تقلید صورت و سرایت عاطفی برای حالات چهره عاطفی پویا و تأثیر آنها بر دقت رمزگشایی

Abstract The present study had the goal to assess whether individuals mimic and show emotional contagion in response to relatively weak and idiosyncratic dynamic facial expressions of emotions similar to those encountered in everyday life. Furthermore, the question of whether mimicry leads to emotional contagion and in turn facilitates emotion recognition was addressed. Forty-one female participants rated a series of short video clips of stimulus persons expressing anger, sadness, disgust, and happiness regarding the emotions expressed. An unobtrusive measure of emotional contagion was taken. Evidence for mimicry was found for all types of expressions. Furthermore, evidence for emotional contagion of happiness and sadness was found. Mediational analyses could not confirm any relation between mimicry and emotional contagion nor between mimicry and emotion recognition.

Introduction It has been suggested that mimicry — the imitation of others’ non-verbal displays by an observer — plays an important role in the communication of affective states [e.g. Freud, 1921, based on a theory by Lipps, 1907 and Bavelas et al., 1986]. For example, Rogers (1957) saw the imitation of a client's non-verbal behavior as a means to communicate empathy and some schools of therapy (see, e.g. Siegel, 1995) advocate imitation as a means of understanding the client's internal state. Facial mimicry in this context is usually conceptualized as an automatic, reflex-like process (see, e.g. Lipps, 1907, Hoffmann, 1984 and Hatfield et al., 1993), with the observer's facial expression matching the observed facial expression. Emotional contagion is a closely related concept that is sometimes defined in overlapping terms (e.g. Hatfield et al., 1993). It is therefore useful to define the specific use of the two terms in the framework of the present study. Specifically, we consider as mimicry the congruent facial reactions to the emotional facial displays of others. That is, mimicry is defined exclusively as an expressive component. In contrast, we define emotional contagion as an affective state that matches the other's emotional display. In a recent review, Hess et al. (1999) conclude that evidence from studies on both adults and infants strongly suggests that, in general, people adopt facial, postural, and vocal behaviors that are congruent with the displays they observe, and that these displays often represent mimicry (see also Dimberg, 1990). However, some examples of counter-mimicry effects (e.g. Lanzetta and Englis, 1989 and Hess, 1998) have also been reported. Specifically, Lanzetta and Englis (1989) found mimicry in a collaborative task situation but counter-mimicry in a competitive task situation. This, and evidence that mimicry may depend on the type of task the participant is engaged in (Hess et al., 1998), suggests that mimicry may not be an automatic, reflex-like mechanism. Furthermore, a number of studies suggest that individuals tend to report emotional states that match the facial emotion displays to which they have been exposed (see, e.g. Hatfield et al., 1993, Strayer, 1993, Laird et al., 1994, Schneider et al., 1994 and Lundqvist and Dimberg, 1995). The two processes, mimicry and emotional contagion, have been suggested to be causally elated. This idea goes back to Lipps (1907) who suggested that the imitated expression leads — via a feedback process — to emotional contagion. As regards the influence of emotional contagion on empathy (the capacity to recognize the emotional state of others), Lipps (1907) as well as Hoffmann (1984) imply that emotional contagion should in turn facilitate emotion recognition. Related ideas have more recently been expressed by Hatfield et al. (1993). Similarly, Cappella (1993), based on evidence in favor of the facial feedback hypothesis (FFH) in particular, proposes that facial feedback from mimicry causes contagion. However, evidence is accumulating that emotional contagion may not be causally related to mimicry (Gump and Kulik, 1996; Blairy et al., 1999). Also, in their review of the literature, Hess et al. (1999) could not find any consistent evidence that mimicry facilitates emotion recognition. Together, these findings throw doubt on the notion that emotion recognition is related to a reflex-like mimicry process via contagion. So why do people mimic at all since this process seems to not be related to either emotional contagion or emotion recognition accuracy? Before answering this question a second look at the evidence reported above is necessary. First, despite the evidence for facial mimicry reported above, it is not clear whether individuals mimic the type of expressions they are likely to encounter in real life. This, because evidence for mimicry in adults is largely based on studies that employed very intense, prototypical facial expressions presented as still photographs. For example, the extensive studies on mimicry and contagion in adults by Dimberg and Lundqvist (e.g. Dimberg, 1990, Lundqvist, 1995 and Lundqvist and Dimberg, 1995) employed stimuli selected from the ‘Pictures of facial affect’ (Ekman and Friesen, 1976), which are a set of highly recognizable and prototypic facial expressions. Such stimuli may in fact elicit a reflex-like response due to their extremity that is not found for less extreme expressions. This notion is supported by the observation that studies finding evidence for the situation dependence of mimicry employed somewhat weaker and more natural expressions (Lanzetta and Englis, 1989; Gump and Kulik, 1996; Hess et al., 1998). Also, McHugo et al. (1991) and Bourgeois and Hess (1999) using video exert of news programs featuring politicians found that mimicry was modulated by the political attitude of the observer. That is, observers