The vagus nerve is a central part of the gut-brain axis and bi-directionally links the vegetative and the central nervous system . A range of cognitive and emotional processes can influence vegetative processes via the vagus nerve, i.e. by changes in heart rate or respiration . The impact of vagus nerve activity on cognitive and emotional processes, i.e., the other direction of information flow, is still largely unknown. Electrical stimulation of the vagus nerve (vagus nerve stimulation, VNS) can be used to study causal effects of vagus nerve activity on functions of the central nervous system. As it constitutes a common treatment for medication-resistant epilepsy, it can be safely applied in humans .
Evidence from rodents [4, 5] and humans [6–8] provide converging evidence that invasive and non-invasive VNS is associated with enhanced reward processing, reinforcement learning and recognition memory. Animal studies indicate that these alterations in behavior are associated with enhanced dopamine release in the midbrain [4, 5, 9]. Based on current evidence, these effects are function-specific, as studies failed to observe effects on working memory , implicit learning  and conflict processing .
The polyvagal theory represents a bio-behavioral model that relates vagus nerve activity to social interaction . Based on phylogenetic reasoning and anatomical findings of vagus nerve connectivity, it implicates the efferent part of the vagus nerve in the expression of social behaviors, e.g., through its projections to laryngeal, pharyngeal and facial muscles essential for verbal and non-verbal communication. However, the role of afferent projections of the vagus nerve to the brain for social cognition is still unclear.
Studies that indirectly assessed vagus nerve activity by means of heart-rate variability (HRV) in humans report that high HRV, and thus parasympathetic tonus, at baseline is predictive for enhanced social engagement and cooperative behavior [14–17]. One meta-analysis of 16 studies demonstrated that compassion, an important emotional aspect of social interaction , positively correlates with HRV . However, the causal relationship between vagus nerve activity and social behavior and the direction of interaction between vagus nerve and the central nervous system during these processes are still unknown.
Several studies show that social interaction recruits similar neural networks as reward processing (for a review see 20). Rewarding social stimuli, e.g. positive facial expressions, own social reputation and positive social feedback, are associated with activity in dopaminergic basal-ganglia-thalamo-cortical circuits similar to non-social rewards [21–25]. In particular, cooperation with other humans is perceived as rewarding and recruits brain regions involved in reward processing, such as the medial orbitofrontal cortex [26, 27], caudate nucleus and anterior cingulate cortex .
Based on human and animal studies demonstrating effects of the VNS on dopaminergic brain circuits and reward processing, one might hypothesize that the stimulation of afferent fibers of the vagus nerve enhances rewards derived from social interactions. On a behavioral level, this might be reflected by enhanced cooperative behavior with other humans.
Recent studies indicate that specific personality traits, in particular neuroticism and extraversion, are associated with social behavior . While personality traits are undoubtedly influenced by multiple neurotransmitter systems, specifically extraversion and neuroticism have been linked to dopamine-dependent reward-processing [30, 31]. Thereby, a highly reactive dopaminergic system, e.g. as measured by dopamine-relevant genes, structural volume of dopamine-rich brain regions or dopamine receptor availability, has been associated with high extraversion, whereas the opposite has been suggested for neuroticism [32–34]. If afferent VNS effects are mediated via the dopaminergic system, one could thus hypothesize individual stimulation effects interact with these personality traits.
Decision making processes can be disentangled into several sub-processes based on choices and reaction time. Drift-diffusion modelling (DDM) constitutes one of the most common methods for the assessment of value-based choices . DDM dissects the decision process into several sub-processes including a starting bias towards response options, the rate of accumulation of information (i.e., the drift rate), the amount of information needed for a decision (i.e., boundary separation) and non-decision operations reflecting perceptual and motor computations. While the drift-diffusion model is commonly used to make inferences on classic perceptual decision making tasks like the flanker  or color discrimination task 37,38. Only few studies have analyzed sub-components of social decision making in humans [39–42]. Previous studies show an association between shifts in starting bias and reward value expectation [38, 43]. These studies report associations between pro-social social behavior and changes in starting bias and drift rate, both relatively early parts of the decision process [39, 42]. Thus, one could hypothesize that VNS effects on cooperative behavior occur at these early stages of the decision process.
Here, we assess the causal relationship between vagus nerve activity and social interaction. To this end, we applied transcutaneous VNS (tVNS) to 19 long-standing seizure-free epilepsy patients, who had never received invasive or non-invasive VNS treatment. We applied auricular VNS in order to stimulate afferent parts of the vagus nerve, while patients performed a computerized version of the well-established prisoner’s dilemma task . Based on trial-by-trial choices of participants, we assessed effects of stimulation on cooperative behavior. Further, we assessed the impact of subject characteristics, e.g., sex and age, on stimulation effects. We also included the big five personality traits into our analysis, i.e., neuroticism, extraversion, openness, agreeableness, and conscientiousness. To understand the impact of tVNS on social decision making in more detail, we used behavioral modelling.