Standardizing Stimulus Parameters for Noisy Galvanic Vestibular Stimulation

Abstract

Noisy galvanic vestibular stimulation (nGVS) delivers band-limited subsensory electrical noise at the mastoids to amplify vestibular signaling via stochastic resonance. This review synthesizes human studies involving healthy participants and proposes a practical stimulation protocol and reporting checklist to promote standardization of nGVS in vestibular research and clinical translation. Evidence supports the use of a bilateral bipolar mastoid montage; 1.75–10 cm² electrodes; zero-mean Gaussian white noise with a 0–30 Hz passband (or 0–640 Hz when replicating prior work or testing cortical or perceptual hypotheses); participant-titrated subsensory intensities (typically 100–400 μA, with fixed protocols often 400–1,000 μA); and task-matched stimulation epochs of 10–30 seconds. Reports should specify the electrode polarity, passband and filter details, amplitude reference (peak, peak-to-peak, or root-mean-square), threshold procedures, and total stimulation dose. Under these conditions, nGVS has been shown to improve vestibulospinal outcomes (reduced postural sway and better gait stability), modulate vestibulo-ocular physiology, and lower vestibular motion detection thresholds. The effects scale with task difficulty and baseline vestibular function; tolerability was favorable at the subsensory level. Priorities for future research include head-to-head tests of montage and frequency, individualized dose–response and after-effect mapping, stratification by baseline vestibular status, and mechanistic biomarkers tied to behavior. Standardized parameters and reporting will enable robust meta-analyses and faster clinical translation.

Introduction

The observation that a small electrical current applied near the ears can directly activate the peripheral vestibular system led to the development of galvanic vestibular stimulation (GVS). Such activation produces measurable physiological and behavioral responses, including nystagmus, postural sway, and involuntary head or body movements, demonstrating that electrical currents can modulate vestibular afferent activity. Historical accounts trace these effects to Alessandro Volta’s battery experiments and to Johannes Purkinje’s reports of imbalance and vertigo with transcranial currents [1–4]. Building on Josef Breuer’s inference of vestibular involvement [5], Goldberg and colleagues [6,7] provided definitive animal evidence with electrodes placed in the perilymphatic and middle ear spaces of squirrel monkeys, GVS altered vestibular afferent firing in the absence of mechanical head motion, with larger currents producing larger deviations from spontaneous discharge. Cathodal currents delivered to the perilymphatic space depolarized and excited afferents often doubling discharge rates, whereas anodal currents hyperpolarized and inhibited them. Notably, irregular afferents were more sensitive than regular afferents. Consistent with these findings, Kim and Curthoys [8] recorded guinea-pig vestibular ganglion activity and showed polarity-dependent, asymmetrical responses for irregular units using both implanted middle ear and surface mastoid electrodes.

Human studies soon confirmed that bilateral mastoid GVS influences perception and balance control, eliciting illusions of motion, postural sway during stance, and trajectory deviations during gait [9–13]. More recently, GVS has been explored for diagnostic and rehabilitative applications across peripheral and central vestibular disorders (e.g., [14]). A key advance was the use of subthreshold electrical noise to exploit stochastic response, thereby lowering the current needed to elicit vestibular effects [15–18]. This “noisy” GVS (nGVS) enables blinding because participants cannot reliably detect stimulation, facilitating rigorous sham-controlled design [19–23]. Because nGVS is typically subsensory, it is also better tolerated, with fewer side effects during prolonged or continuous delivery [18,24–26]. Early demonstrations used random-noise GVS to drive frequency-specific sway and muscle activity [27,28]. Inspired by stochastic response related benefits of mechanical noise on stance [15] and autonomic effects of electrical noise [29], Yamamoto, et al. [18] extended nGVS exposure in neurodegenerative conditions, reporting faster transitions from rest to activity, consistent with meaningful motor effects.

Contemporary studies indicate that nGVS can benefit both impaired and healthy vestibular systems [26,30–33], yet results vary widely. Differences in current amplitude, frequency bandwidth, duration, electrode size and placement, and intensity normalization (e.g., peak vs. root mean square [RMS], absolute μA vs. % threshold) hamper direct comparison and reproducibility. As emphasized in a recent parametric review focused on vestibular impaired populations [34], many reports lack sufficient detail to enable replication. This review synthesizes nGVS stimulation parameters used in healthy participants. We consolidate current practice, map parameter choices to vestibular behavioral and physiological outcomes, and propose a standardized protocol to improve reproducibility and enable cross-study comparisons in future nGVS research.