were more likely to mimicry a politician if they shared his political beliefs than when not. In sum, studies finding clear evidence for facial mimicry and emotional contagion tend to employ prototypical, high intensity, still photographs as stimulus material, whereas those studies that found evidence of situational influences on mimicry employed more naturalistic, less prototypical or weaker, facial stimuli. Thus, the question of whether individuals mimic the type of expressions they are likely to encounter in everyday life deserves further investigation. As regards the lack of evidence for the facilitation of emotion recognition by mimicry, it is possible that the use of prototypical stimulus material may not suffice to uncover subtle improvements in decoding accuracy due to mimicry. Specifically, the process described by Lipps (1907) is relatively elaborate and demands both a certain empathic ability and introspection. Yet, emotion displays can be decoded by using other sources of information. The sender's emotion displays, the facial, vocal, postural, etc., expressions emitted by the sender can be used to draw inferences regarding the presumed emotional state of the sender using a pattern-matching approach (e.g. Buck, 1984). For example, the presence of upturned corners of the mouth and of wrinkles around the eyes can be interpreted as signaling happiness whereas eyebrows drawn together in a frown may signal anger. This process should be especially useful for the decoding of the prototypical, highly recognizable expressions often used in studies on mimicry and contagion. In those circumstances, the additional information provided by the process described by Lipps may not in fact add to the already high level of decoding accuracy. Thus, facial mimicry may be expected to have a facilitative effect on emotion recognition mainly in situations where the emotion displays are relatively weak and non-prototypical and where the participants do not know the sender and have no previous attitude towards the sender. Finally, it may also be that contagion, but not mimicry facilitates emotion recognition. The present study is aimed to address these issues. Given the preceding considerations, it was considered important to use expressions that were not preselected according to their correspondence to an emotion stereotype. Furthermore, to enhance ecological validity dynamic facial expressions were chosen. Specifically, participants were asked to decode a series of video clips of emotional expressions of happiness, anger, sadness, and disgust. These facial expressions were recorded during an emotional imagery task and represent the spontaneous, idiosyncratic expressions of the expressor. Participants’ own facial expressions were measured using facial electromyography and their emotional state was assessed using an unobtrusive self-report measure. Mediational analyses (Baron and Kenny, 1986) were used to assess the influence of mimicry on emotional contagion and on emotion recognition.

Results 3.1. Facial mimicry The presence of facial mimicry implies a pattern of facial activity in response to the emotional display of others. To assess facial mimicry, we therefore verified first whether a distinct pattern of facial activity emerged in response to the stimuli. For this, the standardized difference scores were analyzed using a one-way repeated measures ANOVA across muscle sites. This analysis across muscle sites is allowable as facial EMG data were previously transformed into z-scores and the data are thus on the same scale. We then used planned contrasts to assess whether the facial activity conformed to the expected patterns specified as a function of the mimicry indices described above. Specifically, for both anger and sadness, mimicry should be indexed by significantly higher levels of Corrugator supercilii than Orbicularis oculi and Levator LAN activity. However, Levator LAN activity should be significantly lower for sadness than for anger mimicry. Mimicry to happy displays should be indexed by significantly higher levels of Orbicularis oculi than Corrugator supercilii and Levator LAN activity. For disgust displays, Corrugator supercilii activity should be significantly lower than Orbicularis oculi and Levator LAN activity. The means for participants’ facial EMG activity are shown in Fig. 1. Note that the data was transformed to z-scores and that the zero level refers thus to the mean across all data points and not to an absence of difference from baseline. Observers’ facial EMG as a function of the emotional facial expression of the ... Fig. 1. Observers’ facial EMG as a function of the emotional facial expression of the sender. Figure options Significant main effects of muscle site, indicating a differentiated pattern of muscle activity across sites, emerged for participants’ facial activity during the decoding of anger, F2,37=7.29, P=0.002, sadness, F2,37=11.85, P<0.001, and happy episodes, F2,37=36.86, P<0.001, but not for disgust episodes. Furthermore, the contrasts specified for the anger, F1,38=9.91, P=0.003, sadness, F1,38=21.99, P<0.001, and happy episodes F1,38=45.41, P<0.001, were also significant. The test comparing Levator LAN activity for anger and sadness episodes was marginally significant in the predicted direction, t=−1.70, P=0.098. Thus, clear evidence emerged for mimicry during the decoding of anger, sadness, and happiness displays. 3.2. Emotional contagion To assess emotional contagion, planned contrasts were conducted on the relevant scales of the well-being questionnaire. As we can not assume that the different emotion scales are scaled identically, the ratings were first z-transformed. Since use of z-scores obliterates information regarding the grand mean for the scales, a separate analysis was conducted on the grand means. Across all conditions, participants reported to be quite cheerful (mean=24.46, S.D.