Relating self–other overlap to ingroup bias in emotional mimicry

ABSTRACT

Humans tend to show congruent facial expressions automatically in reaction to their partners which is defined as emotional mimicry. Although occurring unconsciously, this tendency has been proven to be modulated by social contextual factors such as group membership. Ingroup bias in emotional mimicryhas been well-documented in previous research; however, few studies have investigated the underlying mechanism. Based on the mimicry-as-social-regulator model, this study explored whether the ingroup bias in emotional mimicry arises from the greater self–ingroup overlap. By recording participants’ facial electromyographic responses while passively viewing dynamic emotional clips performed by either racial ingroup or outgroup actors, Study 1 validated the presence of ingroup bias in the mimicry of happiness, but not anger. Using asimilar procedure in Study 2, anew sample was employed (N = 37), and a measurement of self–other overlaps via the Inclusion of the Other in the Self Scale was added. The results of Study 2 reproduced the ingroup bias in happy mimicry, and further demonstrated that the effect of group membership on emotional mimicry was mediated by the self–other overlap. In summary, this study provides evidence that the level of interpersonal closeness predicts emotional mimicry.

KEYWORDS:

• Self–other overlap
• emotional mimicry
• ingroup bias
• interpersonal closeness

Emotional mimicry refers to the unconscious copying of another person’s emotional expression (Hatfield et al., 1992; Hess & Fischer, 2013). The process is operationally defined in empirical studies as “the congruent facial muscular activations in response to an emotional facial expression” (Seibt et al., 2015). This kind of emotion-specific facial mascular activation can be recorded by facial electromyography (EMG), a technique that measures electrical activity during facial muscle contractions (Hess et al., 2017). For example, an EMG pattern of increased Zygomaticus Major (ZM) activity and decreased Corrugator Supercillii (CS) was found to be related to happiness and a reverse pattern, for anger expression. Despite its ubiquity in daily life, emotional mimicry is modulated by plenty of social contextual factors such as group membership (Hess & Fischer, 2013, 2014; Seibt et al., 2015).

Not all emotional mimicry is equatable. Prior research has demonstrated that mimicry of facial expressions varies for different group members (Ardizzi et al., 2014; Bourgeois & Hess, 2008; Hühnel, 2015; Kuszynski, 2015; Mondillon et al., 2007; Peng et al., 2020; Van der Schalk et al., 2011). For example, Bourgeois and Hess revealed a higher level of mimicry for ingroup (versus outgroup) angry and sad faces (Bourgeois & Hess, 2008). By manipulating group membership as different subject or ethnicity in two studies, van der Schalk and colleagues consistently showed that ingroup anger and fear displays were mimicked to a greater extent than outgroup displays of these emotions (Van der Schalk et al., 2011). Age was another frequently used manipulation factor of group membership. A series of studies indicated a same-age advantage in emotional mimicry. Ardizzi et al. (2014) suggested that teenagers exhibited greater facial EMG responses to peers’ facial expressions. Hühnel (2015) found that younger participants showed more happy mimicry to younger (versus older) confederates when talking about a happy event. In a similar way, Kuszynski (2015) demonstrated that mimicry of smiles was reduced within intergenerational dyads compared with same-generation dyads during a face-to-face interaction. A recent study provided initial evidence for ingroup bias in happy mimicry among participants from Eastern cultures (Peng et al., 2020). In summary, these studies have collectively shown the presence of ingroup bias in emotional mimicry.

Why do people mimic more ingroup (versus outgroup) expressions? The mimicry-as-social-regulator model (Hess & Fischer, 2013) assumes that the level of interpersonal closeness predicts emotional mimicry. It suggests that emotional mimicry is not based on mere perception but on the interpretation of signals as emotional intentions in a specific context; therefore, emotional mimicry is affected by a series of social contextual factors. Accordingly, we suggest that a higher degree of ingroup (versus outgroup) mimicry might be attributable to the higher level of self–ingroup (versus outgroup) interpersonal closeness. Self–other overlap depicts the relationship between the self and the other and can be measured by the Inclusion of Other Scale (IOS; Aron et al., 1992). The social identity theory suggests that the concept of a “social self” describes the overlap in mental representations of the self and the ingroup (Ellemers & Haslam, 2012), as evidenced by a series of empirical studies (Otten & Epstude, 2006; Scheepers et al., 2013; Tropp & Wright, 2001). Moreover, the greater self–ingroup (versus outgroup) overlap, by default, accounts for ingroup bias in many areas, such as empathy (Eisenberg & Sulik, 2012). Research has shown a link between self–other overlap and emotional mimicry; specifically, greater self–other overlap induces more unconscious behavior imitation (Maister & Tsakiris, 2016). For example, younger people who were included by an elder in a ball-tossing game displayed increased overlap between the self and elder, which in turn elevated mimicry of emotional expressions of the other-age group (Hühnel et al., 2018). Affectively deviant smiles (e.g., smiles at funerals) can be mimicked only when people feel close to the expresser (Kastendieck et al., 2021). Our recent study (Peng et al., 2020) found that rTPJ, a neural area for self–other processing (Donaldson et al., 2015), modulated ingroup bias in emotional mimicry. Taken together, previous studies have suggested that self–other overlap is correlated with group membership and emotional mimicry. Thus, it is plausible that it mediates the relationship between group membership and emotional mimicry. In this light, this study explored the role of self–other overlap in ingroup bias in emotional mimicry.

