Psychology by BogdanovaThe Encyclopedia of Psychology for the Artificial Era

RESEARCH DOSSIER

Neurocognitive Recovery and Neural Plasticity

Abstract. Neurocognitive recovery after acquired brain injury combines spontaneous neurological changes, experience-dependent plasticity, relearning, compensation, and the reorganization of everyday functioning. This research monograph examines foundational definitions, the experimental history of postinjury plasticity, contemporary neural mechanisms, methods of longitudinal assessment, and comparative intervention evidence across motor recovery, memory, attention, executive functioning, and aphasia. Randomized rehabilitation trials, Cochrane evidence syntheses, biomarker studies, and international stroke rehabilitation guidelines are evaluated with particular attention to treatment timing, delivered dose, recovery trajectories, measurement validity, and transfer to daily activities. The scientific synthesis differentiates neural adaptation, restitution, compensatory performance, and socially supported participation, identifies limitations in prognostic inference and clinical generalization, and proposes empirically tractable priorities for future recovery science.

Keywords: neurocognitive recovery; neural plasticity; stroke rehabilitation; acquired brain injury; cognitive rehabilitation; neurorehabilitation; motor recovery; memory rehabilitation; aphasia; executive dysfunction; rehabilitation transfer; neural reorganization; neuropsychological assessment; functional participation

Neurocognitive Recovery as a Scientific Problem

Neurocognitive recovery is the change over time in cognitive, perceptual, communicative, and action-related functioning after injury or disruption of the nervous system. Its explanatory subject is the relationship among neural events, observable behavioral capacities, learned strategies, and participation in everyday life. Improvement may involve restitution of an impaired process, compensation by alternative procedures, spontaneous biological changes, structured practice, environmental support, or combinations of these processes. A participant who scores better on a memory test may have retained more information, learned the test procedure, improved attention, or adopted an external reminder. These alternatives predict different behavior outside the testing situation. The scientific purpose of this dossier is to distinguish such mechanisms and examine the evidence that links interventions to meaningful longitudinal outcomes. The shared definitions developed by the Stroke Recovery and Rehabilitation Roundtable emphasize that changes in impairment, task performance, and real-world participation require distinct measurements Bernhardt et al. (2017).

Neural plasticity concerns activity- and experience-dependent changes in nervous-system structure, function, connectivity, or responsiveness. Changes can follow ordinary learning, injury, altered sensory input, developmental transitions, and rehabilitation. Cramer et al. (2011) defined plasticity broadly and reviewed the translational problem of converting laboratory mechanisms into successful treatments. Plastic responses may facilitate recovery, preserve ineffective patterns, or reflect processes incidental to performance changes. Therefore the observation of a changed functional magnetic resonance imaging signal, cortical motor map, or electrophysiological response cannot alone certify clinical improvement. A satisfactory explanation pairs neural measurement with appropriately defined behavior, relevant control conditions, and a temporal account of what changed first. The same caution applies when a brain network becomes more distributed during the recovery period: greater activation may indicate additional recruitment, task difficulty, or vascular and measurement differences.

Stroke and acquired brain injury provide the most developed human evidence because their time course can be studied from an identifiable event through rehabilitation and community reintegration. Yet these conditions are heterogeneous: lesion location, tissue damage, age, comorbidity, prior skill, treatment exposure, mood, social circumstances, and access to care all shape observed trajectories. A comprehensive monograph must examine mechanisms at the synaptic and circuit levels, the person-level organization of cognition and action, the rehabilitation environment, and the research methods used to infer causality. The aim is an evidence-based synthesis, rather than an assertion that any single biological theory explains every recovery pattern. The systematic clinical recommendations of Winstein et al. (2016) and later Canadian guidance Salbach et al. (2025/2026) demonstrate how this multilevel understanding translates into coordinated care, while also exposing continuing uncertainties.

Definitions of Restitution, Compensation, Adaptation, and Participation

The vocabulary of recovery affects what experiments claim to show. Restitution denotes return toward a premorbid pattern of performing a function using the affected capability, while compensation denotes achieving a task goal through a changed method, another body part, an external device, or altered environmental demands. Behavioral adaptation encompasses acquisition and adjustment of strategies; participation concerns engagement in social and everyday roles. Bernhardt et al. (2017) explicitly distinguished restitution and compensation because a task completion score can improve even while the impaired movement or cognitive operation remains substantially altered. A person may use a checklist to complete a complex sequence successfully while still experiencing a measurable deficit in prospective memory. That successful strategy has substantial functional value, but it supports a claim about supported task performance rather than restoration of unaided recall.

Neural restoration is still another construct. Reperfusion, resolution of edema, recovery from diaschisis, synaptic remodeling, altered excitability, and strengthening of surviving pathways have different biological meanings and timescales. Neural plasticity should be operationalized as a measurable alteration in function or structure with a plausible causal relation to experience or injury. Neurocognitive recovery should be operationalized as change in specified cognitive and functional outcomes over defined observation periods. The distinction permits empirical results such as enhanced independence without improved impairment scores, improved impairment without occupational participation, or changed imaging without demonstrable activity gains. None of those patterns is contradictory once the outcomes are specified. Mechanistic accounts must explain which observed changes follow natural recovery, which follow intervention, and which reflect learning to perform a particular assessment.

The International Classification of Functioning, Disability and Health offers a practical framework for evaluating recovery across impairment, activity, participation, and contextual supports. The 2025 update of Canadian stroke recommendations uses this perspective to organize rehabilitation planning and the transition to community life Salbach et al. (2025/2026). Research that reports only a cognitive composite can miss its significance for medication management, financial decisions, employment, social participation, and self-directed living. Conversely, participation can improve with environmental modifications while the underlying impairment remains stable. Investigators should predefine which construct is the primary endpoint, monitor the others as differentiated outcomes, and report the conditions under which the change occurred. This conceptual discipline is central to avoiding inflated claims about neuroplasticity.

Historical Foundations of Recovery Research

A long tradition of neurological observation has associated focal lesions with disturbances of language, sensation, and movement, but modern recovery science emerged through experimental studies of neural reorganization and measurable rehabilitation outcomes. Research in the late twentieth century increasingly examined how skilled practice changes the functional organization of surviving brain tissue. Nudo et al. (1996) provided influential animal evidence by studying how rehabilitative training affected motor representations after experimentally induced ischemic injury. Their work linked experience after injury with changes in cortical organization and motor performance, supplying a mechanistic premise that subsequent clinical trials would test under far more heterogeneous human conditions. The study is historically important because it made postinjury experience an experimental variable rather than a passive accompaniment to recovery.

Experiments on cortical representation developed alongside research on learning and environmental enrichment. The synthesis by Kolb et al. (2003) situated brain plasticity within experience-dependent changes involving development, hormonal state, aging, stress, disease, and behavior. Later Kleim & Jones (2008) organized principles including specificity, repetition, intensity, timing, salience, interference, and transfer. Those principles are heuristics derived from substantial basic science, not universal prescriptions that guarantee better clinical outcomes at any intensity or time. Rehabilitation researchers increasingly confronted a translational problem: therapies can produce measurable task learning without proving restoration of damaged circuits, while animal models permit direct physiological investigations that human clinical settings usually cannot reproduce.