Stochastic Resonance in nGVS

Before detailing stimulation parameters, we first explain why adding noise via stochastic resonance can enhance vestibular responses and outline the mechanisms underlying nGVS. Stochastic resonance is a statistical property of nonlinear systems whereby the addition of noise can amplify a subthreshold signal, yielding a higher output signal-to-noise ratio than at the input when noise is turned to an optimal level [35]. With band-limited Gaussian noise, a weak stimulus that would otherwise remain undetectable can cross threshold and become perceptible; if the noise level exceeds the optimum, the signal is masked and the signal-to-noise ratio deteriorates [36]. Fig. 1 illustrates this canonical “inverted-U” relationship, which has been demonstrated in physical, chemical, and biological systems, including animals and humans [37].

Fig. 1.

Stochastic resonance occurs when adding noise enables detection of a subthreshold vestibular signal. Each panel shows a sinusoidal input, added zero-mean noise on top of the input, a firing threshold (dashed line), and the resulting output spikes (top trace). A: No noise. The input remains below the vestibular system threshold and no spikes are generated. B: Optimal noise. An optimal amount of subthreshold noise intermittently raises the summed signal above threshold near the sinusoid peaks, yielding time-locked spikes and improved detectability. C: High noise. Large noise causes frequent, non-phase-locked crossings that mask the signal and degrade detectability, which may provoke vestibular symptoms. Amplitudes are in arbitrary units and not to scale.

Biological demonstrations of stochastic resonance span multiple sensory domains. In crayfish mechanoreceptors, adding noise to a weak stimulus increased single-neuron firing rates [38]. Within the ear, white noise added to vowel stimuli enhanced formant clarity in sciatic nerve recordings from toads [39]. In the vestibular periphery, Indresano, et al. [40] observed stochastic resonance in the bullfrog saccule, where low-level noise paired with near threshold stimuli improved neural signal-to-noise ratio. They further proposed a hair cell contribution to this benefit, noting that adding Cd²⁺, which delayed postmortem hair cell decline, caused slower degradation of stochastic resonance-related improvements in neural signal-to-noise ratio [40]. Together, these findings establish that stochastic resonance is present in biological systems and, specifically, within the inner ear.

Human studies similarly support stochastic resonance. Collins, et al. [41] showed that increasing levels of mechanical or electrical noise improved stimulus detection, and that subthreshold mechanical or electrical noise applied to the soles of feet or knees reduced quiet standing sway [15,41,42]. In the vestibular domain, adding noise to GVS lowered the current needed to induce body motion in approximately 90% of participants [17]. Not all datasets, however, exhibit the classic bell-shaped response across noise intensities, as Assländer, et al. [43] reported participant-specific “optimal” intensities without a consistent bell curve at the group level. In contrast, a more recent study found stochastic resonance like behavior for white noise nGVS in 55% of participants, with fewer stochastic resonance consistent responses to pink noise [44]. Despite mixed patterns across studies, the preponderance of evidence indicates that stochastic resonance can, at least in part, account for nGVS-related reductions in vestibular thresholds (e.g., [17]) (Fig. 1).

Traditional GVS has long been used to probe vestibular function with minimal confounds from other senses [45]. Leveraging stochastic resonance, nGVS extends this approach by using subsensory noise to enhance vestibular signal processing in both healthy individuals and those with impairment [25,46]. Importantly, nGVS is not universally effective as in bilateral vestibulopathy, vestibulospinal thresholds were lowered only in patients with residual vestibular function [47], consistent with the requirement, implicit in stochastic resonance, that a functional transduction to afferent pathway remain intact [30]. At the neural level, irregular vestibular afferents are more sensitive to galvanic currents than regular units, providing a plausible substrate for noise-assisted detection [6–8]. In humans, coherence between nGVS and vestibulo-muscular activity is strongest within roughly 2–20 Hz, aligning with natural head-motion dynamics [48]. Together, with stochastic resonance theory, these observations support a 0–30 Hz passband as a vestibulo-peripheral default for posture, gait, and reflex outcomes, while a broader 0–640 Hz band remains reasonable when replicating prior protocols or proving cortical or perceptual hypotheses that may benefit from wider spectral drive. Thus, stochastic resonance provides a principled framework for selecting nGVS parameters (noise type, bandwidth, intensity normalization) and for interpreting heterogeneous outcomes across studies. In subsequent sections, we use this framework to organize the literature and propose parameter standards that maximize the likelihood of operating near the stochastic resonance optimum.

