Transcutaneous auricular vagus nerve stimulation reduces error amplitude during force field learning in healthy adults Authors: Hiroki Obata, Hirofumi Sekiguchi Published: 2025 ABSTRACT AbstractTranscutaneous auricular vagus nerve stimulation (taVNS) has gained increasing attention as an expected new clinical application for motor disease due to its synergistic effect with post-stroke rehabilitation. On the other hand, there are still few reports regarding the effects of taVNS on motor learning in healthy subjects. In the present study, we investigated how taVNS affects a reaching movement task under a force field condition (i.e., force field learning). A total of 32 healthy, right-handed subjects were divided into a taVNS and a sham group. Subjects were asked to make reaching movements with their right arm while holding the handle of the manipulandum against a velocity-dependent curl force field. taVNS was applied for 3 s at the start of the reaching movement. The results demonstrated that taVNS reduced the maximum rightward displacement (i.e., error amplitude) during force field learning, and the final values were significantly smaller in taVNS groups than in sham groups. These results suggest that taVNS matched with task initiation has an online effect on force field learning.Subject terms: Health care, Medical research, Neurology, Neuroscience Introduction Vagus nerve stimulation (VNS), which electrically simulates the left cervical bundle of the vagus nerve through implanted stimulating electrodes, was developed as a neuromodulation technique to treat intractable epilepsy1,2. Subsequently, its antidepressant effect was demonstrated in treatment-resistant depression3,4. Moreover, VNS has been applied to a variety of diseases, such as pain treatment, cardiovascular disease, diabetes, and obesity5. One of the systems developed in response to requests for the convenience of VNS is transcutaneous auricular vagus nerve stimulation (taVNS), a non-invasive method of VNS that transcutaneously stimulates the auricular branch of the vagus nerve6. Potential clinical applications of taVNS have expanded to many diseases7–9. In recent years, an expected new clinical application of VNS has been its synergistic effect with post-stroke rehabilitation. First, in animal stroke models, some studies have reported that VNS applied during upper extremity movement for motor recovery training significantly restored function as compared to training alone or VNS applied after training10,11. Additionally, in a human study, Dawson et al. reported that the combination of VNS and post-stroke upper extremity rehabilitation promotes the recovery of function12. Subsequent follow-up and replication studies have been done by other groups13. Similarly, several studies have shown that taVNS combined with conventional rehabilitation training improved upper and/or lower limb motor functions after stroke better than rehabilitation training alone14,15. With these studies as a background, VNS and taVNS are becoming promising options for post-stroke rehabilitation. Although the detailed neural mechanisms as to why VNS promotes post-stroke rehabilitation have not been fully elucidated, the following points have been suggested: (1) VNS induces neuroprotective effects such as the suppression of inflammation and angiogenesis16, (2) VNS enhances the reorganization process of the motor cortex15,17, and (3) VNS increases a brain-derived neurotrophic factor that regulates synaptic plasticity18. The latter two explanations suggest that VNS possesses metaplastic properties that promote neuroplasticity in the brain. If stimulation of the vagus nerve promotes neural plasticity in brain, VNS and taVNS may also influence motor learning in healthy subjects, as neuroplasticity in motor-related areas is involved in the motor learning process. In previous studies, a few researchers investigated whether taVNS facilitates motor learning. Chen et al. reported that offline taVNS (offline means that taVNS is applied before or after training) enhanced motor sequence learning19. On the other hand, Pierre and Shinohara reported that online taVNS (online means that taVNS is applied during training) at nonspecific timing during training compromised motor learning of a complex force exertion task20. However, due to the limited number of studies, the effects of taVNS on motor learning are not yet systematically understood. It remains unclear which types of motor learning taVNS effectively influences, as well as the optimal timing and stimulus parameters required to facilitate motor learning. In the present study, among