from individuals high or low on the self–report dimensions ...from individuals high or low on the...

34
PENTON-VOAK ET AL. PERSONALITY JUDGMENTS FROM FACIAL IMAGES PERSONALITY JUDGMENTS FROM NATURAL AND COMPOSITE FACIAL IMAGES: MORE EVIDENCE FOR A “KERNEL OF TRUTH” IN SOCIAL PERCEPTION Ian S. Penton–Voak University of Bristol Nicholas Pound Brunel University Anthony C. Little University of Liverpool David I. Perrett University of St Andrews In addition to signaling identity, sex, age, and emotional state, people fre- quently use facial characteristics as a basis for personality attributions. Typi- cally, there is a high degree of consensus in the attributions made to faces. Nevertheless, the extent to which such judgments are veridical is unclear and somewhat controversial. We have examined the relationship between self–re- port and perceived personality using both faces of individuals and computer

Upload: others

Post on 05-Apr-2020

1 views

Category:

Documents


0 download

TRANSCRIPT

Page 1: from individuals high or low on the self–report dimensions ...from individuals high or low on the self–report dimensions of agreeableness, extraversion, and, for male faces only,
Page 2: from individuals high or low on the self–report dimensions ...from individuals high or low on the self–report dimensions of agreeableness, extraversion, and, for male faces only,

from individuals high or low on the self–report dimensions of agreeableness,extraversion, and, for male faces only, emotional stability.

It would be easy to start this article with a quote from a novel or playrelating an individual’s facial features to his or her personality attrib-utes or character, since the belief that faces portray character is ubiq-uitous. This persistence of lay theories of physiognomy is hard tocredit, given the long–standing assumption that such attributionsare considered to be erroneous by “orthodox” science. Nonetheless,in 1974 Ligget reported that 90% of undergraduate students believedthat the face is a valid guide to character. By 2000, this figure was stillaround 75% (Hassin & Trope, 2000), and the importance of this beliefcan hardly be overstated. A recent article demonstrated that infer-ences of competence made from politicians’ faces predicted U.S. con-gressional election outcomes at levels far above chance (the most“competent” looking candidate won in nearly 70% of the 2004 senateraces; Todorov, Mandisodza, Goren, & Hall, 2005). Within psychol-ogy, it is generally thought that the trait inferences made from faces,although consistent, have very low validity. Here, however, webriefly review the growing literature suggesting that somewhat ac-curate perception of personality is possible from facial characteris-tics alone. In the two studies reported here, we assess the accuracy ofresponses made to individual faces and to computer compositefaces.

There is a considerable literature on the nature of the attributions wemake when encountering other people. Some of this work has investi-gated the inferences that are made to target faces associated with cer-tain behaviors, typically by pairing face stimuli with brief descriptionsof behavior (Todorov & Uleman, 2002, 2004). These studies suggestthat these spontaneous trait inferences become an integral part of therepresentation of a person, rather than mere associations caused bythe co–occurrence of faces and behaviors, as suggested by earlier work(e.g., Skowronski, Carlston, Mae, & Crawford, 1998). Given that socialperception functions to guide our own actions and predict the actionsof others, incorporating trait attributions into representations of indi-vidual conspecifics seems judicious.

Nevertheless, the willingness of people to make personality attri-butions in the absence of behavioral cues (i.e., from facial informa-tion alone) is perhaps somewhat surprising. We seem to make suchinferences about personality traits spontaneously and without re-

608 PENTON-VOAK ET AL.

Page 3: from individuals high or low on the self–report dimensions ...from individuals high or low on the self–report dimensions of agreeableness, extraversion, and, for male faces only,

flection or deliberation (Hassin & Trope, 2000; Todorov et al., 2005).In short, many aspects of person perception could be considered tobe “intuitive” in a two-system framework (Kahneman, 2003).

Social–cognitive approaches to person perception often stresseconomy in information processing. Perceptual cues may act as trig-gers for categorization of individuals, which may in turn lead to theactivation of stereotypes (Cloutier, Mason, & Macrae, 2005). In manycases, the cues to category membership (sex or age, for example) arethemselves accurate, even if the stereotypes then activated are not. Itis intriguing to consider the possibility that there may be perceptualcues that accurately inform judgments of personality. Should suchcues exist, a number of questions emerge in person perception, be-cause trait attribution ceases to be a generally inaccurate cognitiveshortcut and becomes instead a potentially adaptive perceptualability.

