Neuroplasticity — A Brain That Builds New Roads Beside the Broken Ones

The Day a Finger First Moved

A person whose one arm could not move at all after a stroke twitches their index finger one morning in the third week. Both they and the family beside them stare at that small movement for a long time. Then they ask: how far does this improvement go, and until when does it continue.

There are more questions. When a therapist suggests practicing everyday movements together with the unaffected hand, people wonder if that means giving up on the weak hand. Rumors also float down the hospital hallway. The saying that no further improvement happens after six months, and the saying that practice is better the more you do it.

The answers to these questions begin inside the brain.

The Brain Works Through Connections

Stroke is a disease where blood flowing to the brain is suddenly blocked or bursts, damaging the brain tissue that received that blood. Cells in the damaged area generally do not revive. Even so, improvements in movement after injury are observed in both humans and animals (PMID 24136129). One reason for this lies in the way the brain works.

Neurons are the basic cells of the brain and nervous system that receive electrical signals and pass them to the next cell. Even for a single action like reaching a hand toward a cup, a network of numerous neurons linked in a row moves together. A synapse is the junction where one neuron passes a signal to the next neuron. Junctions that are used together frequently get better at passing signals, while junctions that are not used for a long time grow weak.

Neuroplasticity is the ability of the brain to reorganize itself by creating new neural connections. Whether an uninjured brain learns a new skill or an injured brain relearns a lost skill, this ability is considered to be the foundation (PMID 18230848). Practice is also the path through which the brain changes. When the same movement is repeated, the junctions used for that movement are used frequently, and frequently used junctions get better at passing signals. Therefore, the rules that govern learning are likely to be very important for relearning after an injury as well (PMID 18230848).

Comparing it to roads, it is like this. If a landslide cuts off one road, it is difficult to restore that road just as it was. Instead, the remaining roads are widened, and side roads are connected that go around the cut-off spot. A similar thing happens in the brain. Brain imaging of humans and tests that stimulate the brain with magnetic fields from outside the skull to observe responses show this. Regaining the use of a hand appears together with the rewiring of networks encompassing the area around the damage and the opposite side of the brain—in other words, reorganization (PMID 15986729).

Analogies have limits. The brain's "new roads" include both newly sprouting connections and changes where the strength of already existing junctions shifts. Furthermore, not all improvements after injury come from new roads. The share of vulnerable surrounding tissue surviving and the share of distant regions with dropped activity coming back to life are also mixed in (PMID 23963341). Nor is a newly connected road always a good one. Reorganization after injury can sometimes help movement and sometimes harm it (PMID 25762917).

The details of these changes have been largely revealed through animal experiments. The work of matching what is found in animals with changes seen in patients is still ongoing today (PMID 25014877). For instance, the conclusion that aerobic exercise increases nerve growth substances inside the brain is like that. Among the studies dealing with this conclusion, only one study involved humans, while the remaining twenty were animal experiments. Therefore, this conclusion practically relies on animal evidence (PMID 30210424).

The process by which an injured brain improves involves changes to remaining connections, and a significant portion of those changes is thought to follow the principles of learning.

Two Meanings of Getting Better

A person who has learned to use their right hand to eat with a spoon again and a person who has learned to use their left hand both eat alone. The way the two people accomplished it is different.

Rehabilitation is treatment and care that helps a person whose everyday functioning has decreased or is at risk of decreasing due to illness, injury, or aging to regain and maintain functions and live again. Function is what the body and mind accomplish, ranging from body movements to everyday activities like getting dressed and walking, to roles within the workplace and family. There are two ways to regain function, and they have separate names.

Recovery is the return of movement or function in the same way as before the injury. This is the side where the right hand lifts the spoon again. Compensation is accomplishing the same task by using another body part or another method instead of the lost way. This is the side of learning to use the left hand for eating with a spoon.

In daily life, "recovery" refers to everything that gets better. Many studies and guidelines also broadly call the entire process of regaining function "recovery." This article uses "recovery" only in the narrow sense above. When conveying results in the broad sense, it is written as "getting better."

