Abstract
Simple reciprocal matching behaviors, such as facial mimicry, appear fundamental to social development and interpersonal processes. Identifying mechanisms and moderators of these reactions to others' behaviors is thus important to understanding basic social–emotional functioning and specific clinical syndromes. This experiment extends early electromyographic (EMG) research (Berger & Hadley, 1975) to explore whether rapid, subtle mimicry involves a general motor-matching mechanism (e.g., the mirror neuron system) or if it is related solely to emotional processes. The EMG measured responses to short, dynamic videos of smiling, scowling, stuttering, and arm wrestling. Although mimicry of emotional stimuli was greater than to nonemotional stimuli, participants matched both nonemotional mouth movements and facial expressions of emotions. Mimicry of arm motions was not significant. Individuals' levels of mimicry of emotional and nonemotional were positively correlated. Findings suggest that both motor and affective processes are involved in producing rapid mimetic reactions to dynamic stimuli.
Matching of others' behaviors, often termed mimicry, is central to theories of social and emotional behavior (Cacioppo, Petty, Losch, & Kim, 1986; Rogers & Pennington, 1991), development (Meltzoff & Moore, 1977), emotional contagion (Hatfield, Cacioppo, & Rapson, 1993), empathy (Bavelas, Beavin-Black, Lemery, & Mullett, 1987), rapport (Lakin & Chartrand, 2003), and social perception (Niedenthal, Brauer, Halberstadt, & Innes-Ker, 2001). Mimicry may be critical for development of typical social functioning (Decéty & Chaminade, 2003; Iacoboni, 2005). Atypical or deficient mimicry is associated with syndromes such as autism spectrum disorders (Beall, Moody, McIntosh, Hepburn, & Reed, 2008; Hepburn & Stone, 2006; McIntosh, Reichmann-Decker, Winkielman, & Wilbarger, 2006; Moody & McIntosh, 2006; Rogers, 2006), disruptive behaviors (de Wied, van Boxtel, Zaalberg, Goudena, & Matthys, 2006), and dismissive-avoidant attachment (Sonnby-Borgstrom & Jonsson, 2004).
Many descriptions of the importance of mimicry suggest that it is an automatic motor response (Hatfield et al., 1993); such a response may lead to these social and emotional outcomes in part through embodiment processes (see Niedenthal, Barsalou, Winkielman, Krauth-Gruber, & Ric, 2005). However, there are numerous possible causes of rapid interpersonal reactions, and the basic mechanisms underlying mimicry are not well understood; both affective and motor processes have been implicated (Hess, Philippot, & Blairy, 1998; Moody & McIntosh, 2006; Moody, McIntosh, Mann, & Weisser, 2007). Further, much of the literature on mimicry uses electromyographic (EMG) measurement of responses to static stimuli (e.g., Lundqvist, 1995; Lundqvist & Dimberg, 1995; McIntosh et al., 2006; Moody et al., 2007; Winkielman & Cacioppo, 2001) or behavioral coding of responses to human interaction or dynamic stimuli (e.g., Chartrand & Bargh, 1999). More recently, studies using EMG responses to dynamic videos (de Wied et al., 2006) and dynamic computer generated clips (e.g., Mojzisch et al., 2006; Schilbach, Eickhoff, Mojzisch, & Vogeley, 2008; Schrammel, Pannasch, Graupner, Mojzisch, & Velichkovsky, 2009) have emerged to obtain more precise measurements to more ecologically valid stimuli. The present study enhances our understanding of basic mechanisms of mimicry by using EMG to assess rapid responses to both dynamic emotional and dynamic nonemotional movement. Directly comparing such responses helps establish how both affective and motor processes are involved in these rapid responses.
Because facial expressions are markers of affective states (Cacioppo, Martzke, Petty, & Tassinary, 1988; Dimberg, 1997; Winkielman & Cacioppo, 2001), facial expressions that match those seen in another may reflect a matching emotional response (happiness causes happiness, reflected on the observer’s face), not a motor-matching response. Just as rapid and subtle facial expressions can be generated by pictures of nonhuman fear eliciting stimuli (Dimberg, Hansson, & Thunberg, 1998), facial reactions to human faces may be mediated by emotional systems. Evidence for an emotional basis of these responses is demonstrated by emotional context altering rapid interpersonal reactions to faces. Participants who are first exposed to a fear induction reacted to angry (but not neutral) expressions with fear expressions (Moody et al., 2007). By showing that rapid interpersonal reactions can reflect complementary emotional responses rather than simple expression matching, these data suggest that mimicry is at least partly the result of the observer’s emotional response.
