Action Observation Combined With Virtual Reality Promotes Motor Recovery After Stroke: A Randomized Controlled Trial

Clinical Question (PICO)

In adults with poststroke hemiplegia and residual unilateral upper-limb movement 3 to 18 months after the acute event (P), does 5 weeks of action observation immediately followed by imitation-based virtual-reality motor training (AO+VR) (I), compared with the same virtual-reality motor training preceded by neutral nature videos (control observation plus VR, CO+VR) (C), improve paretic-hand dexterity and broader motor and functional outcomes at treatment completion and 6 months (O)? The prespecified primary comparison was the between-group difference in change in Box and Block Test (BBT) performance for the paretic hand.

Bottom Line

In this small, assessor-blinded randomized trial, adding goal-directed action observation to VR training produced greater paretic-hand dexterity gains than VR alone: the between-group difference in change was 7.8 blocks immediately after treatment (95% CI, 7.1–7.9) and 10.8 blocks at 6 months (95% CI, 10.6–10.9). The combination improved the study’s targeted dexterity outcome, while strength, spasticity, disability, and activities of daily living improved similarly in both groups.

Design

Trial type: Multicenter, randomized, assessor-blinded, parallel-group clinical trial; participants were not blinded.
N: 48 randomized; 24 AO+VR and 24 CO+VR.
- AO+VR: 24
- CO+VR: 24
Randomization: 1:1 computer-generated allocation with concealed, randomly sized permuted blocks, stratified by baseline hand use (minimal versus discrete).
Setting: Two inpatient stroke-rehabilitation centers in Italy: Centro Cardinal Ferrari in Parma and Clinical Institute Quarenghi in Bergamo.
Enrollment: January 2022 to September 2024.
Mean follow-up: Outcome assessment at posttreatment (within 72 hours; 5 weeks after baseline) and 6 months after treatment.
Analysis: Intention-to-treat generalized estimating equations for repeated measures; no per-protocol analysis. Three control participants missed follow-up assessments and were handled by mean imputation.
Primary outcome: Paretic-hand BBT dexterity—the number of 25-mm wooden blocks transferred in 60 seconds. The study used 7 blocks as the smallest detectable difference, not as a formally established minimal clinically important difference.

Population

Inclusion Criteria

Adults with a clinical diagnosis of stroke 3 to 18 months after the acute event.
Primarily motor symptoms with unilateral upper-limb paresis.
Residual paretic-limb movement, including a Medical Research Council index of 3 and minimal to discrete active use.
Sufficient cooperation and cognitive understanding to participate.

Exclusion Criteria

Severe cognitive impairment (Mini-Mental State Examination score <20).
Severe unilateral spatial neglect (Bells Test cutoff ≥50%), severe ideomotor apraxia, or severe anosognosia.
Severe language-comprehension deficit, untreated psychiatric disorder, or sensory/uncorrected central visual impairment that hindered participation.
Drug-resistant epilepsy.

Baseline Characteristics (overall or representative arm)

The 48 participants were hospitalized patients with poststroke hemiplegia and residual upper-limb movement. Groups were reported as similar in age, sex, time from stroke, affected side, hand-use level, treatment duration, and baseline BBT, Motricity Index, modified Ashworth Scale, modified Rankin Scale, and modified Barthel Index scores.
Of 164 people assessed, 65 were ineligible, 33 declined, and 18 were excluded for other reasons. All randomized participants completed the assigned intervention; 24/24 AO+VR and 21/24 CO+VR participants completed both follow-up assessments.

Interventions

AO+VR: Twenty 45-minute sessions, four times weekly over 5 weeks, with tolerance for up to five missed sessions. Each session began with approximately 1.5 minutes of silent videos showing an actor performing unimanual or bimanual goal-directed actions from a lateral view, immediately followed by imitation in the VR environment. Patients used sensor-equipped physical objects and completed at least three repetitions of each action; 52 exercises increased in complexity over the program.
CO+VR: The same session frequency, duration, VR platform, therapist support, physical setting, and progressive unimanual/bimanual motor exercises, but preceded by approximately 1.5 minutes of neutral natural-scene video without motor content. This isolated the added action-observation/imitation component as far as possible.
Both groups: Continued the centers’ standardized multidisciplinary inpatient rehabilitation, including physiotherapy, occupational therapy, and task-oriented functional training, delivered independently of study allocation.

