The Many Faces of a Neutral Face - Head Tilt

Embed Size (px)

DESCRIPTION

Face analysis in nonverbal communication

Citation preview

  • Journal of Nonverbal Behavior 27(2), Summer 2003 2003 Human Sciences Press, Inc. 111

    THE MANY FACES OF A NEUTRAL FACE:HEAD TILT AND PERCEPTION OF DOMINANCE AND EMOTION

    Alain Mignault and Avi Chaudhuri

    ABSTRACT: Based on the premise that human head tilt is homologous to animaldominance displays, we hypothesized that when a head is bowed, the face shouldbe perceived as submissive, sad, displaying inferiority emotions (i.e., shame, embar-rassment, guilt, humiliation, and respect) and, paradoxically, as contracting the zy-gomatic major muscle. Conversely, a raised head should be perceived as moredominant and displaying greater superiority emotions (i.e., contempt and pride). Weconducted two experiments showing 3-D models of faces to 64 participants. Theresults confirmed our hypotheses and also showed that a raised head connoteshappiness. In addition, we found a significant influence of the actors sex on partici-pants perception, such as a bias towards perceiving stronger upward contraction ofthe mouth in female than male actors when the head is tilted. We discuss thesefindings within the context of evolution and social behavior.

    KEY WORDS: dominance; facial emotion displays; head angle; sex; smile.

    Social animals often have dominance hierarchies. Within a hierarchy,animals perform dominance and submission (or appeasement) displays.The hypothesized functions of these dominance relationships are to mini-mize conflicts and increase social cohesion. Whereas an appeasement dis-play represents recognition of a dominance relationship, a dominancedisplay, on the other hand, may be impressive enough to dissuade a chal-lenger from attacking (Preuschoft & van Schaik, 2000). Furthermore, domi-nance displays can act as a self-fulfilling prophecy by reinforcing dominantstatus, which in turn leads to priority in access to resources (Tiedens,2000). Resource priority, of course, has a significant impact on the chancesof surviving and mating. The efficacy of a dominance/appeasement display

    Alain Mignault and Avi Chaudhuri, Department of Psychology, McGill University.Address correspondence to Alain Mignault, Department of Psychology, McGill Univer-

    sity, 1205 Dr. Penfield Avenue, Montreal, Quebec, Canada H3A 1B1; e-mail: alainego.psych.mcgill.ca; URL: http://www.psych.mcgill.ca/labs/cvl/.

  • 112

    JOURNAL OF NONVERBAL BEHAVIOR

    depends heavily on how it is perceived by others. The perception of domi-nance/appeasement displays is the focus of this paper.

    Appeasement and dominance gestures are especially frequent in pri-mates. For example, subordinate chimpanzees greet dominant ones bycrouching and bowing their upper body almost to the ground, while thedominant animal adopts an erect posture and raises its body hairs to in-crease its stature (de Waal, 1989). As a dominance display, the rhesusmonkey sometimes quickly jerks its head up and then back to the restingposture while staring (Redican, 1982). More generally, many animals in-crease their apparent size in dominance and reduce their apparent size insubmission (Darwin, 1872/1965; Lorenz, 1966; Preuschoft & van Schaik,2000; Redican, 1982). Like other researchers (e.g., Castelfranchi & Poggi,1990; Keltner & Anderson, 2000), we believe that these ritualized domi-nance displays are not limited to monkeys and apes. We hypothesize thathead inclination in humans is also a component of dominance displayshomologous to those of other primates.

    Many instances exist across different cultures of an association be-tween head angle and dominance or emotion. For example, to connotecontempt, the English language contains expressions such as to turn upones nose at someone, while Italian contains con il naso all insu`(with the nose up). To describe pride, French has garder la tete haute(to keep the head high) and Hindi has shir uchcha rakna (keep oneshead high). To describe shame, Chinese speakers use the expression ditou ren zui (to bow ones head in pleading a wrong doing). The Germanexpression la den Kopf nicht hangen (dont let your head hang low)implies avoidance of sadness or submission, as does the Spanish expres-sion levantar cabeza (to lift the headmeaning: to get back on onesfeet). To describe respect, Japanese speakers say atama ga sageru (tolower ones head).

    A widespread method to emphasize a persons importance is to use araised platform. Judges, royalty, clergy, and dictators often appear raisedabove everyone else. Similarly, Hollywood stars are often shown descend-ing a high stairway. Conversely, soldiers and police often force prisoners toadopt positions in which their body is bowed or lowered. In filmmaking,the use of camera angles to increase or decrease a persons apparent heightand power has long been common lore in the profession (Mandell & Shaw,1973).

    In this study, we explore issues relating head angle, sex differences,dominance, and emotions. We are specifically interested in the followingquestions: Is the perception of dominance linked to head angle? Whichemotions are linked to head angle? Are we more likely to perceive a

  • 113

    ALAIN MIGNAULT, AVI CHAUDHURI

    bowed head as smiling? Are the two sexes perceived similarly? And, whatevolutionary ties link these phenomena?

    Head Angle and the Perception of Dominance

    Only a few studies have examined the influence of head angle on percep-tion of dominance in humans. Most have confirmed that participants per-ceive a head viewed from below as more dominant and from above asmore submissive. Henley and Harmon (1985) found that standing whileinteracting with a sitting person is perceived as a dominant gesture. Usingone newscaster and 143 participants, Mandell and Shaw (1973) found thatperceived potency increased as the height of the camera decreased. How-ever, Kappas, Hess, Barr, and Kleck (1994) found no significant influenceof angle on the perception of dominance. They showed 72 dynamic facialstimuli at 40, 0, and 40 degrees in viewing angle, and asked 16 partic-ipants to rate these faces on 13 scales including submissive/dominant. It ispossible that their non-significant results may be due to a small samplecombined with a conservative multivariate statistical design.

