Cortical neuromodulation by vibration-induced illusion of movement is observed in older individuals, but not in individuals with acute stroke: a cross-sectional fNIRS study Authors: Brieuc Léger, Pascal Auzou, Élodie Fourdrinoy, Mathilde Sarrazin, Sylvine Celot, Céline Gay, Barbara de Dieuleveult, Clara Cohen, Canan Özsancak, Stéphane Perrey Published: 2026 ABSTRACT AbstractBackgroundTogether with the aging of the world population, stroke incidence is expected to increase in the coming years. Stroke is frequently associated with sensorimotor impairments of the upper limb, including proprioceptive deficits, that substantially impair motor control and quality of life. Understanding how physiological aging and acute stroke affect proprioceptive processing is therefore crucial for developing targeted rehabilitation strategies.MethodsVibration-induced illusion of movement was administered to 28 young, 30 older, and 23 individuals with acute stroke to the distal tendon of the triceps brachii. In healthy participants, vibration was applied to both arms, whereas in participants with acute stroke, it was applied to the paretic left arm. Vibration-induced illusion of movement was delivered either with or without visual feedback of the vibrated arm. Each condition consisted of six 15-s vibrations at 80 Hz. The illusion was quantified using the Standardized Kinesthetic Illusion Procedure scale, while cerebral hemodynamic activity within frontoparietal and sensorimotor areas was assessed using fNIRS through changes in (de)oxyhemoglobin concentrations. Subsequently, activation profiles were established, and group cerebral hemodynamics and functional neural efficiency were compared.ResultsOlder adults exhibited more extensive bilateral activation in the frontoparietal and sensorimotor cortices compared with young adults, suggesting age-related changes in neural recruitment during proprioceptive processing. In contrast, participants with acute stroke showed no significant task-related cortical activation and displayed an overall reduction in cerebral activity compared with both healthy groups. Despite this reduced activation, neural efficiency in participants with acute stroke was comparable to that observed in older participants.ConclusionTogether, these findings suggest that preserved compensatory mechanisms may still be engaged early after stroke, and suggest potential implications of vibration-induced illusion of movement to limit age-related proprioceptive decline and to support early rehabilitation after stroke.Clinical Trial registration: NCT06218563—2024-01-12.Graphical Abstract Supplementary InformationThe online version contains supplementary material available at https://doi.org/10.1186/s12984-026-02129-w.Keywords: Vibration, Illusion, Proprioception, fNIRS, Acute stroke, Aging Introduction As the global population continues to age [1], predictive models indicate that approximately one in four individuals over the age of 25 will experience a stroke during their lifetime [2]. Additionally, the incidence of ischemic strokes is projected to rise among both men and women of all ages during the current decade [3]. Furthermore, 77.4% of the affected individuals exhibit upper limb motor deficits [4], with approximately 50% retaining moderate to severe motor deficits after a period of four years [5]. In addition, proprioceptive deficits of the upper limbs have been reported in up to 63% of individuals with a stroke [6]. Together, these persistent sensorimotor impairments substantially contribute to post-stroke disability and reduced quality of life. Indeed, 67% of individuals report the non-use or disuse of their affected arm as a major problem [5]. In this context, tendinous vibration emerges as a promising proprioceptive stimulation approach with neuromodulatory effects, capable of preventing alterations of the cerebral sensorimotor network after limb immobilization [7–9] and promoting motor function recovery [10]. Indeed, tendinous vibration has been demonstrated to elicit a stimulatory effect on the intrafusal group 1a fibers of the muscle spindle [11]. Subsequently, a muscle stretch signal is transmitted to the sensorimotor and frontoparietal cortical regions, which have been identified as the mediators of both voluntary and illusory movements, the sensation of movement, and the urge to move [12–16]. This mechanism has been demonstrated to create an illusion of movement when the tendon is vibrated, at an optimal frequency of 70–80 Hz [13, 17, 18]. However, the vividness of the illusion can be mitigated when the vibrated limb is visible to the subject, as vision provides more reliable information and is deemed to dominate proprioception [19, 20]. This phenomenon is of considerable interest, as evidenced by the findings of numerous studies suggesting that the vibration-induced illusion of movement (VIM) elicits greater and more extensive activations predominantly in the contralateral sensorimotor and frontoparietal regions when compared to the effects of vibration without perceived illusion [15, 21, 22]. Furthermore, the velocity and intensity of the illusion have been found to be directly related to the degree of activation [15, 21, 23, 24]. These activations exhibit a right frontoparietal dominance, with motion sensation suggested to originate in the right inferior parietal lobule [12, 23, 25]. However, it was shown that the extent to which the visual stimulus attenuated the illusion can be directly proportional to the degree of activation observed in the posterior somatosensory associative area [13]. Moreover, VIM has the potential to improve the immediate performance of a reaching task in individuals with chronic stroke, suggesting a possible modification of brain activity [26]. A case study further indicates that a lesion in motor regions alters the illusion of movement up to three weeks post-stroke [27], whereas after six months the illusion experienced by the person with a stroke becomes comparable to that of healthy individuals. Although these findings suggest that the perception and neural processing of VIM may vary across post-stroke stages, the neural correlates of VIM during the acute stage remain unknown. Determining whether VIM-related sensorimotor and frontoparietal activations are preserved or altered in individuals with acute stroke is therefore essential to evaluate the potential integration of VIM into neuromodulation strategies aimed at promoting upper-limb motor recovery after stroke [28, 29]. To investigate the neural correlates of VIM, functional near-infrared spectroscopy (fNIRS) appears to be a suitable method in neuromodulation [30]. By emitting near-infrared light (760–850 nm), fNIRS enables indirect quantification of localized brain activity by measuring cortical hemodynamic changes in the brain in a similar fashion to functional magnetic resonance imaging (fMRI) [31, 32]. Concentration variations of oxyhemoglobin (HbO) and deoxyhemoglobin (HbR) can be determined a posteriori by utilizing the modified Beer-Lambert Law [33]. Indeed, fNIRS measures cortical brain activity with greater temporal resolution than fMRI [34], and higher spatial resolution than electroencephalography (EEG) [35]. Furthermore, the affordability, portability, and practicality of the tool serve to reinforce its status as a preferred instrument. Consequently, fNIRS facilitates the execution of experimental conditions in a more ecologically relevant environment (i.e., sitting/standing positions) and provides a superior framework for sensorimotor experiments [36]. Therefore, the objective of this study was to investigate how visual feedback modulates the neural correlates of VIM in healthy young individuals, in the context of physiological aging, and after acute stroke using fNIRS. To this end, a series of vibrations were applied to the distal tendon of the left and right triceps brachii, with and without visual feedback of the vibrated arm. Individuals with acute stroke were exposed exclusively to vibrations applied to the (left) paretic arm. We hypothesized that, compared with visual feedback, vibrations without visual feedback would induce heightened activation of the contralateral sensorimotor and frontoparietal networks, consistent with the perception of the illusion. Conversely, visual feedback of the immobile limb was expected to attenuate the illusion and reduce activation within the sensorimotor and frontoparietal networks while promoting activity in contralateral somatosensory regions, likely due to visuo-proprioceptive sensory conflict. Finally, compared with healthy individuals, we hypothesized that individuals with acute stroke would exhibit altered and reduced activation of the contralateral sensorimotor and frontoparietal networks because of impaired sensorimotor processing following stroke. Materials and methods Study characteristics The present cross-sectional study was reported in