucinogenic experience through the development of non-hallucinogenic analogues[53,54]. The success of such candidates sharpens a key question: to what extent is the subjective experience necessary for durable functional gains? The answer is likely condition-specific[55]. While the phenomenological journey may be integral to psychotherapeutic reappraisal in mental health, it may be less critical, or even undesirable, in domains like motor rehabilitation. A practical hybrid model can accommodate this spectrum by treating the subjective state not as a universal requirement, but as one potential training signal among others, allowing the plasticity-enabling mechanism to be leveraged according to clinical need and context.

A critical distinction remains regarding whether structural plasticity alone, as promoted by non-hallucinogenic psychoplastogens, is sufficient for all bypass objectives[53,54,56]. While motor rehabilitation might rely primarily on the “bottom-up” reopening of critical periods and spinogenesis, recovery from conditions involving high-level cognitive constraints, such as learned non-use or chronic pain, may require the “top-down” phenomenological experience to effectively relax the precision weighting of maladaptive priors[28,38,51,57]. Future hybrid protocols must therefore match the agent to the pathology: using sub-perceptual plasticity enhancers for fine-motor iBCI tuning and hallucinogenic doses when the primary bottleneck is a rigid, maladaptive internal model.

Translating these principles into a pragmatic first-in-human protocol, however, requires an agent that balances plasticity-promoting potential with established human safety data and regulatory familiarity. In this context, S-ketamine and racemic ketamine represent the most immediate, clinically established human precedent for hybrid paradigms. As a Food and Drug Administration (FDA)-approved N-methyl-D-aspartate (NMDA) receptor antagonist with well-characterized psychoplastogenic properties, driving rapid dendritic spinogenesis via brain-derived neurotrophic factor (BDNF) release and mTORC1 activation, ketamine is already routinely administered in perioperative and surgical environments, eliminating major safety unknowns regarding acute physiological monitoring in neurosurgical populations[20,58]. Furthermore, clinical protocols already pair ketamine with structured psychotherapeutic or cognitive training during its post-acute window to consolidate functional gains[16]. Incorporating sub-anesthetic ketamine as an initial pharmacological probe in iBCI-implanted patients presents a substantially lower regulatory, safety, and logistical barrier than classical 5-HT2A agonists such as psilocybin, providing an ideal, pragmatic “first-in-human” testbed for pharmacologically assisted neuroprosthetic co-adaptation.

iBCIs as extrinsic bypass engineering

iBCIs operate as an extrinsic bypass by creating a direct functional link, a new input–output chain, that circumvents damaged neural pathways[11,59,60]. Their invasive nature is foundational, achieved through surgically implanted electrodes that provide the high spatial resolution and signal fidelity necessary to decode complex movement and speech intentions[13,61–63]. This characteristic, however, introduces significant translational considerations, including the need for neurosurgery, long-term device biocompatibility, and the clinical infrastructure for maintenance. Together, these considerations define both their high-bandwidth potential and their practical challenges[64–66].

The most direct manifestation of this bypass is in motor restoration. Landmark studies have demonstrated that intracortical signals can control robotic limbs for reach-and-grasp functions in individuals with tetraplegia[67]. Crucially, this concept extends beyond external devices to biological actuators; research has successfully restored volitional arm and hand movements by using decoded cortical activity to electrically stimulate paralyzed muscles[13]. In these cases, the iBCI functions as a literal bridge, constructing an alternative pathway from intention to action.

Similarly transformative progress has been made in communication restoration. iBCIs can now decode attempted speech into text and synthesized audio at accelerating rates, enabling real-time communication for individuals with paralysis[68–70]. This highlights that the extrinsic bypass is not limited to motor functions. It can restore a fundamental channel for language and social connection, directly impacting identity and quality of life.

Reaching high-level performance with an iBCI is a dynamic learning process shaped by co-adaptation[71]. Effective use rarely appears immediately. Instead, the user gradually discovers reliable, repeatable ways of modulating population activity while the decoding algorithm tunes to neural features and compensates for signal drift[72]. As established in Section 2 (A unifying language: plasticity-gated learning under constraints), learning efficiency is governed by manifold constraints: decoder adjustments within the user’s native subspace are acquired rapidly, whereas mappings requiring patterns outside that manifold require extensive training. Over repeated closed-loop practice, user strategies and decoder mappings converge, consolidating dependable control over the external device.

From a materials perspective, the electrode–tissue interface is a first-order constraint on iBCI performance. Traditional rigid electrodes (e.g., silicon-based arrays, metal microwires) exhibit a profound mechanical mismatch with brain tissue (electrode elastic modulus > 100 GPa vs. brain tissue ~1–3 kPa), triggering chronic neuroinflammatory responses and glial encapsulation that drive progressive signal degradation[64,65,73]. Recent advances in hydrogel-based neural probes with tissue-matching mechanical and chemical properties offer a promising solution, demonstrating continuous multi-week brain signal monitoring and substantially reduced microglial activation in vivo[73,74]. These materials-level innovations are directly relevant to hybrid neurorehabilitation protocols, as the recording interface determines the fidelity and longevity of the training signals available to guide plasticity-biased learning. Looking further ahead, utilizing the inherent invasiveness of the iBCI to integrate multimodal neurochemical sensing, capable of tracking neurotransmitters, metabolites, or inflammatory markers alongside electrophysiological activity, could provide a direct, real-time readout of the brain’s receptivity to learning during the critical window, enabling adaptive, closed-loop optimization of training parameters[75,76].

Deep commonalities and a comparative framework

The most productive comparison is not “psychedelics versus devices,” but how each paradigm amplifies learning under constraint. iBCIs supply an external scaffold that renders specific functions learnable by providing immediate, quantifiable feedback and a controllable mapping between neural activity and an output channel[11,59]. Psychedelics, by contrast, induce an internal permissive state that renders certain reorganizations learnable by temporarily shifting the rules and thresholds that govern plasticity itself[16,30]. Within this framework, population manifold dynamics (Section 2) and ECM structural brakes (Section 3: Psychedelics as intrinsic bypass enablers) represent complementary descriptions of network constraint.