The two terms have various meanings even among researchers and clinicians. Therefore, a proposal was made to define them separately across three tiers: the neural cell tier, the movement pattern tier, and the everyday activity tier (PMID 19118128). An international consensus meeting of scholars studying recovery after stroke also organized common terms and definitions to be used in research (PMID 28697708).

There are forms of compensation at each of the three tiers. At the everyday activity tier, using the left hand to eat with a spoon is compensation. At the movement pattern tier, reaching for a cup by tilting the shoulder and torso when reaching out the arm is compensation. Therefore, even if someone lifts a spoon with their right hand, if they lift it by tilting their torso, it is compensation in terms of movement pattern.

There is also compensation inside the brain. If the primary pathway that used to send movement commands is injured, other pathways can be used more heavily to fill the vacancy. Pathways used as substitutes in this way can also be involved in stereotyped movements that settle in over time. These are movements where multiple muscles exert force all at once in the same fixed pattern. These reorganized pathways can sometimes help movement and sometimes harm it (PMID 25762917).

Where to place the weight between the two has not yet been settled. One review considers it important to actively use the injured-side network rather than relying on compensation strategies. The reason is that maladaptive reorganization can solidify compensation strategies, preventing the return of original movement. However, this review also notes that most of the evidence comes from short- to medium-term studies (PMID 41967694).

There is also the observation that what was learned first can make subsequent learning or correction difficult. The same observation also notes that these limitations can be overcome by creating a suitable environment (PMID 22415920).

Compensation also has its uses. Compensation strategies do not reduce impairment—that is, the extent to which the body's movements itself have decreased due to injury. Instead, they get people back home as early as possible. Reviews emphasizing recovery also note this, and consider that the recovery side, which reduces impairment, may last longer (PMID 31920570). Even when the ability to plan and judge is injured, compensation strategies such as using reminder devices or learning problem-solving sequences can be helpful. The evidence is limited but positive (PMID 22436364).

Recently, a proposal has emerged that dividing the two as if cutting them with a knife is itself ambiguous. The idea is to view getting better based on how deeply it reaches from overtly manifested actions down to the inner layers of the brain. In this framework, compensation is getting better at shallow layers. If overtly manifested outcomes are the same, the side that has improved down to the inner layers is considered more complete (PMID 41877523). Distinguishing the two is not easy to begin with. A 2012 paper discussing walking assessment pointed out that the assessments used in clinical practice at the time could not distinguish between the two (PMID 22333921).

Recovery, where the old way returns, and compensation, accomplishing it by a different way, are both getting better, and researchers aim for even deeper recovery wherever possible.

When — The First Few Months, and Beyond

The speed of improvement after an injury is fastest at first, drawing a curve that becomes more gradual as time goes on (PMID 23963341). This curve contains the share contributed by practice, along with the share that happens regardless of treatment (PMID 23963341).

Spontaneous recovery is improvement that happens on its own as time passes after an injury, regardless of treatment. Four things are intertwined in the process of getting better after an injury. First, surrounding tissue that was in danger due to a lack of blood survives between the first few days and a few weeks. Second, distant regions connected to the injured site temporarily drop in activity and then revive again. Third is the reorganization of connections, and fourth is compensation. The four influence one another, and each operates at a different time (PMID 23963341).

Looking at the degree of initial symptoms, spontaneous recovery 3 to 6 months later can be predicted quite well in the vast majority of patients (PMID 37895442). The debate surrounding the so-called "Proportional Recovery Rule"—the idea that a fixed proportion of what was lost comes back—is covered in "Does 70% of What Was Lost Come Back?".

During those same first few months, the brain also responds much more strongly to practice. Improvement in which the damage itself decreases occurs mostly in the first 1 to 3 months in both humans and animals. Two things overlap during this period: the brain spontaneously rewiring itself, and the responsiveness to training increasing (PMID 24136129). A sensitive period is a not-so-long period after an injury when the brain reacts uncommonly strongly to training. Using the analogy of a road, it is the time right after construction when the ground is soft and new paths form well.

The evidence for sensitive periods first emerged from rat experiments. When rats train on specific tasks timed to this period, they recover almost completely (PMID 34544853).