accordance with the STROBE guidelines [37]. The behavioral data from this study have been previously analyzed in [38]. Healthy young and older adults, as well as individuals with acute stroke, were recruited over the course of two time periods: from January 2024 to April 2024, and from February 2024 to June 2025, respectively. All experiments were conducted in the Neurology and Vascular unit of the University Hospital Center of Orléans (France). The Institutional Review Board (CPP Sud-Méditerranée 1) approved the study protocol on June 14 2023, and it was subsequently registered under the number 2023-A01062-43 (Jardé law, RIPH2). The study was prospectively registered on ClinicalTrials.gov, under the identifier NCT06218563. All participants provided free and informed written consent in accordance with the Declaration of Helsinki. Participants Healthy individuals were recruited through a flyer distribution campaign and word of mouth. These individuals received financial compensation of 50€ for participating in the study. Participants with acute stroke were recruited after being admitted to the unit. Participants with acute stroke had to have experienced an ischemic or hemorrhagic stroke in the right hemisphere for the first time, and therapists had to note a left motor deficit in the upper limb. Additionally, therapists were required to confirm the absence of significant cognitive impairment and aphasia, though this requirement was not quantified. Participants were evaluated using the NIHSS [39] upon admission to the unit and within 24 h of the experiment. During this period, the following tests were used to assess motor ability and spatial neglect: the Fugl-Meyer Assessment for the Upper Extremity (FMA-UE, [40]), the Bells Test [41] and the Line Bisection Test [42]. All participants had to be 18–85 years old and were screened for right-handedness (score > 40) using the Edinburgh Handedness Inventory [43]. In addition, participants underwent the Thumb Localizing Test (see [44]). As in [44], the individual median score was utilized to compute group characteristics (see the Section “Demographic data”). Procedures The participants were seated in a quiet, dedicated room, in a manner that ensured comfort. The subject’s left or right arm was then gently placed in a movable support strap, which exposed the distal triceps brachii tendon (see Fig. 1). This configuration was selected to prevent any disruption to the sensation of movement [45]. The Vibramoov PHYSIO vibrator device (Techno Concept, Mane, France) was placed by flexing the participant's arm and palpating the apex of the elbow to use it as a reference point. Then, the device was placed a few centimeters above this point, on the distal tendon of the triceps brachii. Participants were instructed to relax to minimize the tonic vibration response and promote the VIM [46]. Participants were also instructed to focus on the sensations experienced in the elbow joint. This approach has the potential to not only enable perception of vibrations but also facilitate discernment of illusory arm movement. Participants received a point for reporting arm flexion, the expected movement, but they were not informed of the expected direction in order to avoid introducing bias. In the visual feedback condition, participants were instructed to keep their eyes on their hand. Tendinous vibrations were applied at 80 Hz for 15 s, with an amplitude ranging from 1 to 2 mm [17, 46]. Concurrently, brain activity was measured using the fNIRS method. The experimenter positioned the fNIRS cap according to the international 10–20 EEG system, meticulously adjusting the hair to promote contact between the sources and detectors and ensuring optimal signal quality [47]. A Picture of the configuration in the quiet, dedicated room. A young participant is seated comfortably in the chair. The movable green strap supports the left arm; a vibrator is placed on the distal tendon of the triceps brachii muscle. Brain activity was simultaneously collected using functional near-infrared spectroscopy. B–E Schematic representation of the four modes of vibration. The vibrator is represented by the blue rectangle and the blue lines show the vibration. B–C Left arm vibration, performed in healthy young and older participants, as well as in participants with acute stroke (paretic arm). D–E Right arm vibration, performed only in healthy young and older participants. B–D Vibration with visual feedback of the vibrated arm, as indicated by open eyes and red dotted arrows pointing from the head to the hand. C–E Vibration without visual feedback of the vibrated arm, as indicated by the closed eyes In healthy participants, vibrations were applied to either the left or right arm, with or without direct visual feedback of the vibrated arm. This resulted in four distinct conditions (Fig. 1): vibration with or without vision and vibration applied to either the left or right arm. These conditions were selected based on prior evidence suggesting that visual feedback can attenuate the VIM [19, 20]. Each experimental condition (Fig. 2) consisted of six vibrations, performed in two blocks of three vibrations, for a total of 24 vibrations (2 × 3 vibrations × vision × arm). Within each group, the order of the conditions was counterbalanced. Given that participants with acute stroke may experience difficulties in maintaining concentration during prolonged tasks and may become fatigued easily, the experiments were conducted exclusively on the left paretic arm. This resulted in a total number of 12 vibrations (2 × 3 vibrations × vision). To account for the delay in the hemodynamic response to return to baseline, as well as muscle conditioning that could influence the sensation of movement and participant anticipation, a resting period varying between 25 and 35 s was selected to prevent periodic physiological noise from synchronizing with the task [46, 48, 49]. Representation of the experimental protocol. A one-minute resting period was initiated at the beginning and end of the acquisition period to ensure that the fNIRS data were not contaminated by the bandpass filter employed during signal processing. Healthy participants were administered right and left arm vibration in a counterbalanced order. Participants with acute stroke were administered vibration to their left paretic arm only. Each condition (A, B, C or D) comprised three trials of 15-s vibration, repeated twice, resulting in six vibrations in total. All vibrations were preceded by a resting period of 25–35 s. The Standardized Kinesthetic Illusion Procedure (SKIP) illusion scores were collected following each condition, after which a 30-s rest period (inter cond.) was established To quantify VIM responses, the Standardized Kinesthetic Illusion Procedure (SKIP) scale was employed [50]. This scale uses a four-point ordinal rating system to evaluate perceived movement at the elbow joint (0 = no illusion perceived, 1 = vague and imprecise, 2 = moderately clear and precise, 3 = perfectly clear and precise). Participants were given an additional point if they perceived movement corresponding to the expected arm flexion, yielding a possible total score of 4. When experiencing difficulty describing the direction of the perceived movement, participants were asked to reproduce it using their non-vibrated arm. SKIP scores were collected following each block of three vibrations, yielding two scores per condition. The mean of these two scores was used for subsequent analysis. The present study exclusively employed the scoring component of the SKIP procedure, which has been shown to differentiate between VIM responses in healthy individuals and in chronic stroke [50]. fNIRS recordings Two portable continuous-wave fNIRS systems (Brite MKII and Brite MKIII, Artinis Medical Systems, The Netherlands) were utilized to record cerebral oxy- and deoxyhemoglobin levels in the frontoparietal and sensorimotor areas. All recordings were conducted at a frequency of 25 Hz (one sample every 0.04 s) using two wavelengths of 760 and 842 nm. Each system comprised 10 LED sources and 8 light receivers, resulting in a montage with 20 LED sources and 16 light receivers. The layout was built with 46 channels (source-receiver pairs), including 42 long channels of 3 cm and 4 short channels (SC) of 1 cm (Fig. 3). Arrangement of the 46 fNIRS channels and sensitivity maps generated with the Monte Carlo photon transport software for the depicted channel arrangement. The color bar represents the sensitivity on a logarithmic scale from 0.01 to 1 mm−1. A Dorsal view of the fNIRS sensitivity profile, showing the 46 channel numbers. Short channels are displayed in bold and are marked with red boxes (channels 2, 15, 27 and 40). Left (B) and right (C) lateral views of the fNIRS sensitivity profile To ensure the correct positioning of the optode layout across frontoparietal and sensorimotor areas, the