By the 2000s and 2010s, multicenter randomized studies changed the evaluation of clinical interventions. The EXCITE trial tested structured upper-limb practice following stroke Wolf et al. (2006), while LEAPS examined locomotor training against an exercise-based comparator Duncan et al. (2011). Later RATULS assessed robotic therapy and an enhanced therapy program in a pragmatic multicenter setting Rodgers et al. (2019). These trials shifted the question from whether neural systems are modifiable to whether a specified approach yields clinically meaningful improvement relative to an appropriate alternative. Concurrently, cognitive rehabilitation reviews evaluated domain-specific training and its transfer to independence and quality of life Cicerone et al. (2019). History thus reveals a progression from demonstrations of plastic capacity toward demanding comparisons of patient-centered effects and practical implementation.

Biological Substrates and Levels of Explanation

Biological substrates of recovery include cellular responses around injury, axonal remodeling, synaptic strengthening and weakening, excitation-inhibition balance, neurovascular responses, and changes in distributed functional connectivity. These processes operate over different spatial and temporal scales and may interact with one another. Krakauer et al. (2012) examined evidence from animal models for a period of heightened poststroke plasticity and discussed why translating its timing to clinical practice is difficult. The major scientific lesson is that an observed change at one scale cannot simply be substituted for an explanation at another. Improvement in reaching accuracy is a behavioral finding; an altered cortical map is a neural finding; a causal mechanism linking the two requires convergent evidence or direct perturbation.

At the network level, the consequences of a focal injury depend partly on which connections are interrupted and what alternate pathways remain functional. The relationship between structure and outcome is therefore conditional on lesion topology and the intact system's existing capacities. For motor recovery, integrity of descending pathways can be important for prognosis, as operationalized in the PREP2 study Stinear et al. (2017). For cognitive functions, distributed circuits and interactions among processing demands complicate one-region explanations. Changes in blood oxygen level-dependent signal may also reflect hemodynamic physiology, vascular injury, or task performance differences. Brain-imaging correlations gain explanatory power when they are longitudinal, appropriately controlled, and combined with physiological and behavioral measures.

At the person-environment level, recovery takes place through actual opportunities to practice and participate. A neural system may retain the capacity for learning without receiving sufficient tailored training, while social arrangements may either enable or prevent newly acquired skills from being used. Lang et al. (2015) stressed that dosage must specify the active practice delivered, not merely the length of a scheduled therapy session. Clinical recommendations consequently integrate structured exercises with everyday goals, family participation, assistive strategies, and transitions between rehabilitation settings Salbach et al. (2025/2026). A multilevel explanation of neurocognitive recovery traces how biological capacity interacts with training exposure and functional demands; it neither assumes a direct translation from a single biomarker nor treats contextual support as secondary.

Methods of Longitudinal Recovery Research

The baseline for a recovery study is itself a methodological choice. Measurements obtained immediately after a stroke may reflect acute physiological disturbance, fluctuating awareness, fatigue, medications, aphasia, and evolving medical conditions. Measurements taken after several weeks omit the earliest phase of change and may select a population that survived or gained access to follow-up services. Repeated assessment should therefore specify time since injury, baseline severity, measurement procedures, concurrent treatment, and relevant clinical events. Longitudinal observational cohorts can describe trajectories and identify predictors, but improvement following treatment in an uncontrolled cohort cannot separate treatment effects from natural recovery, practice effects, and regression to the mean. Well-designed randomized comparisons address some of those uncertainties at the cost of narrower eligibility and a need for careful intervention reporting.

Repeated neuropsychological assessment creates additional difficulties. Attention, memory, language, and executive tests contain measurement error and often have ceiling and floor effects. Test familiarity can create gains unrelated to broader functional recovery, while repeated forms may differ in difficulty. Investigators need normed measures, appropriate language adaptations, reliable-change analysis, blinded raters where feasible, and explicit handling of attrition. A meaningful outcome battery includes performance-based cognitive indicators and real-world activities, preferably with patient-centered measures of participation. Bernhardt et al. (2017) advocated shared definitions across stroke recovery studies; without standardized endpoints, trial comparisons can conflate changes in impairment, activity, and compensation. Functional transfer requires testing situations that differ from the trained exercise.

Neuroimaging, transcranial magnetic stimulation, electroencephalography, lesion mapping, and other physiological methods can characterize residual pathways and changes during recovery. Each is a measurement instrument with domain-specific limitations rather than a direct picture of the underlying causal process. Imaging differences after a therapy must be evaluated against controls for head movement, task difficulty, vascular reactivity, and repeated exposure. Prognostic algorithms such as PREP2 Stinear et al. (2017) provide a concrete example of integrating clinical and physiological measures, but their performance must be validated in independent patient populations and alternative care settings. The appropriate standard is triangulation among patient behavior, function, controlled interventions, and physiological indicators, together with transparent reporting of missing data and uncertainty.

Spontaneous Recovery and the Temporal Organization of Plasticity

The first weeks after a brain injury may include rapid changes as edema, perfusion, connectivity, and excitability evolve, alongside the earliest learning and adaptation. Consequently, the steepest segment of a behavioral recovery curve can occur before structured therapy has exerted much of its cumulative effect. The distinction between spontaneous and intervention-associated change is conceptually useful but empirically difficult: people begin adapting immediately, rehabilitation exposures vary, and the physiology of injury never proceeds in complete isolation from experience. Krakauer et al. (2012) reviewed animal evidence suggesting a temporally heightened period of poststroke plasticity, while emphasizing that the timing and intensity of experience are consequential. A clinical curve documents the net result of interacting processes; it does not identify how much improvement was caused by any one component.

The relationship between time and treatment can be nonlinear. Intense intervention introduced too early may exceed medical or neurological tolerance, while delaying every opportunity for appropriate practice can allow disuse and avoidable skill loss. The AVERT randomized trial found that a specific protocol of more frequent, higher-dose mobilization within the first twenty-four hours after stroke produced lower odds of favorable three-month outcome compared with usual care AVERT Trial Collaboration Group (2015). This finding directly challenges the simplistic view that earlier and more intense activity is invariably preferable. It does not imply that medically appropriate rehabilitation should be postponed indefinitely. Modern guidance emphasizes stability, individualized intensity, and adjustment to tolerance, with task-specific activities as patients become able to participate Salbach et al. (2025/2026).

Temporal research must distinguish an individual's time since injury from calendar time, duration of treatment, accumulated meaningful practice, and developmental or aging processes. A study that classifies participants simply as acute or chronic may conceal substantial heterogeneity within each group. Delayed gains after treatment can arise through continued practice, consolidation, environmental adaptation, and progressive participation; plateau on one measure need not mean learning has ended. Longitudinal models should estimate individual trajectories and report variability rather than assuming one universal recovery curve. The most useful question is when a particular treatment ingredient changes a specific functional outcome for a defined patient subgroup, compared with an appropriate alternative and over a sufficiently long period.

Experience-Dependent Plasticity and Task-Specific Learning

Practice changes performance through mechanisms including perceptual discrimination, error correction, feedback integration, motor planning, memory consolidation, and strategy selection. After injury, the surviving neural system can support new learning even when some functions remain impaired. The experience-dependent framework of Kleim & Jones (2008) treats specificity, repetition, salience, and timing as interacting determinants of neural change. Those propositions direct research toward detailed descriptions of the practice itself: what response was trained, how often it was attempted, whether successful repetitions occurred, which errors were informative, and how practice progressed. Merely counting hours in a rehabilitation program can obscure large differences in actual skill-relevant activity.

