Immersive VR helps healthy adults learn rare words faster than a tablet method, but does not outperform a structured tablet method for aphasia anomia rehabilitation
How this was rated
Two-experiment design with within-subject crossover in the clinical arm (n=16 PWA) - a strong design that controls for between-patient variability in aphasia recovery. Pre-specified GLMM analysis with subject and item as random effects. Matched word exposures (7 per word per method) make the iVR-vs-DSL comparison rigorous. The 16-patient sample limits power to detect small effects, and the chronic-stroke window (3-96 months post-onset) is broad. The neurotypical Experiment 1 (n=32) provides clear external validation that iVR has a real word-learning effect in unimpaired adults; the null in aphasia suggests the mechanism does not straightforwardly extend to anomia rehabilitation under matched-exposure conditions. Open Access CC BY 4.0; ethics-approved by University of Geneva CUREG and Bordeaux research ethics committees; pre-registered data deposit at Yareta. Funded by the Swiss NSF NCCR Evolving Language.
Ratings use a simplified four-tier scheme (High, Moderate, Low, Very Low) informed by the GRADE working group. Learn more about how studies are rated.
Two within-subject experiments using a virtual marketplace iVR application against a matched-exposure tablet (digital static learning, DSL) method. In 32 neurotypical French adults learning rare French words, iVR significantly outperformed DSL by Day 3 (z = 4.556, p < .0001). In 16 people with mild-to-moderate post-stroke aphasia learning frequent French words in a crossover design, both methods produced significant learning gains across Day 1, 5, 12, and 19 (p < .001), but iVR was NOT significantly better than DSL on accuracy (estimate 0.025, p = .704).
An open-access two-experiment study (Exp 1: 32 neurotypical adults; Exp 2: 16 people with chronic mild-to-moderate aphasia, 3-96 months post-stroke) testing whether immersive VR adds value over a controlled tablet-based anomia treatment with matched word exposures. In healthy adults learning unfamiliar words, iVR was clearly better. In aphasia, both methods produced robust learning effects across the 19-day protocol - but iVR was NOT significantly superior to a well-controlled tablet method. The result is non-inferiority, not superiority. Clinicians considering iVR for chronic anomia should not expect a treatment-effect bump from immersion alone if the tablet method is structured and controls exposures; iVR's value in this clinical context may come from acceptability, generalization, or ecological transfer rather than accuracy at the trained items.
Key findings
- Experiment 1 (32 neurotypical native French adults learning rare French words): iVR significantly outperformed digital static learning (DSL) by Day 3 (z = 4.556, p < .0001), with the iVR advantage emerging after the first learning session
- Experiment 2 (16 people with mild-to-moderate post-stroke aphasia, mean age 65, 3-96 months post-onset): both iVR and DSL produced significant accuracy gains across Day 1, 5, 12, and 19 (main effect of day: chi-squared(2) = 310.80, p < .001) - learning effects were robust
- Mean naming accuracy in PWA: Day 1 = 43.35%, Day 5 = 58.8% (+14%), Day 12 = 64.56% (+21%), Day 19 = 68.92% (+25%) - both methods drove substantial improvement
- iVR did NOT significantly outperform DSL in the aphasia arm (Method[iVR] estimate = 0.025, SE = 0.065, p = .704). The crossover design controls for between-patient variability in aphasia recovery, so the null is not explained by sample heterogeneity
- Both methods controlled the number of word exposures (exactly 7 per word per method), included written form, required oral production, and provided speech-recognition feedback - meaning the comparison isolates immersion + ecological context as the iVR-specific contribution
- iVR's mechanism is hypothesized to operate through sense of presence, sense of agency, embodiment, and ecologically rich semantic context (virtual marketplace with avatar interactions and object manipulation) - factors that demonstrably aided learning in the neurotypical arm but did not translate to a measurable accuracy advantage in chronic aphasia
- The clinical sample had spared semantic system per BECLA aphasia battery - so the null finding cannot be attributed to semantic processing impairment masking an iVR advantage
Background
Anomia - the difficulty in finding or producing words - is one of the most common and persisting symptoms in post-stroke aphasia. Despite decades of treatment research, responses to anomia treatment vary substantially between patients and few patients achieve full recovery. The authors propose that immersive virtual reality (iVR) might be a relevant new approach because (a) word learning is known to be highly contextual and multimodal, (b) the lexical-semantic network is sensitive to rich, embodied context, and (c) iVR has been shown to outperform less-immersive methods for second-language vocabulary learning in healthy participants. However, no prior published study had tested immersive VR (as opposed to semi-immersive or non-immersive variants) for word relearning in aphasia, and the earlier iVR-vs-control studies that exist in the broader rehabilitation literature have generally compared iVR to traditional face-to-face therapy - a comparison that confounds immersion with method altogether.