sources and receivers were localized in the Montreal Neurological Institute (MNI) space using a 3D digitizer (FASTRAK, Polhemus, USA). This allowed the location of each channel to be extracted onto Brodmann areas (MRIcro) using NIRS-SPM [51]. Coregistration was performed on a single representative participant prior to data acquisition. Due to slight interhemispheric inconsistencies introduced by small head movements during digitization, and despite the strictly symmetrical optode layout, the spatial coordinates of the left-hemisphere channels were mirrored to the right hemisphere to enforce geometric symmetry of the montage. Furthermore, a sensitivity profile (Fig. 3) was generated using a Monte Carlo simulation involving 10 million photons through the tMCimg software in AtlasViewer (MatLab R2017b, The MathWorks Inc., USA) to estimate the probabilistic path of photons through the cortical surface [52]. The combined sensitivity profile and Brodmann area projections confirmed that the montage covered frontoparietal and sensorimotor areas (Fig. 3, Table 1).The Brodmann areas were grouped in six regions of interest as follows: Frontal (BAs 44, 45), dorsolateral prefrontal (dlPFC, BAs 9, 46), motor (BAs 4 and 6), primary somatosensory (S1, BAs 1, 2, 3), somatosensory association (S2, BAs 5, 7), and inferior parietal areas (BAs 39 and 40). The BAs reported in Table 1 correspond to those that exhibit the largest overlap and are all part of the same region. Table 1Spatial localization of the 46 channels obtained with NIRS-SPM from the 3D digitizerChannelsSourcesReceiversMNI coordinatesBrodmann areas% Of overlapXYZ111− 70.3− 67.6245.5339—Angular gyrus, part of Wernicke's area40—Supramarginal gyrus part of Wernicke's area0.850.152*21− 75.09− 65.6529.9639—Angular gyrus, part of Wernicke's area40—Supramarginal gyrus part of Wernicke's area0.510.1351− 76.62− 50.1245.4740—Supramarginal gyrus part of Wernicke's area0.96412− 53.98− 70.4768.6739—Angular gyrus, part of Wernicke's area40—Supramarginal gyrus part of Wernicke's area0.470.30532− 37.2− 66.1183.747—Somatosensory Association Cortex0.98642− 46.45− 56.7182.347—Somatosensory Association Cortex0.64733− 22.38− 50.294.65—Somatosensory Association Cortex7—Somatosensory Association Cortex0.470.1843− 32.33− 37.5392.641—Primary Somatosensory Cortex3—Primary Somatosensory Cortex0.300.279103− 20.09− 14.8794.634—Primary Motor Cortex6—Pre-Motor and Supplementary Motor Cortex0.260.741014− 64.6− 57.6864.3839—Angular gyrus, part of Wernicke's area40—Supramarginal gyrus part of Wernicke's area0.100.901144− 54.05− 44.2279.551—Primary Somatosensory Cortex2—Primary Somatosensory Cortex3—Primary Somatosensory Cortex0.230.350.191254− 70.17− 38.1562.8540—Supramarginal gyrus part of Wernicke's area0.861394− 62.92− 25.9771.721—Primary Somatosensory Cortex2—Primary Somatosensory Cortex3—Primary Somatosensory Cortex0.380.030.541465− 53.1746.4639.645—pars triangularis Broca's area0.5915*75− 42.8954.4744.789—Dorsolateral prefrontal cortex46—Dorsolateral prefrontal cortex0.370.601666− 65.342.4416.245—pars triangularis Broca's area11767− 59.330.4346.8744—pars opercularis, part of Broca's area45—pars triangularis Broca's area0.610.191887− 48.8621.0366.779—Dorsolateral prefrontal cortex0.591997− 57.7510.2163.156—Pre-Motor and Supplementary Motor Cortex0.822048− 44.02− 30.7586.824—Primary Motor Cortex0.632188− 40.691.6781.286—Pre-Motor and Supplementary Motor Cortex12298− 49.44− 11.7379.784—Primary Motor Cortex6—Pre-Motor and Supplementary Motor Cortex0.560.4223108− 30.81− 8.4889.456—Pre-Motor and Supplementary Motor Cortex12411937.2− 66.1183.747—Somatosensory Association Cortex12513953.98− 70.4768.6739—Angular gyrus, part of Wernicke's area40—Supramarginal gyrus part of Wernicke's area0.350.292614946.45− 56.7182.347—Somatosensory Association Cortex0.7327*121075.09− 65.6529.9639—Angular gyrus, part of Wernicke's area40—Supramarginal gyrus part of Wernicke's area0.480.0628131070.3− 67.6245.5339—Angular gyrus, part of Wernicke's area40—Supramarginal gyrus part of Wernicke's area0.820.1729151076.62− 50.1245.4739—Angular gyrus, part of Wernicke's area40—Supramarginal gyrus part of Wernicke's area0.010.9830111122.38− 50.294.65—Somatosensory Association Cortex7—Somatosensory Association Cortex0.510.0731141132.33− 37.5392.641—Primary Somatosensory Cortex3—Primary Somatosensory Cortex0.270.2632191120.09− 14.8794.634—Primary Motor Cortex6—Pre-Motor and Supplementary Motor Cortex0.300.7033131264.6− 57.6864.3839—Angular gyrus, part of Wernicke's area40—Supramarginal gyrus part of Wernicke's area0.130.8734141254.05− 44.2279.551—Primary Somatosensory Cortex2—Primary Somatosensory Cortex3—Primary Somatosensory Cortex0.200.310.2835151270.17− 38.1562.8540—Supramarginal gyrus part of Wernicke's area0.7736201262.92− 25.9771.721—Primary Somatosensory Cortex2—Primary Somatosensory Cortex3—Primary Somatosensory Cortex0.360.030.5437161348.8621.0366.779—Dorsolateral prefrontal cortex0.6538181359.330.4346.8744—pars opercularis, part of Broca's area45—pars triangularis Broca's area0.530.1939201357.7510.2163.156—Pre-Motor and Supplementary Motor Cortex0.7240*171442.8954.4744.789—Dorsolateral prefrontal cortex46—Dorsolateral prefrontal cortex0.490.5141181453.1746.4639.645—pars triangularis Broca's area0.5242181565.342.4416.245—pars triangularis Broca's area0.9743141644.02− 30.7586.824—Primary Motor Cortex0.6744161640.691.6781.286—Pre-Motor and Supplementary Motor Cortex145191630.81− 8.4889.456—Pre-Motor and Supplementary Motor Cortex146201649.44− 11.7379.784—Primary Motor Cortex6—Pre-Motor and Supplementary Motor Cortex0.540.43Channels marked with an asterisk correspond to the four short-separation channels (two per hemisphere: one anterior and one posterior) Spatial localization of the 46 channels obtained with NIRS-SPM from the 3D digitizer 39—Angular gyrus, part of Wernicke's area 40—Supramarginal gyrus part of Wernicke's area 0.85 0.15 39—Angular gyrus, part of Wernicke's area 40—Supramarginal gyrus part of Wernicke's area 0.51 0.1 39—Angular gyrus, part of Wernicke's area 40—Supramarginal gyrus part of Wernicke's area 0.47 0.30 5—Somatosensory Association Cortex 7—Somatosensory Association Cortex 0.47 0.1 1—Primary Somatosensory Cortex 3—Primary Somatosensory Cortex 0.30 0.27 4—Primary Motor Cortex 6—Pre-Motor and Supplementary Motor Cortex 0.26 0.74 39—Angular gyrus, part of Wernicke's area 40—Supramarginal gyrus part of Wernicke's area 0.10 0.90 1—Primary Somatosensory Cortex 2—Primary Somatosensory Cortex 3—Primary Somatosensory Cortex 0.23 0.35 0.19 1—Primary Somatosensory Cortex 2—Primary Somatosensory Cortex 3—Primary Somatosensory Cortex 0.38 0.03 0.54 9—Dorsolateral prefrontal cortex 46—Dorsolateral prefrontal cortex 0.37 0.60 44—pars opercularis, part of Broca's area 45—pars triangularis Broca's area 0.61 0.19 4—Primary Motor Cortex 6—Pre-Motor and Supplementary Motor Cortex 0.56 0.42 39—Angular gyrus, part of Wernicke's area 40—Supramarginal gyrus part of Wernicke's area 0.35 0.29 39—Angular gyrus, part of Wernicke's area 40—Supramarginal gyrus part of Wernicke's area 0.48 0.06 39—Angular gyrus, part of Wernicke's area 40—Supramarginal gyrus part of Wernicke's area 0.82 0.17 39—Angular gyrus, part of Wernicke's area 40—Supramarginal gyrus part of Wernicke's area 0.01 0.98 5—Somatosensory Association Cortex 7—Somatosensory Association Cortex 0.51 0.07 1—Primary Somatosensory Cortex 3—Primary Somatosensory Cortex 0.27 0.26 4—Primary Motor Cortex 6—Pre-Motor and Supplementary Motor Cortex 0.30 0.70 39—Angular gyrus, part of Wernicke's area 40—Supramarginal gyrus part of Wernicke's area 0.13 0.87 1—Primary Somatosensory Cortex 2—Primary Somatosensory Cortex 3—Primary Somatosensory Cortex 0.20 0.31 0.28 1—Primary Somatosensory Cortex 2—Primary Somatosensory Cortex 3—Primary Somatosensory Cortex 0.36 0.03 0.54 44—pars opercularis, part of Broca's area 45—pars triangularis Broca's area 0.53 0.19 9—Dorsolateral prefrontal cortex 46—Dorsolateral prefrontal cortex 0.49 0.51 4—Primary Motor Cortex 6—Pre-Motor and Supplementary Motor Cortex 0.54 0.43 Channels marked with an asterisk correspond to the four short-separation channels (two per hemisphere: one anterior and one posterior) fNIRS (pre)processing Initially, the signals underwent a preprocessing stage within the MatLab environment (MatLab R2024b, The MathWorks Inc., USA). A visual investigation was conducted, during which channels were excluded when the observed signals of light intensity (760 or 842 nm) were deemed to be of substandard quality. Moreover, the quality testing of near infrared scans (QT-NIRS) function was employed to identify poor-quality channels based on the provided quality index [53]. All QT-NIRS parameters were utilized with their recommended default values, which are as follows: bpFmin = 0.5, bpFmax = 2.5; windowSec = 5; windowOverlap = 0; quality_threshold = 0.75. The quality threshold is defined as the ratio of the