Outcomes

Primary Outcome (change in paretic-hand BBT score):

The paretic-hand time×group interaction favored AO+VR (GEE Wald χ²=3263.1; P<0.001). Mean baseline-to-posttreatment improvement was 8.0 blocks (SD 5.1) with AO+VR versus 2.6 blocks (SD 4.0) with CO+VR. The reported between-group difference in change was 7.8 blocks (95% CI, 7.1–7.9) at posttreatment and 10.8 blocks (95% CI, 10.6–10.9) at 6 months; both comparisons were reported as significant.
The AO+VR mean change exceeded the prespecified reliability benchmark of 7 blocks, whereas the CO+VR mean change did not. Because a formal stroke-specific minimal clinically important difference has not been established for the BBT, this is evidence of change beyond measurement error, not proof of a patient-important threshold.
Prespecified subgroup/moderator analysis: Randomization was stratified by hand-use level, but no significant treatment×baseline-motor-function interaction was found. An age×time-from-stroke interaction was exploratory rather than a prespecified efficacy subgroup: younger participants and those treated earlier, particularly before 120 days, tended to improve more (Wald χ²=16.1; P<0.001), without a reported treatment-specific effect estimate for each subgroup.

Secondary Outcomes:

Motricity Index improved over time in both groups (Wald χ²=73.5; P<0.001), without a significant treatment×time interaction.
Modified Ashworth Scale scores decreased over time in both groups (Wald χ²=35.45; P<0.001), without a treatment×time interaction.
Modified Rankin Scale scores improved over time (Wald χ²=74.5; P<0.001), without a group or treatment×time effect.
Modified Barthel Index scores improved over time (Wald χ²=23.6; P<0.001), again without a group or treatment×time effect.
BBT performance of the nonparetic hand improved over time in both groups (Wald χ²=21.1; P<0.001), with comparable gains. This was exploratory and may reflect bilateral engagement or general motor practice rather than a specific AO+VR effect.

Adverse Events / Safety:

No intervention-related adverse events were detected, and intervention compliance was reported as 100%.
Three CO+VR participants did not complete follow-up because of worsening clinical conditions, including accidental trauma and unrelated medical treatment; no AO+VR participant dropped out. The study was not powered to establish safety differences.

Figures

CONSORT flow diagram of the trial — *Figure 1.* Flow diagram of the clinical trial according to CONSORT guidelines. All randomized participants were included in the intention-to-treat analysis; missing posttreatment or follow-up data were handled by mean imputation.

Experimental setup and action-observation protocol — *Figure 2.* Experimental setup and study protocol for the action observation plus virtual reality and control observation plus virtual reality interventions. The figure shows the shared experimental setting, an example goal-directed action video followed by imitation, and the natural-scene control video.

Primary and secondary outcome changes over time — *Figure 3.* Mean change over time in the primary and secondary outcome measures: paretic- and nonparetic-hand Box and Block Test, Motricity Index, modified Ashworth Scale, modified Rankin Scale, and modified Barthel Index at baseline, posttreatment, and 6-month follow-up. Error bars represent mean standard error.

Age and time-from-stroke interaction — *Figure 4.* Exploratory interaction effect between age, time from stroke onset, and paretic-hand Box and Block Test improvement. Error bars represent mean standard error.

Criticisms

The sample was very small (48 randomized versus 94 planned), drawn from two experienced Italian inpatient rehabilitation centers; this limits precision, external validity, and reliable subgroup analysis.
Participants knew their allocation, and the experimental videos were more directly related to the intended motor task than the neutral nature videos. Expectancy, attention, novelty, and engagement effects could therefore contribute alongside action observation itself.
Three control follow-up losses were handled by mean imputation. Although the losses were attributed to unrelated clinical events, mean imputation can understate uncertainty and is weaker than modern multiple-imputation or mixed-model sensitivity analyses.
The primary outcome was a dexterity test closely aligned with the intervention. The authors did not include core upper-limb measures such as the Fugl-Meyer Assessment or Action Research Arm Test, and no formal stroke-specific BBT minimal clinically important difference exists.
The strikingly narrow reported confidence intervals around the between-group differences should be read alongside the small sample and the reporting choices; the paper does not establish that a 10.8-block advantage translates into a comparable improvement in everyday arm use.
Most secondary and functional outcomes improved similarly in both groups. The trial therefore supports a targeted dexterity signal, not superiority of AO+VR for overall disability, independence, strength, or spasticity.

Funding

The study was funded by an Italian Ministry of Health Starting Grant for young researchers to Dr Errante (SG-2019-12370506). The funder peer-reviewed the proposal but had no role in study design or manuscript writing. The trial was registered at ClinicalTrials.gov as NCT05163210. The authors reported no disclosures.

The paper

Authors. Errante et al.
Title. Action Observation Combined With Virtual Reality Promotes Motor Recovery After Stroke: A Randomized Controlled Trial.
Journal. Stroke.
Year. 2026.
DOI. 10.1161/STROKEAHA.125.054101
PMCID. PMC13117561

Deep Dive — click to expand

What this is

This is a 48-person inpatient rehabilitation trial asking whether the cognitive-motor step before practice matters: does watching a purposeful action and then imitating it add anything to the same VR exercises performed without action observation? AO+VR produced a larger paretic-hand BBT gain than VR alone—8.0 versus 2.6 blocks from baseline to treatment end—with the reported between-group advantage persisting at 6 months. The cleanest interpretation is a promising, durable dexterity signal in a selected subacute-to-chronic rehabilitation population, not proof that VR-plus-observation improves global recovery.