Despite this shared learning-dependent route to functional bypass, iBCIs and psychedelics diverge fundamentally in where the bypass is instantiated (Table 1). iBCIs construct an external, hardware-dependent bridge spanning electrodes, decoders, and effectors[11,59,67]. Psychedelic-enabled interventions target endogenous circuit remodeling to yield self-sustaining network updates that persist after drug clearance[14,16,56].

This distinction also clarifies what is being bypassed. Psychedelics tend to bypass maladaptive internal constraints, such as rigid threat models, depressive rumination, compulsive habits, or learned non-use, by enabling corrective experience and reconsolidation during periods of heightened flexibility[77,78]. The obstruction lies within inference, valuation, or control systems that block exploration even when biochemical plasticity is available[49]. iBCIs, by contrast, bypass structural transmission failures by routing intent-related signals around damaged pathways and directly coupling preserved neural representations to an output channel[13,59]. In both cases, learning is the engine, but the dominant bottleneck differs: psychedelics loosen constraints on updating, whereas iBCIs replace a broken conduit while leaving internal inference largely intact[79].

The paradigms also differ in operating principle, namely precision versus broad self-organization. iBCIs can be engineered for high functional specificity, mapping neural activity to discrete outputs such as cursor trajectories, phonemes, or multi-degree-of-freedom limb kinematics, which makes their effects direct and measurable[11,62]. Psychedelics are broad-spectrum plasticity modulators that lower constraints on learning across multiple systems simultaneously. This breadth is double-edged because plasticity is not inherently therapeutic. It amplifies whatever learning occurs, making context, task design, and psychosocial environment central determinants of outcome[4,78,80].

Timescales and dependency further separate the two. iBCIs can yield usable function within sessions after calibration and can provide on-demand capability with continued use, but they impose ongoing dependency on external hardware and clinical or technical infrastructure[59]. Their long-term durability depends jointly on tissue–device stability and decoder robustness[81]. Psychedelic-enabled change follows a delayed consolidation logic: a brief window of enhanced flexibility opens, learning occurs during and after that window, and durable gains crystallize over subsequent days to weeks[16]. The explicit aim is to reduce long-term dependence by cultivating self-sustaining internal circuitry.

Clinically, this distinction calls for objective stratification tools. Functional and structural neuroimaging can quantify the integrity of residual pathways: diffusion tensor imaging (DTI) can assess the continuity of corticospinal tracts, while resting-state or task-based functional magnetic resonance imaging (fMRI) can map the viability of motor-related functional networks. For a patient with a complete spinal cord transection, DTI-confirmed absence of anatomical continuity would argue decisively for an extrinsic iBCI-first approach. Conversely, in a stroke survivor with partially preserved fibers but hemiparesis, fMRI may reveal maladaptive interhemispheric inhibition or pathological activation patterns indicative of a “policy-limited” state, suggesting that an intrinsic plasticity-based approach, potentially hybridized with iBCI-guided training, could reactivate the dormant network. Emerging blood- and cerebrospinal fluid-based biomarkers of neuroplasticity may further refine patient selection by indexing the brain’s intrinsic capacity to respond to plasticity-promoting interventions. Incorporating such biomarkers into a clinical decision algorithm would transform the conceptual bypass taxonomy into a personalized treatment paradigm. Furthermore, while fluid biomarkers (such as BDNF or MMP-9) provide valuable retrospective biochemical confirmation, their slow turnover limits real-time clinical utility. Direct electrophysiological readouts recorded via the iBCI itself—such as local field potential (LFP) spectral slope, aperiodic exponents (1/f dynamics), or real-time E/I balance metrics—offer immediate, continuous proxies for tracking the precise onset and closure of the pharmacologically opened plasticity window.

Finally, the approaches carry distinct risk and burden profiles that shape responsible deployment. Psychedelics entail physiological and psychological risks and, critically, context-dependent adverse outcomes, making screening, protocol control, and guided environments integral rather than auxiliary[82,83]. Invasive iBCIs introduce surgical risks, chronic interface challenges, long-term maintenance demands, and unresolved issues of neural data privacy and security[84,85]. These differences do not undermine their shared learning foundation. They explain why the same conceptual engine, plasticity-gated learning driven by feedback, yields different practical realities depending on whether the bypass is built within neural circuits or sustained as a persistent external loop.

The choice between these paradigms depends heavily on the nature of the neurological lesion. In cases of complete structural discontinuity, such as total spinal cord transection, the external conduit of an iBCI is non-negotiable as there is no remaining biological substrate to remodel[13]. Conversely, in “policy-limited” disabilities such as post-stroke hemiparesis or focal dystonia, where the physical pathways are partially preserved but functionally silenced by maladaptive learning, the intrinsic psychedelic bypass may take precedence[4,57]. Recent work supports this view by revealing that stroke recovery involves complex compensatory dynamics, including contralesional hyperexcitability and aberrant interhemispheric inhibition, that can be reversed through plasticity-based interventions[7]. Similarly, the efficacy of psychedelic therapy in depression may stem from its ability to disrupt ingrained brain–body feedback cycles and re-establish flexible, context-appropriate autonomic regulation[8]. Beyond these two categories, neurodegenerative conditions add further complexity: molecular constraints on plasticity, such as disease-induced splicing dysregulation or chronic neuroinflammation, can render anatomically intact pathways functionally inaccessible, thereby blurring the boundary between structural and policy-limited categories. Hybridization, i.e., using an iBCI as an “artificial scaffold” to guide the broad, pharmacologically-induced plasticity, is most potent for the middle ground, partially preserved networks in which intrinsic and extrinsic strategies can synergize to drive functional recovery.

Do psychedelics have unique advantages as bypass enablers?