number of signal windows (5 s) that meet the criteria for an acceptable scalp coupling index (> 0.8) and peak spectrum power (> 0.1) to the total number of windows. Participants were excluded from the analysis if more than 15/46 (> 33%) of their channels had been pruned previously. The signals were then imported into the NIRSTORM toolbox [54], within the MatLab environment. The light intensity signals were converted to optical density, and movement artifacts were corrected with the Temporal Derivative Distribution Repair [55]. An infinite impulse response Butterworth bandpass filter of 3rd order (0.01–0.08 Hz) was then applied to remove the slow fluctuations (< 0.01 Hz), Mayer waves (~ 0.08–0.1 Hz), respiration (~ 0.25–0.35 Hz) and heartbeats (~ 1–1.3 Hz) from the signals [32, 48, 49, 56]. Nonetheless, given that these particular filters exhibit a transition band and that Mayer waves frequencies coincide with the hemodynamic signal of interest, along with the potential for signal contamination by task-evoked hemodynamic signals, short-channel signals were utilized to eliminate residual superficial noise through the implementation of an ordinary least squares approach (see [54, 57–59]). Finally, the individual differential pathlength factors were computed using the formula from [33], and the optical density data were converted into HbO and HbR concentrations using the modified Beer-Lambert Law [60]. The HbO and HbR trials were corrected relative to their 5-s baseline, and averaged for further analysis. In accordance with the prevailing consensus on the reporting of fNIRS data, and in an effort to avoid bias, both the HbO and HbR chromophores were reported (see [49, 61, 62]). Given the absence of a consensus [61–63] in the existing literature regarding the preference for the utilization of HbO [64–66] or HbR [67–70] to assess neurovascular coupling (dominant in the hemodynamic response), a channel was considered as "activated" when a significant increase in HbO and/or a significant decrease in HbR was observed during the task. Statistics The initial sample size (n = 30 per group) was the same as in our previous publication reporting the behavioral results of this study [38]. This was determined based on previous studies investigating the VIM, which included between 15 and 31 participants per group [50, 71, 72], as well as studies examining its neural correlates, which included 5–22 participants per group [13, 15, 21, 24, 25]. Considering that some participants could be excluded due to poor signal quality inherent to fNIRS recordings [47], this recruitment target was considered appropriate to obtain a sample size comparable to previous neuroimaging studies investigating VIM neural correlates while accounting for potential data loss. All statistics were performed in MatLab R2024b. Parametric statistical tests were used, as these methods have been shown to be robust for sample sizes greater than five, even in cases of non-normality and unequal variances [73]. Initially, group ages were compared using a one-way ANOVA followed by a post-hoc Tukey–Kramer procedure [74] using the function multcompare. The Tukey–Kramer procedure was applied to all pairwise comparisons between the three groups. Subsequently, the channel activations were determined in each condition and group using beta values derived from a generalized linear model using a simple boxcar function that did not consider the response pattern. Since it is known that there is a ~5-s delay time after a stimulus for the hemodynamic response to rise to its maximum value and to return to baseline, the analysis was conducted on the 25-s interval following the onset of the vibration. One-sample t-tests were therefore conducted on the beta values. The resulting p-values were corrected for multiple comparisons using the false discovery rate correction based on the Benjamini–Hochberg procedure [75] with the function mafdr, across the tested channels, separately for each group, experimental condition, and chromophore (HbO and HbR). The effect sizes were computed using the meanEffectSize function, which provides an unbiased estimate of Cohen’s d, also known as Hedges’ g. The Hedges’ g effect sizes were interpreted as follows: small (g = 0.20–0.50), moderate (g = 0.50–0.80) or large (g > 0.80) [76]. Beyond activation localizations, group and vision hemodynamic differences were compared separately for the left and right arms. Beta values were averaged from the contralateral regions of interest to the vibration, comprising the sensorimotor and frontoparietal areas and encompassing all channels. Subsequently, two-way mixed ANOVAs were conducted on these values, with group as the between-subject factor and vision as the within-subject factor. Effect sizes were calculated for all ANOVAs using partial eta squares (ηp2), and were interpreted according to the following classification: small (0.01–0.06), moderate (0.06–0.14), or large (> 0.14). Following significant ANOVA effects, post-hoc pairwise comparisons were performed using the Tukey–Kramer procedure, which accounts for unequal group sizes and controls the family-wise error rate. For each vision condition, the Tukey–Kramer procedure was applied to all pairwise comparisons between the three groups. Pairwise effect sizes were expressed using Hedges’ g. Consistent with the hypothesis that reduced neural activity reflects more efficient information processing for a similar behavioral outcome [77, 78], we compared group cerebral efficiency in generating the perception of the illusory movement using a functional efficiency ratio (fER) for each group. A rescaling procedure (rescale function) was first applied to the averaged contralateral beta values obtained during the vibration without vision condition, constraining them to the range 0–4. This was done to rescale beta values to a comparable range with the SKIP score, ensuring that neither neural nor behavioral measures disproportionately drive the ratio due to scale differences. This ratio was only investigated in the conditions without vision, as the illusory movement was expected to be maximal in the absence of visual feedback. Subsequently, the fER was derived using the following formula: The constant 1 was added to both the numerator and denominator to prevent the ratio from consistently equaling − 1 when the SKIP score was 0 (i.e., an absence of illusion). Importantly, this additive constant also ensured that variations in beta values (HbO and HbR) continued to contribute to the ratio rather than being overshadowed by the SKIP score alone. The fER group differences for the left-arm vibration were assessed using a one-way ANOVA followed by the Tukey–Kramer procedure, applied to all pairwise comparisons between the three groups. For the right arm, fER differences were examined using a Student’s t-test, as only two groups were compared (young vs. older). Finally, the fER was multiplied by − 1 for HbR to ensure consistent interpretation with a positive scaling. Results Demographic data A total of 29 healthy young participants, 30 healthy older participants and 26 participants with acute stroke were included in the behavioral analysis [38]. However, following signal quality assessments and channel rejection, one healthy young participant (37 channels) and three participants with acute stroke (19, 44 and 46 channels) were excluded from the fNIRS analysis. This reduced the final samples to 28 healthy young, 30 healthy older participants and 23 participants with acute stroke. Participants with acute stroke had an impaired sensorimotor network; see the Supplementary materials Tables 1 and 2 for further details. All participant characteristics are presented in Table 2. Statistical comparisons of baseline characteristics between groups are provided in the Supplementary materials Table 3. Young participants had significantly more rejected channels than older participants (p < 0.001), and significantly lower DPFs than both older participants (p < 0.001) and participants with acute stroke (p < 0.001). In addition, participants with acute stroke had significantly higher Thumb Localizing Test scores than both healthy young (p < 0.001) and older (p < 0.001) participants. Finally, no group differences were found for sex (Young=Older: X2 = 0, p = 1; Young=Stroke: X2 = 0.65, p = 0.42; Older=Stroke: X2 = 0.35, p = 0.56) or handedness (F(2,78) = 2.05, p = 0.14). Table 2Characteristics of healthy young and older participants, as well as participants with acute strokeCharacteristicsYoungMean ± SDOlderMean ± SDAcute strokeMean ± SDSample Size (n)283023Years of age ± SD (range)26.29 ± 6.59 (19-39)63.23 ± 7.88 (49-77)67.48 ± 10.63 (48-85)Sex (n) (female/male)14/1414/168/15Handedness Laterality Index ± SD86.59 ± 14.5292.43 ± 11.4492.57 ± 10.97Channels rejected (n) ± SD7.45 ± 8.781.01 ± 1.663.88 ± 4.44Differential Pathlength Factors ± SD 760 nm6.19 ± 0.197.19 ± 0.207.29 ± 