    Head Angle and Emotion

    Many emotion researchers have suggested that a specific head angle is acomponent in the expression of certain emotions. A bowed head has beenproposed as a component of shame (Darwin, 1872/1965; Ekman & Oster,1979; Izard, 1971, 1991; Keltner & Harker, 1998; Tomkins & McCarter,1964; Wallbott, 1998; Wiggers, 1982), humiliation (Tomkins & McCarter,1964), embarrassment (Keltner, 1995; Keltner & Anderson, 2000; Keltner &Buswell, 1997), sadness (Darwin, 1872/1965; Ekman & Friesen, 1976;Frijda, 1986), disgust (Wallbott, 1998), and respect (Morris, 1977). A hori-zontal (i.e., erect) head posture has been associated with pride (Darwin,1872/1965). A raised head has been linked to contempt (Izard, 1971,1991; Morris, 1977; Rosenberg & Ekman, 1995; Wiggers, 1982), pride(Wallbott, 1998), joy (Wallbot, 1998), and boredom (Wallbott, 1998).

    To study the influence of head angle on the perception of Ekmansbasic emotions (Ekman, 1992), researchers showed neutral facial expres-sions at different angles. This method is likely to reveal participants per-ceptual biases. However, the results are contradictory. In a first task,Kappas et al. (1994) found, using films of four human actors, that a headseen from above was perceived as sadder than when seen from below.

  • 114

    JOURNAL OF NONVERBAL BEHAVIOR

    However, in a second task using wire-frame virtual actors they obtainedthe opposite results. Lyons et al. (2000) found with one female actor and aforced choice paradigm that Japanese and British participants perceivedmore happiness than sadness in a bowed head. The virtual stimuli used byKappas et al. (1994) did not include shading and showed lines connectingthe points, which may have impeded the holistic perception of sadness.They suggested replicating their experiment with a better 3-D model. Weimplemented their suggestion in our study.

    Emotions and Dominance

    We hypothesized that some human expressions of emotion operate as sta-tus displays homologous to greeting behaviors in non-human primates.Frijda (1986) proposed that pride, contempt, and scorn involve dominatingaction tendencies, while humility, shame, guilt, remorse, and devotion in-volve submitting action tendencies. Tiedens (2000) proposed that angerand pride are linked to dominance whereas guilt, sadness, and gratitudeare linked to submission. Many researchers have described embarrassmentas a submission display (e.g., Castelfranchi & Poggi, 1990; Keltner & An-derson, 2000; Leary & Kowalski, 1995). Barrett (1995) wrote that the func-tion of shame is to communicate deference and submission to otherswhereas guilt communicates a submission to social standards. Finally,many studies have found that expressions of fear, surprise, and sadness areperceived as being submissive (e.g., Etcoff, 1984; Hess, Blairy, & Kleck,2000).

    To test whether some emotions function as status displays, we decidedto focus our attention on the emotions whose predominant function is mostlikely to signal superior or inferior status. This criterion eliminates emotionssuch as sadness and happiness. The second criterion was to retain onlyemotions that researchers had previously associated with head angle (asreviewed above). This criterion eliminates emotions such as anger, fear,gratitude, and surprise. Combining these criteria, we obtained two catego-ries that are nearly identical to the emotions related by Frijda (1986) to anaction tendency of dominating or submitting. Thus, our operating hypoth-esis was that participants should be more likely to perceive inferiority emo-tions (i.e., embarrassment, shame, humiliation, guilt, and respect) in abowed head and more superiority emotions (i.e., pride and contempt) in araised head.

  • 115

    ALAIN MIGNAULT, AVI CHAUDHURI

    Head Angle and the Perception of the Mouth

    Although early theorists (e.g., Darwin, 1872/1965) stated that sadness islinked to a bowed head, Kappas et al. (1994) hypothesized that when thehead is bowed the U-shape of the jaw leads the observer to perceive aneutral face as if the corners of the mouth were pulled up, while a raisedhead gives the impression that they are pulled down. Lyons et al. (2000)described this as a conflict between the message sent by a bowed headand one sent by the mouth. However, we believe that the Kappas et al.(1994) hypothesis has never been properly tested because previous studiesasked participants about perceived emotions. To avoid ambiguity, we ex-plicitly asked participants to evaluate mouth contraction.

    Method

    Design

    We conducted two experiments. Because these two experiments aresimilarly designed, we present them together. For each task, a four-factordesign was employed with participants sex as the between-persons factorwhereas the within-persons factors were actor, actors sex, and head angle.Note that in the present paper we use the term actor to refer to the personwho appears in the picture shown to participants. Actors and head anglesare fully crossed with participants.

    Stimuli

    Stimuli for both experiments were the same. We had the choice be-tween a set of 3-D models of faces with eyes closed and another set witheyes open. We chose the set with eyes closed, because staring is related todominance and gaze aversion to submission (for a review, see Ellyson,Dovidio, & Brown, 1992). With a 3-D model showing opened eyes, therewould have been a direct gaze only at 0 degrees. We would have beenunable to determine whether perceived dominance was due to gaze orhead angle. To avoid this potential confound, we chose to eliminate gazeby using actors with closed eyes.