The intrinsic nature of psychedelic-enabled bypass suggests plausible advantages: the possibility of durable network change without permanent hardware, greater integration with motivation and meaning, and potentially broader generalization across contexts because the learned function is encoded within endogenous circuitry rather than within a device-specific mapping[14]. The social critical-period reopening results also suggest a privileged interface with interpersonal learning contexts, which may matter for rehabilitation adherence and identity reconstruction after injury[16]. In addition, accumulating evidence suggests that psychedelics may also exert therapeutic effects on phantom limb pain and related chronic pain states, indicating that their utility may extend beyond “policy-limited” functional impairments to patients with somatic loss or sensorimotor representational mismatch[86–88].

At the same time, these strengths imply unique responsibilities. If plasticity is globally amplified, maladaptive learning can also be amplified, especially under stress, unsafe environments, or poorly designed training regimes[77]. For example, in an iBCI context, heightened plasticity could inadvertently strengthen off-target neural patterns associated with compensatory but inefficient strategies, leading to unstable control or the consolidation of aberrant motor programs. This risk is not merely hypothetical; it has been observed in animal models of excessive plasticity induction where uncontrolled environmental input generated persistent dystonic postures[89]. For intrinsic bypass strategies, “set and setting” should be treated not as soft psychology but as hard experimental design: it defines what the nervous system samples during a period of enhanced learnability[82,83].

A roadmap for hybrid “plasticity-gated iBCI rehabilitation”

A hybrid strategy, currently a conceptual proposal that remains to be empirically validated, can synergize their distinct strengths (Table 1). The core concept leverages a powerful complementarity: iBCIs excel at providing structured, high-frequency error feedback and precise performance metrics, while psychedelics can temporarily relax the brain’s biological constraints on reorganization (Fig. 1C). In this framework, the pharmacologically opened plasticity window increases network malleability, while the closed-loop iBCI serves as a precision instructor to steer structural remapping toward functional target manifolds. To transform this conceptual framework into an actionable research program, we present a structured translational roadmap anchored by a concrete preclinical paradigm, explicitly falsifiable hypotheses, and proposed clinical stratification principles.

An ideal initial paradigm involves evaluating iBCI skill acquisition in a rodent model under a closed-loop perturbation task. Adult rats implanted with high-density neural probes (e.g., Neuropixels or multi-shank hydrogel arrays that can simultaneously track at least 80–100 simultaneous single- and multi-units across motor cortical layers) can be trained to execute a center-out reaching task via a neural decoder[90,91]. By introducing a within-manifold or outside-manifold decoder perturbation, the paradigm can directly test whether pharmacological plasticity facilitates adaptation to control policies that are difficult to acquire under baseline manifold constraints. The task becomes substantially more difficult to learn within a single session[27]. Animals would receive either a classic psychedelic (e.g., psilocybin, 0.3 mg/kg intraperitoneally [i.p.]), a non-hallucinogenic psychoplastogen (e.g., tabernanthalog, 50 mg/kg i.p.), or vehicle[16,53]. Crucially, to prevent acute motor discoordination and population statistical non-stationarity from destabilizing decoder calibration, intensive decoder training is decoupled from peak acute drug effects and delivered during the post-acute metaplastic window. While online adaptive decoding is employed during initial co-adaptation and calibration phases, decoder parameters must be held strictly fixed during performance evaluation test blocks. Holding the decoder fixed during testing isolates genuine neural learning and manifold reorganization from algorithmic co-adaptation, thereby preventing a “moving target” confound where adaptive algorithms merely track transient neural noise or acute drug-induced non-stationarities. These doses should be treated as provisional starting points and would require pilot dose–response studies to identify behaviorally non-disruptive windows, with locomotor activity, motivation, arousal, and task engagement measured as control endpoints. Population dynamics (evaluated via cross-validated PCA and GPFA under rate-matched control sub-sampling within individual sessions) along with primary behavioral metrics (such as learning curve slope, trial-to-expert velocity, and retention after a 14-day drug-free washout), will directly adjudicate whether the pharmacological intervention accelerates learning by facilitating traversal within pre-existing subspaces or by expanding manifold geometry. To ensure sufficient statistical power for low-dimensional manifold estimation, rodent implementations must strictly maintain this high single-unit recording yield across sessions. While high-density recordings in rodents offer a practical high-throughput screening platform, rodent motor cortex exhibits significantly lower dimensional complexity than that of primates. Consequently, non-human primates (NHPs) remain the essential translational gold standard for demonstrating true out-of-manifold geometric shifts and higher-dimensional manifold expansion prior to initiating human clinical trials.

Building upon this benchmark, our first roadmap hypothesis is that pharmacologically induced plasticity accelerates out-of-manifold decoder acquisition by relaxing local excitation-to-inhibition constraints. The hypothesis is directly testable in closed-loop animal models, in which learning rates, achievable decoder rotations, and manifold geometry can be quantified before, during, and after a plasticity intervention to determine whether the network transiently explores new states or consolidates a lasting change. If psychedelics transiently expand the neural manifold, animals receiving psychoplastogens should exhibit improved acquisition of decoder mappings that require neural activity outside the intrinsic manifold, accompanied by measurable increases in manifold dimensionality or expansion of the accessible neural state space. Conversely, if psychoplastogen-treated animals show improved learning only for decoder mappings that remain within the pre-existing manifold, without measurable changes in neural manifold geometry, the data would support facilitated traversal of existing neural subspaces rather than manifold expansion.

