The temporal course of vicarious embarrassment: An electrophysiological study.

Vicarious embarrassment as a social pain emotion has been studied from cognitive and neuroimaging perspectives. However, the dynamic time course of vicarious embarrassment remains unclear. We conducted an event-related potential (ERP) study to investigate the temporal dynamics of vicarious embarrassment and examine the role of perspective-taking in an emotional judgment task. The ERP results showed that vicarious embarrassment evoked more negative MFN (250–310 ms) and N400 (400–520 ms) components than the neutral condition over the frontal region. The time-frequency analysis results showed that, in the 200–1,600 ms time window, greater alpha power was elicited when participants imagined themselves rather than others in an awkward situation; in the 500–1,900 ms time window, for both groups, vicarious embarrassment involved greater theta oscillations than the neutral condition over the frontal region. These results extend previous findings that vicarious emotion involves mentalizing processes, and demonstrate that people first detect and evaluate the target's misconduct and negative outcomes and then experience the vicarious embarrassment.

Keywords: Vicarious embarrassment; mentalizing; perspective-taking; ERP

Imagine the following scenario: You are attending an awards ceremony. The host announced your award, and you go to the podium to receive the award – but then the host realized that there was a mistake; it is not you who won the award, but someone else. How is your emotional state at this time? This scene happened at the 2017 Oscars ceremony: "There's a mistake. Moonlight, you guys won Best Picture." Not only the guests present but also the audience in front of the TV felt deeply embarrassed.

Embarrassment, shame, and guilt, as self-conscious emotions, are generally evoked during negative evaluation after norm violation (Heerey et al., [24]; Kilian et al., [29]; Paulus et al., [53]; Takahashi et al., [59]; Zhu et al., [68]). Embarrassment evoked by personal social transgressions is a driving force for maintaining interpersonal relationships and repairing one's own social pratfalls (Jankowski & Takahashi, [27]; Melchers et al., [45]; Muller-Pinzler et al., [48]). The experience of embarrassment is caused not only by one's own misconduct but also by the evaluation of others' behavior, which is described as vicarious Embarrassment (VE). VE can be triggered by others' social norm violation, and even innocent bystanders can suffer embarrassment (Melchers et al., [45]; Miller, [46]). We can also feel embarrassment for another even without the person themselves experiencing embarrassment. For example, in a lecture, if you saw that the speaker's jeans zipper was unzipped, you would feel very embarrassed, even though the speaker was not embarrassed because he was unaware of the situation. It has also been suggested that the somatovisceral response pattern of VE is similar to the first-person experience of embarrassment. VE was found to elicit shortened heart period, finger vasoconstriction, and skin conductance responses (Miller, [46]; Muller-Pinzler et al., [48]; Shearn et al., [57]).

Previous studies have tried to explain how the protagonist's emotion can transmit to bystanders. Most of this research has focused on the phenomenon of empathy. It seems that individuals can automatically share others' feelings (Hawk et al., [23]; De Vignemont & Singer, [12]). However, empathy cannot fully explain vicarious emotion because the observer may feel embarrassment even when the protagonist does not realize he is in an awkward situation. Based on self-identity theory, some researchers have proposed that the observer's social identity would be threatened by others' norm violation if they came from the same social group (Melchers et al., [45]). For example, Miller ([46]) found that people were less affected by the embarrassment of relative strangers because strangers' feelings are less salient to us or create fewer problems for us to resolve. Furthermore, compared to a disliked person, one would experience a higher degree of empathic embarrassment and empathic concern when a liked person was embroiled in an embarrassing situation (Stocks et al., [58]). The ability to image oneself to the same situation as someone else and attribute their mental state to understand others' beliefs, desires, and intentions is referred to as theory of mind (ToM) or mentalizing (Keysers & Gazzola, [28]; Paulus et al., [54]). Based on mentalizing, researchers found that observers could experience embarrassment through imaging and the feelings others would experience in a social norm violations situation. There is also an effect of perspective-taking on empathic embarrassment: Imagining a target's thinking and feeling in a socially compromising situation enhances empathic concern, whereas imagining oneself in the target's situation enhances empathic embarrassment (Stocks et al., [58]).

