The vagus nerve is the body’s longest cranial nerve, wandering from the brainstem down to the heart and gut, and its auricular branch provides a rare accessible entry point for external stimulation. When electrodes stimulate the cymba concha of the left ear, signals travel along afferent vagal fibers to the nucleus tractus solitarius in the brainstem, which in turn activates the locus coeruleus and basal forebrain. These nuclei trigger the release of norepinephrine and acetylcholine, two neuromodulators critical for arousal regulation, synaptic plasticity, and cognitive control. In animal models, vagus nerve stimulation has been shown to enhance cognitive flexibility and memory, and to consolidate the synaptic plasticity produced by learning itself. The Dallas team, led by F. M. Filbey, wanted to translate those preclinical findings into a controlled human model that assessed multiple domains of executive function rather than the narrow learning paradigms that have dominated most human taVNS studies to date.

Forty-two healthy volunteers aged 18 to 54 were recruited from the community and assigned to either active or sham stimulation using an alternating allocation procedure stratified by age and sex, a quasi-experimental design in which participants were blinded to their condition but assignment was not fully randomized. After completing mood questionnaires and having their heart rate variability monitored with a validated chest strap, participants received one hour of cognitive training through the BrainHQ program while wearing the NEMOS taVNS device. In the active condition, electrodes were placed on the cymba concha of the left ear, targeting the auricular vagal branch; in the sham condition, they were placed on the center of the ear lobe, which lacks vagal innervation. Stimulation intensity was individually calibrated just below each participant’s pain threshold, with pulses of 200 to 300 microseconds delivered at 25 Hz in cycles of 30 seconds on and 30 seconds off.

BrainHQ is an adaptive, computer-based training program whose algorithm continuously adjusts task difficulty based on moment-to-moment performance, ramping up challenge when participants succeed and easing off when they struggle. The participants worked through three tasks in fixed sequence: Mind Bender, a set-shifting exercise; Divided Attention, which trains discrimination of specific criteria in competing images; and Mixed Signals, a Stroop-based task matching auditory and visual stimuli. This continuously adaptive design was central to the study’s logic. Because the algorithm keeps performance near an individual’s capacity, it continuously taxes cognitive control processes rather than allowing them to plateau, potentially making the training sensitive to subtle transient changes in neural state that static tasks might miss.

After the training hour and a rest break, participants were fitted with a 64-channel EEG cap and underwent simultaneous taVNS and EEG recording during rest with eyes closed and eyes open, followed by a monetary incentive delay task measuring responses to reward and punishment. The researchers then collected a battery of executive function assessments targeting the three core components of the classic tripartite model: cognitive flexibility via the Dimensional Change Card Sort, inhibitory control via the Flanker test, and working memory via the List Sorting test, along with a modified Stroop Color Word task. Spectral power and coherence were calculated across delta, theta, alpha, beta, and gamma bands, with electrodes grouped into frontal, central, parietal, and occipital regions in each hemisphere to map how cortical communication differed between groups.

The results painted a nuanced picture. Unadjusted multivariate analyses suggested a significant overall effect of taVNS on executive function measures, and follow-up comparisons showed that the active group outperformed the sham group on BrainHQ, earning more total stars with a moderate-to-large effect size. Yet when the models were adjusted for age, sex, and education, the evidence for an independent group effect became unclear, and the BrainHQ difference did not survive Bonferroni correction. No group differences emerged on the monetary incentive delay task, the standard Stroop, or the NIH Toolbox measures. The researchers suggest that the adaptive nature of BrainHQ may have made it more sensitive to transient neural modulation, whereas the static difficulty of the other assessments, combined with a healthy sample performing near ceiling, may have lacked sufficient sensitivity to capture momentary changes in cognitive capacity.