Second, iBCIs could provide the structured, dense error signals required to guide the broad, non-specific plasticity induced by psychedelics. Because psychedelic plasticity amplifies whatever learning occurs, the quality of the concurrent experience is paramount[92]. An iBCI is uniquely positioned to fulfill this need by delivering continuous, quantifiable feedback, adaptive task difficulty, and precisely timed reinforcement[71]. Thus, within a hybrid protocol, the iBCI transcends its role as a mere output device to become an instructional system that actively sculpts network reorganization, steering the brain away from maladaptive attractors and toward functional control policies during the critical window. This hypothesis predicts that psychoplastogen-treated animals will acquire stable decoder control more rapidly than controls under identical reinforcement contingencies, while preserving the specificity of learned motor strategies. If decoder performance does not improve despite pharmacologically enhanced plasticity, or if learning becomes more variable, unstable, or maladaptive, the proposed role of psychedelics in facilitating reinforcement-guided optimization would not be supported.

A third, long-term hypothesis is that psychedelics may enhance the consolidation and partial internalization of device-mediated skills. Beyond proficient device operation, a key rehabilitation goal is the development of durable, internalized control policies that reduce dependency on external calibration. Psychedelic-enabled plasticity could amplify the natural consolidation processes observed in long-term iBCI use, leading to more robust neural remapping[14,79]. This would manifest as reduced reliance on continuous decoder adaptation and increased resilience to neural signal drift. Even partial internalization, where a user becomes less dependent on a fragile set of neural features, would represent a meaningful step toward bridging an external interface with an endogenous capability. If psychedelics facilitate consolidation, decoder skills acquired during the period of enhanced plasticity should exhibit greater long-term retention and reduced dependence on external assistance after the pharmacological effects have resolved. If performance improvements disappear once drug exposure or decoder assistance is withdrawn, or if long-term retention is indistinguishable from controls, the hypothesis that psychedelics promote durable consolidation would not be supported.

Potential translation to human populations would require careful pathological stratification rather than broad generalization. Two populations that may warrant future investigation, owing to their partially preserved neural substrates, are chronic ischemic post-stroke upper-limb hemiparesis and motor-incomplete spinal cord injury (American Spinal Injury Association Impairment Scale [AIS] C or D). In both scenarios, the primary barrier to recovery may include maladaptive learning in addition to residual structural disruption (e.g., learned non-use or contralesional hyperexcitability) rather than absolute anatomical discontinuity. A future exploratory hybrid study could potentially incorporate objective patient characterization using approaches such as DTI to evaluate residual tract integrity and fMRI to assess functional network viability; administration of a behaviorally non-disruptive dose of a non-hallucinogenic psychoplastogen, or a carefully supervised low-dose psychedelic regimen, to promote plasticity while minimizing acute cognitive disruption; and intensive, closed-loop iBCI-guided neuromodulation targeting the ipsilesional motor networks[73,74,93]. This combination leverages the extrinsic iBCI to deliver high-fidelity, repeatable error signals, while the pharmacologically opened intrinsic bypass increases the likelihood that these training signals are consolidated into more durable and self-sustaining internal control policies.

Translating these hypotheses into theoretical clinical protocols highlights practical considerations centered on timing, safety, and agent selection. To eliminate dosing timing confounds and prevent acute population instability from degrading decoder performance, experimental protocols must explicitly decouple safety monitoring from skill acquisition across two separate operational arms: (1) an acute/sub-acute phase (0–4 h post-dose) dedicated to physiological monitoring, safety interlocks, and mapping acute state-space drift without forcing complex task co-adaptation; and (2) a post-acute metaplastic phase (12–72 h post-dose) dedicated to intensive closed-loop iBCI training during the peak structural plasticity window, fully free from acute perceptual or motor disruption. During this post-acute window, clinicians can utilize real-time electrophysiological signatures derived directly from the BCI interface (e.g., LFP spectral slope and E/I ratio metrics) alongside systemic biomarkers to dynamically track plasticity state transitions and optimize the timing and duration of intensive training blocks. Proposed frameworks would align this post-acute plasticity-enhancing window with structured training blocks, followed by protected consolidation periods where sleep plays a key role[94–96]. During acute metaplastic states, task selection prioritizes safety by pairing conservative motor tasks with closed-loop safety interlocks within the BCI control loop[82]. Specifically, implementing conservative decoder learning rates and automated task-difficulty scaling prevents heightened malleability from inadvertently reinforcing noisy or sub-optimal compensatory strategies[97]. Furthermore, agent selection could be tailored to the specific rehabilitation target: non-hallucinogenic plasticity enhancers may be preferable for targeted motor training where minimal phenomenological disruption is ideal, whereas subjective experience may remain relevant for domains requiring cognitive reappraisal[53]. The overarching principle is a co-designed framework in which the pharmacological compound and the BCI training regimen are systematically matched to the rehabilitation objective.

A rigorous evaluation of this hybrid roadmap requires anticipating scenarios where theoretical synergy breaks down. First, the pharmacological intervention may alter global arousal or cortical gain without expanding population state-space dimensionality, leaving out-of-manifold decoder acquisition unimproved over baseline. Second, unsteered metaplasticity risks accelerating the consolidation of maladaptive compensatory movements, task-specific dystonias, or pathological attractor states. Third, acute drug-induced synaptic remodeling can trigger severe neural representation drift. If adaptive decoders fail to track this non-stationarity, real-time closed-loop control will collapse under elevated decoding error noise. Finally, disease-specific blunting from severe neurodegeneration or chronic neuroinflammation may impair key signaling cascades, such as 5-HT2A or TrkB–mTOR pathways, preventing pharmacological critical-period reopening altogether. Under these alternative outcomes, functional recovery remains strictly bottlenecked by pre-existing circuit constraints or entirely dependent on extrinsic hardware compensation.

Safety, ethics, and neurorights

Hybrid interventions combine the risks of both domains: surgical and device-security risks on one side, altered-consciousness risks and psychiatric screening on the other[98]. Governance frameworks already emphasize responsible innovation in neurotechnology, including values such as safety, privacy, and stewardship, and the need for oversight and societal deliberation[99]. Reports by the Nuffield Council on Bioethics (nuffieldbioethics.org/publication/novel-neurotechnologies-intervening-in-the-brain/) and UNESCO have further highlighted the ethical and social issues raised by technologies that intervene in the brain, particularly questions of agency and identity, which are especially salient for BCIs[100]. These existing frameworks provide a foundation, yet the convergence of iBCIs and psychedelics raises unique challenges that demand additional ethical specificity.