1. Shadow Audit

The headline is about better dexterity, but the broader rehabilitation story is deliberately less dramatic. Motricity Index, spasticity, modified Rankin Scale, and modified Barthel Index all improved over time without a treatment×time advantage. The nonparetic hand also improved similarly in both groups. Thus the incremental effect appears concentrated in the test most tightly coupled to the intervention’s target—rapid block transfer with the paretic hand—rather than in strength, tone, disability, or activities of daily living. The study also planned 94 participants but randomized only 48, and it used mean imputation for three control follow-up losses. Those facts matter more than the mechanistic language about stable plastic change.

2. Inversion Engine

To invert the conclusion, the added AO component would have to lose its advantage once the trial is repeated at the planned scale, with a less expectancy-sensitive control and outcomes that generalize beyond BBT. Numerically, the immediate mean change difference was roughly 5.4 blocks (8.0 versus 2.6), while the study’s reliability benchmark was 7 blocks. A replication showing an AO+VR-minus-control difference below about 7 blocks—or a confidence interval crossing zero on a prespecified functional outcome—would remove the claim of reliable individual-level dexterity change. The present design cannot tell whether the observed separation is action observation, greater attention to a motor-relevant video, therapist enthusiasm, or a combination.

3. Second-Order Catalyst

If the result is real, the first change should occur in inpatient and outpatient neurorehabilitation programs that already own a VR platform and can produce short, silent, task-specific videos. The protocol is operationally light: add roughly 1.5 minutes of lateral-view purposeful action observation, immediately followed by imitation, before the existing 45-minute session, four days per week for five weeks. The subgroup most likely to move first is the patient with residual movement, sufficient comprehension, and a treatment start near the subacute window—especially before 120 days—although that timing signal was exploratory. A service should pilot the workflow with blinded BBT, Fugl-Meyer or Action Research Arm Test, and real-world arm-use measures rather than assume the BBT result transfers automatically.

4. Asymmetric Leverage

The leverage is not a large effect across every outcome; it is a modest additional pre-practice ingredient with a potentially large denominator. A one-to-two-minute observation phase could be added to many existing therapy sessions at low equipment cost, while a small dexterity gain may matter disproportionately for tasks requiring grasp, release, and manipulation. But the denominator is currently small: 24 people per arm, two highly experienced centers, and a selected population that excluded severe cognitive, language, neglect, apraxia, sensory, and motor impairments. The asymmetric upside is therefore a cheap, scalable adjunct; the asymmetric downside is institutionalizing an attractive ritual that has not yet shown superiority on participation or daily-life outcomes.

5. Paradigm Destroyer

The paper challenges the reflex that rehabilitation technology is only a delivery channel for repetitive motor practice. It suggests that the representational cue immediately before execution may be an active ingredient, at least for paretic-hand dexterity. The protocol update for tomorrow morning would be: “Before selected VR upper-limb tasks, show a brief goal-directed action from the patient’s viewpoint or a lateral model view, then ask for immediate imitation; track dexterity separately from global disability.” It should not be: “Replace conventional rehabilitation with VR,” because both groups received standard rehabilitation and only the targeted BBT outcome separated.

MVP — Minimum Viable Proof

The minimum bedside-changing statement is: in a larger, independently replicated randomized trial using an attention-matched control, adding brief goal-directed action observation before the same VR exercises improves paretic-hand dexterity by at least 7 BBT blocks over control and also improves a validated functional upper-limb measure or observed daily-life arm use, without unacceptable burden or harms.

Best Combination

Combine this finding with the established principles of task-specific, intensive, salient, and progressively challenging practice rather than treating action observation as a standalone therapy. Use the AO+VR sequence as a front end to meaningful unilateral and bimanual tasks, preserve conventional physiotherapy and occupational therapy, and measure both impairment-level dexterity and participation-level outcomes. Meta-analytic evidence supporting action observation and VR can establish plausibility; this trial adds a direct comparison of AO+VR against VR alone and a 6-month signal, but its small sample and narrow outcome profile make it better viewed as a component-specific trial than as a definitive rehabilitation strategy.

Overvalue Warning

Do not equate the 7.8- and 10.8-block reported between-group differences with a proven patient-important improvement: the BBT has no formally established stroke-specific minimal clinically important difference, and the study’s 7-block threshold was a smallest detectable difference.
Do not generalize the exploratory finding that younger, earlier-treated participants improved more into a validated treatment-response rule. The analysis was small, exploratory, and did not supply a treatment-specific effect estimate for each age or timing subgroup.

Educational use notice. The Clotbust Review is dedicated to medical education. Figures from published articles are embedded for teaching purposes, with attribution to the original publication. We do not claim copyright over any embedded figure; all rights remain with the original publisher and authors under the article's published license. If you are a rights holder and would like a figure removed, please contact the editorial team. See our editorial notice for the full policy.