Nineteen (n = 19) volunteers were included in this protocol (9 females and 10 males, mean age 26.6 ± 5.3 years). All participants were right-handed (mean score 0.8 ± 0.2) as determined by the Edinburgh Handedness Inventory35. None of the volunteers had any neurological, psychological, or musculoskeletal diseases. The protocol was approved by CPP Ile de France IV (number 2021-A03219-32, and ClinicalTrials NCT05315726), and all participants provided verbal informed consent, in accordance with the revised 2013 Declaration of Helsinki.
Visits 1 (V1) and 2 (V2) each included two measurement blocks (PRE and POST, see Fig. 1), separated by a 20-minute rest at V1 and a 20-minute LMV of the wrist flexors at V2. V1 and V2 were 11 days apart, during which participants were instructed to maintain their normal lifestyle. V1, V1 and V2 were used to assess data reproducibility over time and served as control conditions.
The LMV protocol started after V2 and finished the day before V3, with participants receiving a total of nine 20-minute LMV sessions. Movement illusions were evaluated subjectively (via ratings) in all sessions and objectively (via EEG) during the second (EEG1) and final (EEG2) sessions, due to the demanding nature of EEG recordings.
In V3 and V4, the same neuromuscular measurements as in V1 and V2 were recorded. Between V3 and V4 (5 days), participants maintained their normal lifestyle. The acute effects of LMV on inducing illusions were assessed by comparing neurophysiological parameters and strength using V1, V1, V2 and V2. The chronic effects of repeated LMV with illusions were assessed by comparing V2 with V3 and V4, with V4 capturing any lasting changes, as previously shown in the literature, up to 2 weeks following the completion of the LMV protocol.
All visits (V1-V4) were consistently scheduled at the same time of day for each participant to minimize potential intra-participant circadian fluctuations, and followed a consistent weekly schedule: V1 (Thursday), V2 (Monday), V3 (Friday) and V4 (Wednesday).
During the neurophysiological measurements (Fig. 2A), participants were comfortably seated upright in a chair, with their right forearm supported in a semi-pronated position on a custom-made support, elbow at 90° flexion, and shoulder at 30° abduction, 15° flexion, and 20° external rotation. During the 20-minute rest (V1) and all LMV sessions (Fig. 2B), they sat in a semi-reclined position to promote muscle relaxation and enhance the likelihood of experiencing illusions. Their arm was hidden from view and supported at the elbow and wrist by malleable cushions, with the elbow semi-flexed, pronated, and the wrist in a neutral position.
Participants completed nine vibratory sessions, one each day over two consecutive weeks, each lasting 20 min of continuous mechanical LMV. The vibration module (Vibramoov, Technoconcept, Manosque, France) was directly applied to the skin, approximately 2 cm proximal to the right wrist crease, and secured around the forearm with a strap (Fig. 2B). To ensure optimal comfort and appropriate pressure, the wristband size was individually adjusted for each participant. Continuous LMV was administered for 20 min at a frequency of 80 Hz with an amplitude of 2 mm (according to the manufacturer's specifications). The participants' view of their forearm was obstructed with a cardboard box to induce movement illusions and diminish the tonic vibratory reflex. Participants were instructed to relax and concentrate on the LMV. The 20-minute duration was chosen based on previous studies demonstrating its effectiveness in inducing movement illusions and in modulating H-reflex and muscle strength. In particular, we drew on the findings of Nito et al. (2021), who reported that corticospinal excitability of the vibrated muscle (FCR) was already significantly modulated after as little as 8 min of LMV.
During each vibratory session, three visual analogue scales were used to rate participants' subjective sensations. These scales, ranging from 0 (non-existent) to 10 (very strong), assessed the intensity of the illusory movement (strength), the duration of the illusion during the last minute (continuity), and the clarity of illusions as if a real wrist extension was performed (vividness). Participants were asked to evaluate their sensations at specific intervals (1 min, 5 min, 10 min, 15 min, and 19 min), providing one score per scale at each time point. The pooled score, calculated as the mean of the three scales, served as the marker for the overall intensity of the illusions.