A study testing this in humans is CPASS (Critical Period Stroke Study). It is a phase 2 randomized trial that gathered 72 patients with mild to moderate impairments in arm function. A trial at this stage is an occasion to first gauge the effect of a treatment, and a randomized trial is a study that divides and compares groups like a raffle. All patients received standard rehabilitation, and the practice group added 20 hours of task-oriented arm practice. Among the practice groups, only the timing at which that practice started differed (PMID 37744191).

The results, compared against the control group that received only standard rehabilitation 1 year later, are as follows. Both the group that started at 2 to 3 months post-stroke and the group that started within 30 days were statistically and markedly ahead. The margin by which they were ahead was slightly larger for the 2- to 3-month side, but the result of comparing the two groups against each other is not in the abstract. The group that started after 6 months showed no distinct difference from the control group (PMID 34544853). The researchers conclude that task-oriented practice was most effective within the first 2 to 3 months after onset. And they view this as empirical evidence that humans, too, have a sensitive period for motor function to improve (PMID 34544853).

What this trial compared was when the same 20 hours were added. Therefore, what the result of the post-6-month group says is simply that those 20 hours did not create a distinct difference from the control group. The conclusion that one cannot get better after that is not contained in this result. The chronic stage is the period typically 6 months past onset. Evidence that improvement happens even in the chronic stage has been built up separately.

There is strong evidence that intensive task practice with high repetition helps at all periods after a stroke (PMID 24505342). This judgment came from a review. A review is a study that gathers studies dealing with one question and synthesizes the results. Out of 467 randomized trials, this review gathered only trials with a low risk of bias and judged each training one by one. "Strong evidence" is a grade according to criteria set by the review itself, and the effects mostly stayed at the level of the trained functions and activities (PMID 24505342).

Evidence that therapy forcing the use of the weak arm increases arm use in chronic patients has also been built up over 20 years (PMID 16917156). In actual clinical practice, too, there are observations that patients' movements continue to improve after the spontaneous recovery period has passed (PMID 37895442; a perspective paper organizing clinical experience).

Then, is the late-arriving improvement recovery, or is it compensation? This question has not yet been solved. One review divides the two by timing. It summarizes that improvement where the damage itself decreases relies on a short sensitive period, whereas compensation can happen at any time after an injury (PMID 24136129). On the other hand, even in studies that performed task-oriented therapy on chronic patients, there are reports that measures assessing motor impairment improved (PMID 39395360; single-group studies without a control group).

Disentangling how much of late improvement is recovery and how much is compensation is counted as a practical challenge (PMID 22415920). The 2012 walking assessment paper we saw earlier also pointed out that the two could not be separated by clinical evaluations at the time (PMID 22333921).

Another review views that evidence is mounting that the improvement curve can be altered by intensive task practice. It also views that it is better to start that practice within 6 months if possible. The same review notes that it is still not well understood how much practice changes the spontaneous improvement process itself (PMID 23963341). To put it in terms of a road, the ground hardens as time goes on, but paths can still be made on hardened ground.

The first few months see the greatest magnitude of improvement and response to practice, and there is evidence that improvement through practice continues even after six months.

What You Practice Grows

Stroke patients who performed strength training generally increased their limb strength, but their comfortable walking speed remained almost unchanged (PMID 40990154). The researchers suggest this may be because the exercise was not similar enough to the task of walking (PMID 40990154).

Relearning or newly learning movements lost after a stroke is a form of learning. Therefore, learning to move is considered a cornerstone of stroke rehabilitation (PMID 38514218).

Motor learning is the process by which movement skill changes to become enduring through repeated practice. Learning has certain properties. Task specificity is the property whereby the exact movements and activities practiced improve the most. The effects of physical therapy after a stroke largely remain confined to the functions and activities actually trained (PMID 24505342). This is also the property the researchers suspected as the reason why strength training barely changed walking speed. Put simply like a path, the paths walked often become wider, and only the paths walked get wider.