Two studies were conducted. Study 1 was a pilot study that explored the racial ingroup bias in facial emotional mimicry. A sample of Chinese participants were recruited to view passively a series of emotional materials (expressing happiness or anger) of either racial ingroup or outgroup actors. Dynamic expressions are well-documented as being more effective in inducing emotions (Rymarczyk et al., 2016; Sato et al., 2008); therefore, emotional clips from a high-resolution 3D dynamic facial expression database (Yin et al., 2008) were employed in this study. EMG techniques were employed to record emotional mimicry intensity. Based on previous theoretical (Hess & Fischer, 2013) and empirical (Peng et al., 2020; Van der Schalk et al., 2011) studies, we hypothesized that participants would show stronger mimicry toward ingroup (versus outgroup) happiness, while the mimicry of anger would not differ between groups. Study 2 – in which self–other ingroup was measured using the IOS – further explored the possible mechanism of racial ingroup bias in emotional mimicry. In Study 2, it was hypothesized that self–other overlap would mediate the effect of group membership on facial emotional mimicry.

Study 1

Participants

The required sample size was 24 (calculated by G*Power with power = 0.8, f = 0.25). Assuming some data loss due to technical issues regarding the use of EMG, 29 college students (5 males, 19.5 ± 1.2 years old) from a Chinese university were recruited to participate in this study. All of the participants had normal or corrected-to-normal vision acuity. This study was approved by the local institutional review board. Written consent was obtained from each participant.

Stimuli and procedure

The stimuli in the current study were derived from the high-resolution 3D dynamic facial expression database developed by Yin et al. (2008), which contains six basic emotions (happiness, anger, fear, sadness, disgust, and surprise), expressed by various ethnic models. In this study, video clips – showing happiness and anger – of 10 East Asians (ethnic ingroup, 5 males) and 10 Whites (ethnic outgroup, 5 males) were included as materials, constituting 40 emotional clips. All clips were edited to 2,000 ms duration, changing from a neutral expression at the beginning to a full-blown emotional expression by the end (Figure 1).

Figure 1. Materials and experimental procedures of Study 1

The study employed a passive viewing paradigm, a well-validated task (Deng & Hu, 2017; Olszanowski et al., 2020), to induce emotional mimicry. In a typical procedure, the participants were told that the experiment was designed to collect sweat gland activity evoked by different videos, to keep the true purpose confidential (Dimberg et al., 2000; Hess & Blairy, 2001). Each trial began with a brief fixation “+”, and was followed by the presentation of one emotional clip for 2,000 ms, during which participants were instructed to view the clip carefully, and their facial muscular activations were recorded simultaneously. The inter-trial interval varied from 1,000 to 1,200 ms (Van der Schalk et al., 2011). The order of emotional clips was randomized for each participant. Following previous research, group membership was emphasized to make the distinction salient (Van der Schalk et al., 2011). Each stimulus was repeated once in another block. At the end of the experiments, participants were fully debriefed. The experiment procedure was written and presented via E-prime 2.0.

Apparatus and data analysis

Biopac system EMG (BIOPAC System, Inc. Santa Barbara, CA) was employed to record ZM and CS activities. Specifically, EMG data were recorded using surface Ag/AgCl bipolar electrodes filled with electrode pastes. Placements of ZM, CS and the reference electrode (left mastoid) were chosen in accordance with the EMG research guideline (Fridlund & Cacioppo, 1986). The EMG activity was continuously recorded at 2,048 Hz.