Signal Characteristics and Stimulus Parameters

We summarize recent nGVS parameters utilized for healthy individuals in Supplementary Table 1 (in the online-only Data Supplement) regarding electrode size and montage, stimulus type, frequency band, intensity, and duration. Below, we synthesize design choices and offer pragmatic recommendations for future work. For clarity, we use the following terms in this review. Polarity refers to the anode and cathode assignment. Current density is the absolute current divided by pad area and reported in mA/cm². Passband is the frequency range of the noise signal. Amplitude reference indicates whether values are reported as peak, peak to peak, or RMS.

Electrode configuration

Electrode configuration is the spatial arrangement and polarity of stimulating and return electrodes. Across nGVS studies, the predominant montage is bilateral bipolar, with the anode on one mastoid and the cathode on the other, which drives current across both labyrinths [26,31,34,45]. Bilateral monopolar implementations also exist, placing same polarity electrodes on both mastoids (e.g., two cathodes) with a single return electrode at a remote site (e.g., forehead); current then flows from each mastoid to the common return and engages both labyrinths [29,49]. In classical GVS, comparisons show that lateral sway components sum under bilateral bipolar stimulation, whereas anteroposterior components tend to cancel [50]. Direct montage comparisons are scarce in nGVS, and several reports similarly find stronger mediolateral than anteroposterior effects [28,43]. Most nGVS papers do not specify which side is anodal versus cathodal, and when reported, lateralization effects are inconsistent [16,44]. Bilateral bipolar should be considered the default montage, and authors should state polarity explicitly, for example, “left mastoid=anode, right mastoid=cathode” (or the reverse).

Electrode size

Electrode size represents the effective skin-electrode contact area through which current is delivered. Electrode size is inconsistently reported. When reported, sizes range from 0.5 to 51 cm². Very small pads (≈0.5 cm²) showed null effects in one study [51], yet pads between 1.75 and 10 cm² frequently yielded significant effects (e.g., 1.75, 3, 5, 9, 10 cm²) [16,20,43, 48,52–56]. Larger pads (≥15 cm²) produced mixed outcomes; critically, a direct comparison showed 3 cm² effective versus 35 cm² ineffective under otherwise similar conditions, plausibly due to higher current density and reduced shunting with smaller pads [57,58]. Consistent with this interpretation, computational modeling predicts that smaller pads yield greater current flow to the target because less current is lost across the skin, helping to explain why significant nGVS effects were observed only with the smaller electrode size [59]. For mastoid nGVS, 1.75–10 cm² electrodes in a bilateral bipolar montage are a practical default, and authors should report pad size and shape explicitly.

Stimulus type

Stimulus type describes the waveform and its statistical properties. In traditional GVS, step currents are interpreted as a proxy for continuous head motion, with cathodal stimulation producing movement toward the ipsilateral side [11, 30,45,60,61]. Sinusoidally modulated currents have been used to approximate head rotations in animals and humans [6,62–65]. In nGVS, the prevailing choice is band-limited, zero-mean Gaussian white noise [17,21,22,66–68]. White denotes a flat spectrum within the reported passband (e.g., 0–30 Hz), Gaussian denotes normally distributed amplitudes, and zero-mean denotes no DC offset. Pink noise (1/f) emphasizes lower frequencies and delivers roughly equal power per octave; although theoretically appealing [29], evidence in healthy cohorts is limited or mixed [44,49]. Unless a specific hypothesis dictates otherwise, use zero-mean Gaussian white noise and report the noise statistics (distribution, mean ≈0) and passband.

