haut de page


























haute de page


































haut de page

mise à jour du
20 novembre 2003
Advances in Experimental Social Psychology
The perception-behavior expressway:
automatic effects of social perception on social behavior
Ap Dijksterhuis, John A. Bargh  
Department of Psychology, University of Amsterdam, The Netherlands
Department of Psychology, New York University, New York, USA
télécharger cet article pdf
Echokinetic yawning, theory of mind, and empathy
Tous les articles sur la contagion du bâillement
All articles about contagious yawning


[...] V. Social perception elicits corresponding behavior

In the following paragraphs, evidence of imitation of observable behavior will be reviewed. The research on imitation of observables can be divided into three domains. First, there is a large literature on imitation of facial expressions. In addition, others have investigated imitation of gestures and movements. Finally, there is evidence of imitation of various speech related variables. The major findings of all three domains will be discussed, starting with facial expressions.

Facial expressions. The evidence for imitation of facial expressions is abundant (e.g., Dimberg, 1982; Vaughan & Lanzetta, 1980; Zajonc et al., 1982). An example of a very contagious facial expression that is familiar to all of us is yawning. If, after a long car or trainride, a person starts to yawn, usually his or her travel companions start to yawn within a few minutes. This tendency to imitate yawning has also been demonstrated empirically. Provine (1986) asked participants to watch a five minute videotape. In one condition, participants watched a video with yawning people, whereas in a control condition participants watched a video with smiling people. As expected, 55% of the participants in the experimental (i.e., yawn) condition started to yawn while watching the video, as opposed to only 21% in the control (i.e., smile) condition. Interestingly, Provine also obtained evidence supporting our claim that activation of the mental representation of an action (which can be the result of perception but also of, for instance, thought) is crucial in eliciting corresponding behavior. That is, one does not have to literally perceive a yawn to engage in yawning. Provine found that reading about yawning or thinking about yawning also caused participants to yawn. Finally, the fact that one of the authors of this paper is yawning right now, can be taken as anecdotal evidence that writing about yawning does the trick as well.

Although no consensus emerged among researchers as to the exact cause of the phenomenon, various investigators have studied imitation of facial expressions among newborns (Anisfield, 1979; Field, Woodson, Greenberg & Cohen, 1982; Jacobsen & Kagan, 1979; Meltzoff & Moore, 1977, 1979, 1983). Meltzoff and Moore (1977; 1979) showed that even one-month-old babies imitate facial expressions. If you look at a baby and open your mouth, the baby will open her mouth. If you stick out your tongue, the baby will often do the same.

An interesting early demonstration of imitation of facial expression among adults can be found in an experiment by O'Toole and Dubin (1968). Their experiment was aimed at investigating mother-child interactions during feeding. They had observed that a mother would usually open her mouth just prior to feeding their infants a spoonful of food. Their intuitive explanation for this finding was that a mother would open her mouth in the hope that her child would do the same and Ðmost importantly- that the food would end up where it is supposed to end up. They put their ideas to a test by watching various mother-infant interactions and observed indeed that both mothers and infants open their mouth. Surprisingly however, in almost 80% of the cases, a mother opens her mouth only after the child does so. In other words, it is the mother who is imitating the child, not vice versa. The child is merely opening his or her mouth upon perceiving the food on its way.

Another example of adult imitation of facial expressions comes from experiments carried out by Bavelas and colleagues (Bavelas, Black, Lemery & Mullett, 1986; 1987). In their experiments, a confederate was the victim of a painful injury that occurred in the presence of the participants. As expected, the participants imitated the expressions of the confederate, that can best be described as a big wince. Interestingly, they also manipulated the visibility of the expression of the confederate. In one condition, the expression of the confederate was easier to see than in a second condition. As a result, the degree to which participants imitated the expression varied as well. More visible expression led to more imitation; that is, the easier it was to perceive the expression the greater the effect on one's own behavior.

Zajonc and colleagues (Zajonc, Adelmann, Murphy & Niedenthal, 1987) reasoned that couples who have lived together for a period of time should have often experienced the same emotions at the same times, and because frequent facial expressions eventually lead to changes in facial lines, they hypothesized that partners should start to look more like each other the longer they are together. In their experiment, they gave participants 24 photographs. These photographs were those of the partners of 12 married couples. Some photographs were made at the wedding, whereas others were made 25 years later. The task of the participants was to assess the degree of resemblance of various pairs of photographs. As predicted, partners who were together for 25 years resembled each other more than random pairs of the same age and than newly-wed couples. Although Zajonc et al. (1987) interpreted these findings in terms of shared emotional experience, these findings are also consistent with the present hypothesis of a direct effect of perception on behavior; that is, it may be that frequent perception of the partner's expression leads one to adopt that same expression repeatedly oneself, producing over time the similarity in facial lines between the two partners (see Bargh, 2000).