Facial EMG data were then pre-processed via AcqKnowledge (Biopac Systems, version 5.0). The signal was visually inspected offline for artifact due to excessive muscle movements, filtered with a 30–500 Hz bandpass filter and 50-Hz filter, and transformed using root mean square. The EMG responses were expressed as change in activity in microvolts from the pre-stimulus 1,000 ms interval. Trials in which the EMG reaction was superior to three standard deviations were excluded from the analysis. In line with prior research (Hess et al., 2017), a mimicry score was calculated to index mimicry of happiness and anger. For happiness, the score was obtained by subtracting the CS from the ZM activity; for anger, the score was obtained by subtracting the ZM from the CS activity.

Results

A Group (Ingroup, Outgroup) by Emotion (Happiness, Anger) repeated measures ANOVA revealed significant main effects of Group: F (1, 28) = 8.96, p = .006, η² = 0.24; and Emotion: F (1, 28) = 65.40, p < .001, η² = 0.70, which were qualified by a Group by Emotion interaction: F (1, 28) = 5.26, p = .030, η² = 0.16. As shown in Figure 2, ingroup happiness induced stronger mimicry than outgroup happiness (t = 3.101, d = 0.81), while mimicry of anger did not differ between groups.

Figure 2. Mean mimicry scores in the pilot study as a function of group and emotion

As a pilot study, Study 1 revealed an ingroup bias in the mimicry of happiness among Chinese participants, consistent with previous research (Peng et al., 2020). Based on the results of Study 1, we further explored the role of self–other overlap in interpreting ingroup bias in emotional mimicry.

Study 2

Participants

Thirty-seven college students (6 males, 21.0 ± 2.1 years old) from a Chinese university participated in this study. All of the participants had normal or corrected-to-normal vision acuity. This study was approved by the local institutional review board. Written consent was obtained from each participant.

Stimuli and procedure

The stimuli used in Study 2 were the same as Study 1. The procedures were the same as those of Study 1, except that after each clip, the participants had to rate the overlap between self and the other (the model they had just seen in the preceding video) on the IOS.

The IOS scale was used to assess self–other overlap by asking participants to rate their relationship with the model. This scale contains seven increasingly overlapping pairs of circles representing the participant and the target. The instructions for the participants were as follows: “The following seven pairs of circles depict different kinds of relationships between two people; please select the corresponding number for your response (1 = two nearly completely overlapping circles, 7 = two non-overlapping circles), to indicate the possible relationship between you and the model you have just seen in the preceding clip.” This scale has been well-validated in previous research (Hühnel et al., 2018; Paladino et al., 2010).

Results

Emotional mimicry

A Group (Ingroup, Outgroup) by Emotion (Happiness, Anger) repeated measures ANOVA revealed a significant main effect of Group: F (1, 36) = 5.88, p = .020, η² = 0.14, and a Group by Emotion interaction: F (1, 36) = 7.31, p = .010, η² = 0.17. In line with Study 1, a Bonferroni-corrected analysis showed that ingroup happiness induced stronger mimicry than outgroup happiness (t = 2.88, d = 0.67), while mimicry of anger did not differ between groups (Figure 3). A follow-up one-sample t-test with the criterion 0 confirmed that this study’s participants mimicked both happiness and anger expressions.

Figure 3. Mean mimicry scores in the formal study as a function of group and emotion

Self–other overlap

The participants’ IOS rating in various conditions had a high reliability (Cronbach’s alpha = .85). A two-way ANOVA revealed significant effects of Group Membership (F [1, 36] = 37.38, p < .001, η² = .51) and Emotion (F [1, 36] = 143.22, p < .001, η² = .80), which were qualified by the two-way interactive effect: F (1, 36) = 6.36, p = .016, η² = .15. The interaction showed that the difference between IOS rating of ingroup and outgroup was greater for happy than for angry faces.