Frequency band

Frequency band specifies the passband over which the stimulus has non-negligible power. Reported bands span 0–2 to 0–640 Hz. Across the corpus, the most prevalent choices were 0–640 Hz (11 studies; 20%), 0–30 Hz (9; 17%), 0–2 Hz (7; 13%), and 0–10 Hz (6; 11%) (Fig. 2). The most common and most replicated across independent groups are 0–640 Hz and 0–30 Hz, selected to cover vestibular-relevant motion. These frequency bands generally improve balance and gait outcomes [17,19,44,68–71]. Very low bands (0–2, 0–10 Hz) also show benefits [21,43,66,67]. Wider bands up to 640 Hz can be effective but are mixed and concentrated within a few labs [20,52–58,72–76]. One study mapped EEG-inspired sub-bands and suggested mid-frequency benefits for reaction time, though this is not vestibular-mechanistic and remains unreplicated [77]. Coherence between nGVS and vestibulo-muscular activity has been observed within 2–20 Hz [48].

Fig. 2.

Frequency bands reported for noisy galvanic vestibular stimulation (nGVS) in human studies. The horizontal lollipop chart shows, for each stated passband, the number of studies that used it; percentages in parentheses refer to the proportion of all included studies. Bands are listed from low to high. “Unknown” indicates that a passband was not reported [85,87]. “Centered at 1,000 Hz” indicates a narrowband stimulus around 1 kHz [16].

Groups selecting 0–2 Hz reported significant effects on vestibular physiological measures including ocular counter-roll gain, baroreflex measures, postural sway, and vestibular motion thresholds, arguing that this range captures natural head-movement dynamics and possible long-range neuronal correlations [21,29,66,67]. Studies using 0–2 Hz reported effects on ocular counter-roll gain, baroreflex measures, postural sway, and vestibular motion thresholds [21,29,49,66,67,78]. For 0–10 Hz, authors aimed to encompass motion signals plausibly encoded by the vestibular periphery and reported improved sway, walking performance, postural stability, and increased ocular vestibular evoked myogenic potential (oVEMP) amplitudes [43,79–83]. Studies using 0–20 Hz often did not state a rationale and showed mixed findings—no change in sway speed on foam in one report versus reduced sway speed on foam (but not solid ground) in another [51,84]. The widely adopted 0–30 Hz band was chosen by several groups to cover the peripheral vestibular frequency range and has been associated with improvements in stabilization, balance, walking stability, gait speed, adaptation, and coordination; only one study reported no significant effects versus control [17,19,22, 44,68–71]. Inukai, et al. [52], selected 0–640 Hz by extrapolating from transcranial electrical stimulation reports showing of high-frequency random noise in a similar range [72]. Subsequent nGVS studies adopted this precedent; however, outcomes across laboratories have been mixed overall [20,52–55, 57,58,73–76]. Together, to standardize while respecting current practice, we propose two prespecified bands: 1) 0–30 Hz as the vestibulo-peripheral default for posture/gait and reflex endpoints, and 2) 0–640 Hz as a broadband option when replicating prior protocols or testing cortical/perceptual hypotheses. Authors should justify their choice and report the exact frequency band and filter characteristics (type/order), along with sampling rate and any ramping used.

Duration

Duration describes the length of each stimulation epoch and the cumulative dose across a session. Reported nGVS durations span 5–30-second epochs for within-task effects to minutes–hours for carryover or therapeutic aims. In practice, 10–30-second blocks are common, and longer continuous stimulations (≥30 min, up to several hours) have generally been well tolerated in healthy and older adults [18,80]. The rationale for duration is rarely stated. One explicit test showed that 5-second epochs produced sway reductions comparable to the more typical 30-second epochs [52]. Consistent short-epoch effects have also been reported at 8.5, 10, 12, and 15 seconds [17,19,51,68,71,82], and at 20–26 seconds for reaction time, balance, and sway speed outcomes [69,76,83]. The 30-second block remains the most frequently used single-epoch duration [20,21,43,44,52–55,66,67,70,79,85], with longer applications (40 sec to 40 min) also reported [29,48,56–58,73–76,78,80,85]. Duration should match task demands; for general laboratory paradigms, 30-second blocks provide a reproducible default. Reports should specify on/off timing, inter-block intervals, and total dose.