The transfer problem arises because learning is commonly specific to practiced stimuli, tasks, and response conditions. Rehearsal of computerized attention exercises may improve that exercise without establishing that a participant can manage a busy street, prepare food safely, or return to work. Similarly, improved reaching with a laboratory apparatus may not produce spontaneous arm use during daily routines. Generalization is an empirical outcome requiring assessment with untrained tasks and ecological observation. Lang et al. (2015) examined the dose-and-timing problem and highlighted the need to measure meaningful active repetitions. Repeated practice can be justified scientifically when its relation to a desired behavior is explicit and its effects are evaluated beyond the training environment.

Practice design also involves motivation, fatigue, reward, feedback, and the person's goals. Feedback can guide correction but excessive dependence on external cues may limit independent performance. Alternating task contexts and gradually increasing complexity may promote flexible use of recovered abilities, although these design features need direct trials in particular populations. Behavioral learning cannot be reduced to a uniformly beneficial neural change: maladaptive repetitions and compensatory habits may also become more efficient through experience. A rigorous intervention describes the targeted process, establishes a comparator, and tests retention and transfer at intervals compatible with the proposed mechanism. The scientific significance of plasticity is therefore linked to the function it supports, and its clinical value depends on what a person can subsequently do.

Upper-Limb Recovery and Constraint-Induced Movement Therapy

Upper-limb impairment after stroke offers a relatively well-defined laboratory for separating movement capacity, use, and participation. A person may retain some movement yet rely consistently on the unaffected arm, thereby reducing opportunities to practice the more affected limb. Constraint-induced movement therapy combines intensive, structured practice of the affected limb with procedures intended to increase its use outside supervised sessions. In the EXCITE randomized trial, Wolf et al. (2006) compared this structured approach with usual care in a selected group several months after stroke. The findings supported improvements in standardized arm performance and self-reported use, while eligibility requirements and the trial's intervention intensity restrict broad generalization to patients with different initial capacities.

The trial's significance is best understood as a test of a specific treatment package. Greater performance improvement after the intervention supports a causal effect of assignment to that package under the study conditions, not an isolated causal attribution to constraint, practice dose, therapist feedback, or neural structural change. The follow-up work described persistence of some gains after therapy Wolf et al. (2008). For rehabilitation theory, this distinction matters because an intervention may succeed through altered habits and consistent arm use even when the underlying biological recovery mechanism remains incompletely characterized. Laboratory movement measures and activity logs can together distinguish improved movement quality from increased reliance on the affected limb.

The findings also expose equity and feasibility questions. Time-intensive practice, the need for adequate residual movement, caregiver assistance, fatigue, and service availability may limit applicability. Comparative effectiveness studies should examine different doses, versions of the treatment, and genuinely active alternatives, including everyday task practice without constraint. Participants who fail eligibility for one intensive protocol still require individualized rehabilitation rather than inference that plasticity is absent. Contemporary recommendations recognize task-specific upper-limb treatment among evidence-informed options, with intervention selection depending on goals, impairment, tolerance, and access Salbach et al. (2025/2026). The more mature research question concerns which components change outcomes for which patients and through what mechanisms.

Gait, Balance, and the Lessons of Locomotor Trials

Walking after stroke requires coordination among strength, sensory feedback, postural control, attention, anticipatory planning, and environmental adaptation. Relearning walking can involve restoration of movement control and compensatory changes in step patterns, assistive device use, or pacing. A mechanistic intervention might focus on repeated stepping with body-weight support, but the hypothesis that such support is superior to alternatives must be evaluated empirically. The LEAPS trial randomized 408 participants to earlier locomotor training, later locomotor training, or a progressive home exercise program managed by therapists Duncan et al. (2011). Its principal finding was that body-weight-supported locomotor training was not shown to be superior to the home exercise approach for the measured walking outcome.

This result remains instructive because the unsuccessful superiority hypothesis does not mean that participants failed to improve or that training has no value. It means that a specialized technology did not provide a superior outcome relative to its studied active comparator. Trials with weak usual-care comparators can make a new treatment appear more distinctive than it is, whereas comparisons with credible alternative rehabilitation can identify the value added by particular equipment or techniques. The gait outcome must also be differentiated from balance confidence, endurance, dual-task walking, community mobility, and falls. An improvement in one measurement domain may coexist with persistent limitations in another.

Transfer is particularly important for walking because clinical test conditions often simplify the world a person must navigate. Laboratory measures may be collected on level surfaces, while community participation requires negotiating uneven ground, visual distractors, crowds, fatigue, and concurrent cognitive tasks. Longitudinal follow-up should test retained walking ability, falls, mobility outside the clinic, and meaningful individual goals. Salbach et al. (2025/2026) emphasizes rehabilitation across functioning and participation rather than a single impairment outcome. Gait studies thereby provide a model for evaluating neurocognitive recovery: the most convincing intervention is one that makes an additional, durable contribution to activities people actually need to perform, under conditions that can be replicated outside a specialized trial.

Robotics and the Difference Between Practice Delivery and Treatment Mechanism

Robotic rehabilitation devices can standardize movement assistance, deliver high numbers of repetitions, record trajectories, and adapt resistance or support. These are practical and measurement capabilities, but the mere presence of a robot establishes neither a unique therapeutic mechanism nor clinical superiority. RATULS, a multicenter randomized trial, compared robot-assisted upper-limb training, enhanced upper-limb therapy based on repetitive functional task practice, and usual care in people with moderate or severe poststroke limitation Rodgers et al. (2019). The central comparison addresses whether technological assistance adds benefits to functional rehabilitation, and how outcomes vary when a robust active therapy is included.

RATULS illustrates the need to distinguish impairment improvements from improvements in activities of daily living. A treatment may alter the performance of movements practiced under support without proportionate gains in independent function. Researchers must consider total dose, task relevance, therapist contact, attention, and baseline impairment before ascribing differences to the technology itself. Trial results should be interpreted with attention to the designated primary outcome, subgroup analyses, and resource costs rather than selectively highlighting secondary measures. Rehabilitation devices can be useful instruments for delivering precisely quantified practice even when the evidence does not support their universal superiority over well-designed conventional therapy.

The methodological lesson extends beyond robotics to virtual reality, computer-assisted cognitive exercises, and stimulation-assisted training. Technology should be assessed as a configurable intervention with specific active ingredients. Investigators need to describe the algorithmic assistance, feedback schedule, adaptation rules, repetitions, adherence, and accessibility requirements. Relevant comparators include therapist-directed practice of equal intensity, alternative devices, and functionally grounded home exercises. Outcomes should include performance in natural environments and the patient's ability to continue using acquired skills. When robotic therapy facilitates access to practice or measurement, that contribution deserves separate evaluation from any claim about stronger biological plasticity Lang et al. (2015) Salbach et al. (2025/2026).

Memory Rehabilitation and the Boundaries of Cognitive Transfer

Memory changes following stroke and traumatic brain injury can disrupt the learning of new information, appointment keeping, prospective intentions, autobiographical continuity, and the reliable performance of everyday tasks. Rehabilitation programs include repeated practice, structured encoding, external reminders, error-reducing procedures, and training in compensatory strategies. Their underlying hypotheses differ. Restoration-oriented exercises seek better functioning of the impaired memory process; compensation-oriented strategies seek safer and more effective activity despite a continuing deficit. In both cases meaningful outcomes should address daily behavior rather than only laboratory scores. The Cochrane review by das Nair et al. (2016) provides an important test of the difference between measured short-term benefits and persistent improvements in functioning.