the many motor learning tasks, we focused on a reaching movement task under a force field condition (i.e., force field learning). This is a typical motor task that is generally used as a novel motor skill learning task and expressed as an error-based motor learning and/or sensory–motor adaptation. The cerebellum is known as a major neural substrate in the learning process21,22. The previous fMRI study showed that, during taVNS, BOLD signals increased in the cerebellum, as well as in other brain regions23, and neural modulation of the cerebellum has been reported to affect sensory–motor adaptation24–26. In stroke research, VNS has been reported to promote the recovery of upper limb motor function21. Therefore, we hypothesized that taVNS affects the process of force field learning that involves upper limb reaching movements. As for the timing of taVNS, based on previous studies that reported a timing-dependent effect of VNS10,11, we assumed that an online effect would be observed when the timing of taVNS and task initiation is matched; therefore, we investigated the online effect with specific timing in the present study. In the present study, we designed a single-blind and sham-controlled experiment to clarify whether taVNS changes the speed and/or retention of force field learning. The present study will help us to understand comprehensively the impact of taVNS on motor learning. Methods Subjects Thirty-two young, healthy subjects (all males, 19.4 ± 1.5 years, mean ± SD) participated in this study. They were all right-handed (laterality quotient = 0.91 ± 0.14; Oldfield, 1971), and they were randomly divided into two groups in order of participation using pseudo-random numbers (block randomization): taVNS (n = 16) and sham groups (n = 16). All participants gave informed written consent prior to participating in the study. The experiments were performed in accordance with the Declaration of Helsinki (1964) and with the approval of the human ethics committee of Kyushu Institute of Technology (Institute of Liberal Arts), Japan. Transcutaneous auricular vagus nerve stimulation/sham stimulation Constant current stimulation was delivered to the left tragus for the taVNS group and to the left earlobe for the sham group, while performing the force field learning task on Day 1. The left tragus was selected as the stimulation site because it is a suitable location where the afferent auricular branch of the vagus nerve is located27. It has been reported that stimulating the left tragus induces significant activation in brain regions associated with the afferent vagal pathway, including the contralateral postcentral gyrus, bilateral insula, frontal cortex, right operculum, and left cerebellum, whereas stimulation of the earlobe activates only the contralateral postcentral gyrus23. The left tragus or earlobe was stimulated using a constant current stimulator (DS7a; Digitimer, USA) at 25 Hz for 3 s after the target was displayed on the screen in each trial; the stimulus waveform was a monophasic square wave (pulse width 0.5 ms) based on a previous study23. The stimulation was administered using custom-developed stimulating electrodes (KT220-004; UNIQUE MEDICAL, Japan), which were round, unipolar stimulation electrodes 1 cm in diameter and made of Ag/AgCl. The clip-shaped electrodes were clamped to the subject’s tragus or earlobe. To reduce impedance between the electrodes and skin, the surfaces of the tragus or earlobe were cleaned with alcohol wipes, and conductive paste (Ten 20; Weaver and Company, USA) was applied to electrode surfaces. The perceptual threshold (PT) was determined at the tragus or earlobe while subjects were seated quietly, and the stimulation intensity was set to 200% of the PT in both the taVNS and sham conditions, in accordance with the previous study23. Experimental paradigm The experiments were conducted over two consecutive days. On Day 1, 50 trials of a reaching task were performed under a null force field condition, and then 150 trials of the task were performed under a force field condition (i.e., force field learning, Fig. 1A–C). On Day 2, 150 trials of the task were performed under the force field condition. The null force field condition included 5 error-clamped trials randomly, and the force field condition included 20 error-clamped trials randomly (Fig. 1D–E). On Day 1, during the force field condition, taVNS or sham stimulation was applied for 3 s from the start of each trial, whereas no stimulation was applied on Day 2. (A) Experimental setup of the force field trials (i.e., force field condition without error-clamped trials). (B) Typical