认知与发展心理学研究室知识库

个体出于某种目的，表现出与内心情绪感受不一致的表情即为伪装表情。探究伪装表情的特征在社会交往、公共安全、临床等诸多应用领域均具有非常重要的研究意义。以往关于伪装表情面部线索与特征的研究，主要关注情绪泄漏、不对称性和表情的持续时间等，并未比较真实表情与伪装表情两者之间动作单元(action units, AUs)的异同，也未有研究考察伪装表情表达的认知神经机制。 同时，在不同情绪状态下，个体自主选择时会倾向于做出何种伪装表情也尚不 清楚。本文采用“情绪性说谎范式”，诱发被试的情绪感受，并要求其做出指 定的或者自主选择做出与情绪感受一致或不一致的表情，通过两个研究(包括 4 个视频材料评定、2 个行为实验和 2 个脑电实验)较为系统地考察伪装表情表达的面部动作特征及其认知神经机制。

研究一包括 2 个视频材料评定和 2 个行为实验，重点考察基于 AUs 的伪装 表情表达的面部动作特征。首先，招募被试对情绪视频片段进行评定(材料评定 A)，并基于评定结果选取了诱发六种基本情绪的视频片段各 6 个，合计 36 个， 用于实验 1 中诱发情绪的材料。实验 1 旨在考察被试在情绪诱发时做出的真实和伪装表情的面部动作特征及其差异。被试在实验室场景下观看不同类型的情 绪性视频片段，并被要求做出指定的与诱发情绪感受一致和不一致的表情。实验 1 的结果发现，不同情绪状态下做出高兴表情共同的 AUs 有嘴角上扬的面部 动作单元(AU12)，同时也有一些特异性的 AUs。例如在诱发悲伤、惊讶和愤怒情绪条件下做高兴表情会出现和表情控制相关的 AU14；与高兴情绪条件下做高兴表情相比，悲伤情绪条件下做出高兴表情的 AU6 的出现频数显著更少。这些发现深化了对“假笑”的认识，表明了可能区分真实和伪装表情的面部线索。

增加中性视频片段，再次招募被试对情绪性视频片段进行评定(材料评定 B)， 并基于评定结果选取了诱发高兴、悲伤、愤怒和中性情绪的片段各 4 个，合计

16 个，用于实验 2 中诱发情绪的材料。实验 2 旨在考察被试在诱发出不同情绪感受时，被要求真实表达与其情绪感受一致的表情以及自主选择表达某种表情来掩饰其情绪感受，真实表情和自主选择表达的伪装表情的面部动作特征及差 异。被试在实验室场景下观看不同类型的情绪视频片段，并被要求做出与诱发 情绪感受一致和不一致的表情(此时被试可以选择做出哪种伪装表情)。实验 2 的结果发现，在诱发悲伤和愤怒情绪条件下个体多倾向于用高兴表情或中性表情来掩饰其情绪感受，在诱发中性情绪条件下个体多倾向于用高兴表情来掩饰其情绪感受；在诱发悲伤情绪和中性情绪条件下做高兴表情时会出现 AU14 的面部动作单元；与真实表情条件相比，伪装表情条件下共同出现的 AUs 之间连接网络更复杂，共同且相互联系紧密的核心 AUs 更多。实验 1 和实验 2 的结果均发现伪装表情条件出现了真实情绪感受对应的表情 AUs 泄露的现象，这表明做伪装表情时可能泄露出真实表情的部分 AUs。实验 2 结果表明，高兴表情通常被用来掩饰负性情绪和中性情绪，除了 AU14 外，真实和伪装的高兴表情的面部线索还可通过 AUs 之间网络连接的复杂性来区分。