That review included thirteen trials with 514 participants and found a short-term benefit for subjective memory complaints after rehabilitation, while evidence for durable effects and improvement in daily independence was not established. The apparent contradiction between improvement in self-reported memory and limited effect on objective or functional measures is informative. Different measures capture different experiences and capacities, and the trial corpus included diverse treatment procedures and small samples das Nair et al. (2016). A small trial cannot reliably resolve whether a strategy is most helpful for a specific lesion profile, task, or context. Nor does an absence of statistically demonstrated long-term benefit prove that a particular individual's external reminder system has no practical value.

A strong memory rehabilitation study describes whether participants can use a strategy without prompting, how often they use it, which untrained tasks benefit, and how outcomes change after formal treatment ends. Repeated verbal list-learning tests are informative about encoding and retrieval under standardized conditions but do not capture all the problems of managing daily intentions. Digital reminders and structured environmental cues introduce another measurement problem: increased task success may depend on ongoing external assistance. Such assistance can legitimately improve functioning and autonomy while leaving the underlying capacity largely unchanged. The cognitive rehabilitation framework summarized by Cicerone et al. (2019) underscores the importance of matching intervention types and evidence strength to actual functional targets.

Attention Training and Domain-Specific Outcomes

Attention is a family of processes including orienting, selective filtering, sustained engagement, and coordination across concurrent demands. Injury may affect these processes differently, with consequences for communication, mobility, work, and safety. Rehabilitation tasks that tax divided attention, visual exploration, or sustained concentration can produce measurable change in the trained domain. To establish broader recovery, however, trials must ask whether improvements persist and whether they generalize to situations in which attention is only one requirement among many. The Cochrane analysis of cognitive rehabilitation after stroke by Loetscher et al. (2019) examined these questions across distinct attention outcomes rather than treating attention as one uniform score.

The review found evidence suggesting a short-term improvement in divided-attention measures, but the evidence was of low quality, with uncertain persistence and no convincing improvement in functional ability or quality of life. It is methodologically important that the result varied with the particular attentional function measured. The absence of a robust finding for alertness, selective attention, or sustained attention prevents a generic claim that training restores the attention system as a whole Loetscher et al. (2019). Performance gains may arise from learning specific stimulus-response rules, increased confidence, better task pacing, or genuine changes in attentional allocation. Those explanations demand distinct tests, including untrained tasks and repeated observation outside therapy.

Attention training also interacts with fatigue, sleep, mood, sensory loss, and the amount of competing information. A patient might perform accurately in a quiet clinic yet become overwhelmed when completing the same activity amid conversation and background noise. The transfer question therefore requires specifying environmental conditions and cognitive load, including dual-task demands. Studies can combine laboratory psychophysics, observer-rated functioning, self-report, and ecological assessment, but they should report disagreement among measures instead of averaging it away. Clinical decisions can be individualized while the general evidence remains limited; the methodological obligation is to distinguish patient goals from empirically demonstrated general efficacy.

Executive Function, Metacognition, and Strategic Rehabilitation

Executive impairment can affect planning, goal maintenance, inhibition, flexible switching, monitoring, and the ability to detect one's own errors. These constructs overlap with attention and working memory, making selective measurement difficult. An intervention that teaches goal-management or structured problem-solving could influence action through several routes: improved executive control, more accessible rules, modified environmental demands, or externally supported sequencing. Neurocognitive rehabilitation research needs to specify which of these is the proposed mediator. A person who follows a written plan more successfully may display a real improvement in independence even if an abstract executive task shows limited change.

The Cochrane review addressing executive dysfunction after acquired nonprogressive brain injury concluded that the available evidence was insufficient for broad generalizations about the effectiveness of cognitive rehabilitation Chung et al. (2021). The review evaluated diverse approaches and comparators with heterogeneous outcome measures. This finding should direct scientific attention toward more precise intervention theories and better-controlled trials rather than toward a binary decision that all executive rehabilitation works or fails. It is particularly important to distinguish experimental success on a structured task from a durable change in complex daily behavior. Performance can be strongly influenced by motivation, language demands, instructions, and the reliability of self-report.

Metacognitive intervention approaches can ask people to predict performance, compare predictions with observed outcomes, monitor errors, and adjust strategies across repeated tasks. The scientific question is whether those learned operations become available during unprompted everyday activity. Studies need repeated measures of self-monitoring accuracy, adherence to compensatory plans, participation, and caregiver-rated behavior, with attention to discrepancies between perspectives. When possible, a matched active control can distinguish the content of strategy training from additional therapist contact. The clinical evidence synthesis of Cicerone et al. (2019) provides a broader context for cognitive rehabilitation recommendations, but individual intervention claims still depend on the population, study design, and observed outcomes.

Language Recovery, Aphasia, and Functional Communication

Aphasia after stroke can affect spoken expression, comprehension, reading, and writing, with consequences extending beyond formal language tests. Everyday communication also requires conversational timing, partner adaptation, social participation, and the ability to convey needs and decisions. Rehabilitation may include impairment-focused language exercises, conversation practice, compensatory tools, and training for communication partners. Different procedures can improve distinct aspects of behavior. The systematic review by Brady et al. (2016) investigated speech and language therapy following stroke and provides a broad evidentiary foundation for evaluating communication interventions rather than assuming that increased practice improves every language domain to the same extent.

Language outcomes are especially susceptible to differences in task selection and generalization. Naming trained pictures does not necessarily show that spontaneous conversation has become more effective. Likewise, an improvement in comprehension under one testing format may not transfer to real-time discussions with multiple speakers. The 2025 European Stroke Organisation guideline evaluated trials of aphasia rehabilitation and differentiated total dose, frequency, intensity, delivery formats, and combinations with brain stimulation Brady et al. (2025). The guideline supports individually tailored speech and language therapy with attention to meaningful total practice, while treating noninvasive stimulation paired with therapy as a research question requiring higher-quality evidence. These distinctions connect the clinical recommendations to the underlying scientific uncertainty.

Functional communication depends on the relationship between the speaker, listener, task, and environment. Communication partners can reduce complexity, allow additional response time, and support alternative modalities; these changes may increase participation without proving biological restitution of every impaired language operation. Conversely, measured language gains may have little significance unless the person can employ them in consequential settings. Rehabilitation trials should therefore report linguistic impairment, communicative effectiveness, patient goals, and participation, along with treatment exposure and retention. Research on language offers a powerful case of socially embedded neurocognitive recovery in which the functional unit of interest may extend beyond an isolated individual performance score.

Cognitive Rehabilitation Across Stroke and Traumatic Brain Injury

Stroke and traumatic brain injury share some rehabilitation questions but differ in typical injury mechanisms, distribution of damage, associated medical complications, and temporal profiles. Focal vascular lesions may disrupt a specific network while traumatic injury can include diffuse axonal damage, focal contusions, and multiple concurrent impairments. The general term cognitive rehabilitation covers highly diverse approaches and should not be interpreted as a single standardized treatment. Cicerone et al. (2019) systematically evaluated rehabilitation literature for traumatic brain injury and stroke, classified evidence by design, and developed evidence-based practice recommendations across cognitive domains. Such reviews provide a more informative starting point than isolated positive trials, though their recommendations necessarily reflect the inclusion criteria and evidence available at the time.

