Continuous theta-burst stimulation of the ventromedial prefrontal cortex reduces Pavlovian go-invigoration and enhances thalamo-striatal reward prediction error signals

The ventromedial prefrontal cortex (vmPFC) represents value, i.e. whether environmental cues are good or bad. Rewarding cue value is aligned with a tendency to act or to inhibit in case of potential punishment, a so-called Pavlovian Bias. Value-related vmPFC activation may therefore impact Pavlovian bias. However, it remains unknown how the activity in the vmPFC modulates the Pavlovian bias. Thus, we applied continuous theta-burst stimulation (cTBS), a non-invasive protocol that transiently reduces cortical excitability, over the vmPFC in a single-blinded, between-subject study. 90 healthy adults performed a motivational Go/NoGo learning task during fMRI after receiving cTBS (n=45) or sham (n=45) stimulation. Behaviourally, cTBS reduced overall Go response rates and slowed reaction times. Computational modelling showed a decrease in the learning rate, selectively for rewarding outcomes. Computational fMRI analysis showed stronger reward prediction error (RPE) signals in the vmPFC, mediodorsal thalamus, and left dorsal striatum after cTBS, without changes in neural signals at cue onset. These findings provide causal evidence that the vmPFC drives the action-invigoration component of the Pavlovian bias. Significance statementWe causally link ventromedial prefrontal cortex (vmPFC) function to motivated action in humans. In a single-blind cTBS-fMRI study (n=90), transient vmPFC inhibition reduced Go responding, slowed reaction times, and selectively lowered the positive learning rate early in learning. At the same time, reward-prediction-error signals increased in mediodorsal thalamus, dorsal striatum, and vmPFC, while cue-onset value/choice signals were unchanged. Together, the results indicate that vmPFC normally speeds value updating and invigorates actions; when inhibited, control shifts toward slower, subcortically supported learning under higher uncertainty.