【作者】 古焱； 杨雪； 胡学平； 陈安涛；

【Author】 Yan Gu;Xue Yang;Xueping Hu;Antao Chen;Faculty of Psychology, Southwest University;

【机构】 西南大学心理学部；

【摘要】 对反馈结果的评价能够优化人类的决策,促进行为调整和环境适应。强化学习理论提出,事件相关电位上的反馈相关负波(feedback-related negativity,FRN),会受到源于中脑多巴胺投射到前扣带皮质(ACC)的奖赏预期误差(reward prediction error, RPE)的调节,反映了实际结果比预期的好(+PE)或者坏(-PE)的程度。然而,先前的研究通常将反馈结果简单地二分为好与坏(正确或错误,赢或输),并没有变化预期误差值的大小,因此,FRN与预期误差的关系仍存在争议。大量的研究还指出,FRN更可能反映的是实际结果与预期的偏离,奖励和惩罚的意外消失都引发了更负的偏转。前人关于FRN的研究主要是基于脑电信号的平均,这种单一波幅的简单效应可能存在其他成分的叠加,使获得的FRN成分已经失真。在本研究中,我们分别记录了金钱收益与损失的脑电活动,并通过变化预期误差值的大小和方向,深入探究预期误差对反馈评价的调节。此外还采用主成分分析(PCA)方法对可能的重叠成分进行了分离。结果发现,当实际的收益比预期的少(-PE)时,在脑电波上产生了更负的偏转;更为重要地是,当实际的损失比预期的少(+PE)时,也产生了更负的偏转。进一步的PCA结果显示,收益条件下的-PE值与损失条件下的+PE值,都与FRN呈显著的正相关,证明收益与损失的反馈评价具有相似性,主要反映了实际结果比预期的多还是少。更多还原

【Abstract】 Evaluation of feedback outcomes can optimize our choice, facilitate behavior adjustment and environmental adaptation. The dominant reinforcement-learning theory contends that the feedback-related negativity(FRN), a frontocentral negative deflection in event-related potentials, is modulated by the reward prediction error(RPE) derived from the midbrain dopamine projecting to the anterior cingulate cortex(ACC), reflecting the degree to which outcomes are better than expected(+PE) or worse(-PE). However, previous studies usually roughly divided the feedbacks into dichotomous good and bad outcomes(right or wrong, win or loss), and the PE size was not varied, making the relationship between FRN and the PEs still controversial. Mounting evidence also point out that the FRN is more likely to reflect the violation of expectancy, for unexpected omissions of reward and punishment both induce more negative deflections. Previous FRN researches were mostly based on the simple averaging waveform, which were susceptible to the overlap from other components and distorted the obtained FRN. In this study, we recorded the EEG signals related to the monetary gain and loss separately, and parametrically varied PE size to further investigate the PE modulation on feedback evaluation. Additionally, the principal components analysis(PCA) was implemented to separate the potential overlying components out. As expected, the results identified a more negative deflection when actual gain was less than expected(-PE); strikingly, less than expected loss(+PE) also yielded more negative deflection. The PCA procedure revealed a positive correlation between FRN and –PE for gain, and + PE for loss. We suggest that feedback evaluation underlying the gain and loss is similar, which may encode "more or less than expected" prediction errors.更多还原