With the rise of social cognitive neuroscience, some researchers have tried to explore the neural basis of vicarious feelings. Using functional magnetic resonance imaging (fMRI), Takahashi et al. ([59]) examined the similarities and differences between the evaluative process of guilt and that of embarrassment. They asked the participants to read sentences with neutral, guilty, or embarrassing content during the scans. The results showed that the emotion conditions activated the medial prefrontal cortex (mPFC), left posterior superior temporal sulcus (STS), and visual cortex. Compared to the guilt condition, the embarrassment condition evoked greater activation in the right temporal cortex (anterior), bilateral hippocampus, and visual cortex. Most of these regions have been implicated in the neural substrate of social cognition or mentalizing (Hein & Singer, [25]). Krach et al. ([33]) found that bystanders could experience VE regardless of whether the protagonist was acting accidentally or intentionally and was aware or unaware of the embarrassing situation. The fMRI results showed that the anterior cingulate cortex (ACC) and the left anterior insula region involved in vicarious feelings of others' physical and social pain were activated. Muller-Pinzler et al. ([49]) examined the neural response of participants while they observed their friend or a stranger in an awkward situation. The results showed that the anterior insula (AI), anterior cingulate cortex, medial prefrontal cortex, and temporal pole were activated, and ACC/AI network activation was higher while observing friends' embarrassment compared to observing strangers. These findings indicate that VE-related neural activity is modulated by social closeness.

Other studies have shown that VE, as a social pain, involves the mPFC, the AI, and ACC. Paulus et al. ([53]) found that sharing others' embarrassment engaged the temporal pole and the medial prefrontal cortex together with the anterior insula and anterior cingulate cortex and additionally stimulated the posterior superior temporal sulcus (STS), which exhibited increased functional integration with inferior parietal and insular cortex areas. Melchers et al. ([45]) studied neural activation patterns in response to video clips from reality TV shows, which are known for presenting the social norm violations, flaws, and pratfalls of its protagonists in real-life situations, thereby qualifying as an ecologically valid trigger for VE. The results showed that participants in the high VE condition (compared to the low VE condition) had higher activation of the middle temporal gyrus, the supramarginal gyrus, the right inferior frontal gyrus, and the gyrus rectus. Functional connectivity analyses confirmed increased connectivity of these regions with the anterior cingulate in the VE condition. These results indicated that high VE content activates brain regions associated with theory of mind, but also with empathic concern and social identity (Melchers et al., [45]).

Accumulating behavioral and fMRI evidence has confirmed that the neural processing of VE is related to the identification of social convention violation and the emergence of vicarious social emotion. However, due to the low temporal resolution of fMRI signals, it does not provide the timing and dynamics information of the process involved. The scalp-recorded EEG and event-related potentials (ERP) have the merit of high temporal resolution, and are widely used in cognitive neuroscience to investigate the dynamic processing of cognition or emotion in the brain. Clarifying the dynamic time processing of VE will help to better understand the information processing of social emotion. To date, there was no ERP study on embarrassment. Recently, other types of social emotions, such as shame and guilt, have been studied through ERP methods. For example, Leng et al. ([40]) found that the feeling of guilt elicited a more negative component than other non-guilt conditions at the frontal electrode sites. This negative component initiated at 350 ms at frontal electrode sites (very similar to the N400 component), may reflect the interactions of self-reflection, condemnation, and negative emotion. Sánchez-García et al. ([56]) replicated and expanded this study using an interpersonal guilt paradigm. They also found that feeling of guilt yielded a frontal negativity between 300 and 500 ms after the onset of stimuli. Zhu et al. ([69]) compared the temporal dynamics of shame and guilt in an interpersonal context. They revealed significant difference in P2 amplitudes and alpha oscillations between shame and guilt; while non-significant difference in the N2, P3 amplitudes, and theta oscillations. Zhu et al. ([69]) indicated that the alpha oscillations over the parietal region are involved in attentional orienting (concerning about others' thoughts and feelings).