Evidence of an emotional component to mimicry, however, cannot rule out other processes causing or influencing mimicry. Other theories of mimicry suggest that these reactions are the result of nonemotional motor matching (e.g., Hatfield et al., 1993). The argument for nonemotional motor-mimicry is supported by Meltzoff and Moore’s (1977) report that neonates match the observed facial movements of a model. These movements were of simple, nonemotional movements such as tongue protrusion and led some to suggest that matching is an innate process needed for social development (Rogers & Pennington, 1991). Although controversy regarding these findings (Koepke, Hamm, Legerstee, & Russell, 1983a, 1983b; McKenzie & Over, 1983a, 1983b) prevents the data from being definitive, they raise the possibility that automatic nonemotional matching exists at a very early age.
More recent research also argues for the existence of a simple, nonemotional motor-mimicry mechanism. Chartrand and Bargh (1999) suggest that matching is the result of a simple perception–behavior link that leads humans to automatically, and unintentionally, match the behaviors, postures, mannerisms, and expressions of others. However, because studies such as this assessed responses over much longer time frames (e.g., bouncing one’s leg when another person was doing the same behavior), it is difficult to compare these results to studies that find very rapid and subtle responses using physiological measures such as with EMG (e.g., Winkielman & Cacioppo, 2001). Further, studies on automatic imitation have shown that participants initiate movements faster when they are compatible with an observed movement than when they are incompatible (Heyes, 2011; Press, Gillmeister, & Heyes, 2006); this suggests that the motor-matching response is primed by observation of others' movements supporting congruent responses but interfering with incongruent ones. However, the findings from studies that rely on these differing methods and outcomes may be the result of separate processes (see Heyes, 2011). Moreover, studies that find mimicry to facial expressions using EMG often confound emotion and motor action (e.g., Oberman, Winkielman, & Ramachandran, 2009). If only angry and happy expressions are shown, the resulting mimicry could be the result of nonemotional motor matching, emotional processes, or both. Therefore, additional work needs to be done to determine whether nonemotional motor and/or emotional matching processes are evident at short time frames and at more subtle levels, and to compare directly emotional and nonemotional matching. Such rapid and subtle matching may be a building block for more visible and extended matching (Moody & McIntosh, 2006; Rogers, 1999); demonstrating that it occurs is thus a necessary step to exploring moderators and causes of the more evident mimicry and imitation.
There is good reason to believe that a nonemotional matching system could cause such subtle mimicry. The Mirror System (MS) first discovered in monkeys and later inferred in humans (Iacoboni et al., 1999; Rizzolatti, Craighero, & Fadiga, 2002) may account for certain types of matching (Bekkering, 2002). Further, early work by Berger and Hadley (1975) suggests a motor mechanism may, in part, cause mimicry. They recorded EMG activity in the arms and lips of observers watching videotapes of arm wrestling and stuttering. They found greater EMG activity over observers' muscles corresponding to the muscles being used by the performers than in muscles not corresponding to muscles being used by the performers. The EMG technology of that time permitted a less detailed and precise analysis than is available today. For example, because they did not have computer assisted analysis much larger muscle reactions were required to be able to observe changes, and they did not examine changes in activity in short time windows. Further, Berger and Hadley (1975) did not use videos of emotional expressions making it impossible to compare mimicry resulting from emotional and nonemotional stimuli.
The present experiment used a method similar to Berger and Hadley’s (1975) to address the question of whether mimicry exists in response to both emotional and nonemotional stimuli as measured by EMG. The present experiment used more sophisticated EMG procedures which permitted exploration of rapid and subtle phasic changes in muscle activity to both emotional and nonemotional stimuli as well as both face and nonface actions. If mimicry is primarily the result of emotional processes, we should expect reaction only to emotional expressions (anger and happiness). If we find mimicry to all stimuli, regardless of emotional content, this would suggest that a motor-mimicry mechanism is also operative. Also, mimicry to nonface actions would suggest that a general motor-matching mechanism exists that is not specific to the face. Finally, the present experiment used dynamic stimuli; dynamic stimuli are advantageous because they may recruit more neural systems than still images (Yoshikawa & Sato, 2006) and are more ecologically valid.