A review of mixed populations must ask whether effects are stable across etiology, severity, time since injury, cognitive domain, and baseline capacity. Pooling studies can obscure genuine differences between individuals who need environmental compensation and those who are able to relearn an impaired skill through repetitive practice. Treatment fidelity matters because two programs labeled memory training or executive rehabilitation may deliver substantially different content. Comparison conditions also matter: an active psychological intervention controls for practice and expectation more effectively than a waitlist, while therapist contact and social support may themselves change functioning. A mechanistically informative study records these dimensions explicitly.

An integrated cognitive rehabilitation program must also address the limitations that determine a person's ability to participate in any training: fatigue, pain, sleep disruption, depression, motor impairment, sensory access, and communication difficulties. These variables can operate as confounders, moderators, or legitimate treatment targets. A longitudinal record of cognitive gains should describe concurrent medical management and rehabilitation exposure, not infer that all improvement arose from one cognitive exercise. The relevant endpoint is the person's ability to manage meaningful tasks safely and sustainably. This outcome is best interpreted through multiple measures and goals, with residual impairment and compensatory dependence reported alongside functional achievements Salbach et al. (2025/2026) Winstein et al. (2016).

Predicting Recovery From Clinical and Neural Biomarkers

A prognostic biomarker is useful only when it improves predictions beyond information already available from clinical assessment and when its prediction changes an important decision. Initial severity, motor evoked potentials, lesion characteristics, age, and time since injury can all contribute information, but their relationships with outcomes are probabilistic. PREP2 used a sequential algorithm combining clinical motor assessment, age, physiological evidence of corticospinal integrity, and additional clinical or imaging information to classify likely upper-limb outcomes Stinear et al. (2017). This work demonstrated the feasibility of combining measurement levels, while its target was a particular clinical outcome after stroke rather than universal prediction across memory, language, mobility, and participation.

The difference between predicting an outcome and predicting a response to treatment is fundamental. A marker can indicate likely natural recovery without identifying the therapy that will produce the greatest additional benefit. To guide treatment selection, investigators need evidence of effect modification: the intervention works differently across biomarker-defined groups under randomized or otherwise credible causal comparisons. Prognostic tools must also be assessed for calibration, discrimination, practical availability, and fairness. A model that performs well in the original cohort may fail when applied to people with different stroke severity, care pathways, demographic composition, or access to imaging. Stinear et al. (2017) illustrates both the promise of bedside-integrated prediction and the need for independent generalizability testing.

Cognitive prognostication introduces further challenges because test performance after injury is influenced by language, arousal, sensory access, premorbid ability, fatigue, and educational context. A model based on one domain should not be extrapolated to personal autonomy or return to employment without validating those outcomes. For communication, the quality of the relationship between lesion characteristics and recovery may vary with network organization and rehabilitation exposure. Predictive studies should preregister outcomes, retain continuous measures where possible, account for missing follow-up, and report uncertainty for individual patients. The operational standard is improved decision quality rather than the appearance of neurobiological precision.

The Proportional Recovery Debate and Statistical Artifacts

The proportional recovery hypothesis proposed that many stroke survivors regain an approximately constant fraction of the function initially lost. Its appeal lies in the possibility of a simple recovery law that could guide expectations and trial design. Yet analyses based on the correlation between baseline impairment and change scores are vulnerable to mathematical coupling because baseline is present in both terms. Ceiling limits can further create patterns that resemble fixed proportional change even when underlying trajectories are heterogeneous. Bonkhoff et al. (2020) examined proportional recovery using Bayesian modeling, while later methodological criticism Bowman et al. (2021) examined inflation associated with mathematical coupling and compression toward scale ceilings.

This debate illustrates why recovery models must be evaluated against plausible statistical null models and the measurement properties of each scale. A strong relation between initial impairment and later change does not by itself identify a single biological process or a treatment-sensitive recovery capacity. If participants approach the maximum score permitted by a test, observed improvement becomes constrained by the instrument rather than only the nervous system. Heterogeneity in severity and intervention exposure can produce apparent subgroups or distort the distribution of individual slopes. Modeling should account for nonlinear change, ceiling and floor effects, reliability, lesion covariates, and uncertainty around predictions, preferably with data from independent cohorts.

The consequences extend beyond a theoretical dispute. A simplistic recovery rule may influence patient counseling, stratification of trial participants, and assumptions about who can benefit from rehabilitation. More complex models require stronger data but can represent distinct trajectories and the possibility of ongoing learning despite a plateau on a chosen measure. Stinear et al. (2017) provides an example of multivariable prediction grounded in clinically meaningful outcomes, whereas Bonkhoff et al. (2020) demonstrates the value of explicit generative and statistical assumptions. The broader lesson is that numerical regularity is a finding about observed measures until alternative mathematical and physiological explanations have been tested.

Noninvasive Brain Stimulation and Causal Mechanism Testing

Noninvasive stimulation methods, including transcranial direct current stimulation and repetitive transcranial magnetic stimulation, are investigated as potential modifiers of cortical excitability and learning conditions. A common experimental idea is that changing the balance between networks could improve the effect of concurrent behavioral training. In a small randomized sham-controlled study, Lindenberg et al. (2010) paired bilateral transcranial direct current stimulation with physical and occupational therapy after chronic stroke and reported better measured motor outcomes in the active group. Such mechanistic trials can identify promising hypotheses, but their small samples and narrow protocols limit the strength of general treatment recommendations.

The mechanism of stimulation remains more complex than a simple increase or decrease in one hemisphere's activity. Lesion location, individual anatomy, current distribution, baseline excitability, task demands, and residual network integrity may all change the effects. Stimulation can alter a physiological measure without showing that the same change mediates durable everyday improvement. Sham procedures may be imperfect, and participant or therapist masking can be challenging. Studies need preregistered primary outcomes, dose characterization, robust sham controls, adverse-event reporting, and follow-up beyond the immediate postintervention session. Reproducibility across independent laboratories is particularly important when an early experiment produces striking results.

The translational evidence differs across disorders and intervention settings. The 2025 European guideline on poststroke aphasia described very low-level evidence for using transcranial direct current stimulation with language therapy and recommended limiting that approach to high-quality trials Brady et al. (2025). This position illustrates how a credible neuroplasticity rationale can coexist with insufficient clinical certainty. Stimulation is best investigated as one potential ingredient within a measured rehabilitation program, with a causal hypothesis linking physiological modulation, task learning, and sustained functional outcomes. Until this chain is supported, a change in neural excitability remains an intermediate finding rather than proof of successful rehabilitation.

Imaging Recovery Networks and Interpreting Reorganization

Functional imaging can document which networks are recruited during tasks after injury and how recruitment changes over time. It is tempting to interpret increased activation in an intact region as compensation, or decreased activation as restoration of efficiency. Those narratives are hypotheses requiring independent evidence. Activation is shaped by performance level, effort, cognitive strategy, movement amplitude, neurovascular responsiveness, and scanner-specific measurement conditions. A participant who completes more of a task may show a different blood oxygen response because the actual behavior changed. Consequently, standardized tasks and repeated imaging do not automatically isolate the causal neural process responsible for recovery Cramer et al. (2011).