0.27 842 nm5.31 ± 0.196.31 ± 0.206.42 ± 0.27Thumb Localizing Test ± SD Left0 ± 00 ± 00.52 ± 0.85 Right0 ± 00 ± 0-Lesion Volume (mL) ± SD––10.56 ± 19.89 Cortical-Subcortical/Mixed (n)––22/1 Nature (n: Ischemic/Hemorrhagic)––22/1 Hemisphere (n: Left/Right)––0/23Days since stroke ± SD (range)––6.87 ± 3.00 (1-12)NIHSS ± SD Admission––7.70 ± 3.69 Experiment––5.04 ± 3.62FMA-UE score ± SD––42.35 ± 22.30Spatial Neglect Line bisection ± SD 5 mm––1.05 ± 1.88 20 mm––9.84 ± 17.18 Bells test ± SD Total omissions––5.61 ± 8.06 L-R omissions––2 ± 2.98 Characteristics of healthy young and older participants, as well as participants with acute stroke The one-way ANOVA revealed a significant difference between the age groups (F(2,78) = 198.70, p < 0.001). Tukey–Kramer tests indicated that the participants in the young age group were significantly younger than participants in the older (p < 0.001) and acute stroke (p < 0.001) groups. The latter two groups were comparable in age (p = 0.10). Behavioral data In brief, the full behavioral sample [38] revealed that healthy young and older participants exhibited significantly higher SKIP total scores without than with vision in both arms (Youngleft: p < 0.001, d = 1.339; Youngright: p < 0.001, d = 1.257; Olderleft: p < 0.001, d = 0.830; Olderright: p < 0.001, d = 0.538). However, no significant difference was observed in participants with acute stroke (p = 0.232, d = 0.240). In addition, participants with acute stroke showed significantly lower SKIP total scores without vision than healthy of both age groups (Young: p < 0.001, d = 1.496, Older: p = 0.049, d = 0.514). Although the present fNIRS sample excluded a few participants due to signal quality issues, the behavioral patterns were consistent with those reported previously [38]. Only a slight difference was noted in the older versus stroke comparison, yielding a p-value of 0.049 in the full behavioral sample, and 0.060 in the present fNIRS sample. Nevertheless, the Cohen’s d values were nearly identical (d = 0.514 and d = 0.511), suggesting that the result is sensitive to sample size rather than representing a qualitative change. fNIRS data Effects of vibration conditions on brain activation for each group All p-values and Hedges’ g are reported for all channels, and for both HbO and HbR in the Supplementary materials Table 4. During vibrations of the right arm with vision, significant activations were observed in the bilateral frontoparietal regions of young participants (Fig. 4). However, no frontal activations were observed in the absence of vision. Instead, activation was observed in the contralateral somatosensory regions. During the left arm vibrations with vision, significant activation of the contralateral frontoparietal and sensorimotor regions was observed, as well as activation of the contralateral somatosensory and inferior parietal regions. Interestingly, young participants exhibited solely activations in the bilateral inferior parietal regions without vision. The plots show the HbO and HbR levels of healthy young participants for each channel and condition. The four different conditions are indicated by the character with left or right arm vibration and eyes open or closed. The red line shows HbO. HbR is shown as the blue line. The vertical black line indicates the start of the vibration, which begins after the five-second baseline period. The plots are displayed for 25 s after the vibration begins. Non-significant channels are semi-transparent, while channels displaying significant increases or decreases in HbO and/or HbR are opaque. Significant changes in HbO and HbR are indicated by a red or blue asterisk above (increase) or below (decrease) the channel's plot. Regions of interest are highlighted with colored backgrounds: frontal areas are blue, dorsolateral prefrontal (dlPFC) areas are dark yellow, motor areas are pink, primary somatosensory (S1) areas are green, somatosensory association (S2) areas are red, and inferior parietal areas are purple. Regions of interest containing at least one significant channel are additionally outlined with colored borders Figure 5 shows that, in older participants, right arm vibration with vision induced significant activation in the bilateral inferior parietal regions and in the contralateral sensorimotor regions. Without vision, more extensive activation was observed in the bilateral frontoparietal and motor regions as well as activations in the contralateral somatosensory region. During left arm vibration with vision, significantly activated channels were observed in the contralateral frontoparietal and sensorimotor regions. Without vision, more extensive activation was observed in these same regions, as well as significant activations in the ipsilateral sensorimotor and inferior parietal regions. The plots show the HbO and HbR levels of healthy older participants for each channel and condition. The four different conditions are indicated by the character with left or right arm vibration and eyes open or closed. The red line shows HbO. HbR is shown as the blue line. The vertical black line indicates the start of the vibration, which begins after the five-second baseline period. The plots are displayed for 25 s after the vibration begins. Non-significant channels are semi-transparent, while channels displaying significant increases or decreases in HbO and/or HbR are opaque. Significant changes in HbO and HbR are indicated by a red or blue asterisk above (increase) or below (decrease) the channel's plot. Regions of interest are highlighted with colored backgrounds: frontal areas are blue, dorsolateral prefrontal (dlPFC) areas are dark yellow, motor areas are pink, primary somatosensory (S1) areas are green, somatosensory association (S2) areas are red, and inferior parietal areas are purple. Regions of interest containing at least one significant channel are additionally outlined with colored borders In participants with acute stroke (Fig. 6), no significant activation was observed when the left arm was vibrated either with or without vision. The plots show the HbO and HbR levels of participants with acute stroke for each channel and condition. The two different conditions are indicated by the character with left arm vibration and eyes open or closed. The red line shows HbO. HbR is shown as the blue line. The vertical black line indicates the start of the vibration, which begins after the five-second baseline period. The plots are displayed for 25 s after the vibration begins. HbO and HbR did not change significantly. Regions of interest are highlighted with colored backgrounds: frontal areas are blue, dorsolateral prefrontal (dlPFC) areas are dark yellow, motor areas are pink, primary somatosensory (S1) areas are green, somatosensory association (S2) areas are red, and inferior parietal areas are purple Group hemodynamics differences During the vibration of the left arm, the two-way ANOVAs revealed that there was no significant group effect (F(2,78) = 1.62, p = 0.20, ηp2 = 0.04), vision effect (F(1,78) = 0.00, p = 0.98, ηp2 = 0.00), or interaction effect (F(2,78) = 1.46, p = 0.24, ηp2 = 0.04) for HbO. However, a significant group effect was identified for HbR (F(2,78) = 4.74, p = 0.01, ηp2 = 0.11, Fig. 7), but no significant vision effect (F(1,78) = 0.50, p = 0.48, ηp2 = 0.01) or interaction effect (F(2,78) = 1.86, p = 0.16, ηp2 = 0.05) were identified. Post-hoc Tukey–Kramer tests revealed significant differences between healthy young participants and participants with acute stroke (p < 0.05, g = 0.74, Fig. 7) but no differences were found between healthy young and older participants (p = 0.73, g = 0.17), or between healthy older participants and participants with acute stroke (p = 0.06, g = 0.96). Violin plots illustrate group-related differences in contralateral hemisphere hemodynamics during left-arm (left panels) and right-arm (right panels) vibration without vision. The top row represents HbO and the bottom row HbR. Young participants are represented in blue, older participants in orange, and participants with acute stroke in green. Post-hoc Tukey–Kramer tests: * p < 0.05 During the vibration of the right arm, the two-way ANOVAs revealed a significant group effect for HbO (F(1,56) = 5.30, p < 0.05, ηp2 = 0.09, Fig. 7). However, no significant vision effect (F(1,56) = 0.004, p = 0.95, ηp2 = 0.00) or interaction effect (F(1,56) = 0.68, p = 0.41, ηp2 = 0.01) were observed. As the group factor included only two levels, the significant main effect reflects the difference between young and older adults. The Tukey–Kramer post-hoc comparison confirmed this effect (p < 0.05) with a Hedges’ g of 0.60 (Fig. 7). Moreover, no significant group effect (F(1,56) = 1.27, p = 0.26, ηp2 = 0.02), vision effect (F(1,56) = 0.13, p = 0.72, ηp2 = 0.00), or interaction effect (F(1,56) = 0.19, p = 0.67, ηp2 = 0.00) was identified for HbR. Functional efficiency ratio In the