The EEG findings were equally intriguing. During the eyes-closed resting condition, the active group showed lower delta-band clustering coefficients than the sham group in the left frontal, central, and parietal regions, with directionally similar but more uncertain effects in left occipital and right-hemisphere regions. Delta oscillations, roughly 1 to 4 Hz, are markers of lower arousal states and large-scale network synchrony, typically linked to deep sleep or low-engagement conditions. A reduction in delta connectivity during active stimulation may reflect a down-regulation of broadly synchronized low-frequency communication and a sharpening of more focal, functionally specific connectivity, consistent with the notion that vagal modulation facilitates flexible reconfiguration of large-scale brain networks. In other words, active taVNS may push the brain toward a higher state of cognitive readiness, reducing reliance on large-scale delta synchrony and enabling more targeted, efficient neural communication. None of these regional effects met the Bonferroni-adjusted significance threshold, however, and no differences appeared in theta, alpha, beta, or gamma bands, or in any band during the eyes-open condition.

Secondary analyses revealed suggestive mood effects. The active group showed an estimated reduction in depression and tension scores from pre to post stimulation compared with the sham group, and delta-band clustering coefficients in left frontal, central, parietal, and occipital regions were positively correlated with depression improvement in the active group, though neither the mood differences nor the correlations survived correction. Notably, heart rate variability showed no clear changes during or after stimulation, aligning with a meta-analysis that found no significant effect of taVNS on vagally mediated HRV. Because auricular taVNS primarily engages afferent rather than efferent vagal pathways, its cognitive effects may be mediated predominantly by central mechanisms such as locus coeruleus-noradrenergic activation without producing detectable peripheral cardiac changes, a central-peripheral dissociation that challenges the use of HRV as a reliable biomarker for these interventions.

The authors are careful to frame their findings as exploratory and hypothesis-generating. The study lacked a pre-stimulation EEG baseline, limiting interpretation to between-group differences during the later stimulation period and precluding distinction between immediate and residual effects of the preceding training session. The quasi-experimental design, modest sample size, and sensitivity of the effects to covariate specification all warrant caution. Nevertheless, the convergence of improved adaptive training performance, reduced left-hemisphere delta connectivity, and suggestive mood benefits provides mechanistic plausibility for the idea that pairing vagal stimulation with continuously adaptive cognitive demands may maximize neuromodulatory effects by repeatedly engaging task-relevant cortical networks. The observed improvements may reflect enhanced neural efficiency emerging from the interaction between vagally mediated neuromodulation and adaptive cognitive training.

What makes this study compelling is its attempt to move beyond the narrow paradigms that have limited the field, evaluating multiple dimensions of executive function while capturing network-level modulation in real time. The divergent effects on adaptive versus static cognitive performance suggest that taVNS-mediated benefits may depend critically on task demand and complexity, a distinction supported by recent literature showing that taVNS effects on cognitive flexibility emerge specifically under high task complexity. If future studies with prespecified outcomes, theoretically justified covariates, and adequate statistical power replicate these patterns, the implications could extend from cognitive enhancement in healthy adults to clinical applications, given preclinical evidence that vagus nerve stimulation facilitates extinction learning and that taVNS ameliorates memory impairment in animal models. For now, the message is one of cautious optimism: a gentle pulse to the ear may indeed tune the brain, but the evidence is still finding its frequency.

Subject of Research: Neuroscience

Article Title: A non-randomized trial of the effects of transcutaneous auricular vagus nerve stimulation on cognitive performance and resting state connectivity

Article References: Filbey, F. M., Short, C. T., Leazure, A., Nekoui, N. E., Brown, T., Kroener, S., & Liu, C. (2026). A non-randomized trial of the effects of transcutaneous auricular vagus nerve stimulation on cognitive performance and resting state connectivity. Frontiers in Human Neuroscience, 20, Article 1801066. https://doi.org/10.3389/fnhum.2026.1801066

Image Credits: AI Generated

DOI: 10.3389/fnhum.2026.1801066

Keywords: taVNS, vagus nerve stimulation, neuromodulation, executive function, EEG, delta band, BrainHQ, cognitive training, fronto-parietal network, neural efficiency, locus coeruleus, mood