Longitudinal imaging becomes more informative when coupled with lesion mapping, stimulation measures, and precise behavioral phenotyping. A network change that precedes improvement is compatible with a contributing mechanism, but it may also reflect an unmeasured common cause or an evolving physiological condition. Experimental perturbation, where safe and feasible, can provide additional causal information by testing whether changing activity in a putatively compensatory region alters behavior. Researchers must also consider attrition: the people able to complete repeated scans may be those with less severe cognitive, communication, or motor impairments. Such selection can exaggerate apparent regularity in recovery networks.

The most useful synthesis connects neural measures to theoretically specified cognitive operations rather than relying on regional labels alone. A language task may depend on listening comprehension, working memory, articulation, and motivation; a mobility task may engage vision, attention, balance, and planning. Network analysis should therefore state which functional capacity is being modeled, what its indicators are, and what differences could arise from changing task demands. Kleim & Jones (2008) and Krakauer et al. (2012) provide mechanistic principles derived from basic science, whereas the clinical reviews of Cicerone et al. (2019) ask whether interventions improve human functioning. Strong recovery neuroscience needs both levels of inquiry and a transparent account of the inferential bridge between them.

Behavioral Dose, Fidelity, and the Implementation Gap

Clinical rehabilitation is often described in terms of scheduled duration even though the active learning ingredient may be a smaller number of meaningful repetitions. An hour of appointment time includes preparation, rest, explanation, transportation, and potentially very different amounts of problem solving or physical practice. Lang et al. (2015) analyzed dose and timing and argued that prescription requires specifying active ingredients, amounts, frequency, and duration. Without these measures, a seemingly negative trial may have delivered too little of the proposed treatment, while a positive trial may have included substantially more therapist attention or practice than its comparator. Intervention fidelity determines whether the tested treatment was actually the one described.

Intensity should also be evaluated against tolerability, fatigue, medical stability, and the biological stage of recovery. The AVERT results show how a specific early high-dose mobilization regimen can produce worse outcomes than the comparison condition AVERT Trial Collaboration Group (2015). Such evidence prevents the mechanical extrapolation of experience-dependent plasticity principles to every clinical setting. The 2025 Canadian recommendations call for individually designed therapy and purposeful task practice within safe and coordinated rehabilitation Salbach et al. (2025/2026). Rehabilitation research can improve dose reporting by counting completed practice attempts, task difficulty, adaptive progression, rest intervals, concurrent interventions, and the amount of activity continued at home.

Implementation is also constrained by workforce expertise, service availability, transportation, language access, caregiver support, and reimbursement systems. An intervention that succeeds in a highly supervised research center may lose effectiveness when its precise components cannot be delivered in routine care. Pragmatic trials such as RATULS Rodgers et al. (2019) are valuable because they confront real treatment pathways, but pragmatic conditions can introduce heterogeneity that complicates mechanistic interpretation. A mature research program combines explanatory trials that isolate causal ingredients with implementation studies that determine how those ingredients can reliably reach people who need them. The final clinical value is established through sustained and equitable gains in function, participation, and quality of life.

Ecological Transfer, Everyday Independence, and Participation

Rehabilitation becomes socially consequential when skills are retained and used in environments that matter to the individual. Everyday independence depends on coordinated cognition, physical functioning, family support, transportation, work demands, financial resources, and institutional accessibility. A laboratory cognitive score can provide an important estimate of impairment yet fail to measure how a person manages medication, navigates a public place, or negotiates a conversation. The Canadian stroke rehabilitation framework explicitly addresses transitions and community participation as part of effective recovery care Salbach et al. (2025/2026). Research should consequently define patient-centered outcomes at the beginning of treatment evaluation, and should not treat a generic quality-of-life questionnaire as an automatic substitute for the activities a participant values.

Ecological transfer should be investigated as a measurable process. Investigators can test skills during therapy, again in untrained simulations, and finally through observations or reports of everyday behavior with the level of assistance documented. Durable transfer means performance persists after prompts, cues, or intensive therapist supervision have been reduced, unless the intervention is explicitly designed as permanent compensatory support. External aids may be appropriate permanent components of independent living, and their usability, reliability, accessibility, and dependence on support deserve assessment. das Nair et al. (2016) and Loetscher et al. (2019) show why immediate gains on particular cognitive measures cannot be assumed to improve independence or life satisfaction.

Participation has a relational dimension that standard neuropsychological batteries frequently miss. Successful communication or return to employment may depend on the flexibility of family members, employers, public institutions, and rehabilitation services. A purely impairment-centered study may incorrectly attribute all difficulties to residual cognitive deficits when organizational barriers contribute significantly. Conversely, increasing environmental support can improve participation without biological restoration, and this change remains an important clinical outcome. Research needs to evaluate person-environment fit through measurable tasks and outcomes without obscuring the psychological mechanisms under study. The scientific challenge is to connect levels of explanation while respecting the distinctive meaning of each level.

Individual Differences, Access, and Equity in Recovery

Recovery trajectories differ according to initial injury, residual pathways, prior skills, age, comorbid conditions, language, mental health, and opportunities for treatment. Clinical cohorts are often selected by the ability to travel to a specialized center, tolerate procedures, understand instructions, and attend repeated appointments. These criteria can systematically exclude people with severe aphasia, limited financial resources, medical complexity, or geographic isolation. Therefore results from a highly selected sample need careful transportability analysis before they are applied to an entire stroke or brain-injury population. Even a well-randomized study can have strong internal validity while its external applicability remains restricted by eligibility and the treatment setting.

Differences in educational experience and language can alter the meaning of neuropsychological scores. A low result on a verbal or culturally unfamiliar instrument may reflect a genuine deficit, measurement mismatch, or a combination of both. Prognostic models should report how these factors were assessed and whether subgroup performance was evaluated. Stinear et al. (2017) represents a promising predictor-development approach, yet such algorithms require scrutiny when the population differs from the source sample. The same question applies to computerized therapy that presumes reliable internet access, digital literacy, or the availability of a caregiver. Treatment efficacy and implementation feasibility are distinct quantities that both determine the real-world benefit of research.

Equitable rehabilitation design means defining outcomes in terms that reflect the diversity of patient goals and social contexts. Return to paid work is not a universal goal, and independence may include appropriate use of assistance rather than absence of every support. Participation measures should allow meaningful cultural and personal variation while retaining interpretable core definitions. Longitudinal studies require retention strategies that do not preferentially lose the most impaired participants. These considerations connect methodological fairness with scientific validity: a recovery model that describes only the most accessible patients will misrepresent the population-level mechanisms and opportunities under investigation Salbach et al. (2025/2026) Winstein et al. (2016).

Evidence Quality, Conflicts, and Clinical Safety

The evidence base contains several distinct types of inference. Animal experiments can demonstrate mechanisms under controlled injury and training conditions, but differ from human stroke in lesion heterogeneity, behavior, and ethical treatment constraints. Small mechanistic human studies can establish feasibility and physiological modulation, yet are commonly underpowered for durable clinical outcomes. Pragmatic multicenter trials can test real-world comparative effectiveness but may be less decisive about isolated mechanisms. Systematic reviews can synthesize the available corpus, although their conclusions depend on study quality, outcome compatibility, publication practices, and the treatment protocols represented. Nudo et al. (1996), Lindenberg et al. (2010), Rodgers et al. (2019), and Cicerone et al. (2019) occupy different positions in this evidentiary hierarchy.