At present, the dynamic time course of the neural processing of VE is still unclear. In addition, previous studies (Stocks et al., [58]) have found the effect of perspective-taking on empathic embarrassment. Stocks et al. ([58]) suggested that imaging others' thoughts and feelings might enhance empathic concern and a desire for future exposure to the person, but not empathic embarrassment. However, as a bystander in real life, we also can feel embarrassed when witnessing others in an awkward situation. The role of perspective-taking on VE needs to be further studied. To bridge these gaps, we attempted to investigate the temporal dynamics of neural mechanisms underlying VE using electroencephalogram (EEG) recording. We asked participants who were assigned to an other-perspective or self-perspective group to complete an emotional judgment task (Takahashi et al., [59]). Based on prior ERP findings, we examined the medial frontal negative component (MFN) and N400 as well as EEG oscillations (theta and alpha), which are associated with the processing of error detection, social norm violation detection, emotion arousal, and perspective-taking. The MFN is generated in the medial frontal cortex, in or near the ACC (Gehring & Willoughby, [20]). This component is associated with deviations from the desired outcome and especially making mistakes in a social context (Boksem et al., [9]). The N400 is well known to be sensitive to semantic integration processes and was also found to be evoked by world knowledge violations and social norm violations (Hagoort et al., [21]; Mu et al., [47]). In the present study, writing sentences were used as stimuli to investigate whether the identification of social convention violation occurs before or after semantic integration processes (N400 component).

ERP in the time domain provide us a high temporal resolution measure of the information processing, but it is difficult to use for measuring brain activity that extends beyond a few seconds (Luck, [43]). EEG does not have this restriction in the frequency domain. In this study, empathic concern appeared in the late stage and lasted for a long time. Time frequency analysis was conducted to provide the frequency domain information. Regarding the frequency domain, theta oscillations are involved in emotion-related salience detection (Başar, [2], [3]). The theta oscillations are also related to memory load in working memory, recall of episodic memory and mentalizing (Bögels et al., [7]; Fuentemilla et al., [19]; Klimesch, [30]; Ole & Tesche, [52]). Alpha oscillations are related to perspective-taking (Woodruff et al., [66]). We examined whether vicarious social emotion is modulated by theta oscillation, alpha oscillation, and perspective-taking.

Forty health volunteers (19 males, all right-handed) aged between 18 and 23 years (mean age = 20.32, SD = 1.25) were recruited from Chengdu University. Participants were randomly assigned to one of the two groups: imagine-other versus imagine-self. Two participants (one assigned to the imagine-other group and one assigned to the imagine-self group) were rejected because of a low accuracy rate of behavioral response and four (one assigned to the imagine-other group and three assigned to the imagine-self group) participants were excluded from further EEG analysis due to substantial electromyographic artifacts. All 34 participants (17 males, mean age = 20.32, SD = 1.25) had normal or corrected-to-normal vision. They gave written informed consent prior to participation in the ERP experiment and were paid 20 RMB (about 3 USD) after the experiment.

According to a meta-analysis 21 ERP studies (Wang et al., [64]) involving N400 components, the effect size (Cohen's d) of early automatic activation and late contextualization generated is −0.41 and −0.36. The significant effect size of N400 for congruent/related target words is −0.52. An a priori power analysis (G*Power 3.1.9.4; Faul et al., [15]) using a medium-large effect size (f = 0.25) suggested that there is an 82% chance of correctly rejecting the null hypothesis of no significant effect of the interaction with 17 imagine-other participants and 17 imagine-self participants for a total 34 participants.