Intensity

Intensity indicates the stimulus amplitude and how it is normalized across participants. Reported amplitudes span 50–1,750 μA. Studies set intensity as 1) a fixed absolute current, 2) a percentage of an individual threshold, and/or 3) participant-specific titration to an optimal response. Thresholds were operationalized as follows: cutaneous=first perception of stimulation or mild tingling at the electrodes; motion/perceptual=first perceived head motion, typically determined with a separate traditional GVS sinusoid (approximately 1 Hz) rather than noise; and nociceptive/pain=used rarely (one study at 90% of a “nociceptive threshold” while participants remained unaware of stimulation). Because authors’ nomenclature varies, reports should state how thresholds were measured and verified, and whether amplitudes are peak, peak-to-peak, or RMS. The prevalence of threshold-based protocols is summarized in Fig. 3A (most commonly 80% and 90% cutaneous), and absolute-current protocols in Fig. 3B (frequent choices around 400 μA, ≤500–700 μA, and around 1,000 μA; asterisks denote unspecified peak/peak-to-peak/RMS).

Fig. 3.

Intensity choices in noisy galvanic vestibular stimulation (nGVS) stimulation studies. A: Studies that set intensity as a percentage of an individual threshold (n=15). Threshold types include cutaneous (first perception/tingling), motion/perceptual (first perceived head-motion in a separate threshold procedure), and nocipeptive/pain as reported by the original authors. B: Studies using absolute current levels (μA; n=36). Bars labeled “≤” denote studies that tested multiple levels and are tallied by the highest level in the sweep; many of these analyzed a participant-specific optimum. Unless otherwise noted, amplitudes are reported as peak; the “≤300 μA” bin includes one study reported in root mean square (RMS); “*” studies did not specify peak, peak-to-peak, or RMS. Parenthetical percentages indicate the proportion of studies within each panel.

Across titration and fixed-level protocols, both 400 and 1,000 μA acutely reduced sway path length [48], whereas 200 μA outperformed 400 μA on several balance metrics in a separate cohort [53]. Overall, participant-specific optima commonly cluster around 100–300 μA for example, means near 127 μA, 179 μA, and 220 μA [67,80–83], with typical optima near 75%–90% of cutaneous threshold [43,81]. Notably, amplitude–response functions are not uniformly bell-shaped across intensities [43]. Keywan, et al. [21] likewise reported optima of 50–300 μA (mean 135 μA).

When intensities are expressed as a percentage of threshold, about 80% cutaneous has been linked to improvements in coordination, sway on foam, and slow-walking stability [17,68,84,86], while about 90% has been used in reaction-time paradigms with mixed results [16,77,78]. Using 80% cutaneous and 50% motion thresholds produced similar gains in gait/stability [71]; motion thresholds were often derived with traditional GVS when nGVS direction was not perceptible [19]. A structured sweep from 50%–110% of cutaneous threshold yielded stochastic resonance-like optima in many participants (mean optimum approximately 281 μA≈80%) [44].

Fixed single levels have also shown benefits at 400–1,000 μA [57,76] and 1,750 μA for reducing cybersickness [56], though 500 μA yields mixed or null findings in some balance tasks [73]. Very high, suprathreshold currents (e.g., 3,000 μA) drive robust muscle activity but are not subsensory [48]. Task-specific effects also occur: 200 μA reduced vestibulo-ocular reflex (VOR) gain, whereas 600 μA increased sway in another protocol [75].

Taken together, when feasible, titrate per participant to a subsensory optimum using objective outcomes (e.g., sway metrics). When individual titration is impractical, 100–400 μA (or 80%–90% of cutaneous threshold) is a defensible default, provided manuscripts clearly specify the amplitude reference (peak/peak-to-peak/RMS), the threshold type and procedure, and whether stimulation was subsensory.

Parameter interactions

Effective stimulation at the labyrinth is jointly determined by the delivered current (μA), the effective pad area (cm²) that sets current density (μA/cm²), and the current path set by montage and polarity, and these factors interact with the stimulus spectrum (passband). Identical currents can therefore have very different effects: 200 μA over 3 cm² is up to 67 μA/cm², whereas 200 μA over 35 cm² is up to 6 μA/cm². To enable interpretation and replication, report absolute current and electrode area, and clearly report montage, polarity, and passband. When mapping dose–response relationships, keep montage and passband constant and vary one parameter at a time. During titration, confirm efficacy with an objective readout such as center-of-pressure sway metrics, video head-impulse test gain, or motion thresholds.