There are several other principles of learning besides this one. One review lists fifteen. Among them are practicing actual tasks as they are, goal-directed practice, practicing while varying conditions, gradually increasing difficulty, and feedback providing results (PMID 31920570). The same review assesses that these principles have not yet been accurately translated into current treatments (PMID 31920570).

It has also been pointed out that the setting where one practices and the setting where one actually uses the skill must resemble each other for skills learned in the hospital to transfer outside the hospital (PMID 11530884). The aforementioned CPASS also selected practice content from meaningful daily activities chosen by the participants themselves (PMID 37744191). Physical rehabilitation mixing multiple elements is likely to improve function and mobility after a stroke, and among them, methods practicing daily tasks can be helpful. However, because subjects and methods vary greatly across studies, the certainty of the evidence is limited (PMID 39932103).

Computer-based cognitive training has been studied in patients with cognitive decline after a stroke. The evidence that it raises global cognition and attention test scores is of moderate-to-high certainty. The effects on memory, language, and motor skills are of low certainty (PMID 40503808). Evidence that activities of daily living improve is still insufficient (PMID 34036682). This means that improvements in test scores and improvements in daily life must be verified separately.

Among 19 studies on computer-based brain training involving healthy older adults, about half failed to properly control for placebo effects (PMID 32589159; systematic review of studies from 2016–2018). A placebo effect is an improvement that occurs solely due to the expectation that there will be an effect. The results from these studies—showing improvements even in abilities that were not trained—form the basis for the broad claims put forward by brain-training game manufacturers (PMID 32589159).

Technology such as robots is also being studied as a tool to assist with this practice. Comparisons where robot training was performed instead of conventional therapy generally showed similar effects, and for patients with severely weak legs, the robot group fared better (PMID 29219877; systematic review of 51 randomized controlled trials). Cases where robot-assisted gait training is added to physical therapy are covered in another installment of this series, "Paralysis and Spasticity".

Because what is practiced improves the most, task-oriented training uses the daily activities one wants to recover directly as practice material.

How Much — The Amount of Practice and Its Limits

In inpatient rehabilitation wards, practicing arm use averaged 4 minutes per physical therapy session and 17 minutes per occupational therapy session. In two studies that counted repetitions, the totals were 23 and 32 per session. Both values combined physical therapy and occupational therapy without distinction (PMID 25568073; a review pooling 10 observational studies). This review concluded that the amount was small (PMID 25568073).

In rehabilitation, dose refers to how much practice is performed, commonly measured by practice time or repetition count, and sometimes alongside how frequently and intensely it was done. In studies measuring therapy that forced intensive use of the weaker arm by repetition count, daily repetitions ranged from 45 to 1,280 (PMID 33884528).

There is also strong evidence that groups who practiced more did better. This was graded by the criteria of the randomized trial review seen earlier, and the size of the difference was small to moderate (PMID 24505342). In an analysis pooling arm function studies, improvement was greatest when task-specific training exceeded 30 hours. However, this analysis pooled single-arm studies without control groups (PMID 39395360).

The saying "more is better," and the evidence for it, both have limits. A 2014 Cochrane review illustrates this. Cochrane is an international research organization that collects and summarizes evidence on treatments. In this review, the group that did relatively more repetitive task practice showed benefits, while the group that did less did not. However, the difference between the two groups was not statistically confirmed (PMID 25387001; moderate certainty).

This means the two groups did not diverge enough to definitively establish that doing more was better. This review concluded that evidence on dose is particularly needed (PMID 25387001).

Increasing the amount does not yield a straight upward climb in effectiveness. For walking and mobility, improvements increased as exercise volume grew, but peaked at a moderate amount. This is an analysis modeling differences in exercise volume across trials, pooling 25 randomized trials involving 788 patients aged 45 and older (PMID 41564822).

Physical condition can also restrict the amount of practice. Cardiorespiratory fitness refers to the ability of the heart and lungs to supply oxygen to the body, and in stroke patients, cardiorespiratory fitness is often lower than the level required for daily activities. This can act as a ceiling that blocks the very ability to practice frequently enough and intensely enough (PMID 27661010). Timing operates alongside it. As seen earlier in CPASS, the same 20 hours yielded different results depending on when they were performed (PMID 34544853).