Chief among these is the paradoxical interplay between “technological embodiment”, the process by which an external device is integrated into the self-schema, and the “ego-dissolution” frequently induced by psychedelics, a transient thinning of the boundaries between self and environment. A hybrid intervention could, in principle, accelerate the embodiment of a robotic limb by relaxing the rigid prior of the biological body-map. This prospect, while therapeutically promising, necessitates strict oversight to ensure that the “augmented self” remains under the user’s volitional control and that the heightened suggestibility inherent to the psychedelic state does not compromise long-term autonomy. Beyond these psychosocial concerns, the hybrid paradigm also confronts a more immediate physiological safety threshold that has never been empirically tested. Crucially, no human patient with an implanted iBCI has ever received a psychedelic agent. A percutaneous pedestal, active surgical wound, or recent craniotomy represents a significant relative contraindication due to risks of altered blood–brain barrier permeability, elevated intracranial pressure during acute autonomic surges, retrograde infection, or cerebrospinal fluid (CSF) leakage. We therefore propose a mandatory minimum post-implantation stabilization window (e.g., ≥ 6–12 weeks post-surgery) to ensure complete dural and scalp healing, stable tissue encapsulation, and baseline electrophysiological signal stationarity prior to initiating any pharmacological plasticity protocol.

Maintaining such control requires rigorous clinical governance. Future clinical investigations would need to incorporate appropriate psychiatric screening (e.g., Mini International Neuropsychiatric Interview (MINI), Structured Clinical Interview for DSM-5 (SCID-5)) to evaluate potential risks associated with psychedelic exposure, including histories of psychosis, bipolar disorder, or uncontrolled cardiovascular conditions. Intraoperative and perioperative risk management must account for both the surgical risks of iBCI implantation and potential drug–device interactions, including altered seizure thresholds or autonomic instability. Continuous physiological monitoring (electrocardiogram [ECG], blood pressure, and pulse oximetry) would likely be an important consideration during future drug-administration studies, together with appropriate medical support. A structured long-term follow-up schedule should include periodic neuropsychological evaluation, device integrity assessments, and surveillance for delayed-onset neuropsychiatric, cardiovascular, or device-related adverse effects; for repeated dosing regimens, receptor-profile-dependent cardiovascular risks should also be considered. These safeguards are integral to the ethical viability of the hybrid approach.

Furthermore, the temporal overlap of acute pharmacological alterations and prolonged, invasive technological embodiment introduces unique challenges to the traditional framework of informed consent. Because a patient’s cognitive, emotional, and suggestible states will naturally fluctuate across different phases of the protocol (baseline clarity, acute psychedelic exposure, and intensive BCI interaction), a static, one-time consent procedure is ethically insufficient. We therefore propose that future studies consider models of dynamic, ongoing consent. This framework requires: an independent pre-enrollment capacity assessment; a detailed advance directive defining acceptable boundaries for decoder co-adaptation and automated training intensity during altered states; continuous intra-session monitoring of neural and behavioral markers of volitional agency; and the integration of pre-arranged “assent signals”—such as a specific, low-threshold BCI command that the patient can deploy to instantly pause the session. Consent is thus operationalized not as a bureaucratic event, but as a continuous, shared decision-making process that safeguards patient autonomy in real time.

Beyond the clinical encounter, the hybrid paradigm raises broader questions of data governance and social justice. Cybersecurity must be treated as a clinical safety issue rather than a purely technical concern, given the risks associated with unauthorized access to neural data or device control[101]. Data governance frameworks must clearly specify ownership, access, sharing, and retention of neural data, recognizing its uniquely intimate and potentially identity-relevant nature. Finally, equity concerns are central rather than ancillary: combining two resource-intensive modalities risks creating a “boutique” therapy unless access, scalability, and distributive justice are incorporated into design and policy from the outset.

Conclusions

The defining difference between psychedelics and invasive iBCIs is where the newly acquired capability ultimately resides. Psychedelic interventions aim to leave behind an internally rewritten network that persists after drug clearance[16,78]. In contrast, iBCIs instantiate function within a coupled human–machine loop. This loop can be refined, partially embodied, and expanded, yet it remains partly external by design. The distinction has immediate implications for indication selection, expectations, and endpoints. When anatomy makes biological output impossible, channel-first solutions may remain essential. When maladaptive dynamics and constrained flexibility represent dominant barriers, plasticity-oriented approaches may represent a hypothesis worth investigating. The hybrid frontier is thus plausible but delicate. It is a theoretical framework that still awaits direct clinical validation, yet one motivated by converging observations from neuroscience, neurotechnology, and plasticity research. iBCI proficiency is, at its core, a plasticity problem, and plasticity-biased states may accelerate both skill acquisition and retention. Yet heightened malleability is a double-edged sword: it can equally amplify harmful or noisy learning signals. The translational opportunity, therefore, lies not in simple combination, but in combination governed by rigorous measurement and control: closed-loop metrics to quantify learning and stability, careful context and training design to shape the teaching signal, and explicit governance to protect agency and neural data.

References

[1] Yang Y, Wang Y, Wang X. Harnessing psychedelics for stroke recovery: therapeutic potential and mechanisms. Brain. 2025;148(6):1862-1865.

[2] Guida L, Sebök M, Wegener S, et al. Flow-augmentation bypass in the treatment of acute ischemic stroke. J Neurosurg Sci. 2021;65(3):269-276.

[3] Becker D, Sadowsky CL, McDonald JW. Restoring function after spinal cord injury. Neurologist. 2003;9(1):1-15.

[4] Lin C, Du X, Wang X. A perspective on Alzheimer’s disease: exploring the potential of terminal/paradoxical lucidity and psychedelics. Mol Neurodegener. 2024;19(1):72.