left vibration without vision condition, the one-way ANOVAs revealed significant differences for both HbO (F(2,78) = 11.70, p < 0.001, ηp2 = 0.23) and HbR (F(2,78) = 10.30, p < 0.001, ηp2 = 0.21). Post-hoc Tukey–Kramer analysis revealed that healthy older participants and participants with acute stroke had smaller fER than healthy young participants (Older: HbO: p < 0.01, g = − 0.93; HbR: p < 0.01, g = − 0.93; Acute stroke: HbO: p < 0.001, g = − 1.27; HbR: p < 0.001, g = − 1.17, Fig. 8). Furthermore, no significant differences were found in the fER between healthy older participants and participants with acute stroke (HbO: p = 0.42, g = − 0.34; HbR: p < 0.75, g = − 0.20, Fig. 8). Violin plots illustrate group-related differences of the functional efficiency ratios in the contralateral hemisphere during left (left panels) and right arm (right panels) vibration without vision. The top row represents HbO and the bottom row HbR. Young participants are represented in blue, older participants in orange, and participants with acute stroke in green. Post-hoc Tukey–Kramer tests (left): ***p < 0.001, **p < 0.01; Post-hoc Student’s t-tests (right): ***p < 0.001, **p < 0.01 In the right vibration without vision condition, Student’s t-tests revealed that the fER was smaller in healthy older participants than in healthy young participants for both HbO and HbR (HbO: p < 0.001, t = − 4.95, g = − 1.28; HbR: p < 0.01, t = − 3.23, g = − 0.84, Fig. 8). Discussion The present study investigated how VIM modulates cerebral activity in response to vibration of the left or right triceps brachii, with or without vision of the vibrated arm, in healthy young and older participants, as well as in the left paretic arm of participants with acute stroke. To this end, an fNIRS device was used to measure cortical hemodynamics in the frontoparietal and sensorimotor areas. Three hypotheses were examined. First, we hypothesized that the absence of visual feedback would increase activation of the contralateral sensorimotor and frontoparietal networks compared with visual feedback. This hypothesis was partially supported. In young participants, activation varied across regions and vibration sides. In older participants, however, VIM reliably activated the contralateral sensorimotor and frontoparietal areas; these responses extended bilaterally in the absence of vision. Second, we hypothesized that visual feedback would attenuate the VIM and reduce activation within the sensorimotor and frontoparietal networks while promoting activity in contralateral somatosensory regions. This hypothesis was also partially supported. At the behavioral level, visual feedback attenuated the VIM in young and older participants, as reflected by lower SKIP scores with vision. At the neural level, however, the predicted reduction in sensorimotor and frontoparietal activity and concomitant increase in contralateral somatosensory activity were not consistently observed. Finally, we hypothesized that acute stroke would be associated with altered and reduced activation of the contralateral sensorimotor and frontoparietal networks compared with healthy individuals. This hypothesis was supported, as participants with acute stroke showed no significant activation in response to VIM under any condition and exhibited lower hemodynamic activity than healthy participants, although this difference did not reach statistical significance when compared with older participants (p = 0.06, g = 0.96). Together, these findings indicate that VIM-related cortical responses are modulated by visual feedback and aging, while acute stroke is associated with a marked reduction in the cortical response to VIM. Beyond these findings, contralateral cerebral activity in older participants in response to right-side vibration was found to be higher than in young participants. In contrast, there were no differences in the cerebral activity between young and older participants during left vibration. Furthermore, young participants demonstrated greater neural efficiency in responding to VIM than both older participants and participants with acute stroke. Effect of vision on cerebral activity related to the VIM The prevailing view regarding visual feedback is that vision generally outweighs proprioceptive input, as demonstrated by studies in healthy participants using the rubber hand illusion [79], the mirror-induced visual illusion [80], and the VIM. In the latter, vision can either enhance the illusion when movement is represented by a virtual avatar [81] or attenuate it when exposed to visual feedback of the vibrated limb [19, 20]. In our study, the results were consistent with this framework, as both young and older participants exhibited stronger VIM responses in the absence of vision, whereas no such effect could be demonstrated in participants with acute stroke [38]. Given that previous studies have suggested that VIM engages a broader cerebral network than vibration without perceived illusion [15, 21, 22], we expected to observe more widespread brain activation when participants could not see their vibrated arm, particularly in healthy young and older adults. In our sample, the results in young adults showed significant activation of the bilateral inferior parietal areas (BAs 39–40), which was consistent across conditions. These regions are classically involved in the VIM [15, 25], multisensory body representation [79, 82, 83], and have also been associated with the conscious experience of movement [12] and the sense of agency [84]. Although vision attenuated VIM responses in young participants [38], the persistence of parietal activation suggests that these areas remain a core neural substrate of the illusory movement and visual awareness. This is likely to reflect the integration of proprioceptive afferences and the predicted sensory consequences of movement. Unexpectedly, (pre)frontal areas showed significant activation only in the presence of vision, with a right-hemispheric dominance. This result contrasts with the classical view that frontal areas constitute a core component of VIM-related networks, particularly under conditions of maximal illusion without vision. However, rather than reflecting reduced frontal area involvement in the illusory experience itself, this pattern may indicate a shift towards higher-order cognitive processes when visual feedback is available. Indeed, (pre)frontal areas have been associated with sensory conflict monitoring [82, 85]. More specifically, the dlPFC has been shown to be involved when participants perform an incongruent task involving a mismatch between their movement and the visual feedback they receive [82]. When vision is present, participants experience an explicit visuo-proprioceptive mismatch (seeing a static arm while feeling movement), which may require additional neural resources to monitor, evaluate, and resolve this sensory conflict. Furthermore, [82] reported a right dlPFC dominance during this mismatch, observing that during a visuo-proprioceptive mismatch involving passive hand movements closely resembling our experimental condition, these activations extended to the ventral frontal areas (BAs 44–45). Furthermore, as the participants were instructed to focus explicitly on the sensation arising from their elbow joint, the observed frontal activation may reflect an enhanced awareness of bodily signals rather than the illusory movement per se. This suggests that VIM under visual feedback may involve not only bottom-up sensorimotor integration but also top-down monitoring of bodily states [82, 86]. In this context, vision may not only attenuate the illusion but also modify its neural underpinnings. This shifts the experience from a primarily parietal-driven perceptual phenomenon towards a more cognitively controlled bodily awareness process. Therefore, further research is needed to investigate and disentangle the complex cerebral network involved in the attenuation of the VIM by vision. Furthermore, the sensorimotor areas responded differently to right- and left-arm vibration, the somatosensory areas being activated only with vision of the vibrated left arm. This finding suggests a complex interaction within the visuo-proprio-lateralized network. However, previous studies have suggested that visual guidance is more important for the right arm, whereas proprioceptive feedback is more important for the left arm [87–89]. This could explain why, during left vibration, a more proprioceptively driven cortical representation may elicit somatosensory activations, whereas such activations may be less necessary with vision of the vibrated right arm due to the dominance of visually guided control. Conversely, older adults exhibited more widespread activation in the frontoparietal and sensorimotor areas when they were not able to see their vibrating arm than when they could see it. With vision, activations were predominantly