Two EEG sessions were conducted during the second and final vibratory session (i.e., EEG1, and EEG2, see Fig. 1). EEG signals were recorded (Icecaps, Neuronaute, BioSerenity, Paris, France) from 21 electrodes, based on the 10-20 system. Electrode impedances were maintained below 5 kω, and signals were recorded at a sampling rate of 500 Hz. The entire EEG assessment was performed with participants' eyes closed. Two 5-minute rest periods were recorded at the beginning and end of the 20 min of LMV to assess the resting state (REST PRE and REST POST). An accelerometer, positioned on the vibrator module and connected to the EEG system, allowed for the detection of when vibrations were on/off for synchronization with the EEG recording.
Research indicates that LMV applied to a resting limb enhances cortical activity, with Mu rhythm ERSP at C3 serving as a marker for this activation. Movement illusions induced by LMV further amplify this cortical activity. Since our LMV conditions were optimized to induce these illusions, Mu desynchronization during the 20-minute LMV, in comparison to the resting period, was used in all participants to quantify brain activity under illusions.
After shaving and dry-cleaning the skin with alcohol, silver chloride surface electrodes (Ag/AgCl) were positioned on the belly of the right Flexor Carpi Radialis (FCR) with an interelectrode distance of 2 cm (center to center). To ensure consistent electrode placements across visits, the distance from the medial humeral epicondyle and the midpoint of the distal wrist crease was measured. Electrodes were positioned at 35% of this distance from the proximal reference point (participants' mean was 9.82 ± 0.83 cm), corresponding to the theoretical optimal electrode site. A reference electrode was placed over the medial epicondyle of the radial styloid. To prevent noise during LMV, the ground electrode was placed on the olecranon process. The EMG signal was amplified (gain 1000) and band-pass filtered (10-500 Hz), digitized at a sampling rate of 2000 Hz, and recorded for off-line analysis (Biopac System Inc., Goleta, CA, USA).
Neuromuscular tests were conducted in a fixed sequence (see Fig. 1) to minimize potential external influences, aside from rest or LMV, on the primary outcome measure, short-interval intracortical inhibition (SICI). SICI assessments were performed last during the pre-test sessions and first during the post-test sessions. This order was designed to respect critical methodological dependencies (e.g., Mmax preceding H-reflex, RMT before MEP and SICI), and to prevent any active muscle contractions from impacting the neurophysiological parameters of interest.
A warm-up phase consisting of 2 min of right wrist and finger flexions starting from low to submaximal intensity preceded the maximum force evaluation. Three (n = 3) maximal grip voluntary contractions (T.K.K.5401 GRIP-D hand-grip dynamometer, Takei Scientific Instruments Co., Ltd, Tokyo, Japan), each lasting 3 s with 30 s recovery in between, were performed in each PRE (V1-V4) and POST (V1 & V2) visits (Fig. 1). The participants' positions were similar across sessions, with their elbows and forearms resting on their right thigh while squeezing the dynamometer. Verbal encouragement was provided during every contraction. The maximal grip strength for each set of contractions was determined as the highest value among the three trials, displayed on the dynamometer screen.
The H-reflex and M-wave at the right FCR were elicited by stimulating the right median nerve. Rectangular 1 ms wave pulses were delivered through bipolar felt pad electrodes, which were secured around the arm using a Velcro band, and connected to a constant-current stimulator (DS7R, Digitimer, Welwyn Garden City, Hertfordshire, UK). The optimal stimulation site, determined on each visit as the location that produced the largest M-wave, was identified near the medial border of the cubital fossa, approximately 2 cm above the medial epicondyle of the humerus. In each session, the stimulation intensity was incrementally increased until the largest H-reflex and the M-wave reached a plateau. Three stimulations were conducted at 120% of the plateau intensity, and the M-wave with the greatest peak-to-peak amplitude was designated as Mmax (mean intensity across visits: 16.0 ± 6.3 mA).
For the H-reflex, the initial visit (V1) was used to determine a reference value. Three stimulations were performed at the intensity eliciting the largest H-reflex; the average of these three peak-to-peak amplitudes was used to calculate Hmax. Subsequently, 10 stimulations were delivered at the intensity evoking 80% of Hmax (H80) within the ascending part of the recruitment curve. During the following PRE and POST visits, to ensure consistent stimulus conditions for evaluating the H80 across these different time points, a stable M-wave was maintained, corresponding to the M-wave associated with H80 (M(H80)); the mean intensity across visits was 6.9 ± 3.4 mA.