[5] He E, Chen K, Liu S, et al. Advances in neuroprostheses: interfaces, materials, and applications. Nano Converg. 2026;13(1):22.

[6] Daly JJ, Wolpaw JR. Brain–computer interfaces in neurological rehabilitation. Lancet Neurol. 2008;7(11):1032-1043.

[7] Fan H, Wang H, Lian Z, et al. Dynamic interactions between hemispheres reveal a compensatory pathway for motor recovery in moderate-to-severe subcortical stroke. J Stroke. 2026;28(1):97-114.

[8] Lian Z, Liu Z, Fan H, et al. Multimodal fusion of brain imaging and proteomics reveals a brain–body pathway linking depression and metabolic dysfunction. Psychol Med. 2026;56:e144.

[9] Vidal JJ. Toward direct brain–computer communication. Annu Rev Biophys Bioeng. 1973;2:157-180.

[10] Vidal JJ. Real-time detection of brain events in EEG. Proc IEEE. 1977;65:633-641.

[11] Hochberg LR, Serruya MD, Friehs GM, et al. Neuronal ensemble control of prosthetic devices by a human with tetraplegia. Nature. 2006;442(7099):164-171.

[12] Bouton CE, Shaikhouni A, Annetta NV, et al. Restoring cortical control of functional movement in a human with quadriplegia. Nature. 2016;533(7602):247-250.

[13] Ajiboye AB, Willett FR, Young DR, et al. Restoration of reaching and grasping movements through brain-controlled muscle stimulation in a person with tetraplegia: a proof-of-concept demonstration. Lancet. 2017;389(10081):1821-1830.

[14] Ly C, Greb AC, Cameron LP, et al. Psychedelics promote structural and functional neural plasticity. Cell Rep. 2018;23(11):3170-3182.

[15] Fawcett JW, Oohashi T, Pizzorusso T. The roles of perineuronal nets and the perinodal extracellular matrix in neuronal function. Nat Rev Neurosci. 2019;20(8):451-465.

[16] Nardou R, Sawyer E, Song YJ, et al. Psychedelics reopen the social reward learning critical period. Nature. 2023;618(7966):790-798.

[17] Vidaurre C, Sannelli C, Müller KR, Blankertz B. Co-adaptive calibration to improve BCI efficiency. J Neural Eng. 2011;8(2):025009.

[18] Abu-Rmileh A, Zakkay E, Shmuelof L, Shriki O. Co-adaptive training improves efficacy of a multi-day EEG-Based motor imagery BCI training. Front Hum Neurosci. 2019;13:362.

[19] Calder AE, Hasler G. Towards an understanding of psychedelic-induced neuroplasticity. Neuropsychopharmacology. 2023;48(1):104-112.

[20] Moliner R, Girych M, Brunello CA, et al. Psychedelics promote plasticity by directly binding to BDNF receptor TrkB. Nat Neurosci. 2023;26(6):1032-1041.

[21] de Vos CMH, Mason NL, Kuypers KPC. Psychedelics and neuroplasticity: a systematic review unraveling the biological underpinnings of psychedelics. Front Psychiatry. 2021;12:724606.

[22] Shenoy KV, Sahani M, Churchland MM. Cortical control of arm movements: a dynamical systems perspective. Annu Rev Neurosci. 2013;36:337-359.

[23] Cunningham JP, Yu BM. Dimensionality reduction for large-scale neural recordings. Nat Neurosci. 2014;17(11):1500-1509.

[24] Gallego JA, Perich MG, Miller LE, Solla SA. Neural manifolds for the control of movement. Neuron. 2017;94(5):978-984.

[25] Churchland MM, Cunningham JP, Kaufman MT, et al. Neural population dynamics during reaching. Nature. 2012;487(7405):51-56.

[26] Martinez CA, Wang C. Structural constraints on learning in the neural network. J Neurophysiol. 2015;114(5):2555-2557.

[27] Sadtler PT, Quick KM, Golub MD, et al. Neural constraints on learning. Nature. 2014;512(7515):423-426.

[28] Oby ER, Golub MD, Hennig JA, et al. New neural activity patterns emerge with long-term learning. Proc Natl Acad Sci U S A. 2019;116(30):15210-15215.

[29] Carhart-Harris RL, Friston KJ. REBUS and the anarchic brain: toward a unified model of the brain action of psychedelics. Pharmacol Rev. 2019;71(3):316-344.

[30] Carhart-Harris RL, Erritzoe D, Williams T, et al. Neural correlates of the psychedelic state as determined by fMRI studies with psilocybin. Proc Natl Acad Sci USA. 2012;109(6):2138-2143.

[31] Ran X, Wang M, Huang J, et al. Mechanistic research and therapeutic prospects of alternative splicing in neurodegenerative diseases. Ageing Res Rev. 2026;118:103133.

[32] Kuang N, Liu Z, Yu G, et al. Neurodevelopmental risk and adaptation as a model for comorbidity among internalizing and externalizing disorders: genomics and cell-specific expression enriched morphometric study. BMC Med. 2023;21(1):291.

[33] Yizhar O, Fenno LE, Prigge M, et al. Neocortical excitation/inhibition balance in information processing and social dysfunction. Nature. 2011;477(7363):171-178.

[34] Isaacson JS, Scanziani M. How inhibition shapes cortical activity. Neuron. 2011;72(2):231-243.

[35] Marder E, Goaillard JM. Variability, compensation and homeostasis in neuron and network function. Nat Rev Neurosci. 2006;7(7):563-574.

[36] Breakspear M. Dynamic models of large-scale brain activity. Nat Neurosci. 2017;20(3):340-352.

[37] Golub MD, Sadtler PT, Oby ER, et al. Learning by neural reassociation. Nat Neurosci. 2018;21(4):607-616.

[38] Hensch TK. Critical period plasticity in local cortical circuits. Nat Rev Neurosci. 2005;6(11):877-888.