Responders vs. non-responders

As a practical guidance, larger nGVS benefits are reported in cohorts with greater postural challenge or reduced vestibular function such as older adults and individuals with bilateral vestibulopathy, whereas low-sway young adults often show smaller or null effects [43,51,71,76,80,81,84]. Cognitive status can also moderate outcomes, with lower baseline spatial working memory predicting larger gains [86]. Residual vestibular function appears necessary: nGVS lowered vestibulospinal thresholds only in patients with remaining function [47]. A pragmatic workflow is to first screen with a brief, low-amplitude traditional GVS to confirm residual function and directionality [74,75], then titrate subsensory nGVS using an objective sway metric (path or velocity) to identify an individual optimum [19,21,44]. Document baseline status with video head impulse test (vHIT) and VEMPs and add patient-reported measures such as the Dizziness Handicap Inventory (DHI) and Activities-specific Balance Confidence (ABC) scale to enable stratified analyses.

Physiological and Functional Effects in Healthy Participants

Outcome domains in nGVS studies span posture/balance, locomotion, vestibular physiology, perception, and select cognitive or autonomic measures; EEG-only studies are beyond the scope here. The most common endpoint is postural sway, with many reports showing reduced sway and trunk motion when nGVS is delivered with suitable parameters [19–21,43, 44,49,51–55,57,71,73,76,79–81,84,85]. Beyond quiet stance, nGVS has been associated with decreased walking deviation [78], shorter sway path/velocity [20,51,81,84], higher gait velocity [71,83], greater walking stability [70], and faster functional mobility [22].

Effects depend on task context and baseline status. Performance gains tend to be smaller during more demanding conditions (e.g., on foam or with cognitive load) [51,54]. Individual differences also matter: participants with lower baseline spatial working memory showed larger benefits [86], and older adults exhibited increased ocular counter-roll gain whereas younger adults did not [78].

Physiological readouts indicate that nGVS can modulate vestibular reflexes: low-level nGVS reduced VOR gain [75], while nGVS increased oVEMP amplitudes without latency shifts [82]. Additional reported effects include changes in muscle activity [51,74], improved spatial learning [86], enhanced vestibular motion perception [21,67], shorter reaction times [16,77], baroreflex modulation [29], and reduced cybersickness during intense virtual reality (VR) [56]. Collectively, these findings suggest nGVS can yield measurable functional benefits in healthy participants, moderated by parameter choices, task demands, and individual baseline function.

For ENT-relevant clinical trials, we recommend standardized endpoints that include computerized dynamic posturography (CDP) (Sensory Organization Test and Motor Control Test), instrumented gait speed and variability and Timed Up and Go (TUG), vHIT gain with covert/overt saccades, ocular and cervical VEMP (oVEMP/cVEMP) amplitudes and thresholds, ocular counter-roll gain, vestibular motion thresholds, and patient-reported outcomes such as the DHI and ABC. nGVS can be layered onto conventional vestibular therapy (e.g., gaze-stabilization exercises, balance/gait training) and VR-based paradigms. We recommend delivering subsensory nGVS during task practice, documenting task difficulty, intensity, and outcomes (CDP, gait metrics, vHIT/VEMPs, DHI/ABC) to assess additive benefits while minimizing ceiling/floor effects.

Clinical Translation and Implications for Clinical Practice

Although the present synthesis is restricted to healthy participants, several near-term clinical scenarios are plausible. First, nGVS can be used as an adjunct during vestibular rehabilitation to modestly reduce sway and improve task performance while patients practice stance and gait; a bilateral bipolar mastoid montage with 2–10 cm² electrodes, zero-mean Gaussian white noise at 0–30 Hz, and a participant-titrated subsensory intensity (100–300 μA) offers a practical starting point. Second, during peri-diagnostic assessments (e.g., posturography, gait evaluation), brief 10–30-second epochs may help reveal latent performance gains that inform prognosis and therapy planning. Third, in older adults at risk of falls, supervised nGVS during balance training may enhance stability and confidence; document baseline performance, titrate below cutaneous threshold, and monitor tolerability. Fourth, for motion intolerance and VR use, subsensory nGVS has shown reduced cybersickness in healthy users and could be explored in carefully screened patients. For suspected unilateral or bilateral vestibular hypofunction, apply extra caution: benefit likely depends on residual function. Consider documenting vestibular responses (e.g., VOR, VEMPs) before use, start at lower currents, and treat any observed effects as hypothesis-generating pending disease-specific trials. Consistent reporting of polarity, pad size, passband and filter details, amplitude reference, thresholding procedures, timing, and total dose will facilitate replication and accelerate clinical learning.

