High-current galvanic vestibular stimulation impairs working memory span, but not other executive functions

Patients with peripheral vestibular dysfunction (PVD) suffer not only from physical problems such as imbalance or vertigo but also from neuropsychological difficulties, including executive deficits. However, it is unclear whether the PVD directly causes executive problems. To examine the causal vestibular influence on executive functions, we induced either high-current (2 mA), low-current (0.8 mA), or sham current (0 mA) galvanic vestibular stimulation (GVS) in 79 healthy participants. Participants solved three tasks, measuring the core executive components (working memory, inhibition, cognitive flexibility) before and during GVS. High-current GVS impaired working memory span, but not inhibition and cognitive flexibility performance. Low-current GVS did not influence executive performance. Results indicate a causal vestibular influence on working memory span. Joint cortical areas of vestibular and working memory processing are discussed. Since high-current GVS in healthy participants serves as a model for an artificial vestibular dysfunction, our results could improve the diagnostics and therapy of patients with PVD.


Introduction
Patients with peripheral vestibular dysfunction (PVD) suffer not only from physical problems such as imbalance or vertigo but also from cognitive deficits. Patients with PVD have impaired visuo-spatial abilities, and recent research showed that they also have impaired executive functions (Agrawal et al., 2020;Hanes and McCollum, 2006;Moser et al., 2016Moser et al., , 2017Popp et al., 2017;Redfern et al., 2004). There are various hypotheses for the cognitive deficits in these patients. Some authors argue that there is a neuronal connection from the vestibular nerve to cortical areas involved in cognitive processing (Ferrè and Haggard, 2020;Popp et al., 2017). Alternative explanations for cognitive deficits in patients with PVD are mediating factors like affective disorders , plasticity processes of the neocortex or the vestibular nuclei after PVD (Lacour and Tighilet, 2010;Smith et al., 2005), or prioritized attentional resources on maintaining balance . Therefore, it is debated why patients with PVD show cognitive deficits and whether the cognitive deficits represent a direct consequence of the PVD.
One of the most frequently used techniques to investigate the direct causal influence of vestibular signals on cognitive functioning is galvanic vestibular stimulation (GVS) (Palla and Lenggenhager, 2014). This technique delivers transcutaneous current to the vestibular nerves via electrodes placed on the scalp covering the mastoid bones. The current simultaneously stimulates all peripheral vestibular afferents (Palla and Lenggenhager, 2014) and activates the vestibular cortex (Lopez et al., 2012). The exact mechanism between the applied current and the vestibular organs is still elusive. Animal studies suggest that hair cells as well as the vestibular afferent fibers are activated by GVS (Gensberger et al., 2016) and NMDA receptors in the vestibular nucleus are down-regulated (Kim et al., 2022). A high-current GVS protocol with a current of more than 1 mA induces a mild vestibular dysfunction in healthy participants, leading to behavioural effects comparable to those of patients with PVD and can thus be used as a patient model (Mac-Dougall et al., 2002Moore et al., 2006). High-current GVS (up to 5 mA) impairs cognitive functions, especially visuo-spatial cognition. While receiving high-current GVS, healthy participants were slower in completing an obstacle course  or solving egocentric mental transformations (Lenggenhager et al., 2007). Worse performance during high-current GVS was also observed in the form of increased error rates in egocentric mental transformations and a visuo-spatial short-term memory task (Dilda et al., 2012). However, high-current GVS did not impair other cognitive functions such as reaction time, dual tasking, object-based mental transformation, or visual-motor coordination (Dilda et al., 2012;Lenggenhager et al., 2007). Dilda et al. (2012) suggested that high-current GVS impairs specific cognitive functions relying on cortical areas that receive vestibular input. Interestingly, they also observed a small adverse effect of GVS on the Stroop test, measuring the executive component of inhibition. High-current GVS might therefore impair executive performance, but no study investigated the influence of high-current GVS on core executive components other than inhibition such as working memory or cognitive flexibility.
Contrary to high-current GVS, it has been shown that low-current GVS has the potential to improve cognitive functions. Cognitively impaired rats improved in a spatial memory task after repeated exposure to low-current noise-enhanced GVS (Ghahraman et al., 2016). Similarly, repeated low-current constant GVS accelerated the recovery of short-term and long-term spatial memory deficits in mice with uni-and bilateral PVD (Nguyen et al., 2021(Nguyen et al., , 2022. In studies with humans, low-current noise-enhanced GVS improved spatial learning (Hilliard et al., 2019) and visual memory recall (Wilkinson et al., 2008). Healthy participants receiving low-current constant GVS improved in visuo-spatial and inhibition performance (Dilda et al., 2012). However, as participants receiving low-current GVS were not compared to a sham group, cognitive improvements in this study rather represent a practice effect. But interestingly, patients with central neurodegenerative disorders improved in an inhibition task during low-current noise-enhanced GVS (Yamamoto et al., 2005). Low-current GVS might improve executive performance, but no study investigated the influence of low-current GVS on core executive components other than inhibition such as working memory or cognitive flexibility.
This study aimed to investigate the causal link between vestibular stimulation and performance in executive functions. We investigated whether a high-current GVS impairs, and a low-current GVS improves executive performance. Therefore, we induced either high-current GVS, low-current GVS or sham GVS in healthy participants and measured changed executive performance from baseline in the three core executive components (working memory, inhibition, cognitive flexibility).