[39] Takesian AE, Hensch TK. Balancing plasticity/stability across brain development. Prog Brain Res. 2013;207:3-34.

[40] Pizzorusso T, Medini P, Berardi N, Chierzi S, Fawcett JW, Maffei L. Reactivation of ocular dominance plasticity in the adult visual cortex. Science. 2002;298(5596):1248-1251.

[41] Zhang J, Lin C, Lv X, Zhao H, Wang X. Psychedelics and the extracellular matrix: rewiring neuroplasticity and metaplasticity for next-generation psychiatric therapies. Biol Psychiatry. 2026;100(6):614-623.

[42] Chollet F, Tardy J, Albucher JF, et al. Fluoxetine for motor recovery after acute ischaemic stroke (FLAME): a randomised placebo-controlled trial. Lancet Neurol. 2011;10(2):123-130.

[43] Dennis M, Mead G, Forbes J. Effects of fluoxetine on functional outcomes after acute stroke (FOCUS): a pragmatic, double-blind, randomised, controlled trial. Lancet. 2019;393(10168):265-274.

[44] Hankey G, Hackett M, Almeida O, et al. Safety and efficacy of fluoxetine on functional outcome after acute stroke (AFFINITY): a randomised, double-blind, placebo-controlled trial. Lancet Neurol. 2020;19(8):651-660.

[45] Lundström E, Isaksson E, Näsman P, et al. Safety and efficacy of fluoxetine on functional recovery after acute stroke (EFFECTS): a randomised, double-blind, placebo-controlled trial. Lancet Neurol. 2020;19(8):661-669.

[46] Goldstein LB, Lennihan L, Rabadi MJ, et al. Effect of dextroamphetamine on poststroke motor recovery: a randomized clinical trial. JAMA Neurol. 2018;75(12):1494-1501.

[47] Sonde L, Nordström M, Nilsson CG, Lökk J, Viitanen M. A double-blind placebo-controlled study of the effects of amphetamine and physiotherapy after stroke. Cerebrovasc Dis. 2001;12(3):253-257.

[48] Olshausen BA. Theoretical neuroscience: computational and mathematical modeling of neural systems. J Cognitive Neurosci. 2003;15:154-155.

[49] Friston K. The free-energy principle: a unified brain theory? Nat Rev Neurosci. 2010;11(2):127-138.

[50] Rao RP, Ballard DH. Predictive coding in the visual cortex: a functional interpretation of some extra-classical receptive-field effects. Nat Neurosci. 1999;2(1):79-87.

[51] Taub E, Uswatte G, Pidikiti R. Constraint-induced movement therapy: a new family of techniques with broad application to physical rehabilitation—a clinical review. J Rehabil Res Dev. 1999;36(3):237-251.

[52] Jones EG. Cortical and subcortical contributions to activity-dependent plasticity in primate somatosensory cortex. Annu Rev Neurosci. 2000;23:1-37.

[53] Cameron LP, Tombari RJ, Lu J, et al. A non-hallucinogenic psychedelic analogue with therapeutic potential. Nature. 2021;589(7842):474-479.

[54] Vargas MV, Meyer R, Avanes AA, Rus M, Olson DE. Psychedelics and other psychoplastogens for treating mental illness. Front Psychiatry. 2021;12:727117.

[55] Yaden DB, Griffiths RR. The subjective effects of psychedelics are necessary for their enduring therapeutic effects. ACS Pharmacol Transl Sci. 2021;4(2):568-572.

[56] Olson DE. Psychoplastogens: a promising class of plasticity-promoting neurotherapeutics. J Exp Neurosci. 2018;12:1179069518800508.

[57] Krakauer JW. Motor learning: its relevance to stroke recovery and neurorehabilitation. Curr Opin Neurol. 2006;19(1):84-90.

[58] Zanos P, Gould TD. Mechanisms of ketamine action as an antidepressant. Mol Psychiatry. 2018;23(4):801-811.

[59] Lebedev MA, Nicolelis MA. Brain–machine interfaces: past, present and future. Trends Neurosci. 2006;29(9):536-546.

[60] Collinger JL, Wodlinger B, Downey JE, et al. High-performance neuroprosthetic control by an individual with tetraplegia. Lancet. 2013;381(9866):557-564.

[61] Schwartz AB, Cui XT, Weber DJ, Moran DW. Brain-controlled interfaces: movement restoration with neural prosthetics. Neuron. 2006;52(1):205-220.

[62] Willett FR, Avansino DT, Hochberg LR, Henderson JM, Shenoy KV. High-performance brain-to-text communication via handwriting. Nature. 2021;593(7858):249-254.

[63] Wairagkar M, Card NS, Singer-Clark T, et al. An instantaneous voice-synthesis neuroprosthesis. Nature. 2025;644(8075):145-152.

[64] Barrese JC, Rao N, Paroo K, et al. Failure mode analysis of silicon-based intracortical microelectrode arrays in non-human primates. J Neural Eng. 2013;10(6):066014.

[65] Salatino JW, Ludwig KA, Kozai TDY, Purcell EK. Glial responses to implanted electrodes in the brain. Nat Biomed Eng. 2017;1(11):862-877.

[66] Gilja V, Pandarinath C, Blabe CH, et al. Clinical translation of a high-performance neural prosthesis. Nat Med. 2015;21(10):1142-1145.

[67] Hochberg LR, Bacher D, Jarosiewicz B, et al. Reach and grasp by people with tetraplegia using a neurally controlled robotic arm. Nature. 2012;485(7398):372-375.

[68] Moses DA, Metzger SL, Liu JR, et al. Neuroprosthesis for decoding speech in a paralyzed person with anarthria. N Engl J Med. 2021;385(3):217-227.

[69] Metzger SL, Littlejohn KT, Silva AB, et al. A high-performance neuroprosthesis for speech decoding and avatar control. Nature. 2023;620(7976):1037-1046.