Guidelines strongly recommend exercise as part of rehabilitation (PMID 42742033). How much to do and in what way are determined by considering timing and physical condition together. The optimal dose is a question that research has yet to answer (PMID 25387001, PMID 42742033).

The amount of practice matters, but simply increasing the amount does not endlessly increase effectiveness; it operates in tandem with timing, content, and physical condition.

In Front of That Finger Again

Let us return to that index finger that first moved in the third week. That was within the first few months when the improvement curve is steep. It is a time when the share that improves on its own and the heightened response to practice exist together. In a trial that added more practice of the same kind to people, practice added within the first 2 to 3 months showed a clear difference from the control group, whereas practice added six months later did not. However, that trial was a small study enrolling 72 patients with mild to moderate arm function.

This sensitive period is not long. The therapist starts from the movement the person can manage right now, and gradually raises the difficulty toward real-world actions like grasping a cup. This approach aligns with two properties of learning. One is the learning principle of gradually raising the difficulty, and the other is the property that the practiced task improves the most, namely task specificity.

Practicing daily activities with the unaffected hand is compensation; even if it does not reduce the impairment, it is a path to returning home earlier (PMID 31920570). On the other hand, there is a review suggesting that relying on compensatory strategies may reduce the extent to which original movement returns, and this review emphasizes therapy that actively uses the injured side. Most of the evidence for this comes from short- to medium-term studies, and where to place the weight has not yet been settled (PMID 41967694). The balancing act between these two practices is covered more deeply in "Able But Unused — An Introduction to Learned Non-Use and CIMT".

We can now also answer the two statements in the hallway. Even if improvement slows down after six months, evidence remains that one can improve through subsequent practice. The amount of practice matters, but simply increasing the amount does not make the effect grow endlessly. How we set goals together is covered in another part of this series, "Rehabilitation".

Questions without answers still remain: what the optimal timing and amount are for each person, and how to distinguish whether the improvement right in front of us is recovery or compensation.

Where to Look Further

• Deeper roots of this article: "How the Central Nervous System Recovers — A Primer on Neuroplasticity"
• Timing: "When Does Recovery Happen — What Time Did, and What Treatment Did", and the debate piece "Does 70% of What Was Lost Come Back? — Stroke's "Proportional Recovery Rule" and the Illusion Statistics Built"
• Recovery and compensation: "How Deep Is the Recovery — Between Compensation and True Recovery", "Able But Unused — An Introduction to Learned Non-Use and CIMT"
• Principles and amount of learning: "Why Therapy Is Designed That Way — An Introduction to Motor Learning Principles", "How Much Is Enough — An Introduction to Dose-Response in Rehabilitation"
• How to measure improvement: "How Do We Measure Recovery — An Introduction to Rehabilitation Outcome Measures"
• Other pieces in this series: "Rehabilitation" (Rehab Team and Goals), "Paralysis and Spasticity" (Weakness, Spasticity, Walking), "Swallowing, Speech and Language" (Swallowing, Speech and Language)