Participants
Initially, 80 healthy participants participated in the study. The a priori power analysis was computed with a power of 90% and an alpha error of 5% in order to detect a middle effect. Expecting outliers due to reaction time analysis we tested 80 instead of the required 54 participants. One participant could not solve the tasks due to technical problems and had to be excluded. Therefore, we analysed data from 79 participants (62 female, 17 male) aged between 18 and 57 years (M = 25.28; SD = 7.40). In the working memory task, one participant had to be excluded due to missing data because of pressing a wrong key during the working memory task consistently. In the inhibition and cognitive flexibility task two participants had to be excluded due to early termination because of dizziness or technical problems during the inhibition and cognitive flexibility tasks. No participant reported a history of vestibular dysfunction and/or hearing loss. All participants gave written informed consent before participation. The study was approved by the ethics committee of the University of Bern (Switzerland).

Design and experimental procedure
We used a 2 × 3 mixed factorial design with the within-participant factor session (session 1, session 2) and the between-participant factor stimulation (high-current, low-current, sham). Participants had to solve the executive tasks twice: first without GVS as a baseline measure (session 1) and then while receiving GVS or sham (session 2). Between the sessions, participants filled in demographic data. The whole experiment lasted approximately 45 min. In session 2 participants were randomly assigned to one of three different stimulation conditions: highcurrent GVS (n = 26), low-current GVS (n = 27), or sham GVS (n = 26). The order of the executive tasks was randomized between participants, but stayed constant within a participant for both sessions. Dizziness sensations during stimulation in session 2 were assessed with a 10-point Likert scale from feeling no dizziness at all to feeling highly dizzy.

Galvanic vestibular stimulation
Galvanic vestibular stimulation was delivered by a bipolar batterydriven current stimulator (NeuroConn DC-Stimulator PLUS). Two circular rubber electrodes (3 cm diameter) were placed upon the participants' mastoid processes with conductive paste (Ten20). A bilateral bipolar stimulation with a sinusoidal signal was used with the amplitude depending on the stimulation condition: 2 mA for the high-current group or 0.8 mA for the low-current group. Since a current of 5 mA for the high-current group has been shown to elicit severe nausea in nearly 20% of participants (Dilda et al., 2011), we chose a current of only 2 mA for the high-current group. Participants in the sham group were stimulated for 15 s prior to task fulfilment with 1 mA accompanied by a fading-in and fading-out of 15 s each (Utz et al., 2010).