[70] Willett FR, Kunz EM, Fan C, et al. A high-performance speech neuroprosthesis. Nature. 2023;620(7976):1031-1036.

[71] Ganguly K, Carmena JM. Emergence of a stable cortical map for neuroprosthetic control. PLoS Biol. 2009;7(7):e1000153.

[72] Dadarlat MC, Canfield RA, Orsborn AL. Neural plasticity in sensorimotor brain–machine interfaces. Annu Rev Biomed Eng. 2023;25:51-76.

[73] Shen Z, Liang Q, Chang Q, Liu Y, Zhang Q. Topological hydrogels for long-term brain signal monitoring, neuromodulation, and stroke treatment. Adv Mater. 2024;36(7):e2310365.

[74] Qin Y, Zhao H, Chang Q, et al. Amylopectin-based hydrogel probes for brain-machine interfaces. Adv Mater. 2025;37(6):e2416926.

[75] Bozorgzadeh B, Schuweiler DR, Bobak MJ, Garris PA, Mohseni P. Neurochemostat: a neural interface SoC with integrated chemometrics for closed-loop regulation of brain dopamine. IEEE Trans Biomed Circuits Syst. 2016;10(3):654-667.

[76] Kim Y, Park S. In vivo multimodal neurochemical interfaces for real-time decoding of brain circuit. Nat Rev Neurosci. 2026;27(3):178-195.

[77] Roseman L, Nutt DJ, Carhart-Harris RL. Quality of acute psychedelic experience predicts therapeutic efficacy of psilocybin for treatment-resistant depression. Front Pharmacol. 2017;8:974.

[78] Nutt D, Erritzoe D, Carhart-Harris R. Psychedelic psychiatry’s brave new world. Cell. 2020;181(1):24-28.

[79] Koralek AC, Jin X, Long JD, Costa RM, Carmena JM. Corticostriatal plasticity is necessary for learning intentional neuroprosthetic skills. Nature. 2012;483(7389):331-335.

[80] Zhang M, Wang Y, Gao TM, Wang X. Psychedelics and consciousness: expanding the horizons of mind and therapy. Research (Wash D C). 2024;7:0495.

[81] Chestek CA, Gilja V, Nuyujukian P, et al. Long-term stability of neural prosthetic control signals from silicon cortical arrays in rhesus macaque motor cortex. J Neural Eng. 2011;8(4):045005.

[82] Johnson M, Richards W, Griffiths R. Human hallucinogen research: guidelines for safety. J Psychopharmacol. 2008;22(6):603-620.

[83] Wang Y, Wang X. Entheogen: an evolutionary medicine for neuropsychiatric disorders. Natl Sci Rev. 2025;12(8):nwaf168.

[84] Klein E, Peters B, Higger M. Ethical considerations in ending exploratory brain–computer interface research studies in locked-in syndrome. Camb Q Healthc Ethics. 2018;27(4):660-674.

[85] Clausen J. Man, machine and in between. Nature. 2009;457(7233):1080-1081.

[86] Kast E. Attenuation of anticipation: a therapeutic use of lysergic acid diethylamide. Psychiatr Q. 1967;41(4):646-657.

[87] Vollenweider FX, Preller KH. Psychedelic drugs: neurobiology and potential for treatment of psychiatric disorders. Nat Rev Neurosci. 2020;21(11):611-624.

[88] Lin C, Wang X, Wang X. Exploring the therapeutic potential of psychedelics in chronic pain management: a new frontier in medicine. Neurosci Bull. 2025;41(4):719-722.

[89] Quartarone A, Siebner HR, Rothwell JC. Task-specific hand dystonia: can too much plasticity be bad for you? Trends Neurosci. 2006;29(4):192-199.

[90] Nicolelis MA, Ghazanfar AA, Faggin BM, Votaw S, Oliveira LM. Reconstructing the engram: simultaneous, multisite, many single neuron recordings. Neuron. 1997;18(4):529-537.

[91] Luo TZ, Bondy AG, Gupta D, Elliott VA, Kopec CD, Brody CD. An approach for long-term, multi-probe neuropixels recordings in unrestrained rats. Elife. 2020;9:e59716.

[92] Carhart-Harris RL, Roseman L, Haijen E, et al. Psychedelics and the essential importance of context. J Psychopharmacol. 2018;32(7):725-731.

[93] Cook DJ, Nguyen C, Chun HN, et al. Hydrogel-delivered brain-derived neurotrophic factor promotes tissue repair and recovery after stroke. J Cereb Blood Flow Metab. 2017;37(3):1030-1045.

[94] Walker MP, Stickgold R. Sleep-dependent learning and memory consolidation. Neuron. 2004;44(1):121-133.

[95] Rasch B, Born J. About sleep’s role in memory. Physiol Rev. 2013;93(2):681-766.

[96] Klinzing JG, Niethard N, Born J. Mechanisms of systems memory consolidation during sleep. Nat Neurosci. 2019;22(10):1598-1610.

[97] Shadmehr R, Smith MA, Krakauer JW. Error correction, sensory prediction, and adaptation in motor control. Annu Rev Neurosci. 2010;33:89-108.

[98] Fins JJ. Rights Come to Mind: Brain Injury, Ethics, and the Struggle for Consciousness. Cambridge: Cambridge University Press; 2015.

[99] Yuste R, Goering S, Arcas BAY, et al. Four ethical priorities for neurotechnologies and AI. Nature. 2017;551(7679):159-163.

[100] United Nations Educational, Scientific and Cultural Organization (UNESCO). Recommendation on the Ethics of Artificial Intelligence. Paris: UNESCO; 2021.

[101] Ienca M, Andorno R. Towards new human rights in the age of neuroscience and neurotechnology. Life Sci Soc Policy. 2017;13(1):5.

Rights & permissions

The Author(s) 2026. This article is published by Higher Education Press at journal.hep.com.cn.