Method
Participants
A total of 41 psychology undergraduates were given extra credit for participating. Data from four were removed from the analyses due to excessive artifacts in the EMG signal. All analyses reported are for the remaining 37 participants (22 female, average age = 19.84 years, SD = 1.4). Participants were treated in accordance with American Psychological Association (APA) ethical guidelines (American Psychological Association, 2002) and procedures were approved by the university Institutional Review Board (IRB).
Stimuli
Video clips of four types of movements were shown to participants. Each movement was modeled by one male and one female actor. Each action was compiled into a block of stimuli that included five male and five female repetitions of each action. Blocking may decrease extraneous orienting responses and has been used successfully in past studies of mimicry (e.g., Dimberg, 1997; Dimberg & Petterson, 2000). Two blocks of clips were emotional facial expressions: a happy smile and an angry scowl. Following Berger and Hadley (1975), the two other blocks were of nonemotional face and body movements: video of a person stuttering the word “pass” and video of the arms of two individuals arm wrestling. Each clip of facial actions began with the actor in a neutral expression, then performing the expression or movement, and finally returning to the neutral expression. The arm-wrestling clip started with two hands grasping each other preparing to arm wrestle, then a brief arm-wrestling match with the actors maximizing their wrist movements followed by the right arm winning the match. To increase ecological validity, the actors performed these actions in a way that was most comfortable to them and thus appeared natural. As a result, duration of the clips varied (happy—female: 5.9 s, male: 7.8 s; angry—female: 3.8 s, male: 6.9 s; stutter—female: 3.2 s, male: 1.9 s; arm wrestling—female arm forward: 5.0 s, male arm forward: 8.7 s). All clips were silent, so any effects were due to visual information only.
Procedures
In a within-subjects design, participants were exposed to each of the four blocks of video clips while EMG recordings were taken over the muscle groups that lead to congruent muscle movements. Participants were seated comfortably while EMG sites were prepared as described in the Measures section. Once appropriate recording parameters were met, participants were given sufficient time to acclimatize to the situation. Next, participants were shown the four blocks of clips. Order of the blocks was balanced by a Latin square method and assigned randomly. Presentation of each set was controlled by the computer to ensure equal and counterbalanced presentation of the male and female examples.
Measures
The EMG was used to record levels of muscle activity over the zygomaticus major (raises corners of mouth), corrugator supercilii (knits brow), orbicularis oris (purses or compresses lips), and carpi radialis and flexor digitorum sublimis muscles of the forearm (flexes forearm). Activity over the zygomaticus muscle indicates positive emotions (Cacioppo et al., 1986); corrugator activity has been shown to be a marker of negative emotions such as anger (Cacioppo et al., 1986). The orbicularis oris and forearm flexor groups were included as measures of reactions to stuttering and arm wrestling, respectively, consistent with Berger and Hadley (1975).
Standard EMG site preparation and electrode placement procedures were followed (Tassinary, Cacioppo, & Geen, 1989). Before electrode placement, skin over the muscle group was cleansed with rubbing alcohol and gently abraded with NuPrep Gel®. Electrodes were 4 mm Ag-AgCl, cup-style electrodes. Muscle activity was continuously recorded using a NeuroScan Labs, SynAmps® model 5083 electroencephalograph amplifier. Activity over each muscle group was recorded using two electrodes placed approximately 1.25 cm apart from center to center, roughly parallel to the length of the muscle. Activity over each muscle was continuously recorded at a sampling rate of 2000 Hz with a 10–500 Hz band pass filter and a 60 Hz notch filter. The EMG signals were immediately amplified at the head-box by a factor of 150 and again by the main amplifier by a factor of 500.
To analyze EMG, each continuous file was first visually inspected for noise and artifacts. Next, the waveform around each stimulus presentation was visually inspected by a research assistant, blind to condition and hypotheses, to look for artifacts and anomalous waveforms. Sweeps that contained clearly anomalous waveforms were dropped from the analyses. No more than 10% of the total number of sweeps for each individual were dropped.