Executive tasks 2.4.1. Working memory
Verbal working memory performance was assessed with an auditive n-back task (BrainTwister, University of Bern). Sounds of letters (C, G, H, K, P, Q, T, W) were presented in a random order with replacement one after the other in a 1-s interval. Participants were instructed to listen carefully to the letters while looking at a white fixation cross on a black background, and to press a key if they heard a letter that was mentioned n trials before. The number of back positions, n, represents the working memory span. All participants started with 2-backs, meaning they had to press the key when the same letter was mentioned 2 trials before (see Fig. 1a for an example). Task difficulty was adaptive: A percentage of correct answers of 90% or higher led to an increased n in the next block (e.g. from 2-back to 3-back), an accuracy of 70% or lower, led to a decreased n (e.g. from 2-back to 1-back) and task difficulty remained the same in case of an accuracy between 70% and 90%. Participants performed four blocks with 20 + n stimuli and could reach a maximal n of 5. Participants received feedback about their performance and were informed about the n to be performed in the following block. For analysis, the number of participants who reached a maximal working memory span of 2-back, 3-back, 4-back or 5-back as a categorically distributed dependent variable was used.

Inhibition
Inhibition performance was assessed with a computerized form of the inhibition condition of the Color-Word Interference Test (D-KEFS, Delis et al., 2001; E-Prime 2 Software, Psychology Software Tools, Pittsburgh, PA, USA). A colour word printed in an incongruent colour appeared on the screen (e.g. the word 'blue' printed in red). Participants had to press the coloured key corresponding to the print colour while inhibiting reading the word (see Fig. 1b for stimuli examples). The task included ten practice trials with feedback and 80 experimental trials. Before performing the task participants solved 80 practice trials with coloured symbols (@@@@@) instead of words to familiarize with the coloured keys on the keyboard. Participants were instructed to respond as fast as possible without making any error. For analysis, number of errors and reaction times were used.

Cognitive flexibility
Cognitive flexibility performance was assessed with a computerized form of the cognitive flexibility condition of the Color-Word Interference Test (D-KEFS, Delis et al., 2001; E-Prime 2 Software, Psychology Software Tools, Pittsburgh, PA, USA). As in the inhibition condition, a colour word printed in an incongruent colour appeared on the screen (e.g. the word 'blue' printed in red). The word appeared normally ('blue') or with two underlines ('_blue_'). If the word appeared normally, participants had to press the coloured key corresponding to the print colour while inhibiting reading the word. If the word appeared with two underlines instead, they had to press the coloured key corresponding to the word's meaning while inhibiting the printed colour (see Fig. 1c for stimuli examples). Therefore, participants had to switch between two tasks (reading and colour naming). The task included ten practice trials with feedback and 80 experimental trials. Participants were instructed to respond as fast as possible without making any error. For analysis, number of errors and reaction times were used.

Data analysis
Statistical analyses were computed with IBM SPSS Statistics 25.0. First, the stimulation manipulation was checked by analysing whether dizziness sensation differed between stimulation conditions. A one-way analysis of variance (ANOVA) with stimulation condition as a between factor and the dizziness sensation rating as a dependent variable was conducted. Subsequently, Bonferroni corrected post hoc t-tests were calculated.
Second, we ensured that executive performance did not differ between stimulation groups in session 1. For this purpose, a Chi-square test for working memory performance and one-way ANOVAs (with stimulation condition as a between-participant factor) for inhibition and cognitive flexibility performance were calculated.
Third, it was analysed whether changed executive performance from session 1 to session 2 differed between stimulation groups. For working memory performance, a multinomial logistic regression, predicting the maximal working memory span, was computed. In addition, a Chisquare test of the distribution in session 2 was performed to analyse whether working memory span differed between groups in session 2. For post-hoc-analysis we looked at the corrected residuals of the distribution. For inhibition and cognitive flexibility performance, 2 (session) x 3 (stimulation condition) mixed-design ANOVAs with accuracy and mean reaction times as dependent variables were calculated. Outliers of reaction times were detected with the Median Absolute Deviation Method as this is the most robust dispersion measure in presence of outliers (Leys Fig . 1. Stimuli examples of the executive tasks to assess the core executive components: a) working memory (n-back task, example of a 2-back condition), b) inhibition (Color-Word Interference Test, inhibition condition), and c) cognitive flexibility (Color-Word interference Test, cognitive flexibility condition). See text for detailed instructions of the tasks. et al., 2013). Outliers were defined as values above or below the median plus or minus 2.5 times the Median Absolute Deviation (Leys et al., 2013).