This study set out to test a tighter question: does iVR add value over a well-controlled digital comparison method with matched exposures?
What the researchers did
The authors built two parallel applications:
-
iVR application: a virtual marketplace where the participant interacts with avatars and manipulates objects in a fully immersive head-mounted display environment. The participant is offered each target object, asked whether they know its name, and either prompted to produce the name (with speech-recognition feedback) or shown the name in auditory + written form.
-
Digital Static Learning (DSL) method: a tablet-based application that presented the same words, same number of exposures (7 per word), same written form, required the same oral production, included the same speech-recognition feedback - but without the immersion, the marketplace context, the avatar interactions, or the object manipulation.
The two methods were rigorously matched on the number of word exposures, the written form display, the requirement to produce the word aloud, and the feedback mechanism. The only systematically varying factors were the iVR-specific features: immersion, ecological communicative context, avatar interaction, and object manipulation.
Experiment 1 (neurotypical adults, n=32)
Thirty-two neurotypical native French-speaking adults learned a list of rare French words across four days. Each participant used both methods (iVR and DSL) on different word lists, in a counterbalanced within-subject design. Picture-naming accuracy was measured at Day 1, 3, and 5. Analysis used GLMM with subject and item as random effects.
Experiment 2 (people with aphasia, n=16)
Sixteen people with mild-to-moderate aphasia (12 male, 4 female; mean age 65, range months-post-onset 3-96) were recruited from the University Hospital of Bordeaux. All were right-handed native French speakers who had had a left-hemisphere stroke, with word-finding difficulty but spared semantic system per the BECLA aphasia assessment battery. Patients used both methods (iVR and DSL) on different lists of common/frequent French words from four semantic categories (musical instruments, fruits, vegetables, tools), in a crossover design across participants. Each treatment list received 3 therapy sessions per method. Picture-naming accuracy and error type were measured at Day 1, Day 5, Day 12, and Day 19.
Statistical analysis used a generalized linear mixed-effects model (GLMM) on accuracy, with method (iVR / DSL) and test day as fixed effects and subject and stimulus as random effects.
What they found
Experiment 1 (neurotypical)
iVR significantly outperformed DSL on word learning by Day 3 (DSL-iVR contrast: z = 4.556, p < .0001), with the iVR advantage emerging after just the first learning session and persisting through Day 5. Both methods produced learning, but iVR was clearly superior in neurotypical adults learning unfamiliar words in their native language.
Experiment 2 (aphasia)
Both iVR and DSL produced robust learning effects in the aphasia group. Mean picture-naming accuracy increased from 43.35% on Day 1 to 58.8% on Day 5 (+14%), 64.56% on Day 12 (+21%), and 68.92% on Day 19 (+25%). The main effect of day was highly significant (chi-squared(2) = 310.80, p < .001).
However, iVR did NOT significantly outperform DSL in the aphasia arm. The GLMM estimated the method effect at 0.025 (SE 0.065, p = .704) - a null finding with a tight confidence interval. Both methods drove substantial gains; neither was better than the other on accuracy.
The error-type analysis (lexical errors, phonological errors, no-responses) also showed no significant method differences.
Why this matters
For clinicians considering iVR for chronic anomia: this study provides the strongest controlled evidence yet that the value of iVR over a well-designed digital tablet method - when word exposures are matched - is not in raw naming accuracy at the trained items. Both methods produce real learning. iVR’s clinical value is more likely to come from elsewhere: acceptability and engagement over longer treatment courses, generalization to untrained items, and ecological transfer to real-world communication situations.
The contrast with Experiment 1 is informative. In healthy adults learning unfamiliar words, iVR is clearly superior - confirming the immersion mechanism is real. The challenge is identifying which aphasia subgroups, treatment intensities, and outcome domains iVR actually benefits in clinical populations.
Limitations
The authors explicitly flag the following:
- Small clinical sample (n=16). The aphasia arm has limited power to detect small or moderate-size effects. A non-significant difference of 0.025 is consistent with iVR being neither better nor worse than DSL; it does not rule out a small advantage that would emerge with a larger sample.