Following visual inspection, EMG data were used to calculate responses to the stimuli. Two windows of activity were used. First, the prestimulus baseline window was the 1000 ms before stimulus onset in which the participants saw a blank screen with a central fixation point. Second, the movement onset window was calculated. Because each stimulus varied in length, a unique postmovement onset window was calculated for each to ensure that the participants' muscle activity was in response to the most pronounced movements being witnessed and responded to. This window was the point from stimulus movement onset to point of maximal stimulus movement. For the happy clips it was 3267 ms; for the angry clips it was 2772 ms; for the stutter clips it was 990 ms; for the arm-wrestling clips it was 3927 ms. Because the time needed to move one’s mouth to speak or stutter is shorter than the time needed to express an emotion or complete an arm-wrestling match, the postmovement onset window for the stutter clips is considerably shorter than for the other movements.
These data were smoothed and rectified, and the integral under the curve for each time window was calculated using CNS Analysis Suite, version 5.51 (The CNS Analysis Suite, 1999). Consistent with other research (e.g., Moody et al., 2007; Winkielman & Cacioppo, 2001), the integral values were next log10 transformed to reduce the impact of extreme values. These values were then standardized within participant and within muscle so meaningful comparisons could be made across muscles and participants. Next, the prestimulus level of activity was subtracted from the poststimulus activity to measure the change in activity caused by viewing each stimulus (i.e., to calculate the change from baseline). Each participant’s mean level of activity for each muscle (zygomaticus, corrugator, orbicularis oris, and the forearm flexor group) was calculated in response to each type of stimulus.
Results
To assess whether the pattern of muscle responses differed depending on the stimulus being observed, a Muscle (4: zygomaticus, corrugator, orbicularis oris, forearm flexor) × Stimuli (4: happy, angry, stuttering, arm wrestling) repeated measures analysis of variance (ANOVA) was run with Gender as a between-participants factor. There were no significant main effects; however, the Muscle by Stimuli interaction was significant, F(9, 171) = 10.14, p < .001, ϵ = .47, η2 p = .35.
The ANOVA demonstrated that there were differences in responses to stimulus type. The next step was to determine whether mimicry occurred for each specific stimulus. We operationalized mimicry as more activation of the congruent muscle compared to the overall activation of all incongruent muscles to the same stimulus. Specifically, we calculated the mean activation of the incongruent muscles for each individual and compared that value to the activation of the congruent muscle (McIntosh et al., 2006). For example, mimicry to smiling is demonstrated by more zygomaticus activation to happy compared to the mean of corrugator, flexor, and orbicularis to happy. A significant paired sample t-tests indicates that there is mimicry for that stimulus.
Because there are theoretically predicted directional hypotheses, these tests are one-tailed. Most comparisons showed evidence of matching. Activation over corrugator in response to angry expressions (M = .31, SD = .55) was higher than the combination of other muscles (M = .12, SD = .25), t(25) = 1.81, p = .04. Activation over zygomaticus in response to happy expressions (M = .69, SD = .62) was higher than the combination of other muscles (M = –.03, SD = .28), t(35) = 6.38, p < .001. Activation over orbicularis oris in response to stuttering (M = .17, SD = .37) was higher than the combination of other muscles (M = .0004, SD = .20), t(32) = 3.19, p = .002. Activation over forearm flexor in response to arm wrestling (M = .13, SD = .16) trended in the direction of being greater than the combination of other muscles (M = .07, SD = .16) but this difference was not significant, t(32) = 1.33, p = .09. Figure 1 displays the means for these comparisons.
Congruent muscle activation versus mean of all other muscle activity in response to stimulus type. Bars represent 1 standard errors
We next compared the amount of mimicry across stimuli. To calculate the amount of mimicry to each stimulus, we subtracted the mean activation of the incongruent muscles from the activation of the congruent muscle. For example, zygomaticus to happy expressions minus the mean of corrugator to happy, orbicularis to happy, and flexor to happy results in the amount of mimicry to happy expressions. A one-way repeated measures ANOVA revealed that there were different amounts of mimicry across stimuli, F(3, 57) = 4.9, p = .02, ϵ = .56, η2 p = .21. Follow-up analyses using paired sample t-tests indicated that there was greater mimicry to happy expressions (M = .71, SD = .76) than to angry expressions (M = .22, SD = .53), t(24) = 2.43, p = .02 (as this and the following contrasts are exploratory, tests are two-tailed), greater mimicry to happy expressions (M = .66, SD = .66) than to arm-wrestling clips (M = .05, SD = .22), t(32) = 5.28, p < .001, greater mimicry to happy expressions (M = .66, SD = .66) than to stuttering (M = .17, SD = .31), t(31) = 3.90, p < .001, and a trend toward greater mimicry to stuttering (M = .15, SD = .27) than to arm wrestling (M = .03, SD = .22), t(28) = 1.79, p = .09. Figure 2 shows the 95% confidence intervals for each differences score allowing comparison among responses to each stimulus.