A pilot experiment was conducted to select appropriate material before the main experiment. Ninety-five pairs of sentences describing a protagonist in a public scenario were collected from the internet. The VE scenario described a social target who was violating a social norm in public and thus threatened his or her social integrity (e.g., "Participating in a speech contest, forgetting words on stage."). Neutral control stimuli described the social target in a public context without violating socially normative standards (e.g., "Participating in a speech contest, doing well on stage."). Nineteen participants (9 males, all right-handed) aged between 20 and 22 years (mean age = 20.53, SD = 1.19) were recruited to rate the sentences. Using a seven-point Likert scale (high ratings indicating strong feeling of VE), participants were asked to indicate whether the presented scene made them feel vicariously embarrassed for the protagonists. Finally, 80 sentences with the highest VE ratings (M = 5.66; SD = 0.48) and 80 sentences with the lowest ratings (M = 2.26; SD = 0.66) were selected for the main experiment. Notably, the scores of the embarrassment sentences were significantly higher than those of the neutral sentences [paired sample t test, t(79) = 40.48, p < 0.05].

Participants sat in a comfortable chair at a distance of approximately 100 cm from the computer screen. The E-Prime 2.0 software package (Psychology Software Tools, Inc.) was adopted to control the program and collect the behavioral data. At the beginning of the experiment, the participants were instructed to imagine themselves as observers in the imagine-other condition or as the protagonist in the imagine-self condition and to perform a situational judgment task. After a practice round of 10 trials, participants performed 160 trials (80 embarrassment scenarios, 80 neutral scenarios) of the situational judgment task. The experimental protocol is shown in Figure 1. Specifically, each trial started with the presentation of a fixation cross in the center of the screen for 800 ms; then, a sentence was presented for 4,000 ms. Participants judged whether they were embarrassed or not when they saw the scenario, and were asked to press the response key as soon as possible. At the end of each trial, a blank screen was presented for 600–800 ms. Participants completed a self-report questionnaire after finishing the experiment. In the questionnaire, they rated how much embarrassment the scene in the EEG recording experiment caused them on a seven-point Likert scale (high scores indicating a strong feeling of VE).

Graph: Figure 1. The sequence of stimulus on an individual trial.

PHOTO (COLOR): Figure 2. Illustration of ERP results in different stimulus conditions at Fz. The MFN and N400 components were elicited by the vicarious embarrassment condition (red solid line) and neutral condition (green dotted line) in the perspective-self (right) and perspective-other (left) groups. Topography maps of difference waves illustrate the locus of MFN and N400. E: embarrassment condition; N: neutral condition.

PHOTO (COLOR): Figure 3. Group-averaged ERSP time-locked to the outcome of VE and neutral conditions at Fz. Topographic maps of difference power are presented for the theta oscillations for selected contrasts. The VE condition elicited larger theta over the frontal and central region. E: embarrassment condition; N: neutralcondition.

PHOTO (COLOR): Figure 4. Group-averaged ERSP time-locked to the outcome of VE and neutral conditions at P6. Topographic maps of difference power are presented for the alpha oscillations for selected contrasts. The perspective-self condition elicited larger alpha over the parietal region than the perspective-other condition. S: perspective-self condition; O: perspective-other condition.

EEG data were recorded by a 64-channel recording system (Brain Products; pass band: 0.01–100 Hz, sampling rate: 1,000 Hz) with international 10/20 montage. The AFz electrode on the cap served as a ground electrode. All channel impedances were maintained below 10 kΩ during the experiment. Vertical electrooculography (VEOG) was performed using electrodes placed below and above the right eye and 10 mm from the lateral canthi of right eye (horizontal EOG).

EEG data were preprocessed and analyzed using the EEGLAB and ERPLAB toolboxes running in the MATLAB environment (Delorme & Makeig, [13]; Lopez-Calderon & Luck, [42]). The average of the bilateral mastoids was used for off-line re-referencing. Continuous EEG data were bandpass filtered between 1 and 30 Hz and segmented into epochs with a time window of 3,000 ms (−1,000 to 2,000 ms relative to stimulus onset). Using the pre-stimulus interval, EEG trials were baseline corrected. Trails contaminated by eye blinks and movements were corrected using an independent component (ICA) analysis algorithm. In all datasets, independent components with a large EOG channel contribution and a frontal scalp distribution were removed (Delorme & Makeig, [13]). For each subject, trials collected at each level of stimulus energy were averaged together, time-locked to stimulus onset.