Imitation of facial expressions has also been studied in the context of emotional contagion (see e.g., Hatfield, Cacioppo & Berntson, 1994). Our facial expressions affect our emotions through a process of feedback elicited by facial muscles (Strack, Martin, & Stepper, 1988). Imitation of facial expressions therefore leads to shared emotions. In concrete terms, the perception of a sad face evokes a sad expression in the perceiver and the perceiver will actually begin to feel sad as well. In the Zajonc et al. (1987) research, the relation between shared facial expressions and shared emotions was obtained in a follow-up study. They had observed variations as to the degree of resemblance of life partners. This led to the intriguing hypothesis that partners who have grown to look like each other more may actually be happier together than those who have not, because their resemblance is due to a greater history of shared emotions. And, in general at least, shared emotions lead to a stronger bond between partners. A questionnaire study indeed confirmed this hypothesis with effects being impressive in size (with a correlation of .49 between resemblance and self-reported happiness).

Behavior matching. The evidence concerning the imitation of movements and gestures is less abundant than the evidence on imitation of facial expressions. Although theorists have always treated the automatic imitation of postures, gestures and movements as a given (e.g., Allport, 1968; Köhler, 1927), early "evidence" was almost entirely anecdotal (see Bavelas et al., 1986; for reviews, see Capella, 1981; Chartrand & Bargh, 1999; Dijksterhuis, 2000; LaFrance, 1979). Later reports, in which posture imitation (or posture mirroring, as it is called more often) was investigated experimentally, suffered from methodological weaknesses (Charney, 1966; Kendon, 1970). Finally, research in the seventies and early eighties was not so much concerned with the occurrence of posture imitation per se, but instead with the relation between imitation and rapport. These studies (e.g., Bernieri, 1988; LaFrance, 1979; LaFrance & Ickes, 1981) speak to the possible function of posture and gesture mirroring in that some experiments clearly show a strong correlation between posture imitation and rapport. However, they do not shed light on how often people spontaneously engage in posture imitation.

The only early investigation we could identify that exceeds the level of mere anecdotal evidence was reported by Eidelberg (1929). In his experiment, participants were instructed to point at their nose upon hearing the word "nose" and to point at a lamp upon hearing the word "lamp." The experimenter, who was clearly visible to the participants, also pointed at his or her nose or at the lamp upon hearing the corresponding instruction. After a while, the experimenter started to make "mistakes," in that he or she pointed at the lamp upon hearing the word nose and vice versa. Interestingly, participants started to make the same mistakes as well. They spontaneously imitated the gestures made by the experimenter, despite the instruction to follow the verbal cues (i.e., the words "nose" and "lamp") and not the behavior of the experimenter.

Bernieri (1988; see also Bernieri, Reznick & Rosenthal, 1988) was the first to provide truly solid evidence for posture imitation. In his studies, a somewhat complicated but nonetheless ingeneous paradigm was used. First, two participants (A and B) were asked to interact. While they interacted, they were videotaped. A little later, both participants A and B were asked to engage in another interaction with a different participant, such that A interacted with C and B would interact with D. Again, both interactions were videotaped. Subsequently, two tapes were constructed on which the gestures and postures of both participants A and B were displayed. One concerned the actual interaction between A and B. The other tape pictured A while interacting with C, and B while interacting with D. Subsequently, judges -who were unaware of which tape displayed the actual interaction between A and B- estimated the degree of posture similarity. If the degree of matching is greater on the first tape (the actual interaction) than on the second, there is evidence for posture matching. Bernieri (1988) indeed obtained this evidence. People do spontaneously mirror the postures of individuals they interact with.

Chartrand and Bargh (1999) replicated and extended these effects. Instead of investigating posture mirroring, they focused on actions such as foot shaking or nose rubbing. In their first experiment, a confederate was instructed to either rub her nose or shake her foot while working with a participant on a task. Importantly, the two were strangers and had only a minimal interaction, greatly reducing the probability that any imitation as motivational in nature Ðsuch as part of an attempt to ingratiate the other person. Their hypothesis, that participants would mimic the behavior of the confederate, was confirmed. Under conditions where the confederate rubbed her nose participants engaged more in nose-rubbing than in foot-shaking, whereas the opposite was true when participants interacted with the confederate who shook her foot. Chartrand and Bargh (1999) replicated and extended this finding in a second study, in which the confederate purposefully mimicked the body posture of the participant. This study obtained clear evidence that mimicry leads to increased liking of interaction partners. The lack of a motivational basis for these findings supports our thesis of an automatic link between social perception and one's own behavior, in a naturalistic interaction context.

Speech related variables. Finally, there is evidence of automatic imitation of various speech related variables. One phenomena that is investigated by several researchers is syntactic persistence, that is, the tendency to use a certain syntax when this syntax is made cognitively accessible. This phenomenon supports the common coding approach to language comprehension and language production postulated by Prinz (1990). Prinz argues that we use the same mental representations for both comprehension and production of speech. According to Prinz (see also Studdert-Kennedy, 1987), language comprehension and production develop at the same time during ontogeny: "…the ability to produce language is of no use when there is no one to listen, and the ability to understand language is of no use when there is no one to produce it" (pp. 177).