Confidence intervals for mimicry for each stimulus type
To directly examine the degree of mimicry to emotional and motor-only stimuli, we averaged the mimicry to happy and angry expressions (emotional), then averaged mimicry to stuttering and arm wrestling (motor-only). These scores were compared using a paired sample t-test, revealing that there was more mimicry to the emotional (M = .52, SD = .47) than motor-only stimuli (M = .12, SD = .23), t(36) = 5.42, p < .001. Moreover, one sample t-tests revealed that there was mimicry to both emotional t(36) = 6.78, p < .001 and motor-only stimuli, t(36) = 3.22, p = .003. Figure 3 shows the 95% confidence intervals for these scores.
Confidence intervals for mimicry for stimuli class (Emotional vs. Motor-Only)
Finally, to assess the within-person association of strength of mimicry to the different classes of stimuli, we computed the correlation between these classes of stimuli. The amount a person mimics emotional stimuli is correlated positively with the amount the person mimics motor-only stimuli, r(37) = .34, p = .04.
Discussion
These data demonstrate that subtle matching of emotional and motor-only movements occurs quickly and to dynamic stimuli. Mimicry occurred both to emotional expressions (happy and angry) and to stuttering. There was a trend toward nonemotional body mimicry, as well. In addition, these data add to the growing body of work demonstrating mimicry to videotaped dynamic stimuli (e.g., de Wied et al., 2006; Mojzisch et al., 2006; Schilbach et al., 2008; Schrammel et al., 2009). By using moving images, the stimuli become more ecologically valid and may recruit more neurological systems. This change may contribute to differences in the degree of mimicry to videos compared to research that relies on static photos; future work should directly compare reactions to these different modes of stimulus presentation.
Although we found mimicry to several facial actions, the differences in congruent versus incongruent movement were not significant when participants were observing arm wrestling. These data suggest that mimicry to facial actions exists regardless of emotional content, but do not demonstrate rapid mimicry to body actions. It may be that body mimicry is simply less rapid than facial mimicry (but see Press et al., 2006, showing rapid effects for hand stimuli). Alternatively, the null findings may be partly due to less attention being focused on the moving elements of the arm-wrestling stimuli. There appears to be a mechanism for preferential looking at faces present from birth (Bednar & Miikkulainen, 2003), which may draw attention to facial features thus facilitating mimicry. During the arm-wrestling clips, participants' attention may have been less consistently focused on the wrist movement and diverted to parts of the clip not relevant to mimicry (e.g., arms, shirt sleeves, arm hair). Body mimicry may be more evident when there is more consistent motivation to focus on another’s specific body movements (as when there are task demands beyond simply watching, e.g., Press et al., 2006). Research specifying conditions under which body matching occurs is needed. One limitation of the present study was the absence of an emotion-relevant body stimulus. Recent work confirms that body configurations communicate emotion (App, McIntosh, Reed, & Hertenstein, IN PRESS), so follow-up research should include such stimuli.
Taken together, our findings extend Berger and Hadley’s (1975) results and suggest that rapid and subtle mimicry may result from a variety of processes. That there was mimicry to emotional and nonemotional facial displays suggests that mimicry is at least partly the result of motor-mimetic mechanisms that may function independently or in conjunction with emotional processes influencing responses to emotional expressions (Moody et al., 2007). Rapid responses that match others' expressions is consistent with involvement of the MS (see Heyes, 2011). More generally, these data add to literature that suggests that mimicry is the result of multiple processes (Lee, Dolan, & Critchley, 2008).