研究二包括 2 个视频材料评定和 2 个脑电实验，考察伪装表情表达的认知神经机制。首先，增加视频片段数量并缩短时长，再次招募被试对高兴和悲伤情绪的视频片段进行评定(材料评定 C 和 D)，并基于评定结果选取了诱发高兴和 悲伤情绪的视频片段各 45 个，合计 90 个，用于实验 3 中诱发情绪的材料。选取 45 段悲伤视频，9 段高兴视频用于实验 4 中诱发情绪的材料。实验 3 旨在考察诱发高兴或悲伤情绪的条件下，被要求做出真实和伪装表情表达的脑电特征的异同。被试在实验室场景下观看高兴或悲伤的情绪性视频片段，并被要求做出指定的与诱发情绪感受一致和不一致的表情。结果发现，在诱发悲伤情绪条件下，与真实表情相比，做出与情绪诱发不一致的伪装表情的伴随性负波(Contingent negative variation, CNV)波幅以及 alpha 的事件相关去同步化(EventRelated Desynchronization, ERD)更强。这些结果表明，和真实表情相比，悲伤情绪诱发下，个体表达伪装表情的认知过程中表现出更强的有意控制和冲突抑制。

实验 4 旨在考察诱发悲伤情绪条件下，自主选择的伪装表情及其脑电特征。被试在实验室场景下观看诱发悲伤的情绪性视频和少量的高兴视频(用于调节情绪，其数据不纳入分析)，并被要求在每次观看后自主选择做出一种与情绪感受不一致的表情。结果发现，个体更倾向于选择做出高兴的伪装表情，且此时自主选择做出的高兴表情比做出除高兴和中性表情外的其他表情(惊讶、愤怒、厌恶、恐惧和轻蔑)诱发更小的 N2 波幅。该结果说明做出高兴伪装表情时可能需要更少的冲突监控和抑制成本。进一步比较实验 3 和 4 中悲伤情绪诱发时做出高兴伪装表情的脑电指标的异同，结果表明自主选择时可能调动更少的认知控 制资源解决冲突。

综上所述，本研究发现了区分真实表情和伪装表情的 AUs 的关键特征和线索，并揭示了伪装表情表达时关于冲突抑制、有意控制等认知过程的神经基础。本研究深化了对伪装表情表达的面部动作特征和认知神经机制的认识，并为识别真实表情和伪装表情以及研发计算机自动识别系统提供了心理学依据。

The expression an individual displays for a certain purpose that does not inconsistent with their emotional state is called "masked facial expression". Investigating the characteristics of masked expressions holds significant research implications across various application domains such as social interactions, public safety, and clinical settings. Previous studies on masked facial expression have primarily focused on facial cues and features, emphasizing emotion leakage, asymmetry, and the duration of expressions. However, there has been a lack of comparison between the action units (AUs) of true and masked expressions, as well as a dearth of research investigating the cognitive neurobiological mechanisms underlying producing masked expression. Furthermore, it remains unclear which masked facial expressions individuals voluntarily choose to display across different emotional states. Therefore, this study employs an "emotional lies paradigm" to elicit emotional states in participants, requiring them to produce specified or voluntarily chosen expressions consistent or inconsistent with their emotional experience. Through two studies comprising four video-based rating, two behavioral experiments, and two electroencephalogram (EEG) experiments, we systematically investigate the facial action features of masked expression and their associated cognitive neural mechanisms.

The study 1 comprising two video-based rating and two behavioral experiments, focuses on investigating the facial action characteristic of masked facial expressions based on action units (AUs). Initially, participants were recruited to rate emotion video clips (Material Rating A), and based on these ratings, six video clips inducing each of the six basic emotions were selected, totaling 36 clips, which were used as materials to induce emotions in Experiment 1. Experiment 1 aimed to investigate the facial action characteristics and differences between true and masked expressions evoked by various emotional states. Participants were required to view various emotional videos in the laboratory scene and produce specific facial expressions that are consistent with and inconsistent with the evoked emotions. The results of Experiment 1 revealed that the expression of happiness across different emotional states shares the facial action unit (AU) of Lip Comer Puller (AU12). However, there are also specific AUs present, such as AU14 related to expression control, observed when expressing happiness in conditions inducing sadness, surprise, and anger. Furthermore, compared with happy expression in happy emotion condition, the occurrence frequency of AU6 in happy expression condition in sad emotion condition was significantly less. The findings enhance our understanding of "masked smiles," revealing facial cues that may differentiate true and masked expressions.