example of movement trajectories for the 5 initial force field trials (gray) and last 5 force field trials on Day 1 (black) in the taVNS group. (C) Typical example of movement trajectories for the 5 initial force field trials (gray) and last 5 force field trials on Day 1 (black) in the sham group. (D) Experimental setup of the error-clamped trials. (E) Typical example of force traces for the 5 initial error-clamped trials (gray) and last 5 error-clamped trials on Day 1 (black) in the taVNS group. (F) Typical example of force traces for the 5 initial error-clamped trials (gray) and last 5 error-clamped trials on Day 1 (black) in the sham group. Please note that these error-clamed trials (E, F) were randomly interleaved in the force field condition. Force field learning task Subjects were asked to reach with their right arm while holding the handle of the manipulandum (PHANToM Premium 1.5; Geomagic, USA). The handle position was displayed as a white cursor (a circle, 6 mm in diameter) on a black background on a horizontal screen located above their hand. The movement of the handle was constrained to a virtual horizontal plane (10 cm below the screen) that was implemented by a simulated spring (1.0 kN/m) and damper (0.1 N s/m). Before each trial, subjects were required to hold the cursor at its starting position (a circle, 10 mm in diameter). After a 2 s holding time, a gray target (a circle, 10 mm in diameter) appeared. After an additional randomly selected holding time (1–2 s), the target color changed to light blue, signaling subjects to start reaching. Subjects were required to move the handle with an amplitude of 10 cm and a peak velocity of 470 ± 45 mms−1 (the target velocity was calculated using the minimum-jerk theory, in which movements are generated to minimize the total jerk—the rate of change of acceleration—along the trajectory, with a movement amplitude of 10 cm and a duration of 0.4 s). A warning message appeared on the screen if the movement velocity of the handle rose above (‘Fast’) or fell below (‘Slow’) this threshold value. To avoid displaying this warning message as much as possible, subjects were asked to practice reaching under a null force condition 30 times before the task on Day 1. At the end of each trial, the handle was automatically moved back to the starting position by the manipulandum. In the force field condition, a velocity-dependent curl force field (clockwise force field) expressed as the following equation was applied: f = Bv, where f = (fx, fy)T is the force to the handle generated by the manipulandum [N], B = (0, 10; − 10, 0) [Nsm−1], and v = (vx, vy)T is the velocity of the handle [ms−1] (x, rightward direction; y, forward direction). In the null field and force field conditions, error-clamped trials, in which the trajectory of the handle was constrained to a straight line toward the target by a virtual “channel,” were randomly interleaved. The virtual channel constrained any motion perpendicular to the target direction by a one-dimensional spring (2.5 kN/m) and damper (25 N s/m). Pain sensation, concentration, fatigue, and alertness A person’s subjective state, such as concentration, fatigue, and alertness, can affect progress and/or retention of motor learning. The sensation of pain can also have an indirect effect by interfering with these states. Therefore, on Day 1 and Day 2, after all tasks were finished, the subjects completed questionnaires to rate their level of concentration (1 = not at all, 7 = very concentrated), fatigue (1 = high level of fatigue, 7 = no fatigue), and alertness (Stanford Sleepiness Scale rating: 1 = highest alertness, 7 = lowest alertness) using a seven-point scale with reference to the previous study 28. Pain sensation was evaluated using a ten-point scale (1 = no pain, 10 = maximal pain). Data recording and analysis Manipulandum motion data were recorded for 2 s at a sampling rate of 500 Hz. Motion data were low-pass filtered using a fourth-order Butterworth filter with a 10 Hz cutoff. To quantitatively evaluate the learning of the force field condition, the maximum rightward displacement of the cursor due to the force field was calculated as the error amplitude value. The average of ten trials immediately after the start of the force field condition was defined as an initial error, and the average of ten trials immediately before the end of the force field condition was defined as the final error. The improvement rate (IR) of the error amplitude on Day 1 or Day 2 was defined as the following ((1) or (2)) and used as the evaluation index: the closer the value is to 1.0, the more the error amplitude value decreases, reflecting accelerated force field learning. The retention rate (RR) of the error amplitude was defined as the following (3) and used as an evaluation index that reflects the retention of motor learning: the negative value indicates forgetting, 0 indicates retention, and a positive value indicates improvement when the initial error on Day 2 was compared to the final error on Day 1. To evaluate changes in force during catch trials in the force field condition, forces exerted at the peak velocity of forward displacement of the cursor were calculated, and the average of the last ten catch trials on Day 1 and Day 2 were defined as the representative values. Statistical analysis Statistical analyses were performed using a commercially available software package (SPSS 21.0, SPSS, USA). A Mann–Whitney U test was used to analyze the effects of the intervention (taVNS vs. sham) on the IR and RR indexes, estimated force during catch trials, stimulus intensity, and numeric rating scales because Shapiro–Wilk normality tests rejected the normality of these indexes. For Mann–Whitney U tests, the r values were calculated as the effect size, and the thresholds for interpreting the r values were respectively set at 0.1, 0.3, and 0.5 for small, medium, and large. The significance level was set at p < 0.05. Data of the IR and FR indexes are presented as the mean and standard error (mean ± SE). Data of stimulus intensity, pain sensation, concentration, fatigue, and alertness are presented as the mean and standard deviation (mean ± SD). Results Stimulus intensities of taVNS and sham No significant difference was found between taVNS (i.e., left tragus stimulation) and sham (left earlobe stimulation) for stimulus intensities on Day 1 (taVNS, 0.90 ± 0.38 mA; sham, 0.86 ± 0.28 mA; Mann–Whitney U test: z = 0.359, p = 0.724, r = 0.063). Pain, concentration, fatigue, and alertness on Day 1 and Day 2 There were no significant differences between the two groups for any of the subjective ratings (pain, concentration, fatigue, alertness) (Table 1). Some people experienced slight pain, but it was not painful enough to interfere with a motor task. No participants withdrew from the experiment due to pain or other reasons. Taken together with the results regarding stimulus intensity, pain or subjective states are unlikely to have affected the present results. Table 1Comparison of subjective ratings between the taVNS and sham groups, along with statistical results of the Mann–Whitney U tests.taVNS (n = 16)Sham (n = 16)z-valuep-valuer-valueDay 1Pain (a.u.)2.7 ± 0.92.6 ± 0.8z = 0.628p = 0.564r = 0.111Concentration (a.u.)5.1 ± 1.05.1 ± 0.7z = − 0.182p = 0.867r = − 0.032Fatigue (a.u.)2.4 ± 0.72.9 ± 1.1z = − 1.466p = 0.196r = − 0.259Alertness (a.u.)3.3 ± 0.82.8 ± 0.9z = 1.741p = 0.102r = 0.308Day 2Concentration (a.u.)5.0 ± 1.25.4 ± 1.0z = − 0.861p = 0.423r = − 0.152Fatigue (a.u.)2.0 ± 0.82.8 ± 1.1z = − 2.109p = 0.051r = − 0.373Alertness (a.u.)2.8 ± 1.02.7 ± 1.1z = 0.315p = 0.780r = 0.056Means ± SD; n = 16 in each group. Comparison of subjective ratings between the taVNS and sham groups, along with statistical results of the Mann–Whitney U tests. Means ± SD; n = 16 in each group. Motor adaptation (Day 1 and Day 2) Figure 2 shows the group means of trial-to-trial changes of the position errors on Day 1 and Day 2. During the null field condition, the position errors remained close to zero in a steady state. During the force field condition on Day 1, the position errors reached steady-state values over time in both the taVNS and sham groups, although the convergent values observed in the last ten trials appeared to be lower in the taVNS group than in the sham group. During the force field condition on Day 2, the position errors did not appear to differ significantly between the groups, but the values in the taVNS group tended to be lower as the trials progressed. Time course of experimental protocols (upper part) and group means of trial-to-trial changes in the position error on Day 1 and Day 2 (lower graph). The red circles and the blue circles represent the position error in the taVNS and sham groups, respectively. Gray areas indicate the first or last ten trials during the force field task on Day 1 and Day 2. Error bar represents the standard deviation of the mean. In Fig. 3, the average improvement rate at the final time points (the last 10 trials) during force field learning was compared between the taVNS and sham groups. A Mann–Whitney U test revealed significant differences between groups on Day 1 (z = 2.638, p = 0.007, r = 0.466), indicating that