- Within-trained-item outcomes only. The primary outcome was accuracy on trained items. The study did not measure generalization to untrained items, untrained semantic categories, or real-world communication contexts - the domains where iVR’s ecological-context affordances would be most expected to matter.
- Matched-exposure design isolates immersion. This is a strength for hypothesis-testing but means the study does not test the real-world question of whether iVR enables MORE total exposures per unit clinician time (which could be a clinical advantage even at equal per-exposure efficacy).
- Custom iVR system. The application was a custom-built virtual marketplace at Fondation Campus Biotech Geneva. Generalization to other iVR systems for aphasia depends on whether they share the relevant affordances (immersion, ecological context, avatar interaction, object manipulation).
- Chronic stroke window 3-96 months post-onset. This is a broad range; iVR responsiveness may differ across acute, subacute, and chronic phases.
- Native French speakers only. Cross-linguistic generalization not tested.
- Both methods used speech recognition feedback. The IBM Watson API-based speech recognition algorithm is a non-trivial component of both methods; its reliability across patients with anomic speech errors may influence the overall pattern of results.
- Acceptability, engagement, and motivation not formally measured. These are the dimensions on which iVR is theoretically expected to add value over longer treatment courses; the study’s design does not address them.
Implications for practice
For chronic anomia rehabilitation, the choice between iVR and a structured tablet method should not be made on the assumption that immersion alone improves accuracy. Both methods produce real learning when word exposures are controlled. iVR's clinical value in this population is more likely to come from (a) acceptability and engagement over longer treatment courses, (b) generalization to untrained items and untrained contexts (not tested here), and (c) ecological transfer to real-world communication situations. Clinicians using iVR for aphasia should design protocols that test these specific affordances rather than treating immersion as a magic ingredient. The result also highlights that well-designed tablet-based digital therapy can produce strong gains - useful framing for clinicians who lack iVR access. Note: this study used a custom iVR marketplace, not Therapy withVR; transfer of conclusions to other iVR systems requires that the comparison method is similarly well-controlled.
Implications for research
Future iVR-vs-control studies in aphasia should (a) include untrained-item generalization tests to detect potential transfer-of-learning advantages that within-trained-item accuracy may miss, (b) include real-world communication transfer measures (e.g., naming in conversation, written generation), (c) report acceptability and engagement metrics over longer treatment courses where iVR's motivational advantages may compound, (d) explore subgroup effects (severity, time post-onset, lesion location) that may modulate iVR responsiveness, and (e) move beyond the matched-exposure framing to test whether iVR enables MORE total exposures per unit clinician time, which could be a real-world advantage even at equal per-exposure efficacy. The clear iVR advantage in Experiment 1 (neurotypical word learning) confirms the immersion mechanism is real - the challenge is identifying which aphasia subgroups, treatment intensities, and outcome domains it actually benefits.
Cite this study
If you reference this study in your work, the canonical citation formats are:
@article{franco2025,
author = {Franco, J. and Glize, B. and Laganaro, M.},
title = {Impact of immersive virtual reality compared to a digital static approach in word (re)learning in post-stroke aphasia and neurotypical adults: Lexical-semantic effects?},
journal = {Neuropsychologia},
year = {2025},
doi = {10.1016/j.neuropsychologia.2025.109069},
url = {https://withvr.app/evidence/studies/franco-2025}
}TY - JOUR
AU - Franco, J.
AU - Glize, B.
AU - Laganaro, M.
TI - Impact of immersive virtual reality compared to a digital static approach in word (re)learning in post-stroke aphasia and neurotypical adults: Lexical-semantic effects?
JO - Neuropsychologia
PY - 2025
DO - 10.1016/j.neuropsychologia.2025.109069
UR - https://withvr.app/evidence/studies/franco-2025
ER - Know of research that should be in this hub? If a relevant peer-reviewed study is not listed here, send the reference to hello@withvr.app. The hub is kept up to date as the literature grows.
Funding & independence
Funded by the Swiss National Science Foundation through the NCCR Evolving Language consortium (Agreement #51NF40_180888). Open Access publication (CC BY 4.0) at Neuropsychologia. The authors declare no conflict of interest. No withVR BV involvement in funding, study design, or authorship. Summary prepared independently by withVR using the published peer-reviewed paper. The iVR application used was a custom marketplace built by the Human Neuroscience Platform at Fondation Campus Biotech Geneva - NOT Therapy withVR or Research withVR. Comparisons to specific commercial iVR systems should not be inferred from this study.