Based on visual inspection of the grand average waveforms, we analyzed the MFN and N400 components. To analyze the mean components of the MFN and N400, we selected the time window of 250 − 310 ms for the MFN component after the onset of the stimulus and the time window of 400 − 520 ms for the N400 component based on the previous studies (MFN, the time window 220 − 320 ms, Boksem & De Cremer, [8]; N400, the time window 300 − 500 ms, White et al., [65]). According to previous studies on social-consciousness emotion and ToM (Cao et al., [11]; Marta M. Kutas & Federmeier, [36]; Leng et al., [40]; Li & Han, [41]) (AF3, AF4, AF7, AF8, Afz, Fz, F1, F2, F3, F4, FCz, FC1, FC2, FC3, FC4) and central (Cz, C1, C2, C3, C4, CP1, CP2, CP3, CP4, CPz) electrodes for the mean amplitudes of MFN and N400 components.

Time frequency analysis was conducted on each trial using the Morlet wavelet transformation to obtain a temporal spectral (power) map running under MATLAB (Vialatte et al., [62]). The wavelet transform is a multiresolution analysis technique that provides a good compromise between time and frequency resolution. The complex Morlet wavelet w (t, f0) has a Gaussian distribution in the time (σt) and frequency (σf) domains around the center frequency f0. Single trial epochs were extracted from −1,000 to 2,000 ms relative to stimulus. Event-related spectral perturbations (ERSP) were computed on the wavelet-transformed epochs for each condition at each time point and wavelet frequency (Makeig et al., [44]).

Based on the visual inspection of the temporal spectral map, time frequency maps encompassing the theta (4–7 Hz) and alpha (8–12 Hz) activities were created within 500–1,900 ms and 200–1,600 ms windows. Power values were normalized with respect to a − 400 to 0 ms pre-stimulus baseline and converted to decibels [10× log (μV2)]. Based on previous studies, we selected the frequency bands and electrode sites (Bögels et al., [7]; Klimesch, [31]; Woodruff et al., [66]; Zhu et al., [69]). The ERSPs in the theta bands (4–7 Hz) during the 500–1900 ms time window at the frontal (AF7, AF8, AF3, AF4, Afz, F1, F2, Fz, FC1, FC2, FCz) and central (C1, C2, Cz, CP1, CP2, CPz) electrode sites were averaged for further statistical analysis. The ERSPs in the alpha bands (8–12 Hz) during the 200–1,600 ms time window at the parietal (P1, P2, P3, P4, P5, P6, Pz) and occipito-temporal (PO3, PO4, PO7, PO8, Poz) electrode sites were averaged for further statistical analysis.

Repeated measures analysis of variance (ANOVA) with a statistical significance level of 0.05 was conducted separately for the mean amplitudes of MFN and N400 components, as well as the theta and alpha bands. Post-hoc tests for multiple comparisons were corrected by the Bonfeeoni method, and all statistically significant effects for all repeated measures ANOVAs were adjusted by the Greenhouse−Geisser method to protect against Type I errors. Effect sizes were presented as partial eta squared (η2).

The mean accuracy of participants' judgments of emotion was 90.66%, and the mean reaction time was 1,813 ms (SD = 377). A 2 (emotion: VE vs. neutral) × 2 (group: imagine-other vs. imagine-self) repeated measures ANOVA was conducted for the accuracy and reaction time. The accuracy rate in the VE condition (M = 0.82, SD = 0.01) was worse than that in the neutral condition (M = 0.94, SD = 0.01), F(1, 32) = 16.86, p < 0.001, η2 = 0.35. The accuracy rate of the imagine-self group (M = 0.93, SD = 0.01) was significantly better than that of the imagine-other group (M = 0.89, SD = 0.01), F(1, 32) = 16.86, p < 0.001, η2 = 0.35. The reaction times in the VE condition (M = 1944.11, SD = 67.72) was longer than that in the neutral condition (M = 1732.19, SD = 64.66), F(1, 32) = 33.12, p < 0.001, η2 = 0.51.