Bock (1986; 1989) reported evidence of syntactic persistence. In one experiment, participants would hear and repeat a sentence such as "The corrupt inspector offered a deal to the bar owner." Later, participants would see a picture of, for instance, a boy handing a valentine to a girl. This picture can be described as "The boy is handing a valentine to a girl" or as "The boy is handing the girl a valentine." As the first sentence has a similar syntactic form as the priming sentence, this is the description participants most often gave. Syntactic structures appear to carry over from one sentence to another.

Whereas in the studies conducted by Bock (1986) participants activated a particular syntax themselves, Levelt and Kelter (1982; see also Schenkein, 1980) investigated syntactic persistence in a social context. In one of their experiments, the experimenter called various shops and either asked "What time does your shop close?" or "At what time does your shop close?". If the former question was asked, shopkeepers more often answered with "Five o'clock", whereas the answer to the latter question was "At five o'clock" in the majority of cases. Importantly, both Levelt and Kelter as well as Schenkein obtained such effects of speech imitation for single words, for clauses as well as for the structural format of entire sentences. Finally, Levelt and Kelter showed that cognitive load did not increase these speech imitation effects (which were already very substantial under normal conditions), suggesting that these effects were automatic in nature.

Recently, Neumann and Strack (2000) obtained evidence for imitation of tone of voice between interaction partners. In one of their experiments, participants listened to an audiotaped speech given by a stranger. While they were listening, participants were asked to repeat what they heard and were audiotaped themselves. It was found that participants adopted the tone of voice of the person on the tape they listened to. A sad tone of voice on the tape elicited a sad tone of voice in the participant, whereas a happy voice led to a happy voice in the participant. These findings are particularly important as they rule out the possibility that participants imitated tone of voice for strategic reasons (e.g., to increase cohesion). They didn't see the person who delivered the speech, they didn't even know who this person was, and no participant was aware of the actual goal of the experiment. Instead, they were successfully led to believe that the experimenters were interested in the reproduction of speech content.

Are emotion and behavior-matching strategic? Bavelas and colleagues (Bavelas et al., 1986, 1987) accounted for their findings with a motivational communicative perspective. They argue that participants imitate in order to show the confederate that they are empathizing with him or her, that they are "feeling their pain." And if there is more eye contact between the confederate and the participant, the participant imitates more because he or she knows that the confederate is better able to see their expression. In other words, they interpret the imitation as a motivated, strategic behavior to create an empathic bond with the other person. This model of imitation (that, according to the division we made earlier between "facilitator-option" and an "inhibitor-option" is an example of a facilitator-option) is the standard account in the field not only of facial mimicry, but of the related phenomena reviewed above of "behavior matching" (La France, 1979, 1982) and "rhythmic synchrony" (Bernieri, 1988; Condon & Ogston, 1966; Condon & Sander, 1974). Most of this research has sought to link behavioral coordination effects with the establishment of rapport and liking between the parties involved, with some researchers viewing empathy as the cause of mimicry and others considering mimicry to be the cause of empathy (see Bernieri & Rosenthal, 1991, and Chartrand & Bargh, 1999, for reviews).

Although it is true that there tends to be greater mimicry when the two individuals like each other than not (e.g., Charney, 1966; LaFrance & Broadbent, 1976), so that rapport between the parties is an important moderator of the effect (see "Moderators" below), this does not mean that the perception-behavior effect requires for its occurrence a motivation or strategy or even positive affect towards the other person as a necessary condition.

After all, the evidence reviewed above shows that the only real precondition of imitation of observable behavior is the perception of the behavior. We would like to emphasize that our explanation of an innate express route between perception and action is supported by this evidence as our explanation would lead one to predict all the reviewed effects to be automatic and non-strategic as opposed to other explanations that claim these effects to be strategic and intentional. There is no evidence at all for the strategic nature of the imitation effects reviewed above, whereas the support for the automatic and unintentional nature of imitation is evident. Meltzoff and Moore (1977) demonstrated a tendency to imitate among newborns. O'Toole and Dubin (1968) showed that mothers tend to imitate their children and there really is no strategic reason to do so. Although Bernieri (1988) showed imitation among people who engaged in an extended interaction (potentially allowing the interactants to engage in motivated imitation), Chartrand and Bargh (1999) showed that even minimal interaction with a complete stranger led to imitation. Finally, Neumann and Strack (2000) obtained evidence for imitation of tone of voice when the person being imitated was not even present.

In sum, there is considerable evidence showing that people automatically imitate observed behavior Ðranging from facial expression and postures to speech patterns. There is no evidence for the strategic nature of the imitation effects, whereas the support for the automatic and unintentional nature of imitation is evident. That is, in the experiments reviewed above, people did not imitate because they wanted to imitate. Instead, they imitated for no other reason than that they are designed to do so.
Response properties of neurons in temporal cortical visual areas of infant monkeys Rodman HR