Table 1
Corrected residuals from the frequency distribution of working memory span in session 2 (stimulation) as post-hoc-analysis for differences between the stimulation conditions (high-current, low-current, sham). Note. Significant corrected residuals (>1.96 or < − 1.96) are marked with an asterisk (*).

Table 2
Means of reaction times and error rates in the inhibition and cognitive flexibility task at session 1 (baseline) and session 2 (stimulation) in the different stimulation conditions (high-current, low-current, sham). Note. Standard errors are presented in brackets.

Discussion
We examined the causal vestibular influence on performance in core executive functions (working memory, inhibition, cognitive flexibility). We induced either high-current, low-current or sham GVS in 79 healthy participants and measured changed executive performance from baseline. High-current GVS impaired working memory span, but not inhibition or cognitive flexibility performance. Low-current GVS did not improve executive performance.

Impaired working memory span
The impaired working memory span due to high-current GVS is in line with a detrimental effect of high-current GVS on memory performance (Dilda et al., 2012). However, Dilda et al. (2012) used a short-term memory and no working memory task. Other studies investigated the effect of low-current GVS on visual or spatial memory, but without a high-current GVS protocol (Ghahraman et al., 2016;Hilliard et al., 2019;Wilkinson et al., 2008). Therefore, this is the first study investigating the effect of high-current GVS on working memory specifically.
One explanation for impaired working memory span due to highcurrent GVS might be joint cortical areas of ascending vestibular information and working memory processing. High-current GVS activates the inferior frontal gyrus (Lopez and Blanke, 2011;Mitsutake et al., 2021), an area involved in working memory processing (Ivanova et al., 2018;Liakakis et al., 2011;Nixon et al., 2004). Additionally, a working memory task comparable to our task activated the superior and middle temporal gyri (Ivanova et al., 2018), areas activated by GVS (Lopez and Blanke, 2011). The hippocampus might be another joint area of vestibular and working memory processing. Some patients with PVD exhibit hippocampal atrophy (Brandt et al., 2005;Göttlich et al., 2016;Hong et al., 2014;Kremmyda et al., 2016;Seo et al., 2016;zu Eulenburg et al., 2010) and vestibular lesions in animals increased the density of NMDA receptors in the hippocampus (Besnard et al., 2012). The hippocampus is involved in working memory processing (Poch and Campo, 2012;Ranganath and D'Esposito, 2001), probably also via a hippocampal-prefrontal pathway (Tierney et al., 2004).
We can exclude two alternative explanations for the impaired working memory performance in the high-current group. First, compared to the low-current or sham group, the high-current group was not the group with the worst working memory abilities, because baseline working memory performance was comparable between groups. Second, worse working memory span in the high-current group is not a consequence of side effects of the stimulation, because high-current GVS did not disrupt inhibition or cognitive flexibility performance.

No impaired inhibition and cognitive flexibility performance
High-current GVS impaired working memory, but not inhibition and cognitive flexibility performance. This finding is consistent with previous results of specific rather than general cognitive impairments due to high-current GVS (Dilda et al., 2012;Lenggenhager et al., 2007). However, a small detrimental effect of high-current GVS on inhibition performance has previously been observed (Dilda et al., 2012). Conflicting results might be explained by different stimulation protocols. Participants in the study by Dilda et al. (2012) received a high-current of 3.5 or 5 mA, while we used only 2 mA. Another finding fostering this assumption is that high-current GVS of 1 mA did not reduce inhibition performance in a Flanker's task when using neutral stimuli (Blini et al., 2020). Therefore, a detrimental effect of high-current GVS on inhibition performance might be subtle and requires a stronger current to unfold. Future studies are needed to replicate whether high-current GVS specifically impairs the working memory component of executive functions.
If high-current GVS disrupts working memory performance specifically, but not inhibition or cognitive flexibility, the maintenance component in working memory tasks can possibly explain this effect. Contrary to inhibition or cognitive flexibility, working memory includes a maintenance component (Shelton et al., 2010). To maintain information in working memory a phonological code is necessary that has to be actively repeated (Baddeley and Hitch, 2019). A crucial brain region for order information of the phonological code is the left supramarginal gyrus (Papagno et al., 2017). Interestingly, this region has been shown to be important for vestibular processing in healthy subjects as well as in patients with PVD (Hüfner et al., 2009;Klaus et al., 2020;Schöne et al., 2022). The supramarginal gyrus is also activated by GVS (Mitsutake et al., 2021). Another brain region that is necessary for phonologically based working memory processing is the inferior frontal gyrus (Nixon et al., 2004), that was described above as joint cortical area of working memory and vestibular processing. Therefore, high-current GVS might disrupt the ascending information from the vestibular nerve to the left supramarginal gyrus and the inferior frontal gyrus leading to a disruption in the phonological code and therefore a specific impairment in working memory.
It is unlikely that a selective working memory impairment is due to the auditory modality used in our study because other studies observed detrimental effects of GVS also in the visual modality (Dilda et al., 2012;Lenggenhager et al., 2007). Moreover, although working memory includes two cognitive components (maintaining and retrieving), there is no reason to assume that it is more demanding than other core domains of executive functions. For example, cognitive flexibility also includes two cognitive components (inhibiting and switching).

