健康与遗传心理学研究室知识库

伪装表情指个体在特定情境下有意识地调控面部表情，使表情的外在表现与内在情绪体验分离的心理行为。通过选择性激活或抑制特定面部动作单元，形成与真实情绪不一致的面部肌肉活动模式。伪装表情的表达与识别受到社会文化规范与情境性动机的双重驱动，在社会交往和临床诊断中具有重要意义。尽管面部动作编码系统为伪装表情的量化分析提供了方法，但伪装表情的具体面部动作单元(action unit, AU)特征，尤其是影响个体对伪装表情识别的因素仍存在一定的研究空白。本研究通过两项递进式实证研究(包括 4 个行为实验)系统探讨伪装表情的面部特征及识别影响因素，重点分析真实与伪装高兴表情的关键区分线索及面 孔区域加工策略。

研究一包括两个行为实验，以 MFED 伪装表情数据库中6种表情类型下的648 段伪装表情视频为材料，使用 OpenFace 进行AU特征提取后分析伪装表情的 AU 特征，发现不同情绪的伪装表情呈现独特的 AU 共现模式。实验1要求被 试完成二择一任务(判断表情是真实或伪装表情)，探究表情类型对伪装表情识别准确性的影响，以及在识别正确率水平不同时，AU特征之间的差异。随后实验2 进一步探究个体能否准确识别伪装表情中被掩饰的真实情绪类型，实验素材与实验 1 相同，被试完成六择一迫选任务，从六种基本情绪类型中选择伪装表情中被掩饰的真实情绪类型，并分析不同情绪类型间是否存在识别混淆的现象，进一步证明 AU 共现模式以及特征的相似性是导致识别混淆的关键因素。

研究二也由两个行为实验组成，在研究一发现“高兴伪装表情具有最高识别正确率”且“愤怒-厌恶和恐惧-惊讶存在识别混淆”的基础上，研究二聚焦于真 实与伪装高兴表情的核心区分线索。实验 3 对真实及伪装高兴表情的 AU 特征以及影响高兴表情识别的因素展开分析。实验中每个被试完成 4 组评定任务，每组评定任务包含 16 个试次(8 真实高兴表情、8 伪装高兴表情)，共 64 个试次。实验材料选自 MFED 数据库中的高兴表情素材，共 16 段真实高兴表情视频，32 段伪装高兴表情视频(8 段悲伤高兴表情、8 段厌恶高兴表情、8 段恐惧高兴表情、8 段愤怒高兴表情)，去除视频素材中与面部表情无关的因素，只保留了情绪面孔。实验 3 结果发现真实和伪装高兴表情之间 AU6 和 AU12 的强度存在差异，并且AU12 和 AU6 的强度变化对识别时的判断及各维度评分产生不同影响，个体对高兴表情的识别存在对嘴部区域尤其是 AU12 的依赖。

实验 4 通过对面孔区域进行分离，将实验材料中的情绪面孔分为上面部、下面部及完整面孔。检验当面部信息被限制在局部区域时，AU6(上面部)和 AU12(下面部)是否可以对高兴表情的识别产生影响。实验素材选自 MFED 数据库中的高兴表情素材，共 16 段真实高兴表情视频，16 段伪装高兴表情视频。对比完整面孔与局部面孔条件下的判断差异，实验结果表明，当下面部和完整面孔可见时， 下面部的 AU12 因视觉显著性成为表情判断的主导线索，而上面部的 AU6 仅在局部面孔时发挥作用。

综上所述，本研究系统揭示了伪装表情中关键面部动作单元的特征差异及表情识别影响判断的因素，尤其深入分析了高兴表情的 AU 特征及 AU6 和 AU12特征对高兴表情识别的影响，结果也表明高兴表情识别受面孔区域可视性调节。研究将进一步深化个体对伪装表情的理解，推动情绪调节和社交行为研究的发展。

Masked expression refers to the psychological behavior in which individuals consciously regulate their facial expressions in specific situations, resulting in a dissociation between the external expression and internal emotional experience. By selectively activating or suppressing specific facial action unit (AU), individuals form facial muscle activity patterns that do not match their true emotions. The expression and recognition of masked expressions are jointly influenced by sociocultural norms and situational motivations, playing a crucial role in social communication and clinical diagnosis. Although the Facial Action Coding System (FACS) provides a methodological basis for quantifying masked expressions, there remains a gap in understanding the specific AU characteristics of masked expressions and the factors influencing their recognition.

This study employed two progressive empirical studies, including four behavioral experiments, to systematically investigate the AU features of masked expressions and the factors that influence their recognition. Particular focus was given to distinguishing cues between genuine and masked happy expressions, as well as facial region processing strategies.

Study 1 consisted of two behavioral experiments using 648 masked expression video clips across six emotion types from the MFED database. AU features were extracted using OpenFace, revealing unique AU co-occurrence patterns across different masked emotions. Experiment 1 employed a two-alternative forced-choice (2AFC) task where participants judged whether an expression was genuine or masked, analyzing how expression type influenced recognition accuracy and AU differences across accuracy levels. Experiment 2 further explored participants’ ability to identify the hidden true emotion behind masked expressions using a six-alternative forced-choice (6AFC) task, and examined confusion patterns across emotions. The results supported the role of AU co-occurrence and similarity in recognition confusion.

Building on findings of experiment 1 that masked happy expressions had the highest recognition accuracy and that anger-disgust and fear-surprise were easily confused, Study 2 focused on the core distinguishing cues of happy expressions. Experiment 3 analyzed the AU features of genuine and masked happy expressions and their effects on recognition. Each participant completed four evaluation blocks (64 trials in total) using 16 genuine and 32 masked happy expression videos from the MFED database. Results showed significant AU6 and AU12 intensity differences between genuine and masked happy expressions. AU12, and to a lesser extent AU6, played distinct roles in recognition and subjective ratings, with participants relying more heavily on the mouth region (especially AU12) in evaluating happy expressions.

Experiment 4 examined whether AU6 (upper face) and AU12 (lower face) influence recognition when facial information is restricted to local regions. Expression stimuli were divided into upper-face, lower-face, and whole-face conditions. Results showed that when the lower face or full face was visible, AU12 served as the dominant cue due to its visual salience. AU6 influenced recognition only when the upper face was presented in isolation.

In summary, this study systematically identified the AU differences in masked expressions and explored the factors affecting recognition judgments. It particularly highlighted the diagnostic roles of AU6 and AU12 in distinguishing genuine and masked happy expressions, and revealed how facial region visibility modulates recognition. These findings deepen our understanding of masked expressions and contribute to the broader fields of emotion regulation and social behavior research.