| When facing the environment,the human brain needs to constantly monitor the surrounding environment and predict the upcoming events,then compare the actual outcome with expected value when the feedback is delivered,which will lead to rapid adjustment of behavior and facilitate environment adaptation.The deviation between the expected value and actual value is usually called as prediction error(PE).Thus,the feedback evaluation,especially the processing of prediction errors,is playing an important role in human cognitive control and behavior adaptation.There has been a long-term focus on feedback processing.The early event-related potential(ERP)study of feedback evaluation found that,a negative deflection would arise in event-related potentials when the feedback implied incorrect performance(i.e.,feedback-related negativity,FRN).The FRN reflects the human brain's perception about the error information of external feedback,being more sensitive to negative feedback.The negative feedback will induce more negative FRN than positive feedback.A dominant reinforcement-learning error-related negativity theory(RL-ERN theory)contends that the FRN is modulated by the reward prediction errors(RPE)derived from midbrain dopaminergic neurons projected to the anterior cingulate cortex(ACC).Therefore,the FRN reflects the human brain's evaluation of the feedback valence(positive feedback and negative feedback),for reward and punishment events,there will produce opposite prediction error signal.Whereas,some researchers believe that the FRN is more likely to reflect the salience of feedback,as they find that unexpected positive feedback and unexpected negative feedback will both induce significant FRN components.In recent years,a growing body of literature has found that the feedback processing may be indeed modulated by the salience prediction error(SPE)signal,and point out that the activities of dopamine neurons are diverse.Furthermore,as the prediction of response-outcome(PRO)theory states,the medial prefrontal cortex(m PFC),particularly the neural activities of ACC,are more sensitive to the deviation of expectation,rather than feedback valence.In addition,a large amount of evidence from neuroimaging has also demonstrated that when outcomes are delivered unexpectedly,the striatum,anterior insula and occipital lobe will emerge significant activities,regardless whether the feedback is positive or negative.Hence,the neural mechanism of feedback processing is still controversial.We also find that reward and punishment have opposite hedonic valence but both are salient events for our survival,so it is difficult to differentiate the salience from valence in those studies investigating feedback processing using only appetitive or only aversive events.Meanwhile,previous studies mainly manipulate the priori probability of the outcomes,or vary the outcome's frequency across blocks of trials,and assume that the outcome is unexpected in low probability or infrequent condition,instead of testing the prediction directly,so that the prediction error's influence on feedback is still unclear.Based on the above problems,in order to clearly explore the neural basis of feedback processing and prediction error's modulation of it,we adopted the functional magnetic resonance imaging(f MRI)approach,including appetitive money and aversive electric shock in the design to investigate whether the neural sources of positive feedback and negative feedback were the same or not.We hypothesized that the brain regions showing activities for both reward and punishment were more likely to reflect the salience of feedback.On this basis,we further varied the prediction errors size and directly compared the FRN component's difference between monetary gain and monetary loss,investigating that the feedback processing is modulated by reward prediction errors,or salience prediction errors.Experiment 1 adopted a probabilistic guessing task and simultaneously applied high spatial resolution f MRI approach to explore the human brain's feedback processing with monetary reward and pain shock in one design,which could reveal the neural sources of positive feedback and negative feedback,and demonstrated that the prediction errors' modulations for reward and punishment conditions were opposite(reward prediction error)or the same(salience prediction error).Different from other studies,we informed the participants of the reward and pain probabilities and explicitly instructed them to make a prediction about outcomes at the cue phase,subsequently press the corresponding buttons to obtain accurate predictions,directly investigating the predictions of participants.Through the direct test of predictions,we could accurately distinguish the unexpected outcomes from expected outcomes,and calculate the prediction errors directly according to the expected values and actual values,avoiding indirectly infer using models with priori hypotheses.Our behavioral results found that the participants' predictions of outcomes were mainly based on the probabilities of cues.Our results showed that bilateral fusiform gyrus,right middle frontal gyrus and left cingulate gyrus were positively modulated by prediction errors under both money and pain conditions,showing the same patterns.While the reward prediction error hypothesis(opposite modulation patterns)did not find any result.This implies that some brain areas of feedback evaluation are more likely to be modulated by salience prediction errors,rather than reward prediction errors.Additionally,we note that the previous studies mainly adopt a time estimation task,probability or gamble task,leading to dichotomous results(right or wrong)and hardly varying the prediction error size.On the basis of experiment 1,we further manipulated the expected value and actual value,requiring the participants to slide in a 0-20 scale at the expected stage and guess the money would be gained or lost.The actual results would be informed at the feedback stage,resulting in the varying prediction errors and studied the prediction error modulation of feedback in such more realistic situation.Similarly,the experiment 2 also contained reward and punishment conditions,but just using monetary gain and monetary loss to make a direct comparison of them.By comparing the FRN components between monetary gain and loss directly,we could demonstrate whether the feedback was modulated by reward prediction error or salience prediction error.Experiment 2 explored the changes in FRN directly by using the ERP techniques.As a result,it was found that,in the case of gain condition,a more negative deflection was produced when actual money was less than expected amount(negative feedback).Importantly,for monetary loss,when actual money was less than expected amount(positive feedback),also produced a more negative deflection.The results of difference wave showed that an obvious FRN was got for gain condition,while a positive FRN for loss condition.Our topographic maps found that the FRN components were similar under the two conditions,which were mainly located in middle frontal site extending to more distributed posterior parietal cortex.The research further proves that the feedbacks of monetary gain and monetary loss are similar and are mainly modulated by salience prediction error.In summary,our research presents the evidences from f MRI and ERP showing that the human brain's feedback processing exerts similar patterns under both reward and punishment conditions,which demonstrates that the feedback evaluation mainly focuses on the salience of feedback,rather than valence,and indicates that the feedback processing is more likely to be modulated by salience prediction error. |
PDF Full Text Request