Neutral video clips were added, and the subjects were recruited to evaluate the emotional video clips (material assessment B). Based on the evaluation results, 4 clips each of happiness, sadness, anger and neutral emotions were selected, and 16 clips in total were used to induce emotions in experiment 2. Experiment 2 aimed to investigate the facial action features and differences between true and voluntarily chosen masked expressions under various emotional states, where participants were required to truly express emotions or voluntarily select a facial expression to conceal their emotional state. Participants viewed various emotionally video clips in a laboratory setting and were instructed to mimic expressions consistent and inconsistent with the induced emotions (at this point, participants voluntarily chose which masked expressions to make). The results of Experiment 2 revealed that individuals tended to mask their emotional states induced by sadness and anger with happy or neutral expressions, and in the case of inducing neutral emotions, individuals leaned towards using happy expressions to conceal their emotional states. When displaying happy expressions under induced sadness and neutral emotion conditions, the facial action unit AU14 was observed. Compared to true expression conditions, the network of connected action units (AUs) was more complex in masked expression conditions, with a greater number of core AUs appearing jointly and interrelated closely. Both Experiment 1 and Experiment 2 found instances of leaked facial action units corresponding to true emotional experiences during masked expression conditions, indicating the potential leakage of some AUs corresponding to true expressions when producing masked facial expression.

The study 2 comprising two video-based rating and two EEG experiments, focuses on investigating cognitive neural mechanism of producing masked facial expressions. Initially, participants were recruited to rate emotion video clips (Material Rating A), and based on these ratings, six video clips inducing each of the six basic emotions were selected, totaling 36 clips, which were used as materials to induce emotions in Experiment 1. Firstly, by increasing the number of video clips and shortening their duration, participants were recruited to evaluate video clips inducing happy and sad emotions (Materials Evaluation C and D). Based on the evaluation results, 45 video clips inducing happy emotions and 45 inducing sad emotions were selected, totaling 90 clips. These clips were utilized in Experiment 3 as materials to induce emotions. 45 video clips inducing sad emotions and 9 inducing happy emotions were selected and these clips were utilized in Experiment 4 as materials to induce emotions. Experiment 3 aimed to investigate the similarities and differences in the neural correlates of participants when asked to display true and masked facial expressions under conditions inducing happy or sad emotions. Participants were situated in a laboratory setting where they view emotionally video clips inducing either happiness or sadness and are instructed to exhibit specified expressions that are either consistent or inconsistent with the induced emotional state. The results revealed that under conditions inducing sad emotions, compared to true expressions, the contingent negative variation (CNV) amplitudes and alpha event-related desynchronization (ERD) were greater when participants produced masked expressions incongruent with the induced emotion. These findings suggest that compared to true expressions, under conditions inducing sad emotions, the cognitive processes involved in expressing masked expressions entail stronger intentional control, conflict inhibition.

Experiment 4 aimed to investigate the voluntarily chosen masked facial expressions and their neural correlates under conditions inducing sad emotions. Participants were situated in a laboratory setting where they view emotionally videos inducing sadness and a small number of happy videos (utilized for emotion regulation purposes and excluded from analysis) and were instructed to voluntarily select a facial expression inconsistent with their emotional state after each viewing. The results indicated that individuals are more tended to choose to display happy masked expressions under conditions inducing sad emotions, and during these instances, voluntarily selected happy expressions elicit smaller N2 amplitudes compared to expressions other than happy and neutral (e.g., surprise, anger, disgust, fear, and contempt). This suggests that displaying masked happy expressions may involve lower conflict monitoring and inhibition costs. Further comparisons of the neural indices associated with producing masked happy expressions under sad emotion induction between Experiments 3 and 4 revealed that voluntary selection may engage fewer cognitive control resources to resolve conflicts.

In summary, this study has identified key clues in distinguishing true from masked facial expressions based on distinctive Action Units (AUs) features, shedding light on the cognitive processes underlying masked expression in terms of conflict inhibition, intentional control, and related mechanisms. This study has advanced our understanding of the characteristics and cognitive neural mechanisms underlying masked facial expressions, providing a psychological foundation for distinguishing true from masked expressions and developing computerized automatic recognition systems.