Mediation analyses

Group membership was negatively correlated with self–other overlap (r_pb = −.25, p = .030) and positively correlated with ZM activity for happiness (r_pb = .40, p < .001), and self–other overlap and ZM activity were significantly correlated (r = −.47, p < .001). To investigate the mediating role of self–other overlap in the effect of group membership on emotional mimicry (for happiness), a mediational analysis was conducted for the mimicry of happiness, in which the predicative variable was group membership; the mediating variable was IOS rating for happy faces, and the dependent variable was ZM activity. To this end, the PROCESS 3.5 Macro for SPSS (Model 4) developed by Hayes (2013) was employed. Figure 4 shows that group membership was a significant predictor for happiness mimicry (β = .25, p = .0001), indicating that ingroup (versus outgroup) faces would induce higher level of mimicry. This association was mediated by self–other overlap (β = −.16, p = .0003). Notably, the predictive effect of group membership on happy mimicry remained significant (β = .19, p = .005) after the inclusion of self–other overlap as a mediating variable. For the indirect effect, 95% CI without “zero” indicated the significant mediation effect. The results showed that the total indirect effect was significant: the standardized effect was .06, 95% CI = [.01, .15]; the ratio of indirect effect to total effect was 24%.

Figure 4. Self–other overlap mediating the relationship between group membership and emotional mimicry

Note: *, **, and *** denote significance at the 5%, 1%, and 0.1% levels, respectively.

Discussion

This study revealed a constant ingroup superiority in the mimicry of happiness, while mimicry of angry expressions did not differ as a function of racial group. In line with the mimicry-as-social-regulator model (Hess & Fischer, 2013), this study demonstrated that ingroup superiority in emotional mimicry is partially attributed to the higher level of self–ingroup (versus self–outgroup) overlap.

Ingroup versus outgroup classification facilitated the acquisition and stabilization of group identity (Tajfel, 1982). Accordingly, the ingroup shared, by default, larger overlapping mental representations with the self (Ellemers & Haslam, 2012). Moreover, Otten and Epstude (2006) stated that the overlapping mental representations between the self and the respective ingroup was achieved by assimilating the individual self into the ingroup rather assimilating the ingroup into the individual self. In line with this idea, the current study showed that relative to outgroup models, the participants reported themselves as overlapping more with ingroup models. This finding implied that group membership (ingroup versus outgroup) was interpreted as different levels of self–other overlap. Such a view might aid in the comprehension of ingroup bias in other areas, such as empathy (Chiao & Mathur, 2010) and altruism (Bernhard et al., 2006).

Unlike some previous studies (Bourgeois & Hess, 2008; Van der Schalk et al., 2011), the present study showed no ingroup bias in anger mimicry, which was a consistent finding for both Study 1 and 2. However, previous studies on the mimicry of anger were inconclusive. For example, ingroup bias in mimicry of anger, as revealed by van der Schalk et al., was not reproduced by subsequent studies (Sachisthal et al., 2016). Other studies also showed that people would not mimic non-affiliative emotions such as anger (Deng & Hu, 2017; Hess & Fischer, 2013, 2014). Consistent with our finding, Hühnel (2015) revealed that in real social interactions, younger participants displayed ingroup bias in the mimicry of happiness, but not anger. More recently, Kastendieck et al. (2021) recorded the mimicry of happiness and sadness in both appropriate (e.g., smiles at weddings and sadness at funerals) and inappropriate (e.g., smiles at funerals, and sadness at weddings) contexts and demonstrated that the mimicry of both appropriate and inappropriate smiles was regulated by the participants’ self-reported interpersonal closeness with the expressers, while the mimicry of appropriate and inappropriate sadness was unrelated with closeness. Therefore, it could be concluded that the mimicry of angry expressions did not always happen and was irrespective of interpersonal closeness.

While most previous studies suggested that emotional mimicry improves self–other overlap (Cooke et al., 2018; Hale & Hamilton, 2016; Paladino et al., 2010), less was known about whether self–other overlap promoted emotional mimicry. This study demonstrated that self–other overlap, serving as a mediator, could partially explain the effect of group membership on emotional mimicry. This finding was consistent with previous research showing that pain and motivation contagions only occurred among people with closer relationships (Martin et al., 2015; Radel et al., 2015). Also, our finding was congruent with past research demonstrating that perspective-taking – a way to increase self–other overlap – leads to increases in both behavioral and emotional mimicry (Hawk et al., 2011; Müller et al., 2011). Altogether, the present study suggested that the relationship between emotional mimicry and self–other overlap was mutually influenced rather unidirectional.

Putting emotional mimicry in an intergroup context, the most important finding was that self–other overlap mediated the effect of group membership on emotional mimicry, echoing the mimicry-as-social-regulator model proposed by Hess and Fischer (2013). Although Hess and Fischer’s model posited that the level of interpersonal closeness may aid in the interpretation of group difference in emotional mimicry, little empirical evidence has been provided. An exception is Hühnel et al. (2018), who observed that younger participants displayed enhanced emotional mimicry as well as self–other overlap toward the other-age group (the elders) after being included by an older player in a Cyberball game. However, these results were derived from two separate studies, making it impossible to determine whether the enhanced mimicry toward the other-age group was due to the increased overlap between them. The current study was a more direct investigation of this issue. Synchronous records of both variables enabled the mediation analysis to reveal the role of self–other overlap in the ingroup bias in emotional mimicry.