No effect of low-current GVS
Low-current GVS did not improve executive performance. In a previous study improved inhibition performance after low-current GVS was interpreted as a practice effect rather than the influence of the stimulation (Dilda et al., 2012). Our results confirm this interpretation while additionally controlling for a practice effect by integrating a sham group in our design. However, our results conflict with a previous study showing that low-current GVS improved inhibition performance in patients with central neurodegenerative disorders (Yamamoto et al., 2005). Contrary to our study, Yamamoto et al. (2005) used a noise-enhanced GVS protocol, and this could be a potential origin of the diverging results. They explain the beneficial effects on inhibition performance by stochastic resonance: a phenomenon in which a random process (noise) added to a subthreshold stimulus can enhance sensory information processing and perception (for a detailed description see Moss, 2004). Compared to conventional GVS, noisy GVS increases activity in the insula, putamen and central operculum (Mitsutake et al., 2021). Future studies should use noise-enhanced subthreshold GVS to examine whether this stimulation condition improves executive performance.

Strengths and limitations
This study has several strengths. It is the first study investigating the causal vestibular influence on the core domains of executive functions. We integrated a sham group to exclude practice effects. Additionally, using a within-design and baseline measurement we could exclude that results were affected by group differences in cognitive abilities between stimulation conditions. One limitation of our study is that we assessed different executive components with tasks covering different modalities and dependent variables. While working memory was assessed in the auditory modality with accuracy being the dependent variable, inhibition and cognitive flexibility were assessed in the visual modality with accuracy and reaction times being the dependent variables. However, as described above, other studies showed detrimental effects of GVS also in the visual modality, and divergent results by measuring working memory in the visual domain are unlikely. Our null findings of high-current GVS on inhibition and cognitive flexibility performance are based on a power analysis to detect a middle effect size. We suggest replicating our findings with a bigger sample size to exclude a small detrimental effect of high-current GVS also on inhibition or cognitive flexibility performance.

Implications
Our results showed that a reduced working memory span is a direct consequence of high-current GVS. Using the artificially induced vestibular dysfunction by high-current GVS as a model for a patient with PVD, our results have clinical implications. That high-current GVS impaired working memory performance is consistent with observed impaired working memory performance in patients with PVD (Hanes and McCollum, 2006;Moser et al., 2016Moser et al., , 2017Popp et al., 2017). If a reduced working memory span is a direct consequence of vestibular dysfunction, patients with PVD should be screened for working memory problems. A screening test for working memory problems could be integrated in the standard neurootological assessment these patients undergo. Possibly, patients with PVD and working memory difficulties could benefit from a working memory training.

Conclusion
High-current galvanic vestibular stimulation impaired working memory span in healthy participants. Our results indicate a causal vestibular influence on working memory span. This could improve the diagnostics and therapy of patients with peripheral vestibular dysfunction.

Declaration of competing interest
None.

Data availability
Data will be made available on request.