Limitations and Challenges

Key issues concern tolerability reporting, heterogeneity of methods and outcomes, and inter-individual variability. Though, GVS is typically well tolerated, explicit safety data remain sparse: some studies report no pain or adverse events even during prolonged stimulation [80], while others document rare, mild symptoms (e.g., metallic taste or tingling during suprathreshold thresholding [19]) or low average ratings of transient nausea/vertigo [22]. Many reports do not systematically query or list side effects [77,82,83]. To avoid under-reporting, studies should include a brief standardized side-effect checklist administered pre/during/post-stimulation.

A second constraint is methodological heterogeneity. Variability in montage, electrode size, noise statistics, passbands, intensity metrics (peak/peak-to-peak/RMS), epoch timing, and outcome measures limits direct comparison and replication. This review proposes standardized reporting (polarity, pad size/shape, noise distribution and mean, passband/filter details, amplitude reference, threshold procedure, on/off timing, total dose) to improve comparability.

A parameter-wise meta-analysis linking effect sizes to montage, passband, intensity, duration, and current density is beyond the scope of this review, and current heterogeneity, including inconsistent amplitude references, threshold definitions, and dosing, limits robust cross study modeling. We therefore present descriptive prevalence summaries and mechanistic rationales, and we encourage future studies to adopt consistent units and definitions so that formal analyses of parameter effects become feasible.

Finally, individual variability remains substantial: despite group-level benefits, non-responders are common [19,22,84]. Some groups pre-screened with traditional GVS to verify sway toward the anodal side [74,75]. Practically, participant-specific titration to a subsensory optimum, clear documentation of baseline function, and task designs that avoid ceiling/floor effects can reduce variability and clarify who benefits most from nGVS.

Safety, Tolerability, and Participant Screening

Subsensory nGVS is generally well tolerated when current density is low (about 100–400 μA through 2–10 cm², roughly 0.01–0.20 mA/cm²), and any symptoms are typically mild and transient, appearing mainly during suprathreshold procedures [19,23]. In practice, screen participants for higher-risk conditions such as active skin disease at electrode sites, implanted cardiac or neurostimulators, poorly controlled epilepsy, pregnancy as a precaution, and acute vestibular crisis or severe migraine on the test day. Before each session, inspect, clean, and dry the skin; establish the cutaneous threshold; confirm a subsensory intensity (for example, 80%–90% of that threshold or a participant titrated optimum) and record electrode size and area to report current density, along with the montage and polarity. During stimulation, pause or stop if nausea or vertigo exceeds 2 on a 5-point scale, if metallic taste or tingling persists, or if headache or skin irritation develops, then reassess at a lower intensity. Document any adverse events with a short, standardized checklist administered before, during, and after stimulation.

Conclusion

nGVS offers a principled, subsensory means to boost vestibular signaling via stochastic resonance, with reproducible benefits across posture, gait, vestibular reflexes, perception, and select cognitive outcomes in healthy participants. Yet heterogeneous methods, including montage, electrode size, noise statistics, passbands, intensity metrics, and dosing, continue to limit comparability and slow translation. Based on the literature reviewed here, we propose a practical baseline protocol: bilateral bipolar mastoid montage; 1.75–10 cm² electrodes with size/shape reported; zero-mean Gaussian white noise with the distribution and mean stated, using either a 0–30 Hz passband as the vestibulo-peripheral default for posture and reflex endpoints or a 0–640 Hz broadband option when replicating prior work or testing cortical/perceptual hypotheses; participant-specific titration to a subsensory optimum (typically 100–400 μA or 80%–90% of cutaneous threshold), with amplitude clearly referenced (peak/peak-to-peak/RMS), thresholds and procedures documented; and task-matched epochs (commonly ~30 s), with on/off timing and total dose reported. Studies should also adopt simple, standardized side-effect checklists to avoid underreporting. Priority gaps include adequately powered, preregistered randomized controlled trials in older adults and in patients with neurological comorbidities, tests of the durability of aftereffects, and real-world fall-risk outcomes, alongside mechanistic biomarkers that link vestibular physiology to behavior. Harmonized, preregistered, and adequately powered trials, paired with open data and code, will enable robust meta-analysis and accelerate clinical translation.

Supplementary Materials

The online-only Data Supplement is available with this article at https://doi.org/10.7874/jao.2025.00514.

Supplementary Table 1.

Summary of recent nGVS studies in human participants