By observing responses within participants to a variety of stimuli, we can comment on similarities and potential differences among these stimuli. Beyond the presence of mimicry to both motor-only displays and emotion displays, there was a significant but moderate correlation between levels of mimicry to these classes. This suggests that there is some association of the processes involved in responses to each type of stimulus, perhaps reflecting within-individual consistency in MS sensitivity. At the extreme, such consistent individual differences in MS responsiveness may be connected to social–emotional deficits seen in autism (Beall et al., 2008; McIntosh et al., 2006; but see Press, Richardson, & Bird, 2010). Another source of individual differences may be early experiences altering emotional responses (Reichmann-Decker, DePrince, & McIntosh, 2009) or learned connections between sensory and motor representation (see Heyes, 2011). However, the moderate size of the correlation leaves ample room for different processes being involved between the two stimulus classes. One interpretation is that the correlation is due to the presence of a consistent motor-mimicry response across emotional and motor-only stimuli, but an additional emotional process that influences the responses to the emotional stimuli.
The possibility of multiple processes is further suggested by differences in the levels of observed mimicry. Mimicry to happy expressions showed the greatest response, and more generally, mimicry to emotional expressions was greater than to motor-only actions. Separate motor and emotional mimicry mechanisms may exist that operate separately or additively. If this is the case, the relatively larger emotional mimicry reactions reported here suggest that the emotional mimicry system may lead to larger reactions than nonemotional mimicry or that the emotional responses are a combination of both processes.
Alternatively, the differing degrees of mimicry could be the result of a variety of neural processes. For instance, Schilbach and colleagues (2008) found that brain regions related to social cognition may be related to mimicry. It is possible that social cognitive processes influenced the mimicry recorded in this study. Further, recent work on the MS suggests several possibilities for why the degree to which we mirror observed actions varies. For example, mirror neuron activation seems to depend on the extent to which the observer identifies with the actor (Oberman, Ramachandran, & Pineda, 2008). The stimuli classes used here may vary by how easily the can be identified with. That is, emotional expressions are by nature nonverbal forms of communication (App et al.,
Explanations based on communication and identification, however, do not explain why reactions to happy expression are so much larger than the other expressions. One possibility is that the different magnitudes of mimicry are the result of the participants' context. The MS responds to the context of the observed action (Heyes, 2009; Umilta et al., 2001), so the variation may be related simply to different levels of motor matching. Indeed, people mimic smiles more of people they like not their scowls (McIntosh, 2006); if this context and the dynamic videos prompted observers to like or feel close to the individuals in the stimuli, smile mimicry may have been differentially enhanced. More generally, there are a number of possible contextual and individual moderators of mimicry. Future research is needed to understand what processes operate in a given context and under what task demands to determine if and how affective and motor systems relate to each other. Directly measuring participants' subjective emotional responses to the stimuli would help clarify the association of emotion to the behavioral responses observed. Future studies should also assess a wider range of facial and bodily stimuli to identify when emotional and nonemotional mimicry occurs to isolate influences on the nature and magnitude of the response.
Regardless of the reasons for the differences, these findings have implications for existing theories that rely on the assumption of behavioral matching being the result of motor-mimicry. For example, one explanation of emotional contagion posits that mimicry reactions are a nonemotional reflex-like behavior that leads to emotional contagion through facial or bodily feedback processes (Hatfield et al., 1993; Hatfield, Cacioppo, & Rapson, 1994; McIntosh, 1996; McIntosh, Druckman, & Zajonc, 1994). The data reported here demonstrate that people do match nonemotional actions, supporting the motor-mimicry assumption. However, that mimicry reactions are also a display of rapid emotional responses (Dimberg, 1997; Moody et al., 2007) adds complexity to a pure motor-mimicry account. If emotional reactions occur simultaneously with nonemotional motor-mimicry, it is difficult to determine whether motor-mimicry is a cause of emotional contagion. Despite these complications, however, our findings of rapid and subtle nonemotional motor-mimicry and other studies supporting this account (e.g., Sigall & Johnson, 2006) suggest that it is a viable and potentially important process. Future research should attempt to determine both how much of emotional contagion is caused by motor-mimicry, and how other mechanisms and variables influence the process.
Footnotes
The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This research was supported in part by a National Research Service Award through the National Institute of Mental Health to the first author: F32MH081409.