Multivoxel Neural Reinforcement Changes Resting-State Functional Connectivity Within the Threat Regulation Network

Using neural reinforcement, participants can be trained to pair a reward with the activation of specific multivoxel patterns in their brains. In a double-blind placebo-controlled experiment, we previously showed that this intervention can decrease the physiological reactivity associated with naturally feared animals. However, the mechanisms behind the effect remain incompletely understood and its usefulness for treatment remains unclear. If the intervention fundamentally changed the brain responses, we might expect to observe relatively stable changes in the functional connectivity within the threat regulation network. To evaluate this possibility, we conducted functional magnetic resonance imaging (fMRI) sessions while subjects were at rest, before and after neural reinforcement, and quantified the changes in resting-state functional connectivity accordingly. Our results indicate that neural reinforcement increased the connectivity of prefrontal regulatory regions with the amygdala and the ventral temporal cortex (where the visual representations of phobic targets are). Surprisingly, we found no evidence of Hebbian-like learning during neural reinforcement, contrary to what one may expect based on previous neurofeedback studies. These results suggest that multivoxel neural reinforcement, also known as decoded neurofeedback (DecNef), may operate via unique mechanisms, distinct from those involved in conventional neurofeedback.