confined to the contralateral hemisphere; without vision, however, they extended to a large part of the ipsilateral frontoparietal and sensorimotor networks, as well as recruiting more channels in the contralateral hemisphere. This phenomenon was observable during vibration of both left and right arms. These findings align with previous studies demonstrating that the frontoparietal and sensorimotor networks are more active during a clear VIM than during vibration without an illusion or an attenuated one [15, 16, 21, 22]. Rather than supporting the emergence of the illusion from a single cortical region, these results suggest that VIM in older adults probably relies on a complex interplay between frontoparietal and sensorimotor areas. It has been suggested that an increase in sensorimotor and frontoparietal networks connectivity occurs in older individuals to compensate for reduced VIM responses [24]. This could also reflect cerebral compensation mechanisms occurring with aging, whereby older adults require greater neuronal resources for a stimulus of the same intensity to achieve equivalent or lower performance than younger adults [90, 91]. Interestingly, in participants with acute stroke, the results could not demonstrate any effect of the vibration with or without vision of the vibrated arm on cerebral hemodynamics. As SKIP scores did not differ significantly with and without vision in this group, likely due to the alteration of the sensorimotor system [38], this may reflect a lack of proprioceptive stimulation for participants with acute stroke. Another factor that may have contributed is post-stroke spasticity, which was not quantified in the present study. In fact, spasticity occurs in approximately 25% of individuals early after stroke [92] and is associated with increased muscle tone and exaggerated stretch reflexes. In such cases, tendinous vibration may elicit involuntary muscle contractions, which could interfere with the perception of the VIM. Previous studies have shown that voluntary muscle contraction can attenuate the illusion [46, 93], suggesting that post-stroke spasticity may represent an additional factor influencing the perception of the VIM after stroke. Furthermore, considering the complex cerebral reorganization occurring in the acute phase of stroke [94], the absence of measured neural responses could indicate a general perturbation in cerebral hemodynamics [95]. Thus, the response may be more diffuse, and the present sample size (n = 23) may have limited sensitivity to detect small effects. To further explore this issue, we conducted a complementary post hoc analysis by averaging effect sizes across contralateral channels and estimated statistical power. As the mean HbR effect size was negligible (g = − 0.070), power estimation was performed only for HbO (g = 0.252). For this effect size, the estimated statistical power was low (0.212), suggesting that small effects may not have been reliably detected with the current sample. Power calculations further indicate that a substantially larger sample would be required to achieve conventional power levels (126 participants with acute stroke for 0.8 power, and 80 participants with acute stroke for 0.6 power). Together, these results suggest that the absence of significant findings may reflect the small effect size and diffuse nature of the response rather than a lack of physiological relevance. Furthermore, the clinical relevance of these neural differences remains unclear, as changes in cortical activation cannot be directly interpreted as functional improvements. Further longitudinal studies combining VIM neurophysiological measures with clinical outcomes are required to determine whether these neural responses may be associated with enhanced functional recovery. Therefore, future studies should consider having a larger sample size for participants with acute stroke. Furthermore, as the absence of measured significant cerebral activity in participants with acute stroke may be due to the absence of enhanced VIM, a protocol incorporating VIM associated with electrical stimulation [96–99] and/or virtual reality [81, 100, 101] to promote VIM responses could provide further insight into the neural correlates of VIM within this pathological condition. How does aging influence the neural correlates of VIM? With aging, the brain is thought to adopt compensation strategies to achieve performance comparable to that of younger adults (CRUNCH model, [90, 91]). Three types of compensation have been suggested: upregulation, selection and reorganization [90]. In our study, older adults exhibited more extended activations than young adults during both left and right vibrations without vision. Furthermore, they had lower VIM responses [38] than young adults during both left and right vibrations without vision. Older adults also demonstrated stronger contralateral cerebral activity than young adults in response to right vibration without vision. These phenomena may therefore be attributable to compensation through upregulation and reorganization associated with the hemispheric asymmetry reduction occurring with aging (HAROLD model, [102]). However, during left vibration without vision, older and young adults demonstrated comparable contralateral cerebral activity. Although there were no significant differences in left and right VIM responses between young and older adults [38], this could be explained by the left arm’s preference for proprioceptive feedback [87–89]. This preference may have required more neural resources in both groups, mitigating discrepancies. Furthermore, a previous study reported faster VIM responses for the right arm of young right-handers, though no differences were observed in older individuals [103]. This suggests that young right-handers may engage fewer contralateral neural resources with right arm vibration than left arm vibration to achieve equal performances with both arms. Interestingly, our results showed that older participants had lower fER than young participants. This supports the idea that older adults may possess compensatory mechanisms involving upregulation to maintain their proprioceptive accuracy and keep their performance as high as possible, even though their VIM responses were still lower than those of young participants. These results align with previous studies on aging demonstrating a peripheral proprioceptive deficit through muscle spindle structural and functional alterations [104, 105], and a central proprioceptive deficit [24, 106] likely related to cortical grey matter reduction [107]. It is also worth noting that our task was passive; participants were only asked to focus on the sensation at their elbow joint. There was no expectation of maximal performance. Nevertheless, this was sufficient to highlight a reduced neural efficiency of the proprioceptive system with aging. Finally, physical activity could prevent the aforementioned proprioceptive decline at both peripheral and central levels. This is characterized at the peripheral level by metabolic and structural adaptations of the muscle spindle [108, 109], as well as an enhanced dynamic sensitivity of muscle spindle afferents following proprioceptive training [110]. Furthermore, this is characterized at the central level through the modulation of gene expression implied in neural function [111], increased connectivity in frontal-related networks [112] and it may attenuate age-related decline in the hippocampus, temporal and frontal regions [113]. In the end, these peripheral and central adaptations could limit the proprioceptive acuity deficit occurring with aging [114–116]. Therefore, investigations of the effect of physical activity on VIM responses and neural efficiency with aging seem highly relevant. How does acute stroke alter neural responses to the vibration? The included stroke participants were in the acute phase (< 14 days) and had a right (sub)cortical lesion (except for one individual, see Table 2) impairing their sensorimotor network (Supplementary materials Tables 1 and 2). Although the cerebral reorganization that occurs during the acute phase of stroke remains unclear [94], it has been documented that general cerebral hemodynamic perturbation can occur [95]. Therefore, we expected to observe decreased cerebral responses to our experimental conditions in participants with acute stroke compared with healthy young and older participants. Indeed, our results showed that participants with acute stroke had lower neural responses to the VIM than healthy young and older participants. While this result was significant when compared with young participants, it was not significant when compared with older participants (p = 0.06, g = 0.96). Nevertheless, Hedges’ g indicated a strong effect of stroke, and this non-significant result may be due to a lack of statistical power related to the acute stroke group’s medium sample