taVNS significantly reduced position errors caused by the force field and facilitated force field learning. On the other hand, the difference did not reach a significant level on Day 2 (z = 1.734, p = 0.086, r = 0.306). Comparison of the improvement rate between the taVNS and sham groups at the final time points during force field learning on Day 2. Red bars and blue bars represent the taVNS and sham groups, respectively. Error bars represent the standard error of the mean. The asterisk indicates significant differences between the taVNS and sham groups (p < 0.05, Mann–Whitney U test). Retention of motor adaptation (Day 2) In Fig. 4, the average retention rates during force field learning were compared between the taVNS and sham groups. The taVNS group showed a negative value, on average, and a Mann–Whitney U test revealed significant differences between groups (z = − 2.864, p = 0.003, r = − 0.506), indicating that taVNS acted to promote forgetting of the learning effect from Day 1 to Day 2, as compared to the sham group. Comparison of the retention rate between the taVNS and sham groups at the initial time points during force field learning on Day 1 and Day 2. The red bar and the blue bar represent the taVNS and sham groups, respectively. Error bars represent the standard error of the mean. The asterisk indicates a significant difference between the taVNS and sham groups (p < 0.05, Mann–Whitney U test). Catch trials (Day 1 and Day 2) The averages of the last ten catch trials in the force field condition were compared between the taVNS and sham groups on Day 1 and Day 2. No significant difference was found between taVNS and sham for estimated force on Day 1 (taVNS, 2.51 ± 0.11 N; sham, 2.26 ± 0.22 N; Mann–Whitney U test: z = 0.641, p = 0.539, r = 0.113) and Day 2 (taVNS, 2.74 ± 0.13 N; sham, 2.68 ± 0.20 N; Mann–Whitney U test: z = 0.038, p = 0.985, r = 0.006). Discussion The purpose of this study was to investigate the effect of taVNS on force field learning in healthy adults. The results demonstrated that taVNS reduced the maximum rightward displacement (i.e., error amplitude) during force field learning, and the final values were significantly smaller in the taVNS group than in the sham group. However, this effect could not be carried over to the following day, and the retention rate showed negative value, meaning taVNS promoted the forgetting of the learning effect. These results suggest that taVNS has online effect on force field learning, whereas there is no effect on retaining the learning. The present study is the first report to clarify the effect of taVNS on force field learning. Two candidates for the neural mechanisms underlying the present result of taVNS could be indicated. First, it is possible that taVNS affected attention during task execution, which in turn affected force field learning. In previous studies of force field learning and split-belt adaptation, appropriate attention is required for the learning process of sensory–motor adaptation, and attention divided by the application of a secondary task has been reported to slow down sensory–motor adaptation using force field 29 and split-belt adaptations 30. As for its neural mechanisms, Taylor and Thoroughman hypothesized that a secondary task can interfere with the feed-forward mechanisms of error correction, which could result in more errors and a slower learning rate 29. While divided attention inhibits motor learning, prior studies in primates have demonstrated that attention enhances the signal-to-noise ratio of the perceptual processing of the attendant stimuli 31,32. It is possible that improving the perceptual processing of the force field resulted in the greater improvement rate of error amplitude in this study. taVNS is assumed to enhance attention by modulating the bilateral locus coeruleus–norepinephrine (LC–NE) system and γ-aminobutyric acid (GABA) neurotransmission, which are essential contributors to attentional processes 33,34. Although the subjective ratings of attentional and arousal levels did not show significant changes with taVNS, it cannot be ruled out that taVNS acted subconsciously to increase the attentional level and affect force field learning. Second, it is possible that taVNS had neuro-modulatory effects on the cerebellum. It is well known that the cerebellum is a major neural substrate in the learning process of error-based motor learning and/or sensory–motor adaptation 21,22. Some studies have reported that cerebellar transcranial direct current stimulation (tDCS) influences motor learning of arm reaching movement and split-belt adaptation; anodal