A 2 (emotion: VE vs. neutral) × 2 (group: imagine-other vs. imagine-self) repeated measures ANOVA on the rating score was conducted. The results showed that the emotion situation did elicit VE experiences in the participants. Participants reported higher embarrassment in the VE condition (M = 5.80, SD = 0.08) than in the neutral condition (M = 1.61, SD = 0.08), F(1, 32) = 1218.86, p < 0.001, η2 = 0.97. There was no significant main effect of group, F(1, 32) = 0.02, p = 0.88, η2 = 0.01.

For the MFN, a 2 (emotion: VE vs. neutral) × 2 (group: imagine-other vs. imagine-self) × 2 (channel: frontal vs. central) repeated measures ANOVA revealed a significant main effect of emotion, F(1, 32) = 5.88, p = 0.02, η2 = 0.16, with higher amplitudes for VE (M = −1.30, SD = 0.53) than for the neutral condition (M = −0.65, SD = 0.41). The main effect of channel was significant, F(1, 32) = 73.80, p < 0.001, η2 = 0.70, with higher amplitudes for frontal (M = −2.59 SD = 0.52) than central electrodes (M = −0.64, SD = 0.46). The main effect of group was not significant, F(1, 32) = 0.12, p = 0.73, η2 = 0.01. None of the interaction effects were significant (all ps > 0.05).

For N400, a 2 (emotion: VE vs. neutral) × 2 (group: imagine-other vs. imagine-self) × 2 (channel: frontal vs. central) repeated measures ANOVA revealed a significant main effect of emotion, F(1, 32) = 11.19, p < 0.001, η2 = 0.26, with higher amplitudes for VE (M = −2.263, SD = 0.30) than for the neutral condition (M = −1.75, SD = 0.27). The main effect of channel was significant, F(1, 32) = 65.34, p = 0.000, η2 = 0.67, with higher amplitudes for frontal (M = −2.99, SD = 0.31, p = 0.00) than central electrodes (M = −1.03, SD = 0.29 μV). The main effect of group was not significant, F(1, 32) = 0.01, p = 0.95, η2 = 0.00. No interaction effects were significant (all ps > 0.05) (see Figure 2).

A 2 (emotion: VE vs. neutral) × 2 (group: imagine-other vs. imagine-self) × 2 (channel: frontal vs. central) repeated measures ANOVA on the theta oscillation revealed a significant main effect of emotion, F(1, 32) = 30.51, p < 0.001, η2 = 0.49, with higher power for VE (M = 8.55, SD = 0.61) than for the neutral condition (M = 6.79, SD = 0.56). The main effect of channel was significant, F(1, 32) = 34.88, p < 0.001, η2 = 0.52, with higher power for frontal (M = 8.76, SD = 0.64) than central (M = 6.58, SD = 0.54) electrodes. The main effect of group was not significant (ps > 0.05). There was no significant interaction effect (all ps > 0.05) (see Figure 3).

For the alpha oscillation, a 2 (emotion: VE vs. neutral) × 2 (group: imagine-other vs. imagine-self) × 2 (channel: parietal vs. occipito-temporal) repeated measures ANOVA revealed a significant main effect of group, F(1, 32) = 5.16, p = 0.03, η2 = 0.14, with higher power for imagine-self (M = −5.41, SD = 1.63) than for imagine-other (M = −0.17, SD = 1.63). The main effects of emotion and channel were not significant (all ps > 0.05). None of the interaction effects were significant (all ps > 0.05) (see Figure 4).

In this study, we aimed to investigate the time course of VE and examine the role of perspective-taking using an emotional judgment task. The participants' choices during the experiment and their ratings after the EEG recording suggested that VE was successfully evoked. This is consistent with previous studies (Paulus et al., [53]). Compared with the neutral condition, the VE condition involved the processing of vicarious emotions, which could lead to longer reaction time and lower accuracy. To the best of our knowledge, this is the first study using an electrophysiological method to explore VE processing. The ERP results showed that the MFN and N400 component was larger in the VE condition than in the neutral condition, and the VE effect was highest in the frontal region. The EEG frequency results showed that VE evoked larger theta oscillation than the neutral condition and had the highest power at the frontal electrode. Particularly, the perspective-self group showed larger alpha oscillation than the perspective-other group.