A cautious assessment of treatment efficacy evaluates the prespecified primary outcome, confidence intervals, missing data, masking, allocation, intervention fidelity, adverse events, and the plausibility of transfer. Secondary improvements cannot erase an unfavorable primary endpoint or limitations created by multiple testing. AVERT Trial Collaboration Group (2015) provides an important safety finding because a more intensive very early mobilization protocol yielded worse favorable-outcome rates. Duncan et al. (2011) and Rodgers et al. (2019) illustrate why comparison with active rehabilitation matters. Loetscher et al. (2019), das Nair et al. (2016), and Chung et al. (2021) document the uncertainty that persists in several cognitive domains despite plausible biological rationales and reports of short-term test improvements.

Clinical recommendations must therefore remain linked to the population and protocol on which they rest. No individual program can be selected from evidence about neural plasticity alone; medical stability, functional goals, trained professional assessment, and comorbidity are essential. The 2025 Canadian recommendations integrate person-centered rehabilitation with evidence-based delivery Salbach et al. (2025/2026), while European aphasia guidance distinguishes supported speech-language therapy from investigational adjunctive stimulation Brady et al. (2025). An editorial synthesis should not imply that it substitutes for specialist care. Clinical safety and credible science converge on the principle that treatment choices need appropriate evidence, monitoring, adaptation to individual capacity, and transparent communication about uncertainty.

Open Problems and a Research Agenda for Recovery Science

A central unresolved question concerns the causal pathway from training to neural change to improved participation. Randomization identifies the effect of a treatment package on a measured outcome, but the intervening biological processes remain difficult to separate. Studies that combine precise behavioral intervention records, repeated neural measures, and rigorous function-based outcomes can test mechanistic mediation, provided they respect the assumptions and limitations of mediation analysis. Another open question is how timing interacts with dose and lesion characteristics: the findings of AVERT Trial Collaboration Group (2015) and the synthesis of Lang et al. (2015) suggest that the answer must be conditional, and intervention studies need sufficiently detailed time and dose measurements to examine those conditions.

A second priority is to strengthen ecological validity. Standardized tests facilitate comparison, but the prediction of practical competence needs direct validation. Research should test whether training gains transfer to new tasks, how long they persist, and what assistance is required for their continued use. Studies should record the costs and burden of intervention, including caregiver labor, specialist time, accessibility, and sustained adherence. Prognostic research requires external validation of models and explicit differences between prediction of general outcome and prediction of treatment response. The debate over proportional recovery illustrates why recovery curves also need measurement-aware mathematical modeling Bonkhoff et al. (2020) Bowman et al. (2021).

A third priority is methodological transparency across international research communities. Harmonized outcomes, accessible protocols, registered analyses, explicit treatment descriptions, and independent replication can reduce selective reporting and enable stronger evidence synthesis. Trials involving severe cognitive impairment or communication difficulty need adapted inclusion and assessment methods so that evidence reflects those populations. The scientific value of a new stimulation device, rehabilitation robot, or digital platform must be established through comparative effects on meaningful functioning rather than the novelty of its technology. These priorities unite mechanistic neuroscience with neuropsychology, rehabilitation psychology, biostatistics, and the study of social participation.

Conclusion

Neurocognitive recovery is a temporally organized interaction among the consequences of neural injury, remaining biological capacities, practice-dependent learning, compensatory strategies, and the environments in which functioning is measured and exercised. Neural plasticity provides a major biological basis for change, but the observation of plasticity alone does not determine whether meaningful recovery has occurred. Restitution, compensation, test-score improvement, and participation are distinct constructs that can follow different trajectories. An adequate account of any individual study identifies which construct was measured, the time since injury, the comparison condition, and the causal inferences justified by its design Bernhardt et al. (2017) Cramer et al. (2011).

Experimental rehabilitation has established that structured practice can improve important outcomes under defined conditions, as demonstrated by Wolf et al. (2006). Other substantial trials show why the scientific expectation of benefit must be disciplined by comparative evidence: specialized locomotor training was not superior to an active home-exercise comparator in LEAPS Duncan et al. (2011), and intervention dose and timing cannot be treated as universally favorable after the AVERT findings AVERT Trial Collaboration Group (2015). The cognitive rehabilitation literature presents more uneven evidence. Short-term gains in memory complaints or attention tasks are distinguishable from demonstrated long-term improvements in activities of daily living das Nair et al. (2016) Loetscher et al. (2019). Language rehabilitation has a stronger set of domain-specific recommendations, while brain stimulation remains investigational in relevant settings Brady et al. (2025).

The strongest synthesis is a multilevel research architecture linking mechanisms to behavior and behavior to daily functioning through transparent measurement and causal designs. Longitudinal cohorts describe recovery trajectories; randomized trials identify treatment effects; lesion and physiological measures inform mechanisms and prediction; ecological assessments establish functional significance; and systematic reviews evaluate the balance and quality of evidence. Future progress will depend on calibrated prediction, precise therapeutic dosing, independent replication, meaningful patient outcomes, and accessibility across diverse populations. This framework places neurocognitive recovery within the broader science of psychological adaptation while preserving the empirical distinctness of stroke, traumatic injury, and the specific functions being rehabilitated.

