Flexible Learning of Cue-Reward Associations in Human Ventromedial Prefrontal Cortex
Material below summarizes the article Dissociating Value Representation and Inhibition of Inappropriate Affective Response During Reversal Learning in the Ventromedial Prefrontal Cortex, published on December 29, 2015, in eNeuro and authored by Zhihao Zhang, Avi Mendelsohn, Kirk F. Manson, Daniela Schiller, and Ifat Levy.
When you are hungry, the smell of a hearty meal is a cue for potential reward. The same smell, however, may not be perceived as much of a cue for reward at the end of a big meal because the meal has reduced the value of food. Other cues, such as the smell of coffee, may become reward-predictive instead. To allow such flexible updating of values, the brain needs to encode current value and inhibit value representations that are no longer valid.
The first function, value encoding, has been linked to the ventromedial prefrontal cortex (vmPFC) in numerous studies. Intriguingly, a separate line of research has associated the same brain area with the inhibition of learned fear responses that are no longer needed, for example when a visual cue that used to predict electric shock ceases to do so. This activation pattern may reflect relative value encoding — a cue that used to predict fear and does not do so anymore has acquired relative positive value — but it could also reflect a separate function, namely the inhibition of prior learning.
To distinguish between these two potential functions of vmPFC, we used an appetitive reversal-learning task in an fMRI experiment with 18 human participants. The task began with an acquisition stage, in which one visual stimulus (a colored square, ‘color A’) was sometimes followed by an image of monetary reward (conditioned stimulus, CS+), and another stimulus (CS-, color B) was never followed by reward. Then, with no explicit signal, the reinforcement contingencies switched (reversal stage), such that color B was now sometimes followed by reward (new CS+), and color A was not (new CS-). Participants were asked to rate, upon each color presentation, how likely they thought it was for a reward image to appear on the next screen. At the end of the experiment participants received the accumulated monetary rewards that they have seen throughout the experiment.
Consider color A. During the acquisition stage, this color predicted reward and therefore acquired positive value. In the reversal stage, color A’s value decreased because it was no longer followed by reward. This enabled us to distinguish between the two potential functions of vmPFC: a region that encodes value would decrease its activity to color A during reversal, reflecting its decreased value; conversely, a region involved in inhibition of prior learning will increase its activity to color A, since the learned reward response for this cue is no longer appropriate.
Participants acquired the appropriate cue-reward associations during the acquisition stage and then successfully reversed those associations during the reversal stage. To look for brain regions whose activity is compatible either with value encoding or with inhibition of prior learning, we first conducted a whole-brain analysis.
This analysis revealed two adjacent regions with distinct activation patterns within vmPFC. The first region was relatively dorsal, and its activity was modulated by reward magnitude, compatible with value encoding. The second region was more ventral, and, during late reversal, exhibited preferential activation to the new CS- (the cue that predicted reward and ceased to do so) compared to the new CS+ (the newly reward-predictive cue).
This activity pattern in the ventral region is consistent with inhibition of prior learning. If this region is involved in inhibition of prior learning, then its activation magnitude should be associated with the strength of behavioral inhibition across individuals.
This is exactly what we found. Participants who had stronger activation to color A during the reversal stage reduced their rating of color A more between the acquisition and reversal stages. We did not find any such association between activity to color B and change in its predictive value (from non-predictive to positively predicting reward), suggesting that the ventral part of vmPFC has a specific role in inhibiting responses to stimuli that used to predict reward, rather than a more general role in updating any responses.
Our study merges two previously separate lines of research showing the co-existence of two distinct functions, value encoding and inhibition of prior learning, in adjacent sub-regions of vmPFC. This result could be important for any condition in which inappropriate reward representations should be suppressed, for example substance abuse or overeating.
It will be interesting to explore the potential role of the different sub-regions of vmPFC in these conditions, and the possibility of training these sub-regions in order to change behavior.
Dissociating Value Representation and Inhibition of Inappropriate Affective Response During Reversal Learning in the Ventromedial Prefrontal Cortex. Zhihao Zhang, Avi Mendelsohn, Kirk F. Manson, Daniela Schiller, Ifat Levy. eNeuro Dec 2015, 2 (6) DOI: 10.1523/ENEURO.0072-15.2015