VE has been found to be associated with brain activity in the ACC (Krach et al., [33]; Melchers et al., [45]; Muller-Pinzler et al., [49]; Paulus et al., [53]), which has been demonstrated to be associated with empathy, especially empathy for pain (Lamm et al., [38]), such as social pain in response to others' mistakes (Krach et al., [33]; Melchers et al., [45]). ERP studies indicated that a group of similar negative components, such as event-related negative (ERN), feedback-related negative (FRN), N2/N200, and MFN, are evoked in the location of the ACC (Boksem & De Cremer, [8]; Boksem et al., [9]; Hajcak et al., [22]; Itagaki & Katayama, [26]; Van Schie et al., [61]; Vincent & Carter, [63]; Yeung et al., [67]). In these components, MFN has been related to social cognition and emotion, which was proposed to reflect a motivational or affective evaluation of negative outcomes in a social context (Boksem & De Cremer, [8]; Boksem et al., [9]). In the present study, we found the MFN component in the 250–310 ms time window over the frontal sites. MFN amplitude was more pronounced for the VE condition compared to the neutral condition and had the highest amplitudes at the frontal electrodes. These results suggest that a bystander would first detect and evaluate a social target's mistakes or negative outcomes in the social context – that is, participants would evaluate the outcomes of "speech contest" and "forgetting words" in the embarrassment scenarios.

Because of the language material used in this study, a classic ERP component of language processing (N400) was found. N400, elicited by the semantic violation, is well known to be sensitive to semantic processes during language comprehension (Brown & Hagoort, [10]; Dikker & Pylkkanen, [14]; Frisch & Schlesewsky, [16]; Kutas & Federmeier, [35], [36]; Proverbio & Riva, [55]). N400 is also sensitive to world knowledge violations (Hagoort et al., [21]). General world knowledge (e.g., "The Dutch trains are white and very crowded") and the semantic violation (e.g., "The Dutch trains are sour and very crowded") were both integrated in the same time frame during sentence interpretation and evoked a similar N400 component. General world knowledge violation sentences mismatch with our representation of the world in memory (e.g., "The Dutch trains are yellow and very crowded"). The N400 has also been found to reflect the ease of integrating a stimulus into a given context. Compared to items that are congruent with the context, incongruent items elicit larger N400 (White et al., [65]). In addition, N400 has been regarded as a potent neural index of the detection of unexpected anomalous stimuli and affective and social incongruent information (Mu et al., [47]). Recently, Mu et al. ([47]) found that N400 serves as a neural marker of detecting social norm violations. In the current study, N400 amplitudes were higher for the VE condition (containing information about social norm violations) than for the neutral condition. This is consistent with previous studies (Hagoort et al., [21]; Mu et al., [47]; White et al., [65]), which suggested that the N400 effect not only is related to semantic violation but also reflects the conflicts with general world knowledge violation and culture in a given context.

Although MFN and N400 are similar in polarity and shape, they are different components. Compared to N400, MFN usually occurs a bit earlier in processing and has a shorter duration and more anterior location (Landi, [39]; Nobre & Mccarthy, [51]). MFN has been located in the ACC, whereas N400 is thought to be generated by left hemisphere language processing regions such as Wernicke's area (Frishkoff et al., [18]; Kutas & Van Petten, [37]; Landi, [39]). The MFN has been found to reflect response monitoring and conflict detection, the motivational value of the signal, and automatic semantic priming, whereas N400 is sensitive to violations of social norms and prediction or expectancy (Frishkoff et al., [17]; Landi, [39]; Tucker et al., [60]; White et al., [65]).