References

  • AVERT Trial Collaboration Group. (2015). Efficacy and safety of very early mobilisation within 24 h of stroke onset (AVERT): A randomised controlled trial. The Lancet, 386(9988), 46–55. https://doi.org/10.1016/S0140-6736(15)60690-0
  • Bernhardt, J., Hayward, K. S., Kwakkel, G., Ward, N. S., Wolf, S. L., Borschmann, K., Krakauer, J. W., Boyd, L. A., Carmichael, S. T., Corbett, D., & Cramer, S. C. (2017). Agreed definitions and a shared vision for new standards in stroke recovery research: The Stroke Recovery and Rehabilitation Roundtable taskforce. Neurorehabilitation and Neural Repair, 31(9), 793–799. https://doi.org/10.1177/1545968317732668
  • Bonkhoff, A. K., Hope, T., Bzdok, D., Guggisberg, A. G., Hawe, R. L., Dukelow, S. P., Rehme, A. K., Fink, G. R., Grefkes, C., & Bowman, H. (2020). Bringing proportional recovery into proportion: Bayesian modelling of post-stroke motor impairment. Brain, 143(7), 2189–2206. https://doi.org/10.1093/brain/awaa146
  • Bowman, H., Bonkhoff, A., Hope, T., Grefkes, C., & Price, C. (2021). Inflated estimates of proportional recovery from stroke: The dangers of mathematical coupling and compression to ceiling. Stroke, 52(5), 1915–1920. https://doi.org/10.1161/STROKEAHA.120.033031
  • Brady, M. C., Kelly, H., Godwin, J., Enderby, P., & Campbell, P. (2016). Speech and language therapy for aphasia following stroke. Cochrane Database of Systematic Reviews, 2016(6), CD000425. https://doi.org/10.1002/14651858.CD000425.pub4
  • Brady, M. C., Mills, C., Øra, H. P., Novaes, N., Becker, F., Constantinidou, F., Flöel, A., Sunnerhagen, K. S., Isaksen, J., Jagoe, C., Jesus, L. M. T., Marangolo, P., Meinzer, M., van der Meulen, I., Campbell, P., Ho, L., Hussain, S., & Hilari, K. (2025). European Stroke Organisation (ESO) guideline on aphasia rehabilitation. European Stroke Journal, 10(4), 1189–1220. https://doi.org/10.1177/23969873241311025
  • Chung, C. S. Y., Pollock, A., Campbell, T., Durward, B. R., & Hagen, S. (2021). Cognitive rehabilitation for executive dysfunction in adults with stroke or other adult non-progressive acquired brain damage. Cochrane Database of Systematic Reviews, 2021(9), CD008391. https://doi.org/10.1002/14651858.CD008391.pub2
  • Cicerone, K. D., Goldin, Y., Ganci, K., Rosenbaum, A., Wethe, J. V., Langenbahn, D. M., Malec, J. F., Bergquist, T. F., Kingsley, K., Nagele, D., Trexler, L., Fraas, M., Bogdanova, Y., & Harley, J. P. (2019). Evidence-based cognitive rehabilitation: Systematic review of the literature from 2009 through 2014. Archives of Physical Medicine and Rehabilitation, 100(8), 1515–1533. https://doi.org/10.1016/j.apmr.2019.02.011
  • Cramer, S. C., Sur, M., Dobkin, B. H., O'Brien, C., Sanger, T. D., Trojanowski, J. Q., Rumsey, J. M., Hicks, R., Cameron, J., Chen, D., Chen, W. G., Cohen, L. G., deCharms, C., Duffy, C. J., Eden, G. F., Fetz, E. E., Filart, R., Freund, M., Grant, S. J., ... Vinogradov, S. (2011). Harnessing neuroplasticity for clinical applications. Brain, 134(6), 1591–1609. https://doi.org/10.1093/brain/awr039
  • das Nair, R., Cogger, H., Worthington, E., & Lincoln, N. B. (2016). Cognitive rehabilitation for memory deficits after stroke. Cochrane Database of Systematic Reviews, 2016(9), CD002293. https://doi.org/10.1002/14651858.CD002293.pub3
  • Duncan, P. W., Sullivan, K. J., Behrman, A. L., Azen, S. P., Wu, S. S., Nadeau, S. E., Dobkin, B. H., Rose, D. K., Tilson, J. K., Cen, S., & Hayden, S. K. (2011). Body-weight–supported treadmill rehabilitation after stroke. New England Journal of Medicine, 364, 2026–2036. https://doi.org/10.1056/NEJMoa1010790
  • Kleim, J. A., & Jones, T. A. (2008). Principles of experience-dependent neural plasticity: Implications for rehabilitation after brain damage. Journal of Speech, Language, and Hearing Research, 51(1), S225–S239. https://doi.org/10.1044/1092-4388(2008/018)
  • Kolb, B., Gibb, R., & Robinson, T. E. (2003). Brain plasticity and behavior. Current Directions in Psychological Science, 12(1), 1–5. https://doi.org/10.1111/1467-8721.01210
  • Krakauer, J. W., Carmichael, S. T., Corbett, D., & Wittenberg, G. F. (2012). Getting neurorehabilitation right: What can be learned from animal models? Neurorehabilitation and Neural Repair, 26(8), 923–931. https://doi.org/10.1177/1545968312440745
  • Lang, C. E., Lohse, K. R., & Birkenmeier, R. L. (2015). Dose and timing in neurorehabilitation: Prescribing motor therapy after stroke. Current Opinion in Neurology, 28(6), 549–555. https://doi.org/10.1097/WCO.0000000000000256
  • Lindenberg, R., Renga, V., Zhu, L. L., Nair, D., & Schlaug, G. (2010). Bihemispheric brain stimulation facilitates motor recovery in chronic stroke patients. Neurology, 75(24), 2176–2184. https://doi.org/10.1212/WNL.0b013e318202013a
  • Loetscher, T., Potter, K.-J., Wong, D., & das Nair, R. (2019). Cognitive rehabilitation for attention deficits following stroke. Cochrane Database of Systematic Reviews, 2019(11), CD002842. https://doi.org/10.1002/14651858.CD002842.pub3
  • Nudo, R. J., Wise, B. M., SiFuentes, F., & Milliken, G. W. (1996). Neural substrates for the effects of rehabilitative training on motor recovery after ischemic infarct. Science, 272(5269), 1791–1794. https://doi.org/10.1126/science.272.5269.1791
  • Rodgers, H., Bosomworth, H., Krebs, H. I., van Wijck, F., Howel, D., Wilson, N., Aird, L., Alvarado, N., Andole, S., Cohen, D. L., Dawson, J., Fernandez-Garcia, C., Finch, T., Ford, G. A., Francis, R., Hughes, N., Price, C. I. M., Pundik, S., Thomson, R. G., ... Shaw, L. (2019). Robot assisted training for the upper limb after stroke (RATULS): A multicentre randomised controlled trial. The Lancet, 394(10192), 51–62. https://doi.org/10.1016/S0140-6736(19)31055-4
  • Salbach, N. M., Yao, J. K., Lindsay, M. P., Nelson, M. L. A., Shi, J., O'Connell, C., Barclay, R., Bastasi, D., Boulos, M. I., Boyce, J., et al. (2026; published online 2025). Canadian Stroke Best Practice Recommendations rehabilitation, recovery, and community participation following stroke. Part two: Delivery of stroke rehabilitation to optimize functional recovery, 7th edition update 2025. American Journal of Physical Medicine & Rehabilitation, 105(2), 151–167. https://doi.org/10.1097/PHM.0000000000002821
  • Stinear, C. M., Byblow, W. D., Ackerley, S. J., Smith, M.-C., Borges, V. M., & Barber, P. A. (2017). PREP2: A biomarker-based algorithm for predicting upper limb function after stroke. Annals of Clinical and Translational Neurology, 4(11), 811–820. https://doi.org/10.1002/acn3.488
  • Winstein, C. J., Stein, J., Arena, R., Bates, B., Cherney, L. R., Cramer, S. C., Deruyter, F., Eng, J. J., Fisher, B., Harvey, R. L., Lang, C. E., MacKay-Lyons, M., Ottenbacher, K. J., Pugh, S., Reeves, M. J., Richards, L. G., Stiers, W., & Zorowitz, R. D. (2016). Guidelines for adult stroke rehabilitation and recovery: A guideline for healthcare professionals from the American Heart Association/American Stroke Association. Stroke, 47(6), e98–e169. https://doi.org/10.1161/STR.0000000000000098
  • Wolf, S. L., Winstein, C. J., Miller, J. P., Taub, E., Uswatte, G., Morris, D., Giuliani, C., Light, K. E., & Nichols-Larsen, D. (2006). Effect of constraint-induced movement therapy on upper extremity function 3 to 9 months after stroke: The EXCITE randomized clinical trial. JAMA, 296(17), 2095–2104. https://doi.org/10.1001/jama.296.17.2095
  • Wolf, S. L., Winstein, C. J., Miller, J. P., Thompson, P. A., Taub, E., Uswatte, G., Morris, D., Blanton, S., Nichols-Larsen, D., & Clark, P. C. (2008). Retention of upper limb function in stroke survivors who have received constraint-induced movement therapy: The EXCITE randomised trial. The Lancet Neurology, 7(1), 33–40. https://doi.org/10.1016/S1474-4422(07)70294-6