size (n = 23). In any case, it appears that acute stroke was associated with reduced contralateral neural responses to vibration. Thus, we hypothesize that acute stroke may have induced compensation by selection [90] and another cerebral network may have been recruited in response to the vibration. Furthermore, although their fER was reduced compared with young participants, it was comparable to that of older participants. This result is noteworthy, as we expected participants with acute stroke to have decreased VIM and cerebral responses. It suggests that, despite having fewer neural resources, participants with acute stroke may preserve an acceptable level of proprioceptive acuity, which may have implications for early sensorimotor readaptation through exercise during the acute phase of stroke. Consistent with this hypothesis, a meta-analysis of Chinese persons with a stroke revealed that rehabilitation within two weeks after stroke improves motor recovery, as demonstrated by enhanced NIHSS, (modified) Barthel Index and FMA scores, compared to rehabilitation two weeks after stroke [117]. However, the authors could not determine the effects of rehabilitation at a very early stage (i.e., within 48 h) after stroke on motor recovery. Additionally, a study in hemiplegic individuals with subacute stroke revealed significant improvements in hand motor performance following a two-week VIM training [10], which persisted one month post-training. However, these authors could not demonstrate significant improvements in EEG cerebral responses at the C3 and C4 loci. More recently, Shurupova et al. (2026) investigated the effects of a three-week intervention including 12 sessions of tendinous vibration (the first two sessions also included muscle vibration) in individuals with subacute stroke (1–6 months), combined with conventional therapy, and compared with conventional therapy combined with robotic mechanotherapy [118]. The authors reported greater improvements in upper limb motor function following the vibration-based intervention. Together, these findings highlight the need for further research to better characterize how tendinous vibration training influences proprioceptive processing and its relationship with motor and cerebral recovery in individuals with acute stroke, especially over long-term outcomes. Methodological considerations To minimize neural responses during the baseline period, we randomly assigned a resting period ranging from 25 to 35 s. However, to maximize the contrast-to-noise ratio, we could have shortened the resting period and task duration to allow for more trials [68]. The hemodynamic response typically peaks approximately five seconds after task onset and takes another five seconds to return to baseline [119, 120]. Therefore, a 10-s rest period between trials may be sufficient to minimize baseline noise. However, choosing appropriate resting and task periods is challenging because of the few seconds needed to perceive the illusion of movement and the short period after the end of vibration when participants may feel their arm returning [46]. Thus, we believe that a vibration duration of 10 s, interposed by resting periods ranging from 12 to 22 s, would be appropriate, as it would enable participants to perform ten trials instead of six. Furthermore, young participants had significantly lower DPFs than both older participants and participants with acute stroke, as well as significantly more rejected channels than older participants. The observed differences in DPFs are unlikely to substantially affect the interpretation of the results, as DPFs vary with age (see [33]). Regarding the rejected channels, only 7 out of 46 channels on average were rejected in the young group, and several channels contributed to each of the six ROIs, suggesting that this difference is unlikely to have substantially affected the overall activation profiles. Additionally, EEG measurements would provide more precise insights into cerebral responses to VIM by investigating concurrent sensorimotor µ-band activity. This is particularly important in individuals with acute stroke, as the absence of an effect may be due to hemodynamic perturbations and to inter-individual variability in lesion volume and stroke severity. Moreover, although the fER was constructed to minimize scale differences between neural (beta values) and behavioral (SKIP score) measures, the distribution of SKIP scores may still have influenced the ratio in some cases. In particular, the larger number of zero SKIP scores observed in older participants and participants with acute stroke may have reduced the relative contribution of beta values to the metric. Furthermore, participants with acute stroke had higher Thumb Localizing Test scores than healthy participants, which may reflect their neurological impairment and could represent an additional factor contributing to the observed differences in fNIRS activity. Finally, it is important to note that cognitive impairments and aphasia were not quantified in the stroke population, as these factors may have influenced the subjective reporting of the illusion of movement. Conclusion In conclusion, our study revealed that VIM is associated with bilateral activation of sensorimotor and frontoparietal areas in older adults. However, acute stroke appears to alter these cerebral responses, as reflected by the absence of significant activations and reduced VIM-related cerebral responses in participants with acute stroke compared with both healthy young and older participants. Despite this reduced activation, neural efficiency in participants with acute stroke was comparable to that observed in older participants. These findings suggest that preserved compensatory mechanisms may still be engaged early after stroke and that VIM training, which directly stimulates the muscle spindle and engages sensorimotor and frontoparietal networks, may be highly relevant for older adults to limit age-related proprioceptive decline and to facilitate recovery after limb immobilization. Altogether, this cross-sectional study suggests potential implications for early rehabilitation after stroke and paves the way for future investigations into the long-term effects of VIM training on motor and cerebral function recovery in individuals with acute stroke. Supplementary Information Supplementary Material 1. Acknowledgements We thank Andreea Aignatoaie and Guillaume Cami for their support in recruiting individuals with acute stroke, as well as Julien Bonnal, Julie Chatelier, Aurélie Mentisano, and Lucie Lebkowski for administering the Fugl-Meyer Assessment of the Upper Extremity and the spatial neglect tests. We are grateful to Ophélie Pila for her valuable feedback and assistance with the illustrations, and to Grégoire Vergotte for his insightful explanations regarding fNIRS data visualization. We also express our profound gratitude to Sophie Apprich from Artinis Medical Systems for her insightful technical support. Finally, we extend our sincere gratitude to all participants for their invaluable contribution to this study. Abbreviations Brodmann areas Electroencephalography Fugl-Meyer Assessment for the Upper Extremity Functional magnetic resonance imaging Functional near-infrared spectroscopy Oxyhemoglobin Deoxyhemoglobin Montreal Neurological Institute Quality testing of near infrared scans Short channel Standardized Kinesthetic Illusion Procedure Vibration-induced illusion of movement Author contributions Conceptualization: C.Ö., P.A., B.L.; Methodology: B.L., C.Ö., P.A.; Investigation: B.L., É.F., M.S., S.C., C.G., C.C.; Formal analysis: B.L.; Data curation: B.L., B.D.; Writing—original draft preparation: B.L.; Visualization: B.L.; Writing—review and editing: B.L., S.P.; Project administration: B.L., B.D., C.Ö., P.A.; Funding acquisition: C.Ö., S.P., P.A.; Supervision: S.P., C.Ö. Funding This work was supported by ANRT via CIFRE (Conventions Industrielles de Formation par la Recherche n°2022/1454). Data availability The fNIRS raw data are available on Recherche Data Gouv via the permanent link: https://doi.org/10.57745/HNZBV9. Declarations Ethics approval and consent to participate The Institutional Review Board (CPP Sud-Méditerranée 1) approved the study protocol on June 14 2023, and was subsequently registered under the number 2023-A01062-43 (Jardé law, RIPH2). The study was prospectively registered on ClinicalTrials.gov, under the identifier NCT06218563. All participants provided free and informed written consent in accordance with the Declaration of Helsinki. Consent for publication Not applicable. Competing interests The authors declare no competing interests. Footnotes Publisher's Note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. References Associated Data Data Availability Statement The fNIRS raw data are available on Recherche Data Gouv via the permanent link: https://doi.org/10.57745/HNZBV9.