Physiognomy has a bad reputation in psychology, and this islargely well deserved: most “studies” in the area have been reso-lutely unscientific, leading physiognomists to be dismissed as char-latans (or worse). Studies carried out in the first half of the 20thcentury demonstrated that measurements of individual facial fea-tures seldom, if ever, correlate with psychological characteristics(e.g., Cleeton & Knight, 1924). Since the 1940s, although work in thisarea has been sporadic, it has generally yielded similar negative re-sults (e.g., Alley, 1988, for review). More recent work, however, hassuggested that there may in fact be some observable relationships be-tween physical appearance and personality that could have formedthe basis for the beliefs, if not the methods, of the physiognomists.One difference between more recent and earlier approaches to judg-mental accuracy is that later research concentrates less on fea-ture-based relationships (e.g., do long noses accurately indicate anyaspect of personality?) and more on configural properties of faces(Hassin & Trope, 2000). Configural approaches are more consistentwith current theories of face processing in general—cognitive repre-sentations of faces seem to be more “holistic” than representations ofother objects such as houses (Tanaka & Farah, 1993). Here the terms“configural” and “holistic” should be taken as referring to the diver-sity of possible cues to personality, including the relative positions ofthe different parts of the face. A configural or holistic approach doesnot ignore individual facial “features”; rather, it does not specify inadvance exactly what constitutes a feature. Previous feature–based

PERSONALITY JUDGMENTS FROM FACIAL IMAGES 609

Page 4: from individuals high or low on the self–report dimensions ...from individuals high or low on the self–report dimensions of agreeableness, extraversion, and, for male faces only,

approaches have considered particular structures as features (e.g.,the nose or the lips). However, a feature can be more broadly consid-ered as anything that is visual and potentially diagnostic. For exam-ple, the spaces between and arrangement of structures moretraditionally considered as “features” (e.g., the space between eyesand eyebrows) may be a cue to some aspects of perceivedpersonality.

Rather than correlating physical measurements with self–reportedpersonality traits, more recent studies concentrate on the relation-ship between strangers’ impressions of a person, configural proper-ties (such as the multiple facial cues that contribute to a “baby–faced”appearance), and the personality that that person self–reports(Zebrowitz, 1998).

STUDIES AT ZERO ACQUAINTANCE

Studies using zero acquaintance paradigms (in which participantsrate the personality of strangers) have found a surprising degree ofcorrelation between self–ratings and stranger ratings on personalitydimensions, often using five-factor models of personality (e.g., Nor-man, 1963). Many trait theorists agree that five trait dimensions (de-rived from factor analyses of ratings of trait adjective pairs) seem torepresent a reasonable compromise that is at least a partially accu-rate measure of people’s personalities (e.g., Barrett & Pietromonaco,1997; McCrae & Costa, 1987; Watson, 1989).

Of particular relevance to the work presented in this article, is pre-vious research employing a five–factor personality model that hasdemonstrated a surprising correlation between stranger ratings andself–ratings of an individual’s personality. Zero acquaintance stud-ies show that people are able to generate consensual impressions ofothers’ personalities. Not only do different judges generate similarresponses to a given target individual, but also these impressions aremore accurate than would be expected by chance (i.e., these impres-sions agree with the target’s self–ratings). Whilethese studies dem-onstrate that trait attribution on zero acquaintance can besurprisingly accurate, few have studied specifically the role of theface alone in these judgments.

In Passini and Norman’s 1966 experiment, small groups of under-graduates were put together for a 15-minute period and asked, with-out verbal interaction, to rate each other using scales corresponding

610 PENTON-VOAK ET AL.

Page 5: from individuals high or low on the self–report dimensions ...from individuals high or low on the self–report dimensions of agreeableness, extraversion, and, for male faces only,

to the “big five” personality factors. They found correlations greaterthan 0.3 between self and others’ ratings for “Extraversion,” “Consci-entiousness,” and “Culture/Openness to experience” factors. Corre-lations for “Agreeableness” (r = 0.15) and “Emotionalstability/Neuroticism” (r = 0.02) failed to reach significance (Passini& Norman, 1966). Clearly, however, the 15 minutes that the groupspent together provided opportunities for personality judgments tobe influenced by nonfacial characteristics, such as clothing, posture,and other nonverbal behavior (eye contact, smiling etc.).

Albright, Kenny, and Malloy (1988) replicated Passini and Nor-man’s findings for the traits “sociable” (extraversion) and “responsi-ble” (conscientiousness), and found that the physical attractivenessof subjects correlated positively with ratings of the traits “sociable,”“talkative,” and “good–natured,” suggesting a halo effect. In addi-tion, there was a strong positive association between raters’ opinionsof targets’ neatness of dress and ratings of conscientiousness, indi-cating that nonfacial cues could be employed to make judgments(Albright et al., 1988). Kenny, Horner, Kashy, and Chu (1992) repli-cated Albright et al.’s findings using videotapes of individuals ratherthan the small-group paradigm, indicating that consensus at zero ac-quaintance is not dependent on interactions between the individualsrating each other.

Passini and Norman’s findings were also replicated by Watson(1989), who used a small-group experiment, but allowed subjects tosay one sentence introducing himself or herself to the rest of thegroup. For individual judges, there was significant self–peer agree-ment for ratings of Extraversion and Conscientiousness, but not forother personality dimensions. By averaging ratings across all judgesthat had rated a target individual, Watson also determined that in-creasing the number of judges increased the size of the self–peer cor-relations for Extraversion (r = .43), Agreeableness (r = .31),Conscientiousness (r = .28), and Culture (r = 0.20, ns) but not Emo-tional Stability ratings (r = –0.04).

Individual differences have been found in both how accurate peo-ple are when judging others and in accuracy when being judged one-self. Ambady, Hallahan, and Rosenthal (1995) found thatindividuals who rate themselves highly on extraversion and agree-ableness are more accurately perceived by others. They also reportedthat women are more accurate judges of strangers’ personalities thanare men, consistent with other literature on nonverbal behavior.

PERSONALITY JUDGMENTS FROM FACIAL IMAGES 611

Page 6: from individuals high or low on the self–report dimensions ...from individuals high or low on the self–report dimensions of agreeableness, extraversion, and, for male faces only,

Although individual differences in zero acquaintance judgmentsdo exist, Albright et al. (1997) demonstrate the cross–cultural natureof consensus of personality attributions. Using a small-group para-digm with Chinese students in Beijing, they reported a pattern of re-sults similar to those found in Western studies. In a further study,American participants judged personality traits of Chinese studentsfrom photographs, and vice versa. They found that Chinese andAmerican participants showed much cross–cultural agreement intheir judgments of personality at zero acquaintance (Albright et al.,1997).

Borkenau and Liebler (1992) attempted to identify the sources ofvalidity for zero-acquaintance judgments using Brunswik’s (1956)lens model, which attempts to define which objective cues (such as asmile or thin lips) are employed (correctly or incorrectly) to perceivepersonality. Judges were presented with film with sound, silent film,a still photo, or just the audiotape of the target reading. The still-pho-tograph condition (most relevant to the studies reported in this arti-cle) shows that even with extremely limited information and notransient facial expression, there were significant correlations be-tween judges’ and self–ratings of Extraversion (r = 0.33) and Consci-entiousness (r = 0.32) while ratings of Agreeableness showed anonsignificant but positive relationship (r = 0.19). The application ofthe lens model was, however, somewhat unsuccessful—measuringdiscrete characteristics seems a less productive research strategythan concentrating on configural cues in this area (Zebrowitz &Collins, 1997).

Other studies essentially replicate the findings reported here (seeKenny, Albright, Malloy, & Kashy, 1994, for a review) suggestingthat, from limited information (even still photographs), judges tendto agree on the personality of other individuals. Furthermore, be-cause these ratings appear to correlate reliably with self–ratedextraversion and conscientiousness in all studies, and self–ratedagreeableness and openness–intellect in a subset of the literature,they are to some extent valid.

ECOLOGICAL PERSPECTIVES ON ACCURATEPERSONALITY PERCEPTION

Studies have also found some evidence for accuracy in social percep-tion using models of personality other than the “big five." Typically,

612 PENTON-VOAK ET AL.

Page 7: from individuals high or low on the self–report dimensions ...from individuals high or low on the self–report dimensions of agreeableness, extraversion, and, for male faces only,

these studies are based within a Gibsonian/ecological theoreticalframework (see Zebrowitz & Collins, 1997, for review). For example,Berry and Brownlow (1989) found that ratings of male facial“babyishness’ were positively correlated with the face owner’sself–reported approachability and warmth, but negatively related toself–reported aggression. For female faces, ”babyishness’ ratingswere associated with low self–reported levels of physical power andassertiveness. Bond, Berry, and Omar (1994) demonstrated that indi-viduals whose faces are rated as appearing “less honest” are morelikely to volunteer for experiments that involve them in deceivingothers than people who are judged to look more honest. Self–reportsof personality attributes are subject to social desirability influences,so it is interesting to note that facial appearance is related to behavioras well as self-reported traits.

Few studies can genuinely claim to be based on physiognomic cuesalone since in many experiments “thin slices” of behavior are pre-sented, providing a multitude of nonfacial cues. Even in studies ofstatic facial photographs, nonfacial cues such as clothing and hair-style are often visible. It is not difficult to see how a neat,well–groomed appearance could accurately signal conscientious-ness, for example, while fashionable hairstyles may be associatedwith extraversion. The current study aims to overcome some of thesepotential problems by employing computer graphics techniques toeliminate non–face-based cues to personality.

USING COMPUTER GRAPHICS TO STUDYPERSONALITY ATTRIBUTION

Computer graphic composites (averages) can be used to extract thedefining characteristics of a group while losing the characteristicsthat make each face look individual (Rowland & Perrett, 1995). Forexample, the average of several young males looks both young andmale, whereas the average of older females looks older and fe-male—yet neither image bears great resemblance to any of theconstituent faces.

Average faces are constructed by delineating feature points (e.g.,the outline of the eyes, the mouth) on each individual face in thegroup. From the delineation data of all the faces in the set, the aver-age position of each feature point on the face can be calculated. Com-puter graphics techniques allow the rendering of average color and

PERSONALITY JUDGMENTS FROM FACIAL IMAGES 613

Page 8: from individuals high or low on the self–report dimensions ...from individuals high or low on the self–report dimensions of agreeableness, extraversion, and, for male faces only,

texture information onto the average face shape, thus generatingphotorealistic composites. These techniques are related in concep-tion, if not in execution, to the photographic composite techniquesused by Galton toward the end of the 19th century (Galton, 1878).

Within an apparently homogeneous population, subgroups can beaveraged to generate visibly different composites. Such techniqueshave been used in studies that test biologically based theories of fa-cial attractiveness (e.g., Penton–Voak et al., 2001; Perrett, May, &Yoshikawa, 1994). Although the differences between subgroup com-posites are subtle, they reliably generate differing social perceptions.This key property of composites suggests that subgroups of individ-uals with particular self–reported personality traits (e.g., extraverts)can be averaged together to generate new faces that embody that per-sonality trait. The facial characteristics shared by individuals whopossess a particular trait should remain in the composite, leading toattributions of that trait (e.g., extraversion) to the composite face. Inthis article, we attempt to generate composite faces that embody thecharacteristics of people who self-report being at the extremes ofeach of the dimensions in the five–factor model of personality (i.e.,extravert and introvert composites; neurotic and stable composites;agreeable and disagreeable composites; conscientious and notconscientious composites and open and closed to experiencecomposites).

The major advantage of using composites generated using thiscomputer graphics technique is that individuating or idiosyncraticcharacteristics not linked to attributions of the personality trait inquestion should be averaged out and not be appreciable in the finalimage. Since the technique is empirical, it does not presuppose anysingle feature or configuration underlying attributions, but it doesassume that information related to personality will be selectivelymaintained by averaging. As such, the technique can potentially lo-cate either individual features that are characteristic of personalitytraits or multiple features that support configurational/holisticjudgments of a trait. Another benefit is that characteristics that arepotentially linked to personality attribution but are temporary (i.e.,hairstyle; facial jewelry; blemishes, not all of which are removablewith careful image editing) will not be present in the final compos-ites. As such, computer graphics techniques are a novel and poten-tially informative way to assess the accuracy of personality

614 PENTON-VOAK ET AL.

Page 9: from individuals high or low on the self–report dimensions ...from individuals high or low on the self–report dimensions of agreeableness, extraversion, and, for male faces only,

judgments based on facial characteristics. In study 1, we replicate theaccuracy found in earlier studies of personality judgment to individ-ual faces, using a large sample of nearly 300 faces. In study 2, we cre-ated composite faces from individuals who self-reported themselvesto be at the extremes of each of the big five dimensions, and assessedwhether raters could accurately we assess the personality of theindividuals who made up the composites.

STUDY 1: ASSESSING ACCURACY OFRESPONSES TO INDIVIDUAL FACES

METHODS

Volunteers (146 male and 148 female) were photographed (in neutralpose) and each completed a 40 item self–report personality question-naire (Botwin, Buss, & Shackelford, 1997). All participants were un-dergraduate students at the University of Stirling aged between 18and 22 years old. A confirmatory factor analysis (independently foreach sex, varimax rotation, constrained to five factors) was used togenerate scores on the big five personality dimensions. For femaleparticipants, the five factors explained 46.3% of the variance in re-sponses; agreeableness (8.8%), conscientiousness (9.7%),extraversion (10%), emotional stability (11.4%) and openness (6.4%).For male participants, the five factors accounted for 39.9% of vari-ance in the questionnaire responses: agreeableness (7.3%), conscien-tiousness (8.4%), extraversion (7.5%), emotional stability (9.6%) andopenness (7.1%)

Photographs of the individuals were then rated on each of the per-sonality dimensions (rating scale = 1–7 where 7 = highest) by groupsof 10 raters. The faces were presented sequentially on a computermonitor in random order, and each rater group rated all the individ-uals of one sex on one of the five personality dimensions. Raters werealso undergraduate students at the University of Stirling (63 femaleand 37 male; mean age 20.8 years) and were instructed to press thespace bar (rather than submit a rating) if they happened to be famil-iar with any of the presented faces. Reliability of these personalityratings was good (Cronbach’s α > 0.70 for all dimensions except maleconscientiousness [α = 0.63]; Table 1).

PERSONALITY JUDGMENTS FROM FACIAL IMAGES 615

Page 10: from individuals high or low on the self–report dimensions ...from individuals high or low on the self–report dimensions of agreeableness, extraversion, and, for male faces only,

RESULTS

For males, there were significant positive correlations between per-sonality ratings based on photographs and self–report scores forextraversion, emotional stability, and openness to experience. For fe-males, there was a significant positive correlation between ratingsand self–report scores for extraversion (Tables 2a & 2b). In addition,for both male and female faces, there were also many significant cor-relations between ratings on the five personality dimensions (Table3). The correlations are significantly positive for socially desirabletraits (high emotional stability, high extraversion, agreeableness,contentiousness, and openness to experience).

Since scores on some rated personality dimensions were not justcorrelated with scores on the congruent self–report dimensions,stepwise linear regressions were carried out to determine whichself–report personality dimensions were the best predictors of rat-ings by observers on each of the big five personality dimensions. Sep-arate analyses were carried out for male and female faces, and foreach of the big five rated personality dimensions. The results of thestepwise linear regressions indicated that for males self–reportscores on congruent dimensions were the best predictors of ratingsof extraversion, emotional stability, and openness. For females,self–reported extraversion was the best predictor of extraversionratings (Table 4).

DISCUSSION

The most robust evidence for accuracy in social perception comesfrom self–stranger agreement in ratings of extraversion across sev-

616 PENTON-VOAK ET AL.

TABLE 1. Reliability (Cronbach’s α) of ratings of 148 female faces and 146 male facesby groups of 10 raters

Female Male

Agreeableness 0.705 Agreeableness 0.747

Conscientiousness 0.775 Conscientiousness 0.625

Extraversion 0.888 Extraversion 0.736

Emotional Stability 0.782 Emotional Stability 0.753

Openness 0.736 Openness 0.735

Page 11: from individuals high or low on the self–report dimensions ...from individuals high or low on the self–report dimensions of agreeableness, extraversion, and, for male faces only,

eral studies and diverse personality instruments (see Zebrowitz &Collins, 1997, for a tabulated review of this literature). Our resultsreplicate these earlier findings, with significant, positive correlationsbetween other–rated and self–reported extraversion of both maleand female targets. Stepwise regression indicated that self–reportedextraversion better predicted extraversion ratings elicited by indi-vidual faces than did scores on any of the other self–report personal-ity dimensions. Previous literature indicates that accurateperception of other traits is less consistent across studies. We foundno other evidence of self–stranger correlations that were consistentacross both target sexes, but emotional stability and openness to ex-perience were also accurately perceived in male target faces. Accu-rate perception of an “openness/culture” dimension has beenreported in zero acquaintance studies (e.g., Passini & Norman, 1966)and for some adjective pairs in response to static facial images (e.g.,“polished–crude” in Borkenau & Liebler, 1992). Borkenau andLiebler also report accurate perception of emotional stability in male,but not female, targets, as we report here. As in other studies, the re-lationship between self–reported personality and stranger–ratertraits is stronger for male than for female target faces.

The observed sex differences in self–reported personality matchsex differences in social stereotypes. Women consider themselves tobe lower in emotional stability and higher in agreeableness than domen (Costa, Terracciano, & McCrae, 2001), so it is essential to analyzethe male and female targets separately. If raters consider womenmore agreeable than men, and targets report the same, significantbut somewhat misleading “accuracy” may be discovered(Paunonen, 1991).

While many studies have found somewhat accurate judgments oftargets’ extraversion, there are considerable inconsistencies acrossstudies in the accuracy of judgments of other traits. This may reflectthe use of different instruments to assess target personality. For ex-ample, Norman (1963) used 20-item 5–factor questionnaires,whereas Costa, Terracciano, and McCrae (2001) employed the60-item Neurotocism, Extraversion, and Openness (NEO) personal-ity inventory. Additionally, varied tasks have been used to investi-gate the raters" attributions of personality (e.g., one item trait scalesas used here and in, for example, Albright et al., 1988, versus the mul-tiple-item trait scales used in some other articles, e.g., Borkenau &Lieber, (1992). In this experiment, we had a relatively large number

PERSONALITY JUDGMENTS FROM FACIAL IMAGES 617

Page 12: from individuals high or low on the self–report dimensions ...from individuals high or low on the self–report dimensions of agreeableness, extraversion, and, for male faces only,

618 PENTON-VOAK ET AL.

TA

BL

E 2

a. C

ross

–cor

rela

tion

s be

twee

n se

lf–

and

oth

er–r

ated

dim

ensi

ons

for

ind

ivid

ual m

ale

face

s).

Sel

f–R

epor

t Per

son

alit

y

Rat

ed P

erso

nal

ity

Agr

eeab

len

ess

Con

scie

nti

ousn

ess

Ext

rave

rsio

nE

mot

ion

al S

tab

ilit

yO

pen

nes

s

Agr

eeab

lene

ss0.

007

–0.1

51–0

.003

–0.0

130.

044

Con

scie

ntio

usne

ss–0

.005

–0.0

050.

168*

–0.0

310.

017

Ext

rave

rsio

n0.

069

0.13

60.

237 *

*–0

.229

**0.

014

Em

otio

nal S

tabi

lity

0.02

9–0

.161

–0.0

080.

182 *

0.04

4

Ope

nnes

s0.

200

0.11

40.

072

0.04

80.

220 *

*

Not

e. C

ongr

uent

rat

ings

und

erlin

ed. *

p<

0.05

; **p

<0.

01.

TA

BL

E 2

b. C

ross

–cor

rela

tion

s be

twee

n se

lf–

and

oth

er–r

ated

dim

ensi

ons

for

ind

ivid

ual f

emal

e fa

ces

Sel

f–R

epor

t Per

son

alit

y

Rat

ed P

erso

nal

ity

Agr

eeab

len

ess

Con

scie

nti

ousn

ess

Ext

rave

rsio

nE

mot

ion

al S

tab

ilit

yO

pen

nes

s

Agr

eeab

lene

ss–0

.024

0.05

10.

127

–0.0

920.

076

Con

scie

ntio

usne

ss0.

076

0.11

50.

148

–0.0

87–0

.068

Ext

rave

rsio

n–0

.030

–0.0

340.

252 *

*–0

.004

0.03

5

Em

otio

nal S

tabi

lity

–0.0

11–0

.048

–0.0

780.

155

–0.0

07

Ope

nnes

s–0

.011

–0.0

120.

120

0.03

30.

132

Not

e. C

ongr

uent

rat

ings

und

erlin

ed. *

p<

0.05

; **p

<0.

01.

Page 13: from individuals high or low on the self–report dimensions ...from individuals high or low on the self–report dimensions of agreeableness, extraversion, and, for male faces only,

PERSONALITY JUDGMENTS FROM FACIAL IMAGES 619

TA

BL

E 3

. Cro

ss–c

orre

lati

ons

betw

een

rate

d p

erso

nalit

y tr

aits

for

ind

ivid

ual m

ale

and

fem

ale

face

s

Agr

eeab

len

ess

Con

scie

nti

ousn

ess

Ext

rave

rsio

nE

mot

ion

al S

tab

ilit

y

Mal

es

Con

scie

ntio

usne

ss0.

591*

**

Ext

rave

rsio

n0.

443*

**0.

430*

**

Em

otio

nal S

tabi

lity

0.50

8***

0.49

6***

0.62

6***

Ope

nnes

s0.

503*

**0.

588*

**0.

747*

**0.

602*

**

Fem

ales

Con

scie

ntio

usne

ss0.

397*

**

Ext

rave

rsio

n0.

391*

**0.

463*

**

Em

otio

nal S

tabi

lity

0.02

60.

094

0.31

2**

Ope

nnes

s0.

130

0.32

0***

0.33

0***

0.16

0

**p

<0.

001;

***

p<

0.00

01.

Page 14: from individuals high or low on the self–report dimensions ...from individuals high or low on the self–report dimensions of agreeableness, extraversion, and, for male faces only,

of targets (nearly 300 faces) rated on a relatively small number oftraits. Other studies have collected more extensive ratings on smallersamples of target faces (e.g., Borkenau & Lieber, 1992, used 100targets; Berry, 1990, employed 40 targets).

Given these differences in methodology, the overall agreement withprevious studies investigating accurate personality perception is ac-ceptable and indicates that our use of single-item trait scales is sufficientto generate findings consistent with earlier studies.

STUDY 2: ASSESSING THE ACCURACYOF RESPONSES TO COMPOSITE FACES

GENERATION OF COMPOSITE IMAGES

The individuals who completed the self–report personality ques-tionnaires were ranked, within sex, on each of the big five personal-ity factors dimension. For each sex, and for each of the five ratedpersonality dimensions, the 10% of individuals (15 faces) who scoredhighest and the 10% who scored lowest were selected and used tomake composite faces. These were made separately for each sex,yielding a total of 20 composites. To construct composites, 219 stan-dard feature points were marked on facial landmarks on each face.The mean × and Y coordinates of each delineated feature point werethen calculated to generate average shape information. “Average”color was generated by rendering color information from each indi-vidual into this average shape, and calculating mean RGB color val-ues across the face set for each pixel location. A wavelet–basedtechnique was then used to generate texture in each composite that is

620 PENTON-VOAK ET AL.

TABLE 4. Results of stepwise linear regressions to find best predictor of personalityratings for individual faces

Best Self–Report Predictor of Ratings

Rating Males Females

Agreeableness NS NS

Conscientiousness Extraversion NS

Extraversion Extraversion Extraversion

Emotional Stability Emotional Stability NS

Openness Openness NS

Note. NS = no significant predictor.

Page 15: from individuals high or low on the self–report dimensions ...from individuals high or low on the self–report dimensions of agreeableness, extraversion, and, for male faces only,

representative of the exemplars (see Tiddeman, Burt, & Perrett, 2001,for technical details of this procedure, and Figures 1 & 2 forexamples).

COMPOSITE RATING

The 20 composites (“high” and “low” for each trait, two sexes) wererated by 23 male and 19 female raters on representative adjectivesfrom the big five and for attractiveness. The adjective–pair anchorsfor the scales were Well–organized—Disorganized (conscientious-ness; loading = 0.74), Retiring-Sociable (extraversion; loading = 0.63),Suspicious-Trusting (agreeableness; mean loading across sexes =0.4), At Ease-Nervous (emotional stability/neuroticism; loading =0.72), Creative-Uncreative (openness to experience/culture; loading= .59), and Very Attractive-Very Unattractive.

RESULTS

Composites generated from individuals with socially desirable per-sonality characteristics were generally rated as more attractive thanthose derived from individuals with socially undesirable personalitycharacteristics. Specifically, composites derived from men whoself–reported high agreeableness, conscientiousness, extraversion,and emotional stability were rated as significantly more attractivethan composites derived from men who were low on each of theseself–report dimensions (Agreeableness: t41= 3.84, p < .001;Conscien-tiousness: t41=2.60, p<0.02; Extraversion: t41= 3.71, p < .001; EmotionalStability: t41= 4.16, p < .001; see Figure 3). Moreover, composites de-rived from women who self–reported high agreeableness,extraversion, and openness were rated as being significantly moreattractive than composites derived from women who were low oneach of these self–report dimensions (Agreeableness: t41

Page 16: from individuals high or low on the self–report dimensions ...from individuals high or low on the self–report dimensions of agreeableness, extraversion, and, for male faces only,

622 PENTON-VOAK ET AL.

FIG

UR

E 1

. Com

posi

tes

der

ived

from

the

10%

of m

en s

cori

ng h

ighe

st a

nd th

e 10

% s

cori

ng lo

wes

ton

eac

h of

the

big

five

sel

f-re

port

per

sona

lity

dim

ensi

ons

Page 17: from individuals high or low on the self–report dimensions ...from individuals high or low on the self–report dimensions of agreeableness, extraversion, and, for male faces only,

PERSONALITY JUDGMENTS FROM FACIAL IMAGES 623

FIG

UR

E 2

. Com

posi

tes

der

ived

from

the

10%

of w

omen

sco

ring

hig

hest

and

the

10%

sco

ring

low

est o

n ea

chof

the

big

five

sel

f-re

port

per

sona

lity

dim

ensi

ons.

Page 18: from individuals high or low on the self–report dimensions ...from individuals high or low on the self–report dimensions of agreeableness, extraversion, and, for male faces only,

tiousness, t41 = 4.2; p < .001 in all cases, Figure 5). Similarly, compos-ites generated from individual females who scored highest onagreeableness and extraversion were accordingly rated higher onthese dimensions than composites generated from females whoscored lowest on these dimensions (Agreeableness: t41 = 6.3, p<0.001;extraversion, t41 = 3.6, p<0.001; Figure 6).

So there is evidence of accurate perception of male agreeableness,extraversion, and emotional stability and female agreeableness andextraversion. However, the differences between composites highand low on each of these dimensions could be nonspecific. For exam-ple, although a composite derived from men who self–report highagreeableness may be rated as significantly more agreeable than onederived from men who self–report low agreeableness, it may also berated as appearing more extravert. Consequently, a series ofdiscriminant analyses were carried out to determine which ratings

624 PENTON-VOAK ET AL.

1.0

2.0

3.0

4.0

5.0

6.0

Agree Consc Extra Stab Open

Self-Report Personality Dimension

Att

ract

iven

ess

Rat

ing

Low High

*** **

**

FIGURE 3 Mean attractiveness ratings (± SEM on a 7-point scale, where 7 = most attractive)for composites derived from the 10% of men scoring highest and the 10% scoring lowest on

each of the big five self-report personality dimensions (*p < .05; **p < .01).Agree = Agreeableness; Cons = Conscientious; Extra = Extroversion;

Stab = Emotional Stability; Open = Openness to experience.

Page 19: from individuals high or low on the self–report dimensions ...from individuals high or low on the self–report dimensions of agreeableness, extraversion, and, for male faces only,

were best able to discriminate between composites that were high orlow on each dimension.

Discriminant analyses (SPSS v13.0) were carried out for ratings ofmale agreeableness, extraversion, and emotional stability compositepairs and ratings of female agreeableness and extraversion compos-ite pairs. For each composite pair (high vs. low), the analysis aimed todetermine which types of rating could best discriminate between sit-uations in which (1) raters were viewing a composite derived fromindividuals high on the dimension or (2) raters were viewing a com-posite derived from individuals low on the dimension. So, for exam-ple, for the occasions when a rater viewed a male compositegenerated from self–report agreeableness ratings, a discriminantanalysis was conducted to determine which rating dimensions(agreeableness, conscientiousness, extraversion, neuroticism, open-ness, or attractiveness) could best discriminate between situations in

PERSONALITY JUDGMENTS FROM FACIAL IMAGES 625

1.0

2.0

3.0

4.0

5.0

6.0

Agree Consc Extra E.Stab Open

Self-Report Personality Dimension

Att

ract

iven

ess

Rat

ing

Low High

*****

FIGURE 4. Mean attractiveness ratings (± SEM on a 7-point scale, where 7 = most attractive)for composites derived from the 10% of women scoring highest, and the 10% scoring lowest

on each of the big five self-report personality dimensions (*p < .05; **p < .01).Agree = Agreeableness; Cons = Conscientious; Extra = Extroversion;

Stab = Emotional Stability; Open = Openness to experience.