Source Papers

• "Critical Period After Stroke Study (CPASS): A phase II clinical trial testing an optimal time for motor recovery after stroke in humans." (2021) — PMID 34544853 — Additional 20 hours of task-oriented arm practice: starting at 2 to 3 months showed the largest benefit, within 30 days was small but significantly improved, and after 6 months showed no significant difference from the control group. Evidence for a human sensitive period for motor recovery (starting from rat studies).
• "The Critical Period After Stroke Study (CPASS) Upper Extremity Treatment Protocol." (2023) — PMID 37744191 — 72 CPASS participants with mild-to-moderate arm function, receiving 20 hours of treatment composed of meaningful activities chosen by the participant.
• "The interaction between training and plasticity in the poststroke brain." (2013) — PMID 24136129 — In both humans and animals, most recovery from damage occurs in the first 1 to 3 months (spontaneous reorganization + increased training response), representing a brief sensitive period. Compensation is possible at any time.
• "Understanding upper limb recovery after stroke." (2013) — PMID 23963341 — Log-type curves can be altered by intensive task practice started preferably within 6 months, with four mechanisms (penumbra rescue, resolution of diaschisis, plasticity, behavioral compensation) operating at different times. The effect of practice on natural mechanisms remains unclear.
• "Stroke Recovery Is a Journey: Prediction and Potentials of Motor Recovery after a Stroke from a Practical Perspective." (2023) — PMID 37895442 — Initial impairment allows prediction of 3 to 6 months of spontaneous recovery in the majority, while actual clinical practice shows continued motor function improvement even after that.
• "What is the evidence for physical therapy poststroke? A systematic review and meta-analysis." (2014) — PMID 24505342 — 467 randomized trials: intensive, repetitive task practice benefits all time periods (strong evidence), higher doses are better (effect size 0.21–0.61), and effects are limited to the trained function.
• "Constraint-Induced Movement therapy: answers and questions after two decades of research." (2006) — PMID 16917156 — Evidence for the efficacy of CI therapy on chronic stroke arm use accumulated over two decades (starting from monkey experiments).
• "Principles of experience-dependent neural plasticity: implications for rehabilitation after brain damage." (2008) — PMID 18230848 — Neuroplasticity is considered the basis for learning in the intact brain and relearning in the damaged brain, and its properties and constraints likely matter for rehabilitation.
• "Reorganization of cerebral circuits in human brain lesion." (2005) — PMID 15986729 — Human imaging and TMS: hand function recovery is linked to network reorganization involving the peri-lesional area and the contralateral hemisphere.
• "Interplay between intra- and interhemispheric remodeling of neural networks as a substrate of functional recovery after stroke: adaptive versus maladaptive reorganization." (2014) — PMID 25014877 — A review aligning animal and patient study findings at the cellular and network levels.
• "Post-stroke BDNF Concentration Changes Following Physical Exercise: A Systematic Review." (2018) — PMID 30210424 — Out of 21 papers (1 human, 20 animal), conclusions on aerobic exercise-induced BDNF changes rely on animal evidence (functional training response remains debated).
• "What do motor "recovery" and "compensation" mean in patients following stroke?" (2009) — PMID 19118128 — Recovery and compensation terms vary widely, proposing definitions at the neural, motor performance, and functional levels.
• "Agreed definitions and a shared vision for new standards in stroke recovery research: The Stroke Recovery and Rehabilitation Roundtable taskforce." (2017) — PMID 28697708 — Consensus on common terminology and working definitions for stroke recovery research.
• "A neuroanatomical framework for upper limb synergies after stroke." (2015) — PMID 25762917 — When primary descending pathways are damaged, alternative pathways compensate by ascending, and reorganization can be harmful or beneficial.
• "Neuroplasticity after stroke: Adaptive and maladaptive mechanisms in evidence-based rehabilitation." (2026) — PMID 41967694 — Intensive, repetitive, task-specific treatment is effective, and actively engaging the damaged network is more important than compensatory strategies.
• "Sensitive periods for language and recovery from stroke: conceptual and practical parallels." (2012) — PMID 22415920 — Learning itself can serve as a stepping stone for subsequent learning and failure to correct, and limits can be pushed past through an appropriate environment.
• "Principles of Neurorehabilitation After Stroke Based on Motor Learning and Brain Plasticity Mechanisms." (2019) — PMID 31920570 — 15 principles of motor learning, compensatory strategies do not reduce impairment but enable early discharge, and clinical application of these principles remains insufficient.
• "Efficacy of executive function interventions after stroke: a systematic review." (2012) — PMID 22436364 — Limited evidence that remedial and compensatory strategies (problem-solving, prompting devices) can help improve executive function.