    We drew on an existing set of 3-D models of faces that could bevirtually rotated in order to ensure that exactly the same expression wasdisplayed at each angle. They were created by scanning over 20,000points on each face with a laser, providing an exact representation of the

  • 116

    JOURNAL OF NONVERBAL BEHAVIOR

    facial surface (for a detailed description of these 3-D models see Bruce etal., 1991; Bruce & Langton, 1994; Hill & Bruce, 1996). We used six maleand six female actors with a neutral facial expression. The virtual camerawas placed one meter away from the face. The projection from the 3-Dmodels to the camera assumed perspective. Rendering was performedusing Geomview 2.0. A Gouraud shading function was used to interpolateshading. The standard OpenGL illumination model was used with a singlelight source at infinity with an angle of 30 degrees and a small amount ofambient light. The rendering process of these images is fully described in aprevious publication from our laboratory (Liu, Collin, & Chaudhuri, 2000).Each face was displayed on a gray background. Note that these images donot contain texture, but only 3-D information. For each actor, we gener-ated views from the following angles: 30, 20, 10, 0, 10, 20, and 30degrees. Therefore, we used a total of 84 stimuli (7 angles 12 actors).Examples of stimuli are provided in Figure 1.

    For the 10 practice trials of each task, we used 3-D models from an-other set. This face database was provided by the Max-Planck Institute forBiological Cybernetics in Tuebingen, Germany. We rendered these imagesusing Matlab 5.1. We did not use texture in order to make the images assimilar as possible to the set of Bruce and colleagues. For the practice trialswe used only an angle of view of 0 degrees. Thus, we had 10 stimuli of 10different actors.

    Participants

    In Experiment 1, there were 31 participants (11 males, 20 females), allMcGill University students. Of this number, 17 received course credits for

    Figure 1. Examples of a female (top row) and male (bottom row) actor at differenthead angles.

  • 117

    ALAIN MIGNAULT, AVI CHAUDHURI

    their participation and 14 received ten dollars. In Experiment 2, we em-ployed 33 participants (12 males, 21 females), also students. Of this num-ber, 16 received course credits and 17 received ten dollars.

    Procedure

    The procedure was the same for both experiments. Participants satwith their eyes at about 52 cm from the 48 cm Macintosh Colorsync moni-tor. Each experiment involved three tasks performed on a Power MacintoshG3 computer, all of which were structured in the same way, with ten prac-tice trials at the beginning of a task followed by two blocks of 42 trials.

    In each trial, a fixation cross was presented for 250 ms, followed by ablank screen for 200 ms, and finally by a facial image. The image appearedfor a maximum of 6 sec or until the participant typed in an answer. Therelevant question was displayed at the same time as the image. The imagesmeasured 15.8 cm in height and 12.3 cm in width. The between-eyes dis-tance of each face was standardized at 4.0 cm. The fixation cross waspositioned to appear exactly in the middle between the eyes. Order ofpresentation of the stimuli was randomized across participants. There wasan inter-trial interval of 1 sec, after which participants had to press a key tostart the next trial.

    Dependent Measures

    In Experiment 1, we measured the perception of age, emotion, anddominance. For the perception of age, participants were asked: How olddoes this person look in this picture (in years)? They could answer anynumber between 0 and 85.1 In the second task, participants were asked thefollowing open-ended question: What is the main emotion expressed inthis picture? (Answer in one word.) They could answer none when theyperceived no emotions. In the third task, participants were asked: Does thisface look submissive or dominant in this picture? Answers were based on aLikert scale from 1 (submissive) to 9 (dominant).

    In Experiment 2, there were three tasks but only the first task is rele-vant to this paper; the other two will not be described here.2 In the firsttask, participants were asked: In which direction are the muscles of themouth contracted? Participants answered on a Likert scale from 1 (verydownward) to 9 (very upward).

  • 118

    JOURNAL OF NONVERBAL BEHAVIOR

    Results

    Outliers

    Participants with atypical response patterns were identified by plottingthe mean responses of all participants as a function of head angle. Therewere no outliers in Experiment 1, but two in Experiment 2. Therefore, weanalyzed the data from 31 participants in each experiment.

    Preliminary Analyses

    As we are not concerned with answers to a particular actor and be-cause this factor is nested under the sex of the actor, we averaged the dataacross actors within each level of the other factors. Other researchers haveused the same averaging technique (e.g., Hess, Blairy, & Kleck, 2000). Pre-liminary analyses were conducted to assess the influence of participantssex on the dependent variables because we suspected, based on other re-searchers results, that it had a negligible influence (e.g., Hess, Blairy, &Kleck, 2000; Kappas et al., 1994; Mignault, 1999). A different three-wayGLM analysis was performed for each task of each experiment except theemotion task. Participants sex was non-significant as a main effect as wellas in interaction with the other variables for all three tasks.3 This factor wastherefore omitted from further analyses.

    Main Analyses

    All tasks except the perception of emotion task were analyzed in thesame manner, that is, with a two-way GLM with repeated measures using a2 (sex of actor) 7 (angle) design. Because the data lacked sphericity,Greenhouse-Geisser corrections were employed for all repeated measures.The results of these analyses are summarized in Table 1. Separate trendanalyses as a function of angle were conducted for actors of either sex. Inthese analyses, we ignored terms higher than cubic as the interpretation oftheir meaning is too complex.

    Perception of Dominance

    The mean results for the perception of dominance as a function ofhead angle are plotted in Figure 2. In this figure, as in the following ones,each point represents the mean of 186 observations (i.e., 6 stimuli 31participants). Most of the variance is explained by a significant main effectof the actors sex and a significant main effect of angle. The main effect of

  • 119

    ALAIN MIGNAULT, AVI CHAUDHURI

    Table 1

    Analyses of Variance of Experiments 1 and 2: Perception of Dominanceand Mouth Contraction as a Function of Head Angle and Actors Sex

    Dominance Mouth contraction

    Source df F p eta2 df F p eta2

    Sex of Actor (S) 1 172.7 .000* .852 1 40.4 .000* .574Angle (A) 1.6 32.3 .000* .519 2.2 59.3 .000* .664S A 3.9 7.4 .000* .199 4.5 7.4 .000* .197

    Male ActorsContrastsLinear 1 19.9 .000* .398 1 74.7 .000* .713Quadratic 1 12.5 .001* .294 1 1.2 .289 .037Cubic 1 10.2 .003* .254 1 28.1 .000* .484

    Female ActorsContrastsLinear 1 59.9 .000* .666 1 84.4 .000* .738Quadratic 1 1.7 .201 .054 1 16.4 .000* .354Cubic 1 3.8 .061 .112 1 18.7 .000* .384

    *p 0.01.

    dominance indicates that perceived dominance varies with head angle.The main effect of the actors sex indicates that women are perceived asless dominant than men. Separate trend analyses revealed that, for maleactors, the linear, quadratic, and cubic terms were significant, whereas forfemale actors only the linear term was (see Table 1).

    Which Basic Emotions Are Associated With Head Angle?

    Participants used 318 different labels to describe the emotions theyperceived in the 84 stimuli. We categorized responses into Ekmans basicemotion categories, that is, anger, contempt, disgust, fear, happiness, sad-ness, surprise plus the category neutral (Ekman, 1992). We added thecategory other for the answers that did not fit into these categories. Asthere were few observations in the categories of contempt, disgust, andsurprise, we transferred these responses into the other category.

    We conducted a Pearson chi-square analysis for two-way tables

  • 120

    JOURNAL OF NONVERBAL BEHAVIOR

    Figure 2. Perception of dominance as a function of head angle for male and femaleactors.

    (Olson, 1987)4 with a significance level of 0.01. Expected values were esti-mated by taking the mean number of observations across angles. Resultsare presented in Table 2. For both male and female actors, we found asignificant influence of head angle only on the perception of sadness andhappiness. Head angle had no significant influence on the perception ofanger, fear, neutral, and the category other.

    In the case of sadness, results are plotted in Figure 3. Seven post-hoc(Bonferonni corrected) chi-square tests for each sex revealed that partici-pants were significantly more likely to perceive sadness when the head isbowed at 30 degrees (for male actors: 2(1) 27.1, p 0.0014; forfemale actors: 2(1) 6.6, p 0.00144) and less likely to perceive itwhen the head is raised at 30 degrees (2(1) 11.4 and 19.2, for maleand female actors, respectively; p 0.0014). For happiness, results areplotted in Figure 4. Post-hoc chi-square tests confirmed that more happi-ness was perceived at a head angle of 30 degrees (2(1) 21.5 and 19.7,for male and female actors, respectively; p 0.0014).

  • 121

    ALA

    IN M

    IGN

    AU

    LT, AV

    I CH

    AU

    DH

    UR

    I

    Table 2

    Number of Responses and Results of Pearson Chi-Square Analyses for the Perception of Ekmans BasicEmotions as a Function of Head Angle and Actors Sex

    Angle Analysis

    Emotion Sex 30 20 10 0 10 20 30 2 p

    Anger Male 10 21 22 21 29 19 21 9.2 .103Female 2 3 4 5 4 6 8 5.2 .401

    Fear Male 6 3 7 8 7 5 4 3.4 .646Female 3 7 4 6 5 2 5 3.9 .576

    Happiness Male 9 9 8 12 25 16 35 38.0 .000*Female 43 35 32 27 36 43 69 27.7 .000*

    Neutral Male 72 71 78 75 62 94 76 7.5 .191Female 53 60 65 74 83 81 61 11.4 .045

    Sadness Male 53 38 28 26 19 11 9 54.7 .000*Female 42 41 37 38 21 14 5 46.6 .000*

    Other Male 36 44 43 44 44 41 41 1.2 .890Female 43 40 44 36 37 40 38 1.3 .888

    The categories neutral and other were added to Ekmans basic emotions. There were too few responses for surprise, contempt, anddisgust, so they were included in the other category.

    *p 0.01.

  • 122

    JOURNAL OF NONVERBAL BEHAVIOR

    Figure 3. Proportion of responses in the sadness category as a function of headangle for male and female actors.

    Influence of Actors Sex on Perception of Basic Emotions

    To study the main effect of actors sex, we conducted tests for thedifference between two proportions. Participants were significantly morelikely to perceive happiness for female than male actors (z 9.46,p 0.01). Participants were significantly more likely to perceive anger formale than female actors (z 8.82, p 0.01). We found no influence ofactors sex on the probability of perceiving fear, neutral, and sadness(z 0.96, 2.05, and 0.78, respectively; p 0.01).

    Perception of Inferiority and Superiority Emotions

    Are bowed heads more likely to communicate inferiority emotionsand raised heads to communicate superiority emotions? In the inferiorityemotions category, we included answers such as shame, shyness, regret,guilt, embarrassment, and respect. In the superiority emotion category, weincluded answers such as pride, self-assurance, feelings of superiority, dis-

  • 123

    ALAIN MIGNAULT, AVI CHAUDHURI

    Figure 4. Proportion of responses in the happiness category as a function of headangle for male and female actors.

    dain, and smugness. The results for inferiority and superiority emotions areplotted in Figures 5 and 6, respectively. Here the small frequencies forcedus to analyze head angle and actors sex separately. We found a significantinfluence of head angle on the perception of both inferiority and superi-ority emotions (2(5) 21.0 and 19.3, respectively; p 0.01). In separateanalyses, we found that participants were significantly more likelyto perceive superiority emotions for male than female faces (z 2.71,p 0.01) and significantly more likely to perceive inferiority emotions forfemale than male faces (z 2.52, p 0.01).

    Perception of the Mouth

    The results of Experiment 2, that is, the perceived mouth contractionas a function of head angle, are presented in Figure 7. Statistical analysesshowed that the main effect of angle explains most of the variance. Sepa-rate trend analyses for each actors sex as a function of angle revealed thatthe linear and cubic terms were significant for male actors, whereas for fe-

  • 124

    JOURNAL OF NONVERBAL BEHAVIOR

    Figure 5. Proportion of responses in the inferiority emotion category as a function ofhead angle for male and female actors.

    male actors the linear, quadratic, and cubic terms were significant (seeTable 1). The corners of the mouth are perceived as contracted upward in abowed head and as contracted downward in a raised head. According tothe post-hoc tests, women are perceived as contracting their mouth up-ward significantly more than men at 30, 10, and 30 degrees (t (4.52) 17.38, 9.09, and 10.80, respectively; p 0.01).

    Discussion

    Perception of Dominance

    We found what common lore would have predicted, that is, perceiveddominance is directly related to head angle. These results extend the find-ings of Mandell and Shaw (1973), but contradict the non-significant resultsobtained by Kappas et al. (1994). The main effect of angle was very largeand unambiguous (see Figure 2). Our results strongly support the hypoth-

  • 125

    ALAIN MIGNAULT, AVI CHAUDHURI

    Figure 6. Proportion of responses in the superiority emotion category as a functionof head angle for male and female actors.

    esis that head angle is a status signal. By using faces with closed eyes, wewere able to measure the specific influence of head angle apart from gaze.We would expect an interaction between head angle and gaze, but this isfor future research to investigate.

    Perception of Basic Emotions

    We found that our participants were more likely to perceive sadness inbowed heads than in raised heads. This result confirms Darwins (1872/1965) hypothesis and Kappas et al.s (1994) findings in their first task, butconflicts with the findings of their second task as well as the findings ofLyons et al. (2000). How can we explain these discrepancies? Lyons et al.(2000) used a forced choice procedure between happiness and sadnesswhereas we used an open-ended design. We found that more than 50 per-cent of the responses were neither in the sadness nor the happiness catego-ries at any angle. Russell (1993) has shown that forced-choice designssometimes yield consensus on the wrong answer and sometimes fail to

  • 126

    JOURNAL OF NONVERBAL BEHAVIOR

    Figure 7. Perception of the direction of mouth contraction as a function of headangle for male and female actors.

    yield consensus on the right answer. Frank and Stennett (2001) have shownthat adding a category none of these terms are correct can prevent atleast the first of these problems. Lyons et al. (2000) did not use such acategory and it may be that their design yielded consensus on the wronganswer for this reason.

    We expected to find an increase in perception of happiness in bothsexes when the head is bowed, given Kappas et al.s (1994) U-shaped jawhypothesis. However, this result was only apparent in females (Figure 4).The increase in happiness perception when the head is raised was unex-pected. As many researchers (e.g., Gross & Levenson, 1995) discriminatebetween high arousal happiness (which we will call joy) and low arousalhappiness (i.e., contentment), we decided to investigate whether these twoemotional states differed in their association with head angle. We foundthat the perception of joy differs drastically from the perception of content-ment. The probability of perceiving contentment varies proportionally withhead angle in the same manner as superiority emotions (Figure 6). How-ever, we obtained a U-shaped curve for the perception of joy in female

  • 127

    ALAIN MIGNAULT, AVI CHAUDHURI

    faces, but a linear relation for male faces. Closed eyes are more salientwhen the head is raised than bowed. We assumed that closed eyes wouldappear neutral but they may be associated with a state of relaxation, whichis close to contentment in emotional space (Russell & Feldman Barrett,1999). Alternatively, some researchers have proposed that one of the char-acteristics of dominance postures is that they are relaxed (Cashdan, 1998;Morris, 1977).

    Perception of Superiority and Inferiority Emotions

    The pattern of our findings supports the hypothesis that inferiority andsuperiority emotions are associated with head angle (see Figures 5 and 6).We recognize, however, that the response frequencies are low. Using afree-response method, Izard (1971) found that participants often confusedshame displays with sadness. Several authors are of the opinion that nega-tive self-conscious emotions are often ignored and go unlabelled or un-acknowledged (e.g., de Rivera, in press; Gans, 2000).

    Perception of the Mouth

    The results of our second experiment support Kappas et al.s (1994)hypothesis that the U-shape of the jaw leads to the perception of a contrac-tion of the corners of the mouth when the head is bowed and a contractionin the reverse direction when the head is raised (see Figure 7). This is thefirst study to measure this relationship directly. Thus, a bowed head leadsto the misperception of smile-related cues. However, like other authors, weobtained contradictory results. For female actors, lowering the head leadsto an increased probability of perceiving both sadness and joy. Our inter-pretation is that both a bowed head and the illusory smile cues that itengenders are components consistent with a submissive (or appeasement)display. Note that the use of faces with closed eyes should have no signifi-cant influence on the perception of the curvature of the mouth.

    Several authors have proposed that smiling is homologous to the pri-mate appeasement display called silent bared-teeth display (Preuschoft &van Hooff, 1995; van Hooff, 1972), though this issue is controversial (Redi-can, 1982). However, there are different kinds of human smiles (Brannigan& Humphries, 1972) and they may connote different levels of dominanceor submission. The type of smile for which evidence of appeasing functionis strongest is the closed smile (Otta, Lira, Delevati, Cesar, & Pires, 1994;Prkachin & Silverman, 2002; Sarra & Otta, 2001). In our opinion, this typeof smile is most similar to the expression perceived in a bowed head.

  • 128

    JOURNAL OF NONVERBAL BEHAVIOR

    Our results may have important implications for the study of smiling.First, the illusory mouth contraction that we found may bias the measure-ment of smiling even when assessed with an apparently objective instru-ment such as Ekman and Friesens (1978) Facial Action Coding System.High inter-rater agreement may result from a shared perceptual illusion.Indeed, a bowed head probably leads raters to perceive the contraction ofthe Zygomatic Major (Action Unit 12) and a raised head to perceive thecontraction of the Triangularis (Action Unit 15). Second, although wefound no significant difference in the perception of mouth contraction atzero degrees between male and female actors, a large difference is per-ceived at other head angles even though the expression itself is fixed. Thisraises questions as to the validity of studies that have found differences inthe amount of smiling between men and women (e.g., Hecht & LaFrance,1998). The sex difference that we found may be due either to a greaterinfluence of stereotypes on the more ambiguous inclined faces or to a sexdifference in mouth shape.

    Sex Differences

    We found that, compared with male actors, females were more likelyto be perceived as happy and as expressing inferiority emotions; less likelyto be perceived as angry and as expressing superiority emotions; and wereperceived as less dominant. These results replicate the findings of otherauthors for the perception of dominance (Keating, 1985; Zebrowitz, Fel-lous, Mignault, & Andreoletti, in press) and anger (Algoe, Buswell, & De-Lamater, 2000). These sex differences are consistent both with socialstereotyping based on womens unequal status and an evolutionary expla-nation based on greater innate aggressiveness in males (similar to mostprimates).

    Conclusion

    Dominance displays are performed in order to be seen by other con-specifics. Adaptive mechanisms have evolved to decode them accuratelyin order to avoid aggression or to gain priority over resources. By studyingneutral faces, we were able to discover perceptual biases and infer theirfunctions. Our results show that a bowed head connotes submission, inferi-ority emotions, sadness, joy in women, and an illusory smile. A raisedhead connotes dominance (especially in women), superiority emotions,happiness (i.e., joy and contentment), and an illusory downward contrac-

  • 129

    ALAIN MIGNAULT, AVI CHAUDHURI

    tion of the corners of the mouth. Some authors (e.g., Smith & Scott, 1997;Turner & Ortony, 1990) have proposed that there are fundamental facialcomponents that are related to different cognitive appraisals. For example,a frown is displayed when identifying an obstacle. In a similar way, wethink that a bowed head is a component whose function is to signal sub-mission, whereas a raised head signals dominance. Because of the sim-ilarity in gesture and context, we think that bowing the head in humans ishomologous to bowing in apes and more generally to body reduction inmany animals (e.g., Keltner & Buswell, 1997). Of course, further cross-cultural and comparative research is needed to confirm the evolutionaryorigin of bowing the head.

    Our results resolve some of the puzzling contradictions in the litera-ture but also raise new questions. Participants perceive both smiling-re-lated cues and sadness in a bowed head, which appears contradictory, butboth are consistent with a submission display. However, why would femalefaces be more likely to be perceived as joyful than male faces when thehead is bowed? Given that bowing the head is a component of severalemotions (i.e., shame, embarrassment, humiliation, guilt, and sadness),have these different emotions evolved from the same submissive display?Similarly, have contempt and pride evolved from the same dominant dis-play? While awaiting answers to those questions, we bow to salute themany facets of the human face.

    Acknowledgments

    The study was supported by research grants from NSERC to AC. The au-thors would like to thank Janet Cleveland for her editorial assistance on themanuscript and Farah Averill for data collection. We received help fromChang Hong Liu in rendering 3-D facial models. We also wish to thank thetwo anonymous reviewers who helped improve the manuscript.

    Notes

    1. We hypothesized that because children view adults from below, adults may associate aface seen from below with dominance and adulthood. Thus, faces seen from below shouldappear older. We found the opposite relation for male actors and a weak support for acubic relation between angle and perceived age in female actors. Therefore, there was notmuch support for our hypothesis and we did not include details about this task in thispaper. However, interested readers are invited to communicate with the first author formore details.

    2. In the second task, participants were asked to evaluate masculinity/femininity and in thethird task they had to evaluate facial maturity.

  • 130

    JOURNAL OF NONVERBAL BEHAVIOR

    3. In both the perception of dominance and the perception of mouth contraction, the three-way interaction between angle, actors sex, and participants sex approached significance(df 2.87 and 2.14; F 2.88 and 2.59; p 0.023 and 0.034; Eta2 0.090 and 0.082,respectively).

    4. We recognize the limits of this analysis because repeated measures violate the assumptionof independence required for using a Pearson chi-square analysis. However, we studiedthe four most common ways in which frequency data with repeated measures are analyzedin literature on facial expressions and all of them have problems. Researchers mostly useanalysis of variance (e.g., Killgore, 2000), chi-square (e.g., Carroll & Russell, 1996), Logitmodels (e.g., Macho & Leder, 1998), or simple t-tests (e.g., Gosselin, 1995). The last threemethods were not designed for repeated measures. In addition, performing multiple t-testsincreases the risk of type I error. Finally, analysis of variance is designed for comparingmeans not frequencies, which usually are not normally distributed.

    References

    Algoe, S. P., Buswell, B. N., & DeLamater, J. D. (2000). Gender and job status as contextualcues for the interpretation of facial expression of emotion. Sex Roles, 42, 183208.

    Barrett, K. C. (1995). A functionalist approach to shame and guilt. In J. P. Tangney & K. W.Fischer (Eds.), Self-conscious emotions: The psychology of shame, guilt, embarrassment,and pride (pp. 2563). New York: Guilford Press.

    Brannigan, C. R., & Humphries, D. (1972). Human non-verbal behaviour, a means of commu-nication. In N. Blurton Jones (Ed.), Ethological studies in child behaviour (pp. 3764).Cambridge, UK: Cambridge University Press.

    Bruce, V., Healey, P., Burton, M., Doyle, T., Coombes, A., & Linney, A. (1991). Recognizingfacial surfaces. Perception, 20, 755769.

    Bruce, V., & Langton, S. (1994). The use of pigmentation and shading information in recogniz-ing the sex and identities of faces. Perception, 23, 803822.

    Carroll, J. M., & Russell, J. A. (1996). Do facial expressions signal specific emotions? Judgingemotion from the face in context. Journal of Personality and Social Psychology, 70, 205218.

    Cashdan, E. (1998). Smiles, speech, and body posture: How women and men display socio-metric status and power. Journal of Nonverbal Behavior, 22, 209228.

    Castelfranchi, C., & Poggi, I. (1990). Blushing as a discourse: Was Darwin wrong? In W. R.Crozier (Ed.), Shyness and embarrassment: Perspectives from social psychology (pp. 230251). Cambridge, UK: Cambridge University Press.

    Darwin, C. (1872/1965). The expression of the emotions in man and animals. Chicago: Uni-versity of Chicago Press.

    de Rivera, J. (2003). Aggression, violence, evil, and peace. In T. Millon & M. J. Lerner (Eds.)Handbook of psychology, volume 5: Personality and social psychology (pp. 569598).New York: Wiley & Sons, Inc.

    de Waal, F. (1989). Peacemaking among primates. Cambridge, MA: Harvard University Press.Ekman, P. (1992). Are there basic emotions? Psychological Review, 99, 550553.Ekman, P., & Friesen, W. V. (1976). Pictures of facial affect. Palo Alto, CA: Consulting Psychol-

    ogists Press.Ekman, P., & Friesen, W. V. (1978). Facial Action Coding System. Palo Alto, CA: Consulting

    Psychologist Press.Ekman, P., & Oster, H. (1979). Facial expressions of emotion. Annual Reviews of Psychology,

    30, 527554.Ellyson, S. L., Dovidio, J. F., & Brown, C. E. (1992). The look of power: Gender differences

    and similarities in visual dominance behavior. In C. L. Ridgeway (Ed.), Gender, interac-tion, and inequality (pp. 5080). New York: Springer-Verlag New York Inc.

  • 131

    ALAIN MIGNAULT, AVI CHAUDHURI

    Etcoff, N. (1984). Perceptual and conceptual organization of facial emotions: Hemisphericdifferences. Brain and Cognition, 3, 385412.

    Frank, M. G., & Stennett, J. (2001). The forced-choice paradigm and the perception of facialexpressions of emotion. Journal of Personality and Social Psychology, 80, 7585.

    Frijda, N. H. (1986). The emotions. Cambridge, UK: Cambridge University Press.Gans, J. S. (2000). The detection of shame in group psychotherapy: Uncovering the hidden

    emotion. International Journal of Group Psychotherapy, 50, 381396.Gosselin, P. (1995). Le developpement de la reconnaissance des expressions faciales des emo-

    tions chez lenfant. [Development of the recognition of facial expressions of emotions inchildren.] Canadian Journal of Behavioural Science, 27, 107119.

    Gross, J., & Levenson, R. W. (1995). Emotion elicitation using films. Cognition and Emotion,9, 87108.

    Hecht, M. A., & LaFrance, M. (1998). License or obligation to smile: The effect of power andsex on amount and type of smiling. Personality and Social Psychology Bulletin, 24,13321342.

    Henley, N. M., & Harmon, S. (1985). The nonverbal semantics of power and gender. In S. L.Ellyson & J. F. Dovidio (Eds.), Power, dominance, and nonverbal behavior (pp. 151181).New York: Springer-Verlag New York, Inc.

    Hess, U., Blairy, S., & Kleck, R. E. (2000). The influence of facial emotion displays, gender,and ethnicity on judgments of dominance and affiliation. Journal of Nonverbal Behavior,24, 265283.

    Hill, H., & Bruce, V. (1996). The effects of lighting on the perception of facial surfaces. Journalof Experimental Psychology: Human Perception and Performance, 22, 9861004.

    Izard, C. E. (1971). The face of emotion. New York: Appleton-Century-Crofts.Izard, C. E. (1991). The psychology of emotions. New York: Plenum Press.Kappas, A., Hess, U., Barr, C. L., & Kleck, R. E. (1994). Angle of regard: The effect of vertical

    viewing angle on the perception of facial expressions. Journal of Nonverbal Behavior, 18,263280.

    Keating, C. F. (1985). Gender and the physiognomy of dominance and attractiveness. SocialPsychology Quarterly, 48, 6170.

    Keltner, D. (1995). The signs of appeasement: Evidence for the distinct displays of embarrass-ment, amusement, and shame. Journal of Personality and Social Psychology, 68, 441454.

    Keltner, D., & Anderson, C. (2000). Saving face for Darwin: The functions and uses of embar-rassment. Current Directions in Psychological Science, 9, 187192.

    Keltner, D., & Buswell, B. N. (1997). Embarrassment: Its distinct form and appeasement func-tions. Psychological Bulletin, 122, 250270.

    Keltner, D., & Harker, L. (1998). The forms and functions of the nonverbal signal of shame. InP. Gilber & B. Andrews (Eds.), Shame: Interpersonal behavior, psychopathology, and cul-ture. (series in affective science, pp. 7898). Oxford: Oxford University Press.

    Killgore, W. D. S. (2000). Sex differences in identifying the facial affect of normal and mirror-reversed faces. Perceptual and Motor Skills, 91, 525530.

    Leary, M. R., & Kowalski, R. M. (1995). Social anxiety. New York: Guilford Press.Liu, C. H., Collin, C. A., & Chaudhuri, A. (2000). Does face recognition rely on encoding of

    3-D surface? Examining the role of shape-from-shading and shape-stereo. Perception, 29,729743.

    Lorenz, K. (1966). On aggression. New York: Harcourt, Brace, Jovanovich.Lyons, M. J., Campbell, R., Plante, A., Coleman, M., Kamachi, M., & Akamatsu, S. (2000). The

    Noh mask effects: Vertical viewpoint dependence of facial expression perception. Pro-ceedings of the Royal Society of London B, 267, 22392245.

    Macho, S., & Leder, H. (1998). Your eyes only? A test of interactive influence in the processingof facial features. Journal of Experimental Psychology: Human Perception and Perfor-mance, 24, 14861500.

    Mandell, L. M., & Shaw, D. L. (1973). Judging people in the newsunconsciously: Effect ofcamera angle and bodily activity. Journal of Broadcasting, 17, 353362.

  • 132

    JOURNAL OF NONVERBAL BEHAVIOR

    Mignault, A. (1999). Connectionist models of the perception of facial expressions of emotion.Unpublished Ph.D. Dissertation, Department of Psychology, McGill University, Montreal,Canada.

    Morris, D. (1977). Manwatching: A field guide to human behavior. New York: Harry N.Abrams, Inc.

    Olson, C. L. (1987). Statistics, making sense of data. Newton, MA: Allyn and Bacon, Inc.Otta, E., Lira, B. B. P., Delevati, N. M., Cesar, O. P., & Pires, C. S. G. (1994). The effect of

    smiling and head tilting on person perception. The Journal of Psychology, 128, 323331.Preuschoft, S., & van Hooff, J. A. R. A. M. (1995). Homologizing primate facial displays: A

    critical review of methods. Folia Primatologica (Basel), 65, 121137.Preuschoft, S., & van Schaik, C. P. (2000). Dominance and communication. In F. Aureli & F. B.

    M. de Waal (Eds.), Natural Conflict Resolution (pp. 77105). Los Angeles: University ofCalifornia Press.

    Prkachin, K. M., & Silverman, B. E. (2002). Hostility and facial expression in young men andwomen: Is social regulation more important than negative affect? Health Psychology, 21,3339.

    Redican, W. K. (1982). Evolutionary perspective on human facial display. In P. Ekman (Ed.),Emotion in the human face (2nd ed., pp. 212280). Cambridge, UK: Cambridge Univer-sity Press.

    Rosenberg, E. L., & Ekman, P. (1995). Conceptual and methodological issues in the judgmentof facial expressions of emotion. Motivation and Emotion, 19, 111138.

    Russell, J. A. (1993). Forced-choice response format in the study of facial expression. Motiva-tion and Emotion, 17, 4151.

    Russell, J.A., & Feldman Barrett, L. (1999). Core affect, prototypical emotional episodes, andother things called emotion: Dissecting the elephant. Journal of Personality and SocialPsychology, 76, 805819.

    Sarra, S., & Otta, E. (2001). Different types of smiles and laughter in preschool children.Psychological Reports, 89, 547558.

    Smith, C.A., & Scott, H.S. (1997). A componential approach to the meaning of facial expres-sions. In J. A. Russell & J. M. Fernandez-Dols (Eds.), The psychology of facial expression:Studies in emotion and social interaction, second series (pp. 229254). Cambridge, UK:Cambridge University Press.

    Tiedens, L. Z. (2000). Powerful emotions: The vicious cycle of social status position andemotions. In N. M. Ashkanasy & C. E. Haertel, (Eds.), Emotions in the workplace: Re-search, theory, and practice (pp. 7281). Westport, CT: Quorum Books.

    Tomkins, S. S., & McCarter, R. (1964). What and where are the primary affects? Some evi-dence for a theory. Perceptual and Motor Skills, 18, 119158.

    Turner, A., & Ortony, T. J. (1990). Whats basic about basic emotion? Psychological Review,97, 315331.

    van Hooff, J. A. R. A. M. (1972). A comparative approach to the phylogeny of laughter andsmiling. In R. A. Hinde (Ed.), Non-verbal communication (pp. 209242). New York: Ox-ford University Press.

    Wallbott, H. G. (1998). Bodily expression of emotion. European Journal of Social Psychology,28, 879896.

    Wiggers, M. (1982). Judgments of facial expressions of emotion predicted from facial behav-ior. Journal of Nonverbal Behavior, 7, 101116.

    Zebrowitz, L., Fellous, J.-M., Mignault, A., & Andreoletti, C. (in press). Explaining the pro-clivity for reading faces: Trait impressions predicted from connectionist modeling offacial metrics. Personality and Social Psychology Review.