Vagus nerve stimulation boosts the drive to work for rewards

Interoceptive feedback transmitted via the vagus nerve plays a vital role in motivation by tuning actions according to physiological needs. Whereas vagus nerve stimulation (VNS) reinforces actions in animals, motivational effects elicited by VNS in humans are still largely elusive. Here, we applied non-invasive transcutaneous auricular VNS (taVNS) on the left or right ear while participants exerted effort to earn rewards using a randomized cross-over design (vs. sham). In line with preclinical studies, acute taVNS enhances invigoration of effort, and stimulation on the left side primarily facilitates invigoration for food rewards. In contrast, we do not find conclusive evidence that acute taVNS affects effort maintenance or wanting ratings. Collectively, our results suggest that taVNS enhances reward-seeking by boosting invigoration, not effort maintenance and that the stimulation side affects generalization beyond food reward. Thus, taVNS may enhance the pursuit of prospective rewards which may pave avenues to treat motivational deficiencies.


Supplementary Methods
During the sessions, we measured pulse, weight, and height as well as waist and hip circumference according to the recommendations of the World Health Organization 1 . Moreover, participants reported their last meal and drink and female participants further reported oral contraceptive use as well as the beginning of their last menstrual cycle. Pulse was also measured after the effort task.
At three time points, participants also responded to state questions presented on a computer screen as VAS using the joystick on an Xbox-360 controller (Microsoft Corporation, Redmond, WA). Items included questions on metabolic state (hunger, fullness, thirst) and mood, which were derived from the Positive and Negative Affect Schedule (PANAS 2 ).
After completing the task block and a repetition of the state VAS, participants received their breakfast and a snack according to the food reward (energy) points earned. We provided a bowl of cereals for breakfast and a snack, both resuming the food reward (energy) points earned during the effort task: First, 100 ml of milk (or almond milk as vegan/lactose free alternative; ~68 kcal) were deducted from the total earned energy points. Second, participants could choose between three different chocolate bars (Twix, Mars or Snickers sticks; ~100 kcal each, Mars Inc., McLean, VA). The remaining points were converted into a serving of cereal. Since several participants had only earned few energy points, they received no additional snack and the volume of the milk was reduced to match the volume of the earned cereal.
Participants received the bowl for breakfast and were instructed that this was their food reward and that they could eat as much as they liked. A 10-min break for breakfast was scheduled, but most of the participants finished eating before the end of the break.
To dissociate effort and rest phases, we segmented the effort trajectories (timelevel data) into work and rest segments. The criteria for segment boundaries were based on the first temporal derivative of r . Work segment onsets were defined as a positive slope at t, combined with a cumulative increase of at least 10 units and no element smaller than -20 during the next second (Equation 2). Rest segment onsets were defined by inspection of the density distribution of the first temporal derivative of r . We determined the local minimum of the distribution on the group level, which indicated a visible distinction between effort and rest phases at a value of ~-20 (2.8

SD) for both levels of task difficulty (Equation 3)
.
For each work segment, we computed an invigoration slope ( I ) from segment onset to its first local peak. Peaks were determined using the MATLAB function findpeaks with default settings (cf. Fig. 1 for illustration). If no peak was found by the algorithm, the endpoint of the slope was set to the first data point of a plateau (second derivative = 0) or, alternatively, to the maximum r value in the segment.
ORDER and STIMSIDE have been centered around the grand mean. To further characterize the decision to maintain effort or take a break, we applied a previously described cost-evidence accumulation model 3,4,5 . In this model, the duration of work and rest segments is used to fit an average amplitude of cost evidence as well as cost-evidence accumulation and dissipation slopes across all work segments. More specifically, the duration of a work segment (TE) is defined as follows: and the duration of rest segment (TR): where A is the shared amplitude of cost-evidence variations and SE and SR are cost-accumulation and cost-dissipation slopes, respectively. Differences in difficulty or reward magnitude could in principle affect all three parameters of the model and are incorporated as linear combinations for each parameter. However, for the current analysis, we used a reduced set of free parameters that were previously reported after extensive model comparisons 3,4,5 . Moreover, we set the mean amplitude, A, to 1 and introduced an additive taVNS effect for each parameter to estimate taVNS-induced changes leading to the following equations: SR = (SR mean + SR mean taVNS * ) + (SR reward + SR reward taVNS * ) *  Note: M = mean, CI = Credible interval, BF10 = Bayes factor in favor of the alternative hypothesis, BF for all parameters were determined using the Savage Dickey density ratio method. BF < 1 can be interpreted as evidence for the null hypothesis (parameter value = 0), while BF > 1 can be interpreted as evidence for the alternative hypothesis (parameter value ≠ 0). Source data are provided as a Supplementary Source Data file.