• "How deep is your recovery? Rethinking true motor recovery versus compensation in neurorehabilitation." (2026) — PMID 41877523 — A depth-based layered model instead of a recovery/compensation dichotomy, where deeper recovery is more desirable when behavioral outcomes are equal.
• "Advancing measurement of locomotor rehabilitation outcomes to optimize interventions and differentiate between recovery versus compensation." (2012) — PMID 22333921 — Current clinical measurements fail to distinguish between neuromotor recovery and compensation.
• "Cardiorespiratory training for people with stroke." (2026) — PMID 42742033 — Chronic phase defined as >6 months, guidelines strongly recommend exercise for rehabilitation, and cardiorespiratory training does not affect mortality or recurrence while slightly improving physical fitness and gait.
• "Resistance training for people with stroke." (2025) — PMID 40990154 — Increases muscle strength but has little effect on comfortable walking speed (possibly due to insufficient task similarity with walking).
• "Motor Learning Following Stroke: Mechanisms of Learning and Techniques to Augment Neuroplasticity." (2024) — PMID 38514218 — Relearning lost motor skills and acquiring new ones, with motor skill learning serving as a cornerstone of stroke rehabilitation.
• "Adaptability and flexibility of the human motor system: implications for neurological rehabilitation." (2001) — PMID 11530884 — Transfer is optimal when the learning context and application context resemble each other.
• "Physical rehabilitation approaches for the recovery of function and mobility following stroke." (2025) — PMID 39932103 — 267 Cochrane trials: physical rehabilitation likely improves function and mobility, functional task training may be useful, and certainty is limited by heterogeneity.
• "The Effectiveness of Computerized Cognitive Training in Patients With Poststroke Cognitive Impairment: Systematic Review and Meta-Analysis." (2025) — PMID 40503808 — Improves cognition, attention, executive function, and quality of life (moderate to high certainty), with low certainty for memory, language, and motor function.
• "Controlling for Placebo Effects in Computerized Cognitive Training Studies With Healthy Older Adults From 2016-2018: Systematic Review." (2020) — PMID 32589159 — Out of 19 papers involving healthy older adults, 9 (47%) had poor placebo controls, and positive far-transfer results formed the basis of manufacturer claims.
• "Efficacy of computerized cognitive training on improving cognitive functions of stroke patients: A systematic review and meta-analysis of randomized controlled trials." (2022) — PMID 34036682 — Improves cognitive function, with insufficient evidence for effects on activities of daily living or depression.
• "Effectiveness of robotic assisted rehabilitation for mobility and functional ability in adult stroke patients: a systematic review." (2017) — PMID 29219877 — 51 papers: robotic training is comparable to standard training, with a robotic advantage only in patients with severely impaired lower limbs.
• "Dose of arm activity training during acute and subacute rehabilitation post stroke: a systematic review of the literature." (2015) — PMID 25568073 — During inpatient rehabilitation, arm activity training consisted of 4 minutes of physical therapy and 17 minutes of occupational therapy per session, across 23 to 32 sessions, showing limited dose.
• "Constraint-induced movement therapy protocols using the number of repetitions of task practice: a systematic review of feasibility and effects." (2021) — PMID 33884528 — CIMT repetition counts ranged from 45 to 1,280 times per day, demonstrating that repetition-based protocols are feasible.
• "Predictive factors and dose-response effect of rehabilitation for upper limb induced recovery after stroke: systematic review with proportional meta-analyses." (2024) — PMID 39395360 — Proportional meta-analysis of single-cohort longitudinal studies: large effects observed above 30 hours, with task-oriented approaches maximizing benefits in both subacute and chronic stages.
• "Interventions for improving upper limb function after stroke." (2014) — PMID 25387001 — Cochrane overview: high-dose repetitive task practice yields benefits (moderate certainty), subgroup differences between doses were not significant, and evidence specifically regarding dose is needed.
• "Effects of overall and modality-specific exercise dose on lower-limb functional mobility in middle-aged and older adults with stroke: A systematic review and bayesian dose-response meta-analysis." (2026) — PMID 41564822 — 25 papers involving 788 people, connecting exercise volume (MET-min/week) across trials using a Bayesian model: lower-limb mobility is non-linear, peaking at moderate doses.
• "Four birds with one stone? Reparative, neuroplastic, cardiorespiratory, and metabolic benefits of aerobic exercise poststroke." (2016) — PMID 27661010 — Low cardiorespiratory fitness may act as a "rehabilitation ceiling" preventing sufficient practice, alongside neurobiological effects of aerobic exercise observed in animals.