=31.55) as well as somewhat irritated (mean=17.13, S.D.=17.38). This emotional state is quite consistent with the experimental situation, which was calm and relaxed but presented the participants with occasional frustrations when they had difficulty with the decoding task. Importantly, this implies that the emotional contagion effects reported below modulated the participants’ affective state rather than changing it. The planned contrasts assessed whether participants reported feeling more cheerful when decoding expressions of happiness, more repulsed when decoding expressions of disgust, more sad/depressed when decoding expressions of sadness, and more irritated/aggressive when decoding expressions of anger. The means are shown in Fig. 2, note that the zero level of the z-scored data refers to the mean across all data points and that negative scores indicate levels lower than the mean and not the use of a bimodal scale. Observers’ self-reported emotional state as a function of the emotional facial ... Fig. 2. Observers’ self-reported emotional state as a function of the emotional facial expression of the sender. Figure options Significant contrasts in the expected direction emerged for self-reported affect during the decoding of happy, F1,40=8.74, P=0.005 and sad faces, F1,40=4.45, P=0.041, for which participants reported significantly more cheerfulness and sadness/depression, respectively. During the decoding of anger expressions the pattern of the means did not conform to the planned contrast as repulsed emerged as the highest rated emotion, and participants reported feeling significantly more repulsed than cheerful, F1,40=7.14, P=0.011. Yet it is interesting to note, that the French word ‘repulsé’ connotes a desire to get away from something. Thus, reporting this feeling in response to an anger expression, while not indicative of empathy, represents an adequate emotional reaction to a threat display (see, e.g. Dimberg and Öhman, 1996). No difference emerged for the decoding of disgust expressions. In sum, the pattern of results is consistent with contagion of happiness and sadness. 3.3. Decoding accuracy Decoding accuracy was 45% for anger expressions, 87% for happiness expressions, 43% for disgust expressions, and 75% for sadness expressions. 3.4. Mediational analyses The notion that mimicry leads to emotional contagion which in turn positively influences decoding accuracy, was assessed using mediational analyses as proposed by Baron and Kenny (1986). For this, decoding accuracy was first regressed on facial activity as assessed by facial EMG (path a). Second, self-reported emotional feeling state was regressed on facial activity (path b). Finally decoding accuracy was regressed on both self-reported feeling state and facial activity (path c).1 The mediational pathway proposed by Lipps would be confirmed if facial activity predicts self-reported emotional state and decoding accuracy, whereas self-reported emotional state predicts decoding accuracy after facial activity is controlled for. Table 1 shows the beta values for the paths. For each emotion expression the self-report on the corresponding scale was employed as well as EMG data from all muscle sites. For this analysis the untransformed difference scores were employed. Image for unlabelled figure Figure options Inspection of Table 1 does not provide evidence for a mediation of decoding accuracy through contagion from mimicry. Considering first the two analyses for the expressions for which evidence for both mimicry and contagion emerged in the previous analyses, that is, expressions of happiness and sadness. For the decoding of happiness expressions, we note no evidence that either higher levels of activity of Orbicularis oculi or lower levels of activity of Corrugator supercilii are associated with either self-reports of cheerfulness or decoding accuracy. Furthermore, no evidence for a link between cheerfulness and accuracy in decoding happy expressions emerges. Regarding the decoding of sad expressions, we note that lower levels of Orbicularis oculi (which is congruent with, but certainly not indicative of a sad expression) are related to self-reported sadness and to decoding accuracy; however, no direct link between emotional state and decoding accuracy emerged once facial activity was controlled for. Similarly, lower levels of Levator LAN are associated with an increase in decoding accuracy. At best, this pattern suggests that lower levels of facial movements incongruent with mimicry are related to an increase in decoding accuracy but no mediation through emotional contagion seems to be involved. The pattern for the decoding of anger is even less supportive of Lipps’ model. Here, the presence of movements that are incongruent with mimicry, specifically, Orbicularis oculi and Levator LAN activity is positively related to self-reports of irritation/aggressiveness. It is possible to speculate, based on the lower decoding accuracy for this expressions (43%), that anger expressions were particularly difficult to decode and that participants therefore showed both some grimacing and reported more irritation/aggressiveness. Yet, this finding is clearly not in support of Lipps’ model. For the decoding of disgust expressions no evidence for either mimicry or contagion emerged. We thus can not expect mediation through these processes. Low levels of both Currugator supercilii and Orbicularis oculi activity were found to be related to higher decoding accuracy, suggesting that lower levels of facial movements that are incongruent with mimicry are related to an increase in decoding accuracy, but again no evidence for mediation via contagion emerges. Furthermore, Levator LAN activity is positively related to self-reported feelings of repulsedness.