TMS was applied to the left primary motor cortex using a 70 mm figure-of-eight coil connected to a monophasic Magstim BiStim stimulator (The Magstim Co., Whitland, UK). The hot spot, defined as the optimal stimulation site producing the largest Motor Evoked Potential (MEP) amplitude for a given intensity, was determined with the assistance of a navigated brain stimulation system (Brainsight TMS Navigation, Brainbox, Cardiff, UK). The hot spot, identified as one of the 12 points on the system map (mean MNI coordinates: -44.7, 13.8, 85.7; Négyesi et al., 2020), was determined individually for each participant during V1 and remained consistent for that participant across all visits. The resting motor threshold (RMT) was defined as the minimum stimulus intensity required to evoke at least 5/10 MEPs with a peak-to-peak amplitude of 50 µV in the relaxed FCR. A block of 15 single-pulse stimuli was delivered at 130% of RMT to measure MEP. Another block of 15 paired-pulse stimulations at 80%-130% of RMT with 3-ms intervals was carried out to study Short Intracortical Inhibition (SICI).
A tailored MATLAB algorithm (MATLAB R2024a; MathWorks, Natick, MA) was developed to extract data from each stimulation (V1-V4), including peak-to-peak amplitudes of Hmax, M(Hmax), H80, M(H80), Mmax, MEP, and SICI. Normalization (% expression) of Hmax, M(Hmax), H(80), M(H80), and MEP was performed relative to Mmax, enabling intra-sessions and interindividual comparisons.
SICI was computed with the following formulae:
MEP is a MEP evoked at 130% of the RMT preceded 3 ms before by a stimulation at 80% RMT. MEPtest is a MEP evoked at 130% of the RMT.
Negative values signify inhibition, while positive values indicate facilitation.
The interquartile range (IQR) method was employed to detect outlier values within each participant, stimulation block, and session (see Supplemental Table A.1). An outlier was identified if the value exceeded Q3 + 1.5 x IQR or was less than Q1-1.5 x IQR. Due to the limited dataset (i.e., only 3 values), this method was not applied to Mmax, Hmax, and M(Hmax).
During LMV, some participants exhibited a small tonic vibratory reflex (TVR). Complementary EMG analysis, similar to the methodology used by Amiez et al. (2024a) was carried out. The results revealed an average TVR of 0.08 ± 0.05%Mmax, ranging from 0.02% to 0.17% of Mmax (averaged from V2 and V3). These findings align with those reported in the illusions group of Amiez et al. (2024a) (0.05 ± 0.03%Mmax) compared to 0.22 ± 0.18%Mmax in their TVR group. Additionally, our EMG analysis during maximal grip strength closely matches their results, as our participants demonstrated an average of 6.01 ± 3.97%Mmax, comparable to the ~ 6% reported in their study.
EEG processing and calculations were performed using the MNE Python library. EEG data were filtered with a 0.5 to 40 Hz band-pass filter. To improve Independent Component Analysis (ICA) decomposition, the initial 180 s of acquisition were excluded to eliminate noise caused by eye movements or other disruptions typically occurring during the transition to a resting state. ICA decomposition was performed, and components related to eye movements and muscle artefacts were manually removed.
ERSP Analysis Event-Related Spectral Perturbations (ERSPs) were calculated using Morlet wavelets with a cycle length that progressively increased from 1 to 30, spanning a frequency range of 1 to 30 Hz across the full experiment duration. An 80-second baseline was applied during the resting state (-110 to -30 s). For analysis, only the mu band (8-14 Hz) was extracted from the ERSP data.
A power analysis was conducted using G*Power (version 3.1.9.4) to determine the required sample size for this protocol. The calculation was based on previous studies from Souron et al. (2018, 2017), which examined chronic strength differences following LMV training. Assuming a medium effect size (f = 0.32), an alpha level of 5%, a statistical power of 90%, and one group with four repeated measurements, the analysis indicated that 19 participants were needed.