stimulation enhances and cathodal stimulation depresses motor learning and adaptation 24–26. Jayaram et al. listed changes in the behavior of Purkinje cells, such as the population and dynamic range of cells involved in the adaptation, as possible mechanisms to explain their result 25. In the fMRI study, taVNS has been reported to increase the BOLD signal in the cerebellum as well as other deep brain regions 23,35,36. While much remains unclear about the extent to which taVNS influences the regulation of cerebellar activity, it is possible that the neuro-modulatory effect of taVNS on the cerebellum affected force field learning like cerebellar tDCS. Limitations This study has some limitations to consider in relation to the experimental paradigm and stimulation protocol. Subjects in this study were all male, and the influence of gender differences on the results cannot be ruled out. The relatively small sample size is also a limitation. While the results showed a statistical difference in the improvement rate on Day 1, further investigations with a large sample size will help to establish practical credibility. Because taVNS is a relatively new technique, appropriate stimulation protocols—such as stimulus duration, timing, frequency, and waveform settings—have not been identified. This limitation is particularly salient with respect to the effect of taVNS on motor learning tasks, for which there are few examples of research reports. Although motor learning tasks differ, Chen et al. reported the offline effect of taVNS 19, and Pierre and Shinohara reported no effect of taVNS at nonspecific timing during tasks. Therefore, the results of this study may vary widely depending on the protocol 20. Combinations of motor learning tasks and stimulus protocols need to be sorted out in future studies. In particular, it is necessary to investigate whether results similar to those of this study can be observed with offline conditioning stimuli, which have been demonstrated to have intervention effects in many other non-invasive brain stimulation studies. No physiological monitoring and/or behavioral makers (e.g., heart rate variability, pupillometry, reaction time, EEG markers) impose limitations on the interpretation of the results of this study. Based on previous studies, we considered that changes in attention could possibly explain the results of this study. However, it is necessary to confirm whether such changes actually occurred under the experimental conditions of this study. In previous studies reporting the effect of divided attention and tDCS on force field learning, changes were observed in error amplitude on the null force conditions after force field learning. On the other hand, this study did not conduct experiments with null force conditions after force field learning to determine the impact of taVNS on the following day, and the effect of taVNS on the force field learning was observed in the error amplitude during force field learning. Taylor and Thoroughman claimed that error amplitude during force field learning reflects changes in feedback control 29, and, if we follow their argument, the present results indicate the effect of taVNS on online correction to the force field. Therefore, further studies are required to reveal the effect of taVNS on the correction of the internal model (i.e., feedforward control). Conclusion and significance In conclusion, the present study showed that taVNS during force field learning reduced error amplitudes, while the retention of its learning effect was not observed on the next day. taVNS is a new intervention for potentially facilitating motor learning, whereas its neurophysiological mechanisms and utilities are unknown. This is the first study reporting an online taVNS effect on error-based motor learning in healthy adults. We expected that the present study would be the basis for applying taVNS as a new neural modulation technique to facilitate motor training and learning for disabled and non-disabled people. Author contributions H.O. wrote the main manuscript text and prepared all figures. All authors reviewed the manuscript. Funding This work was supported by JSPS KAKENHI Grant Number 21K19719 and 25K03508. Data availability The experimental data used for figures are available as open data via Kyushu Institute of Technology’s online data repository. http://hdl.handle.net/10228/0002001971. Declarations 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 experimental data used for figures are available as open data via Kyushu Institute of Technology’s online data repository. http://hdl.handle.net/10228/0002001971.