EEG rhythms in theta and alpha bands have been related to the perception and understanding of others' actions and mental/emotional states (Woodruff et al., [66]). Our results showed that the theta oscillations (500–1,900 ms) were significantly larger for the VE condition than for the neutral condition. Previous studies have demonstrated that theta rhythm is related to memory and language processing. The frontal theta rhythm has been found to be involved in encoding and recalling of the information from long-term memory (Klimesch, [30]; Ole & Tesche, [52]) or working memory processes (Fuentemilla et al., [19]; Klimesch, [30]; Ole & Tesche, [52]). During language processing, theta responses have been implicated in lexical-semantic retrieval (Bastiaansen et al., [4]). For example, Hagoort et al. ([21]) found that the theta rhythm of the frontocentral midline was sensitive to semantic violations in sentences. In another study about language processing, the neural basis of mentalizing was investigated using magnetoencephalography (MEG). The results showed that the dynamics of theta oscillations involved episodic retrieval as well as mentalizing. In the present study, we used written materials to induce vicarious embarrassment and neutral conditions. Similar to previous studies, our findings showed that theta oscillations were elicited over the frontal region were elicited, possibly because VE involves language processing, episodic memory retrieval, and mentalizing. Mentalizing involves imaging oneself in the target's situation (Paulus et al., [54]). By processing language information and projecting themselves into the position of the target, people retrieve related episodic memory.

In addition, the theta event-related synchronization has also been found to be modulated by emotional processing and the valence dimension (Aftanas et al., [1]; Bekkedal et al., [5]; Knyazev et al., [32]; Krause et al., [34]). Knyazev et al. ([32]) compared the event-related low frequency synchronization of emotion information processing by presenting emotional or neutral stimuli. Their results showed that emotional stimuli yielded stronger theta and delta synchronization than neutral stimuli. Recently, Zhu et al. ([69]) compared the temporal dynamics of emotional processing, such as shame, guilt, and happiness, and found similar results to previous studies, with emotional stimuli eliciting theta synchronization over the frontal region. Moreover, they also found that shame and guilt, as social pain, involved a higher theta oscillation than happiness. In our present experiment, VE, as a social pain emotion, could modulate the theta oscillation, which resulted in it having larger theta rhythm than the neutral condition. Our results also suggested that bystander participants grasped the protagonist's emotions as if they were their own and were more emotionally aroused compared to the neutral condition.

We also showed that the alpha oscillation (200–1,600 ms) was modulated by perspective-taking. When participants imagined themselves in a situation, the ERSPs of alpha were larger over the parietal and occipital region than when they imagined others in a situation. The alpha oscillations over the parietal and occipital regions have been related to attentional processing, arousal, and perspective-taking abilities (Niedermeyer, [50]; Woodruff et al., [66]; Zhu et al., [69]). Right parietal cortex alpha power increases have been linked to internal attention (Benedek et al., [6]). Woodruff et al. ([66]) found that greater alpha suppression in the occipital electrodes was negatively correlated with empathy. Thus, the present findings suggest that people are inclined to orient their attention to their internal self when they imagine themselves in a situation. At the same time, people will have more empathy for others when they imagine others in a situation.

In the current study, we employed the high temporal resolution of the electrophysiological method to investigate the time course of vicarious embarrassment and examine the role of perspective-taking. The results revealed a difference between vicarious embarrassment and neutral emotion in an early stage (MFN, 250–310 ms; N400, 400–520 ms) and a late stage (theta oscillations, 500–1,900 ms). These findings suggest that vicarious embarrassment may be composed two different neural processes. In the early stage, the brain detects and evaluates a social target's mistakes or negative outcomes in the context; in the late stage, the brain grasps the target's emotion and produces the vicarious embarrassment emotion. Moreover, the difference in alpha oscillation between taking one's own perspective and taking another's perspective reflects the modulating role of perspective-taking on attention and empathy. However, the perception of vicarious embarrassment is not modulated by perspective-taking.

Our study warrants further research in the future. At present, there have been some studies on vicarious embarrassment, but is there a difference between "first-hand" embarrassment and vicarious embarrassment in neural processing. Further study should be investigated to compare the neural processing of vicarious embarrassment and "first hand" embarrassment using ERP or fMRI.

We thank Nantong Wang for data collection and LetPub (http://www.letpub.com) for its linguistic assistance during the preparation of this manuscript.

No potential conflict of interest was reported by the authors.