Statistical analyses were conducted using JASP (version 0.19.0.0 ; University of Amsterdam, The Netherlands), and "rmcorr R package" for repeated-measures correlations. Data normality was verified using the Shapiro-Wilk test.
One-sample t-tests (comparing with the zero '0' value) were performed to confirm that the protocol successfully induced SICI during the control conditions (V1, V1, and V2), as described and illustrated in the supplemental Figure A.1.
To verify that our LMV protocol successfully induced sensory illusions, we calculated the average pooled score from the three visual analogue scales over the 20-minute session and compared this score to the zero '0' value (one-sample t-tests), for the second LMV session (EEG1).
We computed an ERSP parameter, which represents the ERSP normalized by the baseline ERSP (during REST PRE) for each participant.
Then, to evaluate the acute effect of LMV on illusion, we tested the within-session evolution of sensory illusion during LMV (i.e., during one session, EEG1), measured both subjectively (pooled score) and objectively (ERSP). Two Friedman tests were performed with a factor Time: 1', 5', 10', 15', 19', for pooled scores; and REST PRE, 1', 5', 10', 15', 19', REST POST, for ERSP. The relationship between subjective (pooled scores) and objective illusion (ERSP) measures during EEG2 was further analyzed through a Spearman's correlation for inter-participants and a repeated measures correlation for intra-participant consistency.
Lastly, to examine the progression of illusions across multiple LMV sessions, both intra- and inter-session changes were analyzed. For subjective pooled scores, a repeated measure ANOVA was used with a Time effect (1', 5', 10', 15', 19'), and a Friedman test for the Session effect (Sessions 1, 2, 3, 4, 5, 6, 7, 8, 9). For the ERSP variable, a Friedman test was used with Time (1', 5', 10', 15', 19') as a factor. The session effect (EEG1, EEG2) was further analyzed using a Wilcoxon test. Finally, the relationship between subjective (pooled scores) and objective illusion (ERSP) measures during the last LMV session (EEG2) was analyzed through a Spearman's correlation for inter-participants and repeated measures correlation for intra-participant consistency.
The acute effects of LMV were assessed using two-factor repeated measures ANOVA for normally distributed variables (RMT and M(H80)), with Visits (V1, V2) and PrePost (PRE, POST) as factors. For non-normally distributed data (SICI, MEP, Mmax, H80, and Grip strength), a Friedman test was performed with a single factor, Visits (V1, V1, V2, and V2). This analysis was followed by a Wilcoxon test to compare the relative changes between PRE and POST in V1 and V2, calculated as follows: V1 = (V1 - V1) / V1, and V2 = (V2 - V2) / V2.
To evaluate the chronic effects of LMV, a repeated measures ANOVA was used for normally distributed variables (RMT, M(H80), and Grip strength) across the factor Visits (V2, V3, V4). For data that were not normally distributed (SICI, MEP, Mmax, and H80), a Friedman test was used with the same factor.
For ANOVA tests, violations of the sphericity assumption were corrected using the Greenhouse-Geisser adjustment when necessary.
Effect sizes were calculated as follows: For ANOVA, partial Eta squared (ղ) was reported and interpreted as small (< 0.01), medium (< 0.06), and large (≥ 0.14). For the Friedman test, effect size was determined using Kendall's W (W), categorized as small (< 0.3), medium (< 0.5), and large (≥ 0.5). For the Wilcoxon test, the rank-biserial correlation (r) was computed, following the same interpretation as Kendall's W.
Post-hoc analyses were applied to all significant results: For ANOVA, pairwise comparisons were corrected using the Holm-Bonferroni method; For Friedman tests, Conover's post-hoc tests with Holm correction were performed. Exploratory correlations (Spearman) reported in the Supplementary Material (Tables A.3 and A.5) were also corrected using Holm-Bonferroni. Omnibus tests (ANOVA and Friedman) corresponding to our a priori hypotheses were reported without correction. Statistical significance was set at p ≤ 0.05 for all analyses.
For all data, normally distributed variables are reported in the text as mean ± standard deviation (SD), while non-normally distributed variables are reported as median with first and third quartiles [Q1 ; Q3].