The current findings offerred complementary support for our prior research showing that rTPJ exerted a modulatory effect in emotional mimicry in different social contexts (Peng et al., 2020). Although it has been confirmed that ingroup bias in emotional mimicry disappeared when the cortical excitability over the rTPJ was temporarily elevated, no conclusion was reached on whether this modulation was achieved by altering the overlapping mental representations for ingroup versus outgroup, given the lack of measurement of self–other overlap. In this study, a straightforward scale for self–other overlap shed light on the behavioral mechanism for the ingroup bias in emotional mimicry and supported our prior research and previous theoretical models (Kraaijenvanger et al., 2017; Wang & Hamilton, 2012) highlighting the role of rTPJ as part of the neural networks relating to the modulatory process of emotional mimicry by social contextual factors (i.e., group membership).

The asymmetric results in the mimicry of happiness and anger was in line with previous studies showing that mimicry of happiness and anger did and did not depend on group membership respectively (Hühnel, 2015; Kuszynski, 2015; Peng et al., 2020), although a previous study has found that mimicry of happiness did not interact with group membership (Van der Schalk et al., 2011). According to Weisbuch and Ambady (2008), emotion-mimicking behavior was determined by the meaning or functions of emotions, and the interpretation of this function was decided by whether the perceiver and the sender share group membership. For example, a happy face might signal “safety” however, when this appears on an outgroup member’s face, it signals strength and dominance of the outgroup relative to the ingroup and the self, and thus, the outgroup member’s happy face elicits less mimicry than the ingroup. Meanwhile, anger is a signal of conflict for both ingroup and outgroup, with a slight difference that outgroup anger connotes relatively stronger conflict than ingroup anger, which is responsible for the ingroup bias in the mimicry of anger in some previous studies (e.g., Van der Schalk et al., 2011), but not in others (e.g., Sachisthal et al., 2016). Moreover, another research suggested that people from China tended to suppress the expression of negative emotions to keep a highly interdependent and harmonious relationship (Wei et al., 2013), which might also explain the current results, in which Chinese participants did not show mimicry of either ingroup or outgroup angry faces.

Several limitations should be noted. First, both studies lacked equal gender distribution of the respondents, which might bias the current findings, considering that previous research has found that emotional mimicry varies between males and females (Schrammel et al., 2009). However, no consistent pattern of gender effects on emotional mimicry has been observed in previous studies (see Seibt et al., 2015 for a review). Therefore, whether the gender of participants affects emotional mimicry remains an open issue, as this requires more evidence. Second, the relatively short presentation of emotional expressions might bias current findings. It should be noted that the clips of Yin’s 3D dynamic facial expression database last around 2,000 ms, which precluded the presentation of longer materials. However, prior research (Achaibou et al., 2008; Deng & Hu, 2017) has revealed that dynamic emotional expressions lasting 1,000 ms are enough to induce emotional mimicry. Third, a subsequent self–other overlap rating was conducted after each emotional clip to enable mediation analysis; however, it is possible that the subsequent rating was driven by the emotional embodied mimicry. The current design cannot preclude this alternative explanation, which is within the framework that group membership and emotional mimicry are bidirectionally related. Finally, although group membership was emphasized to make the ethnicity salient, the relative static design might weaken the ecological validity of the research. Thus a real and interactive paradigm is encouraged in future research to explore the group membership effects on facial emotional mimicry and its underlying mechanisms.

In conclusion, this study revealed that compared with outgroup happiness, there is greater tendency to mimick ingroup happiness, while the mimicry of anger was independent of racial group membership. Furthermore, this study confirmed that greater interpersonal closeness with ingroup versus outgroup members was associated with this ingroup bias in emotional mimicry, which supported the mimicry-as-social-regulator model. Future studies to explore the neural mechanism underlying the effects of group membership as well as other social contextual factors on emotional mimicry, using multimodal methods and designs that are more interactive and ecological were recommended.

Disclosure Statement

No potential conflict of interest was reported by the author(s).

Additional information

Funding

This work was supported by the National Social Science Foundation of China (Major Program) [19ZDA021].