ISSN 1671-3710
CN 11-4766/R
主办:中国科学院心理研究所
出版:科学出版社

心理科学进展 ›› 2022, Vol. 30 ›› Issue (4): 834-850.doi: 10.3724/SP.J.1042.2022.00834

• 研究前沿 • 上一篇    下一篇

预期错误在恐惧记忆更新中的作用与机制

李俊娇1, 陈伟2,3,4, 石佩2,3,4, 董媛媛2,3,4, 郑希付2,3,4()   

  1. 1广东第二师范学院教师教育学院, 广州 510303
    2华南师范大学心理学院, 广州 510631
    3华南师范大学心理应用研究中心, 广州 510631
    4广东省心理健康与认知科学重点实验室, 广州 510631
  • 收稿日期:2021-06-25 出版日期:2022-04-15 发布日期:2022-02-22
  • 通讯作者: 郑希付 E-mail:zhengxifu@m.scnu.edu.cn
  • 基金资助:
    国家自然科学基金项目(32000752);国家自然科学基金项目(31970996);广东省哲学社会科学规划项目(GD19YXL01);广东省教育科学“十三五”规划项目(2019JKDY025);教育部人文社会科学研究项目(20YJC190009)

The function and mechanisms of prediction error in updating fear memories

LI Junjiao1, CHEN Wei2,3,4, SHI Pei2,3,4, DONG Yuanyuan2,3,4, ZHENG Xifu2,3,4()   

  1. 1College of Teacher Education, Guangdong University of Education, Guangzhou 510303, China
    2School of Psychology, South China Normal University, Guangzhou 510631, China
    3Center for Studies of Psychological Application, South China Normal University, Guangzhou 510631, China
    4Guangdong Key Laboratory of Mental Health and Cognitive Science, Guangzhou 510631, China
  • Received:2021-06-25 Online:2022-04-15 Published:2022-02-22
  • Contact: ZHENG Xifu E-mail:zhengxifu@m.scnu.edu.cn

摘要:

依据错误驱动的学习理论, 行为预期结果与实际结果之间的不匹配即预期错误(Prediction error, PE)是学习产生的驱动因素。作为显著性信息中的一种, 预期错误和物理显著性、惊讶、新异性等存在信息加工阶段的不同, 与记忆更新的关系也有差异。近年来, 记忆再巩固干预范式(reconsolidation interference)被证明可用于人类条件性恐惧记忆的更新, 其中记忆提取激活阶段所包含的预期错误起到了引发记忆“去稳定”、开启记忆再巩固的关键作用。在促进恐惧记忆更新的行为机制上, PE被认为是记忆去稳定的必要非充分条件。记忆提取必须包含适量的PE, 但其引发的是记忆去稳定、消退还是中间状态, 还需结合记忆本身性质确定。在促进恐惧记忆更新的神经机制上, 杏仁核、导水管周围灰质(PAG)、海马均在PE探测和计算过程中具有重要作用; 前额叶皮层(PFC)及其亚区在PE开启记忆再巩固过程中扮演了重要角色。上述过程又受到神经系统中特定神经递质的重要调节, 尤其是多巴胺能和谷氨酸能。未来研究应进一步探索基于PE计算模型的量化研究, 整合PE与其他边界条件的交互作用, 考察不同类型显著性在记忆再巩固中的作用等; 并亟待使用多学科手段探索PE在恐惧记忆更新中作用的神经与分子机制。同时, 需进一步开展PE作用的个体差异研究, 促进研究结果向临床应用转化。

关键词: 预期错误, 条件性恐惧, 记忆更新, 再巩固, 提取干预范式

Abstract:

The error-driven learning theory believes that the reinforcement brought by the stimulus must be surprising or unpredictable for the individual to form learning. The mismatch between the expected consequence of behavior and the actual result, known as prediction error (PE), is the driving factor of learning, according to this theory. The Rescorla & Wagner model, the Pearce-Hall model, and the temporal difference (TD) model are the three most common models for calculating prediction error. The RW model and the TD model, in particular, have had a significant impact on the research of prediction error-driven learning and memory. Under different learning models, prediction error is classified as reward or punishment prediction error (RPE or PPE); positive or negative prediction error; and singed or unsigned prediction error (SPE or UPE). As a type of salience, PE is different from other types of saliences. Salience includes stimulus novelty, valence evaluation, stimulus rareness and other salience. Physical salience, surprise (unexpected novelty), and expected novelty are all types of novelty, but only unexpected novelty can promote dopamine release; physical salience with no direct rewards can only cause a short spike in dopamine. Prediction error, on the other hand, are mostly related to the recognition, result perception, and valence evaluation processes.

A large body of work investigated the role of prediction error in the formation and updating of fear memory. Firstly, prediction error is considered to be a necessary factor in the process of fear acquisition. Negative PE is the source of successful fear extinction. Secondly, under the framework of Reconsolidation Interference of conditioned fear memory, prediction error is demonstrated as a necessary condition of memory destabilization. Prediction error plays a key role in fear memory reconsolidation: (1) PE during memory reactivation is an important boundary condition for memory destabilization; (2) PE is a necessary but not sufficient condition for the triggering of reconsolidation. The degree of prediction error determines whether or not the memory can become unstable. PE's role in memory updating has progressed from a qualitative to a quantitative examination, which has become an important measure of such research development; (3) The size of PE required for memory destabilization is proportional to the memory’s original strength. Thus, it is critical to take both sides into account when evaluating a retrieval manipulation. We propose an integrated model of retrieval boundary conditions and memory features for the reconsolidation of fear memories based on these studies and previous models.

However, until recently, the neural mechanism underlying the involvement of prediction error in fear memory update has remained largely elusive. Recent work has revealed the brain areas involved mainly include the amygdala, ventrolateral peri-aqueduct gray matter (vlPAG), hippocampus, and prefrontal cortex (PFC). The PFC, in particular, is a distinct area that may distinguish the fear extinguish with or without reconsolidation. While a range of neurotransmitters are linked to the role of PE in memory destabilization in terms of neuromodulation in brain circuits, the most significant of which is dopaminergic. However, glutamate’s participation in the same process is also worth mentioning.

We propose that in the future direction of the research on fear memory updating, further exploration should be made on quantitative research based on the PE calculation model, integrating the interaction between PE and other boundary conditions, and investigating the role of different types of saliences in memory reconsolidation. Importantly, multidisciplinary methods are urgently need to be used to investigate the neural and molecular mechanisms of PE's role in fear memory renewal. Individual differences in the effects of PE, on the other hand, must be investigated in order to facilitate the translation of studies from bench to bedside.

Key words: prediction error, fear conditioning, memory updating, reconsolidation, reconsolidation interference paradigm

中图分类号: