7
 Facial movement varies by sex and is related to attractiveness Edward R. Morrison a, 4 , Lisa Gralewski  b , Neill Campbell  b , Ian S. Penton-Voak a a  Departmen t of Experiment al Psycholo gy, University of Bristol, Bristol BS8 1TU, UK  b  Departmen t of Compute r Science, University of Bristol, Bristol BS8 1TU, UK Initial receipt 9 October 2006; final revision received 5 January 2007 Abstract Facial movement has received little attention in studies of human attractiveness, yet dynamic displays are an important aspect of courtship in many species. This experiment investigated whether facial movement could be used to identify sex, and whether the ease of identification was associated with attractiveness. We removed shape cues to sex by applying movement from individual faces to a standardised facial model. Participants were able to distinguish between male and female animations of this model at levels above chance. Furthermore, there was a positive association between ease of sex identification and attractiveness for female, but not male, faces. Analysis of facial movement suggested several behaviours that are more frequent in women than men (blinking, tilting, nodding, shaking, and amount of movement). Although some of these behaviours may be cues to sex identification, none alone was directly linked to attractiveness. Our findings suggest that feminine motion is attractive in female faces, but sexually recognisable movement has no clear influence on male attractiveness, in agreement with work on stati c faces employi ng compos ite images. D 2007 Elsevier Inc. All rights reserved.  Keywor ds:  Facial attractiveness; Dynamic stimuli; Sex differences; Mate choice 1. Introduction Biological ap  proaches to mate choice hold  that cues such as symmet ry (ll er &  Thornhill, 1998)  and sexual dimorp hism are attra ctive ( Anders son, 1994 ). In humans, studie s of facial attrac tivene ss have argued that averagene ss and symmet ry ma y si gnal deve lopment al stab ili ty  and res istance to disease (Thornhil l & Møller , 1997) and heteroz ygosi ty (Gangestand & Buss, 1993) and, hence, be consid ered attrac tive. Sexual dimo rphism is attrac tive in female faces, perhaps because it signals youth  and fertility, which are valuable traits in a potential mate ( Perrett et al., 1998). These structur al cues may signal useful information to prospective mates (Fink & Penton-Voak, 2002). Sexually dimorphic characteristics in male faces are not unequivo- cally attractive but can be in circumstances when putative heritable  quality Q  is paramount in importance, such  as in short-t erm as opp ose d to lon g-t erm att ract ive nes s ( Little, Jones, Penton-V oak, Bur t, & Perr ett , 2002).  Tw o recent reviews summarise this literature admirably (Gangestad & Scheyd, 2005; Rhodes 2006). De spit e the we al th of st udie s in the ar ea of fa ci al attractiveness, the general reliance on static stimuli in the research to date is a concern.  Rubenstein (2005)  reported a low corr ela tio n bet wee n the att rac tiv eness of ind ivi dua l faces pre sented in sta tic and dynami c condit ion s (r =.19 and  r =.21,  p N .18), a findin g that suggests that conclusion s from experiments with static faces should be treated with caution. A static face is, in many ways, a poor stimulus, since faces are always dynamic in real social interactions. Faci al moti on is known to convey cues about identi ty (Bassil i, 1978; Bruce & V alenti ne, 1988; Kni ght & Johnst on, 1997; Lander , Chr ist ie, & Bruce, 1999; Pik e, Kemp, Towell, & Phillips,  1997; Thornton & K ourtzi, 2002) and emoti onal expression (Bassili, 1978, 1979).  Rubenstein (2005)  showed that emot ion is an important cue to att ractiv eness in dyn ami c faces, suggestin g tha t stati c images may lack important social cues relevant to attrac- tivene ss. Indeed ,  Rig gio, Wid aman, Tucker, and Salinas (1991)  used rat ings of beh aviour from video sequences to argue that dynamic informati on was an important  compon ent of overall attract ivenes s indep endent of facial (static) beauty from photographs, although the study could not specify what aspects of movement were important in these judgements. 1090-5138/$ – see front matter  D 2007 Elsevier Inc. All rights reserved. doi:10.1016/j.evolhumbehav.2007.01.001 4  Corresponding author. Tel.: +44 0117 954 6621.  E-mail address:  [email protected] (E.R. Morrison). Evolution and Human Behavior 28 (2007) 186–192

Facial Movement Varies by Sex and is Related to Attractiveness

Embed Size (px)

DESCRIPTION

nonverbal

Citation preview

  • x a

    ib,y, Uni

    niver

    inal re

    attra

    ent

    by

    ale a

    ractiv

    n me

    n, no

    ogni

    Scheyd, 2005; Rhodes 2006).to argue that dynamic information was an important

    component of overall attractiveness independent of facial

    (static) beauty from photographs, although the study could

    Evolution and Human BehavKeywords: Facial attractiveness; Dynamic stimuli; Sex differences; Mate choice

    1. Introduction

    Biological approaches to mate choice hold that cues such

    as symmetry (Mller & Thornhill, 1998) and sexual

    dimorphism are attractive (Andersson, 1994). In humans,

    studies of facial attractiveness have argued that averageness

    and symmetry may signal developmental stability and

    resistance to disease (Thornhill & Mller, 1997) and

    heterozygosity (Gangestand & Buss, 1993) and, hence, be

    considered attractive. Sexual dimorphism is attractive in

    female faces, perhaps because it signals youth and fertility,

    which are valuable traits in a potential mate (Perrett et al.,

    1998). These structural cues may signal useful information

    to prospective mates (Fink & Penton-Voak, 2002). Sexually

    dimorphic characteristics in male faces are not unequivo-

    cally attractive but can be in circumstances when putative

    heritable bqualityQ is paramount in importance, such as inshort-term as opposed to long-term attractiveness (Little,

    Jones, Penton-Voak, Burt, & Perrett, 2002). Two recent

    reviews summarise this literature admirably (Gangestad &

    Despite the wealth of studies in the area of facial

    attractiveness, the general reliance on static stimuli in the

    research to date is a concern. Rubenstein (2005) reported a

    low correlation between the attractiveness of individual

    faces presented in static and dynamic conditions (r=.19

    and r=.21, pN .18), a finding that suggests that conclusionsfrom experiments with static faces should be treated with

    caution. A static face is, in many ways, a poor stimulus,

    since faces are always dynamic in real social interactions.

    Facial motion is known to convey cues about identity

    (Bassili, 1978; Bruce & Valentine, 1988; Knight &

    Johnston, 1997; Lander, Christie, & Bruce, 1999; Pike,

    Kemp, Towell, & Phillips, 1997; Thornton & Kourtzi, 2002)

    and emotional expression (Bassili, 1978, 1979). Rubenstein

    (2005) showed that emotion is an important cue to

    attractiveness in dynamic faces, suggesting that static

    images may lack important social cues relevant to attrac-

    tiveness. Indeed, Riggio, Widaman, Tucker, and Salinas

    (1991) used ratings of behaviour from video sequencesFacial movement varies by se

    Edward R. Morrisona,4, Lisa GralewskaDepartment of Experimental Psycholog

    bDepartment of Computer Science, U

    Initial receipt 9 October 2006; f

    Abstract

    Facial movement has received little attention in studies of human

    in many species. This experiment investigated whether facial movem

    was associated with attractiveness. We removed shape cues to sex

    model. Participants were able to distinguish between male and fem

    was a positive association between ease of sex identification and att

    suggested several behaviours that are more frequent in women tha

    Although some of these behaviours may be cues to sex identificatio

    that feminine motion is attractive in female faces, but sexually rec

    agreement with work on static faces employing composite images.

    D 2007 Elsevier Inc. All rights reserved.1090-5138/$ see front matter D 2007 Elsevier Inc. All rights reserved.

    doi:10.1016/j.evolhumbehav.2007.01.001

    E-mail address: [email protected] (E.R. Morrison).nd is related to attractiveness

    Neill Campbellb, Ian S. Penton-Voaka

    versity of Bristol, Bristol BS8 1TU, UK

    sity of Bristol, Bristol BS8 1TU, UK

    vision received 5 January 2007

    ctiveness, yet dynamic displays are an important aspect of courtship

    could be used to identify sex, and whether the ease of identification

    applying movement from individual faces to a standardised facial

    nimations of this model at levels above chance. Furthermore, there

    eness for female, but not male, faces. Analysis of facial movement

    n (blinking, tilting, nodding, shaking, and amount of movement).

    ne alone was directly linked to attractiveness. Our findings suggest

    sable movement has no clear influence on male attractiveness, in

    ior 28 (2007) 186192not specify what aspects of movement were important in4 Corresponding author. Tel.: +44 0117 954 6621.

    these judgements.

  • independent, allowing estimation of school-level effects.

    E.R. Morrison et al. / Evolution and Human Behavior 28 (2007) 186192 187The role of motion in facial attractiveness deserves more

    research, and a biological perspective may prove useful: the

    importance of dynamic behaviours in mate choice has been

    demonstrated in several species. For example, female fruit

    flies choose males on the basis of their ability to bdance,Qwhich signals their neuromuscular condition (Maynard-

    Smith, 1956). Likewise, male funnel-web spiders that sway

    their abdomens at high frequency are more successful at

    mating (Singer et al., 2000). Video editing and computer

    animation techniques have allowed experimental manipu-

    lation of such dynamic behaviours while keeping other cues

    constant. For example, Rowland (1995) edited video

    footage of male sticklebacks to manipulate the tempo of

    their courtship displays and found that females prefer

    sequences that were speeded up. Computer animation has

    been used to generate artificial sticklebacks in mate

    preference experiments (Kunzler & Bakker, 1998). In

    humans, motion capture methods have been used to animate

    a standard figure in order to study motion while controlling

    for variation in static cues. In one recent study, dances by

    symmetrical men were rated as most attractive, suggesting

    that dynamic cues can signal underlying quality indepen-

    dently of static cues, which were constant across all stimuli

    (Brown et al., 2005).

    Specifying characteristics of movement is problematic,

    but sex typicality in facial movement is a potentially

    identifiable trait relevant to mate choice: sex-typical traits

    are linked to mating success in many species (Andersson,

    1994). Some work has investigated sex differences in

    human movement. For example, a meta-analysis of studies

    using point-light walkers shows that people can recognise

    sex from dynamic cues, with accuracy around 70% (Pollick,

    Kay, Heim, & Stringer, 2005). However, there are clear

    anatomical differences between men and women that

    influence gait, especially the wider female pelvis. Anatom-

    ical differences on this scale are not present in the face, but,

    nonetheless, some evidence suggests that facial movement

    of men and women may differ: mens movement may be

    more asymmetrical (Alford & Alford, 1981; Alford, 1983),

    and womens, more animated (Hall, 1985). A few studies

    have tried to present dynamic faces to assess whether sex

    can be identified from motion alone. Berry (1991) reported

    that adults and children could identify sex from point-light

    displays of dynamic faces. However, she was unable to

    remove all structural cues to sex (indeed, adults could

    identify sex from the static faces, although not as well as in

    the dynamic condition). Hill and Johnston (2001) also

    explored the role of facial motion on sex identification by

    using motion-capture technology to animate an androgy-

    nous head with the movement from one of four men or four

    women. People could successfully identify sex in this

    experiment, and nonrigid motion (movement of the internal

    features, rather than larger scale head movement) was

    particularly useful for doing so. The same result was found

    in similar experiments with point-light animations (Hill,

    Jinno, & Johnston, 2003). In these studies, severalThis approach also allows explanatory variables to be added

    at either level (e.g., pupils socioeconomic status at level 1

    and whether the school was coeducational or mixed at

    level 2). In our case, we nested animations (level 1) within

    the identity of the actor from which they came (at level 2).

    This approach also allowed us to investigate whether any

    relationship between sexually identifiable motion and attrac-

    tiveness applied across different actors and across different

    sequences of movement within actors. Given that dynamic

    cues to attractiveness are likely to be ephemeral in nature,

    investigating between- and within-target effects is important.

    2. Methods

    2.1. Facial animations

    We obtained facial motion data from 1-min video

    sequences for each of five men and five women who were

    discussing the same topic (their upcoming Christmas

    holidays). The volunteers were all young Caucasian

    undergraduate students recruited at Stirling University.

    The sequences were labelled semiautomatically with 30

    landmark points per video frame (1500 frames in total;

    Gralewski, Campbell, Morrison, & Penton-Voak, 2006).

    The landmarks were clustered around the mouth and

    eyebrows since these areas are the main components of

    nonrigid facial motion (Gralewski et al., 2006) and are

    important for perception of identity (Hill & Johnston, 2001)

    and sex (Bruce & Valentine, 1988). These points were then

    used to produce a vector, which provides a shape descriptor

    of the face. The face data were scale-normalised to remove

    size differences due to varying camera depth.

    Since we were interested only in dynamic cues, facial

    motion was extracted from the scale-normalised data foranimations were generated from the same actor, meaning,

    that the ratings made on them are not strictly independent

    and could result in pseudoreplication if the analysis does not

    take this lack of independence into account.

    To our knowledge, no one has investigated whether facial

    movement varies in its attractiveness independently of static

    cues. In this experiment, we aimed to use a different method

    to replicate the experiment of Hill and Johnston (2001) by

    applying male or female movement to the same face model

    and also test whether recognisability of the sex of movement

    was attractive. We statistically controlled for the lack of

    independence in our data using hierarchical linear modelling,

    which is a form of regression that does this by explicitly

    modelling the variance shared by the units of analysis. In

    educational research, for example, pupils from the same

    school may have exam scores that are more similar than

    pupils from different schools because of the effects of the

    school on performance (Bryk & Raudenbush, 1992). A two-

    level hierarchical linear model nests pupils (the unit of

    analysis at level 1) within schools (at level 2) and, hence,

    controls for the fact that pupils exam scores are not all

  • Each animation was then split into six consecutive 10-s

    2.91). The model also revealed that the variation among2

    E.R. Morrison et al. / Evolution and Human Behavior 28 (2007) 186192188clips, resulting in a total of 60 clips. The study was

    approved by the ethics committee of the Department of

    Experimental Psychology, University of Bristol.

    We analysed the results using a two-level hierarchical

    linear model (using HLM 6.02). This model can be

    understood as linear regression looking at the effect of the

    sex of motion on the rated sex of each clip, controlling for

    the lack of independence between movement sequences

    from the same actor. For the judgments about sex, the

    outcome variable was the rated sex of each clip, which was

    analysed as a binomial variable with 108 trials (i.e., number

    of times each stimulus was rated male out of 108). The

    level 1 predictor was sex of the movement in each clip,

    which was uncentered as it was dummy-coded as 0 for

    female and 1 for male. Level 2 units were the identities of

    the 10 people whose movement was used, and there were no

    level 2 predictors. The level 2 intercept term was treated as a

    random effect (because the 10 actors represent only a

    sample of all people), whereas the slope term was a fixed

    effect (because male and female are the only categories into

    which the predictor can fall). This type of analysis was

    therefore a random coefficients model, in which the

    relationship of interest lay only at level 1. The presence of

    units at level 2 was merely to statistically control for the

    nonindependence of stimuli derived from the same actor.

    Consequently, the degrees of freedom are derived from the

    number of units of analysis (60), although they are not

    independent, because their shared variance has been taken

    into account.

    For the attractiveness ratings, the same random coef-

    ficients model was used, but now, attractiveness was the

    normally distributed continuous outcome variable. The

    level 1 predictor was the number of times each clip was

    rated male from the first experiment, treated as a random

    effect. This model allows testing of two different issues:

    whether those individuals whose sex is more identifiable

    seem more attractive and whether movement from a

    particular individual that is sexually identifiable is also

    attractive. Two models were runone for female faces and

    one for male. This model can be understood as linear

    regression looking at the effect of ease of sex identification

    on the attractiveness of each clip, controlling for the lack

    of independence.

    3. Sex identification

    3.1. Participants and procedures

    Sixty-eight female (mean age=22.6, S.D. 6.5 years,

    range=1853) female and 40 male (mean age=27.1,each individual and used to animate an androgynous line

    face, which was generated as the average of the 10 male and

    female faces in the videos, thus removing the structural cues

    to sexual dimorphism. This produced 10 1-min animations

    with motion specific to each individual (see Appendix A).level 2 units was significant [v (9)=156.6, pb .0001],meaning that clips from different actors varied in their

    mean ratings. This result suggests that analyses that fail to

    control for nonindependence of multiple stimuli derived

    from one target are confounded.

    4. Attractiveness

    4.1. Participants and procedures

    For this task, the same clips were blocked by sex, and

    participants were informed of the sex of the faces. Con-

    sequently, a different set of 28 women (mean age=20.1,

    S.D.=1.8 years, range=1828) and eight men (mean

    age=20.7, S.D.=3.0 years, range=1827; age of rater had

    no effect in any analyses) viewed the clips in random order

    within block. The reason for this design, as the first

    experiment showed, was that people were not very accurate

    at identifying sex (see Section 3.2). This design prevents

    possible misclassifications of sex influencing attractiveness

    judgements. If the criteria for attractiveness differ across

    male and female dynamic stimuli (as they do across male

    and female static stimuli), then misclassification of sex (e.g.,

    erroneously thinking a male clip were female) would lead to

    potentially uninterpretable attractiveness judgements. Be-

    cause participants were informed of sex, they could judge

    attractiveness in the context of rating opposite sex or same

    sex faces. Participants rated the faces for attractiveness on a

    scale of 1 (very unattractive) to 7 (very attractive). Interrater

    reliability for attractiveness ratings was high (Cronbachs

    a=0.84, n=108; for men a=0.68, n=8; for women a=0.86,S.D.=6.3 years, range=1840) participants took part in

    the sex identification experiment. Stimuli were presented in

    random order using Eprime experimental software. Partic-

    ipants were instructed to judge whether each animation was

    male or female after viewing it once. Participants responded

    using the keyboard and did not receive feedback. Partic-

    ipants were not told how the stimuli had been created until

    they had finished the experiment.

    3.2. Results

    Mean accuracy was 57% (S.D.=0.5%), with no differ-

    ence between male and female raters. Sex of target,

    however, did influence accuracy, which was significantly

    higher for male (one-sample t test vs. 50%: mean=61%,

    SD=0.5%, t108=11.1, pb .0001) than for female (one-sample t test vs. 50%: mean=54%, SD=0.5%, t108=4.3,

    pb .0001) clips (paired-samples t test for the two groups:mean difference=7%, t108=5.7, pb .0001). These effects didnot interact with participant sex (i.e., men were no better or

    worse than women at identifying male or female clips).

    Male and female stimuli differed in the number of times

    they were categorized as male (multilevel regression,

    c=0.60, t58=2.62; p=.012). Male stimuli were 1.8 times aslikely to be judged male as female stimuli (95% CI=1.15

  • n=28). Most participants were undergraduate students who

    received course credit for participation.

    4.2. Results

    In this analysis, we operationalized the sex typicality of a

    given stimulus as the number of times that the sex of the

    given clip had been correctly identified by the participants

    in the first task detailed above (i.e., a clip that was correctly

    identified in 100/108 trials is considered to be higher in sex

    typicality than a stimulus identified correctly on 70/108

    occasions). For female faces, sex typicality defined in this

    way predicted attractiveness (multilevel regression, c=1.02,t4=7.08, pb .0001). There was significant variation in theerror term for the level 2 intercept [v2(4)=15.9, p=.004],but not for the slope [v2(4)=3.7, pN .500], meaning thatdifferent actors varied in their mean attractiveness, but

    correct identification of sex was associated with attractive-

    ness in the same way across actors.

    For male faces, sex typicality had no effect on

    attractiveness ratings (c=0.31, t4=1.60, p=.184). Nei-ther source of level 2 variation was significant [intercept

    v2(4)=5.9, p=.204; slope v2(4)=6.2, p=.181]. Fig. 1 shows

    58

    E.R. Morrison et al. / Evolution and Human Behavior 28 (2007) 186192 189Fig. 1. The relation between attractiveness and correct sex recognition of

    movement for female (upper) and male (lower) faces. Data are z-standardised,

    and each symbol represents one actor.between level 2 units was significant for these two

    variables. In other words, those clips that blinked and

    moved more were more likely to be rated female. Thisthe scatter plot of attractiveness and rated sex typicality for

    each level 2 identity of both sexes.

    5. Movement analysis

    We further investigated which specific aspects of

    movement might differ between men and women and,

    hence, be potential cues to sex-typical motion and/or

    attractiveness. To simplify the task of defining facial

    movement (which, like all biological motion, has complex

    spatial and temporal properties), we concentrated on five

    aspects of movement that could be objectively measured:

    blinks, bshakes,Q bnods,Q btilts,Q and total movement. Blinkswere defined simply as number of blinks made during each

    sequence. bShakes,Q bnods,Q and total movement weredefined objectively by measuring the movement made by

    the centroid (the imaginary average vector position of all

    landmark points) in each animation. Graphs of these

    movements over time were smoothed (filtered) and the

    number of turning points counted. A shake was defined as a

    movement that exceeded a left and right threshold3.8% of

    the total range of horizontal movement across all individ-

    uals. Likewise, a nod was defined as movement that

    exceeded an upper and lower threshold8.0% of the total

    range of vertical movement. A tilt was defined as movement

    of the line between the nose and forehead landmarks that

    exceeded two angle thresholds (9.7% of the total range), and

    total movement was simply the distance moved by the

    centroid from frame to frame. These particular thresholds

    were used to strike a balance between every slight

    movement becoming a nod or tilt and allowing a reasonable

    and realistic number of such movements to be recorded. For

    example, the rate of nodding was 3.6 times every 10 s for

    women and 2.0 for men; the respective rates of tilting were

    2.6 and 1.4.

    Each of these behaviours was more common in

    women than men when entered as a normally distributed

    outcome variable into a hierarchical linear model as

    used in Section 3.2. Blinks (c=1.9, t58=25, p=.019),shakes (c=1.38, t58=2.4, p =.02), nods (c=1.7,t58=2.1, p=.039), tilts (c=0.11, t58=4.1, p=.003),and total movement (c=1.19, t58=2.9, p=.006) were allpredicted by sex (female was dummy-coded as 0 and male

    as 1, hence the negative coefficients). Variation among the

    level 2 units was significant for all five variables, again

    suggesting that simple linear regression would be inappro-

    priate. We tested the possibility that these cues could be

    used to identify sex by entering each variable as a predictor

    of the number of times each clip was classified as male by

    participants in the sex identification task. Two variables

    predicted number of times rated male: blinks (c=0.14,t58 =2.4, p = .04) and total movement (c=0.23,t =2.5, p=.03), whereas the others did not. Variation

  • interpretation of our findings could be that since the stimuli

    E.R. Morrison et al. / Evolution and Human Behavior 28 (2007) 186192190lacked variation in all cues except motion, attractiveness

    judgments are driven by sex identification as the only source

    of variation, rather than as an ecologically meaningful cue to

    attractiveness. The lack of such a relationship in male faces,

    however, does not support this interpretation.

    Our findings are strikingly similar to those from

    studies of static faces, in which femininity is attractive in

    female faces (Perrett, May, & Yoshikawa, 1994; Perrett

    et al., 1998), but the relationship between masculinity

    and attractiveness in male faces is less simple, at least with

    composite faces (Rhodes, 2006). Although Rubenstein

    (2005) argues that what is attractive in static faces is not

    necessarily attractive in dynamic faces, we suggest that at

    least one important principle is the same: sexual dimorphism

    in static faces and sex typicality in dynamic faces are

    attractive (for females).relationship held when sex was included in the model.

    Adding the numbers of nods, shakes, and tilts resulted in a

    composite variable, bfacial movements,Q which was close topredicting number of times rated male (c=0.04,t58=2.1, p=.06) but not attractiveness when controllingfor blinking and sex.

    Odds ratios indicate that the male clips blinked 87%

    (95% CI 7799%) as much as female clips and moved 79%

    (6497%) as much. None of the five aspects of movement

    predicted attractiveness ratings. When the same analyses

    were run separately for each sex, the only significant

    relationship was that total movement predicted number of

    times females were rated male (c=0.30, t58=2.9,p=.046), i.e., female clips that moved more were more

    likely to be identified as female.

    6. Discussion

    These data suggest that men and women differ in the

    way they move their faces, and that people are sensitive to

    these differences. Furthermore, movement that was more

    often identified as female was considered more attractive

    in female faces, but no such effect was evident for male

    faces. The task was by no means easy, however, as

    reflected in the fairly low overall accuracy, and informally,

    many participants reported finding the task difficult. Our

    value of 57% is quite close to the accuracies of 68% in

    Berry (1991) and about 60% in Hill and Johnston (2001),

    despite differences in the stimuli and presentation details.

    Male animations were more easily identified than female

    animations, although accuracy for both was above chance.

    Overall, there was a slight tendency for participants to rate

    a face as male (53%).

    Although the line-drawn stimuli here are evidently not

    naturalistic, raters seem able to rate them for attractiveness

    in a meaningful way. Interrater reliability in attractiveness

    judgements was as high as most studies of photographic

    stimuli (although see Honekopp, 2006, for a discussion of

    interrater reliability in attractiveness studies). A simpleIn studies with static faces, masculine faces are often

    perceived as signalling negative personality traits such as

    coldness and dishonesty (Penton-Voak et al., 2001). Should

    these personality attributions extend to dynamic properties of

    faces, it may, in part, explain why sex typicality is not

    attractive in our dynamic clips. Work with composite

    computer graphic images has shown that womens prefer-

    ences also vary over the menstrual cycle, being higher for

    masculine male faces during the fertile phase (Jones et al.,

    2005; Penton-Voak et al., 1999; Penton-Voak & Perrett,

    2000). Preferences also vary according to mating context,

    masculine faces being more attractive as short-term or extra-

    pair partners (Perrett et al., 1998), and according to individual

    traits such as self-perceived attractiveness (Little, Burt,

    Penton-Voak, & Perrett, 2001). Womens preferences may

    therefore represent a strategy to trade off the benefits of good

    genes signalled by a masculine face with the benefits of a

    prosocial partner signalled by a less masculine face. Although

    data about the role of masculine facial shapes in real faces

    sometimes conflict with findings from these computer

    graphic studies (Rhodes, 2006), work in other modalities

    and using other techniques indicates that masculinity is

    not unequivocally preferred by women in all situations

    (Gangestad, Simpson, Cousins, Garver-Apgar, & Christensen,

    2004; Roney, Hanson, Durante, & Maestripieri, 2006).

    Dynamic facial cues may be particularly important in

    investigating these tradeoffs because of the potential for

    more information to be conveyed. For example, facial

    motion is important in personality attribution as it improves

    accuracy over static cues alone (Borkenau & Liebler, 1992;

    Watson, 1989) and carries particular information about

    emotion (Bente, Feist, & Elder, 1996). Dynamic information

    may also signal important aspects of an individuals

    underlying biology, which may be relevant to attractiveness.

    Static facial masculinity is related to current levels of

    circulating testosterone in men (Penton-Voak & Chen,

    2004), while static facial femininity is related to current

    levels of circulating oestrogen in women (Law Smith et al.,

    2005). Body movement may signal symmetry and hormonal

    profiles in women (Brown et al., 2005; Grammer, Filova, &

    Fieder, 1997; Grammer, Fink, Mller, & Thornhill, 2003).

    Because people can change the way they move their faces,

    whereas static cues are invariant (at least over the short

    term), more temporally precise information can be signalled.

    For example, men and women might be expected to alter

    their facial movement in different social situations, such as

    when interacting with an attractive member of the opposite

    sex compared with a rival of the same sex. Our results

    demonstrate a relationship between attractiveness and sex

    identification both between (some individuals move more

    attractively than others across episodes) and within targets

    (the attractiveness of a given individual depends on transient

    dynamic cues) and support the possibility of these strategic

    mate choice behaviours.

    Our preliminary analysis of specific aspects of facial

    motion identified several sex differences. Blinking, nodding,

  • perhaps because a more animated face signals extraversion,

    ODonohue and Crouch (1996), for example, also showed

    to understanding facial perception gave way to configural

    E.R. Morrison et al. / Evolution and Human Behavior 28 (2007) 186192 191approaches (e.g., Tanaka & Farah, 1993). In many ways, the

    technique we applied here (based on counting of discrete

    behaviours) is more similar in concept to a feature-based

    approach than a configural approach. More sophisticated

    analysis of the spatial and temporal aspects of facial motion

    may be necessary to define the properties of movement that

    are associated with attractiveness ratings.

    Computer animation techniques allow experimental

    manipulation of the multiple signals in static and dynamic

    human facial cues independently. Similar techniques have

    proved fruitful in teasing apart the separate elements of

    animal signals in studies of species as diverse as spiders

    (Uetz & Roberts, 2002), sticklebacks (Kunzler & Bakker,

    2001), and Jacky dragons (Johnstone, 2006; Ord, Peters,

    Evans, & Taylor, 2002). The work reported here represents a

    first step towards understanding dynamic cues to attractive-

    ness in human faces.

    Acknowledgments

    The authors thank Helen Chang for collecting the raw

    video footage used in this study. EM is supported by an

    ESRC PhD studentship.

    Appendix A. Supplementary data

    Supplementary data associated with this article can

    be found in the online version at www.doi:10.1016/

    j.evolhumbehav.2007.01.001.

    References

    Alford, R. D. (1983). Sex differences in lateral facial facility:

    The effects of habitual emotional concealment. Neuropsychologia, 21,

    567570.

    Alford, R. D., & Alford, K. F. (1981). Sex differences in asymmetry in the

    facial expression of emotion. Neuropsychologia, 19, 605608.that women smile and gesture more than men when speaking

    to a romantic partner. None of our objectively defined

    behaviours predicted attractiveness in either sex. This is

    perhaps not surprising: our understanding of static face

    processing advanced rapidly when feature-based approacheswhich is an attractive personality characteristic (Fink, Neave,

    Manning, & Grammer, 2005). These differences in facial

    movement echo earlier findings in both body movement and

    facial expression. Frable (1987) found that women move

    their bodies more than men in expressive movement.shaking, and tilting were all more frequent in women than in

    men, and the total amount of rigid movement was higher for

    female clips. Some of these cues could be used to identify

    sex in the absence of structural cues, although only blinking

    and total movement predicted the actual ratings made. Only

    total movement was related to attractiveness in female faces,Andersson, M. (1994). Sexual selection . Princeton, NJ7 PrincetonUniv. Press.

    Bassili, J. N. (1978). Facial motion in the perception of faces and of

    emotional expression. Journal of Experimental Psychology: Human

    Perception and Performance, 4, 373379.

    Bassili, J. N. (1979). Emotion recognition: The role of facial movement and

    the relative importance of the upper and lower areas of the face. Journal

    of Personality and Social Psychology, 37, 249258.

    Bente, G., Feist, A., & Elder, S. (1996). Person perception effects of

    computer-simulated male and female head movement. Journal of

    Nonverbal Behavior, 20, 213228.

    Berry, D. S. (1991). Child and adult sensitivity to gender information in

    patterns of facial motion. Ecological Psychology, 3, 349366.

    Borkenau, P., & Liebler, A. (1992). Trait inferences: Sources of validity at

    zero acquaintance. Journal of Personality and Social Psychology, 62,

    645657.

    Brown, W. M., Cronk, L., Grochow, K., Jacobson, A., Liu, K., Popovic, Z.,

    & Trivers, R. (2005). Dance reveals symmetry especially in young men.

    Nature, 438, 11481150.

    Bruce, V., & Valentine, T. (1988). When a nods as good as a wink: The role

    of dynamic information in facial recognition. In: M. M. Gruneberg,

    Morrison P. E., &, R. N. Sykes (Eds.), Practical aspects of memory:

    Current research and issues (pp. 169174). Chichester, UK7 Wiley.Bryk, A., & Raudenbush, S. (1992). Hierarchical linear models. Newbury

    Park, CA7 Sage Publications.Fink, B., Neave, N., Manning, J.T., & Grammer, K. (2005). Facial

    symmetry and the dbig fiveT personality factors. Personality andIndividual Differences, 39, 523529.

    Fink, B., & Penton-Voak, I. (2002). Evolutionary psychology of

    facial attractiveness. Current Directions in Psychological Science, 11,

    154158.

    Frable, D. E. S. (1987). Sex-typed execution and perception of expressive

    movement. Journal of Personality and Social Psychology, 53, 3916.

    Gangestad, S. W., & Buss, D. M. (1993). Pathogen prevalence and human

    mate preferences. Ethology and Sociobiology, 2, 8996.

    Gangestad, S. W., & Scheyd, G. J. (2005). The evolution of human physical

    attractiveness. Annual Review of Anthropology, 34, 523548.

    Gangestad, S. W., Simpson, J. A., Cousins, A. J., Garver-Apgar, C. E., &

    Christensen, P. N. (2004). Womens preferences for male behavioral

    displays change across the menstrual cycle. Psychological Science, 15,

    203207.

    Gralewski, L., Campbell, N., Morrison, E., & Penton-Voak, I. (2006).

    Analysis of facial dynamics using a tensor framework. Journal of

    Multimedia, 1, 1021.

    Grammer, K., Filova, V., & Fieder, M. (1997). The communication paradox

    and possible solutions: Towards a radical empiricism. In: A. Schmitt, &

    K. Schafer (Eds.), New aspects of human ethology (pp. 91120). New

    York7 Plenum Press.Grammer, K., Fink, B., Mller, A. P., & Thornhill, R. (2003). Darwinian

    aesthetics: Sexual selection and the biology of beauty. Biology Review,

    78, 385407.

    Hill, H., & Johnston, A. (2001). Categorizing sex and identity from the

    biological motion of faces. Current Biology, 11, 880885.

    Hill, H., Jinno, Y., & Johnston, A. (2003). Comparing solid-body with

    point-light animations. Perception, 32, 561566.

    Hall, J. A. (1985). Nonverbal sex differences. Baltimore7 Johns HopkinsUniversity Press.

    Honekopp, J. (2006). Once more: Is beauty in the eye of the beholder?

    Relative contributions of private and shared taste to judgments of facial

    attractiveness. Journal of Experimental Psychology: Human Perception

    and Performance, 32, 199209.

    Johnstone, R. A. (1996). Multiple displays in animal communication:

    dBackup signalsT and dmultiple messagesT. Philosophical Transactionsof Royal Society of London: Series B, 351, 329338.

    Jones, B. C., Perrett, D. I., Little, A. C., Boothroyd, L., Cornwell, R. E.,

    Feinberg, D. E., Tiddeman, B. P., Whiten, S., Pitman, R. M., Hiller, S. G.,

    Burt, D. M., Stirrat, M. R., Law Smith, M. J., & Moore, F. R. (2005).

  • Menstrual cycle, pregnancy and oral contraceptive use alter attraction to

    apparent health in faces. Proceedings of the Royal Society of London,

    Series B, 272, 347354.

    Knight, B., & Johnston, A. (1997). The role of movement in face

    recognition. Visual Cognition, 4, 264273.

    Kunzler, R., & Bakker, T. C. M. (1998). Computer animations as a tool in

    the study of mating preferences. Behaviour, 135, 11371159.

    Kunzler, R., & Bakker, T. C. M. (2001). Female preferences for single and

    combined traits in computer animated stickleback males. Behavioral

    Ecology, 12, 681685.

    Lander, K., Christie, F., & Bruce, V. (1999). The role of movement in the

    recognition of famous faces. Memory Cognition, 27, 974985.

    Law Smith, M. J., Perrett, D. I., Jones, B. C., Cornwell, R. E.,

    Moore, F. R., Feinberg, D. R., Boothroyd, L. G., Durrani, S. J.,

    Stirrat, M. R., Whiten, S., Pitman, R. M., & Hillier, S. G. (2005).

    Facial appearance is a cue to oestrogen levels in women.

    Proceedings of the Royal Society of London, Series B, 273,

    135140.

    Penton-Voak, I. S., Perrett, D. I., Castles, D. L., Kobayashi, T., Burt, D. M.,

    Murray, L. K., & Minamisawa, R. (1999). Menstrual cycle alters face

    preference. Nature, 399, 741742.

    Perrett, D. I., Lee, K. J., Penton-Voak, I. S., Rowland, D. R., Yoshikawa, S.,

    Burt, D. M., Henzi, S. P., Castles, D. L., & Akamatsu, S. (1998). Effects

    of sexual dimorphism on facial attractiveness. Nature, 394, 884887

    (doi:10.1038/29772).

    Perrett, D. I., May, K. A., & Yoshikawa, S. (1994). Facial shape and

    judgements of female attractiveness. Nature, 368, 239242.

    Pike, G. E., Kemp, R. I., Towell, N. A., & Phillips, K. C. (1997).

    Recognizing moving faces: The relative contribution of motion and

    perspective view information. Visual Cognition, 4, 409437.

    Pollick, F. E., Kay, J. W., Heim, K., & Stringer, R. (2005). Gender

    recognition from point-light walkers. Journal of Experimental Psychol-

    ogy: Human Perception and Performance, 31, 12471265.

    Rhodes, G. (2006). The evolutionary psychology of facial beauty. Annual

    Review of Psychology, 57, 199226.

    Riggio, R. E., Widaman, K. F., Tucker, J. S., & Salinas, C. (1991). Beauty

    is more than skin deep: Components of attractiveness. Basic and

    E.R. Morrison et al. / Evolution and Human Behavior 28 (2007) 186192192perceived attractiveness influences human female preferences for sexual

    dimorphism and symmetry in male faces. Proceedings of the Royal

    Society of London, Series B, 268, 16.

    Little, A. C., Jones, B. C., Penton-Voak, I. S., Burt, D. M., & Perrett, D. I.

    (2002). Partnership status and the temporal context of relationships

    influence human female preferences for sexual dimorphism in male

    face shape. Proceedings of the Royal Society of London, Series B

    10951100.

    Maynard Smith, J. (1956). Fertility, mating behaviour and sexual selection

    in Drosophila subobscura. Journal of Genetics, 54, 261279.

    Mller, A. P., & Thornhill, R. (1998). Bilateral symmetry and sexual

    selection: A meta-analysis. American Naturalist, 151, 174192.

    ODonohue, W., & Crouch, J. L. (1996). Marital therapy and gender-linked

    factors in communication. Journal of Marital and Family Therapy, 22,

    87101.

    Ord, T. J., Peters, R. A., Evans, C. S., & Taylor, A. J. (2002). Digital video

    playback and visual communication in lizards. Animal Behaviour, 63,

    879890.

    Penton-Voak, I. S., Jones, B. C., Little, A. C., Baker, S., Tiddeman, B.,

    Burt, D. M., & Perrett, D. I. (2001). Symmetry, sexual dimorphism in

    facial proportions and male facial attractiveness. Proceedings of the

    Royal Society of London, Series B, 268, 16171623.

    Penton-Voak, I. S., & Chen, J. Y. (2004). High salivary testosterone is

    linked to masculine male facial appearance in humans. Evolution and

    Human Behavior, 25, 229241.

    Penton-Voak, I. S., & Perrett, D. I. (2000). Female preference for male faces

    changes cyclically: Further evidence. Evolution and Human Behavior,

    21, 3948.Applied Social Psychology, 12, 423439.

    Roney, J. R., Hanson, K. N., Durante, K. M., & Maestripieri, D. (2006).

    Reading mens faces: Womens mate attractiveness judgments

    track mens testosterone and interest in infants. Proceedings of

    the Royal Society of London Series B, Biological sciences, 273,

    21692175.

    Rowland, W. J. (1995). Do female sticklebacks care about male courtship

    vigour? Manipulation of display tempo using video playback. Behav-

    iour, 132, 951961.

    Rubenstein, A. J. (2005). Variation in perceived attractiveness. Psycholog-

    ical Science, 16, 759762.

    Singer, F., Riechert, S. E., Xu, H., Morris, A. W., Becker, E., Hale, J. A., &

    Noureddine, M. A. (2000). Analysis of courtship success in the funnel-

    web spider Agenenopsis aperta. Behaviour, 137, 93117.

    Tanaka, J. W., & Farah, M. J. (1993). Parts and wholes in face recognition.

    Quarterly Journal of Experimental Psychology Human Experimental

    Psychology, 46, 225245.

    Thornton, I. M., & Kourtzi, Z. (2002). A matching advantage for dynamic

    faces. Perception, 31, 113132.

    Thornhill, R., & Mller, A. P. (1997). Developmental stability, disease and

    medicine. Biological Reviews of the Cambridge Philosophical Society,

    72, 497548.

    Uetz, G. W., & Roberts, J. A. (2002). Multisensory cues and multimodal

    communication in spiders: Insights from video/audio playback studies.

    Brain Behaviour and Evolution, 59, 222230.

    Watson, D. (1989). Strangers ratings of the five robust personality factors:

    Evidence of a surprising convergence with self-report. Journal of

    Personality and Social Psychology, 57, 120129.Little, A. C., Burt, D. M., Penton-Voak, I., & Perrett, D. I. (2001). Self

    Facial movement varies by sex and is related to attractivenessIntroductionMethodsFacial animations

    Sex identificationParticipants and proceduresResults

    AttractivenessParticipants and proceduresResults

    Movement analysisDiscussionAcknowledgmentsSupplementary dataReferences