Page 20: from individuals high or low on the self–report dimensions ...from individuals high or low on the self–report dimensions of agreeableness, extraversion, and, for male faces only,

which (1) raters were viewing a high-agreeableness composite or (2)raters were viewing a low-agreeableness composite.

Each analysis procedure generates a discriminant function basedon the linear combination of the predictor variables that provides thebest discrimination in the two situations (the predictor variables be-ing ratings of agreeableness, conscientiousness, extraversion,neuroticism, openness, and attractiveness). For each predictor (rat-ing dimension), the analysis generates a standardized canonicaldiscriminant function coefficient that indicates the partial contribu-tion of the rating dimension to the discriminant function. This coeffi-cient can be used to assess each rating dimension’s uniquecontribution to the discriminant function. The larger the standard-ized canonical discriminant function coefficient, the better thepredictor is at discriminating between the two conditions of interest.

For the composite dimensions where there was evidence of accu-rate perception (i.e., male agreeableness, extraversion, and emo-

626 PENTON-VOAK ET AL.

1.0

2.0

3.0

4.0

5.0

6.0

Agree Consc Extra Stab Open

Self-Report Personality Dimension

Av

erag

eC

on

gru

ent

Rat

ing

Low High

**

**

**

FIGURE 5. Mean ratings (± SEM on a 7-point scale, where 7 = most socially desirable)on congruent dimensions of male composites made from individuals low (bottom 10%)

and high (top 10%) on self-report personality dimension.Ratings by 19 male and 23 female raters. (*p < .05; **p < .01).

Agree = Agreeableness; Cons = Conscientious; Extra = Extroversion;Stab = Emotional Stability; Open = Openness to experience.

Page 21: from individuals high or low on the self–report dimensions ...from individuals high or low on the self–report dimensions of agreeableness, extraversion, and, for male faces only,

tional stability and female agreeableness and extraversion), sociallydesirable and undesirable composites did differ in attractiveness(Figures 3 & 4). The discriminant analyses, however, indicated thateven with attractiveness ratings included as a predictor, ratings oncongruent personality dimensions were best able to discriminate be-tween situations in which a subject was viewing a composite high orlow on the dimensions being tested (Tables 5 & 6).

These results indicate that, for both male and female faces, ratingsof agreeableness are best able to discriminate between compositesderived from individuals who score high on self–report agreeable-ness and those that score low on that dimension. Moreover, a similarpattern exists for extraversion in males and females and for emo-tional stability in males, where again ratings on the congruent per-sonality dimension are best able to discriminate between compositesderived from individuals high and low on each self–report dimen-sion. Given the pervasive nature of halo effects, it is perhaps surpris-

PERSONALITY JUDGMENTS FROM FACIAL IMAGES 627

1.0

2.0

3.0

4.0

5.0

6.0

Agree Consc Extra Stab Open

Self-Report Personality Dimension

Av

erag

eC

on

gru

ent

Rat

ing Low High** **

FIGURE 6. Mean ratings (± SEM on a 7-point scale, where 7 = most socially desirable)on congruent dimensions of female composites made from individuals

low (bottom 10%) and high (top 10%) on self-report personality dimensionRatings by 19 male and 23 female raters. (*p < .05; **p < .01).

Agree = Agreeableness; Cons = Conscientious; Extra = Extroversion;Stab = Emotional Stability; Open = Openness to experience.

Page 22: from individuals high or low on the self–report dimensions ...from individuals high or low on the self–report dimensions of agreeableness, extraversion, and, for male faces only,

ing that ratings of attractiveness are fairly poor discriminators of the“high” and “low” composite images on these dimensions.

DISCUSSION

Participants’ judgments of composites constructed from individualfaces with certain self–reported personality traits are as accurate as, ifnot more accurate than, ratings of the individual faces that comprisethem. Raters accurately perceive the self–reported composites of in-dividuals who are at the extremes of the agreeableness, extraversion,and emotional stability dimensions. In study 2, more traits were ac-curately attributed to both male and female target composite faces,unlike the individual faces reported in Study 1.

One trait, “Openness to Experience” (assessed with the adjectivepair creative–uncreative), was perceived accurately in individualmale faces but not in composites of either sex. A plausible but un-tested explanation for this finding is that a cue derived from outsidethe face such as hairstyle (or clothing visible around the neck) influ-ences judgments of “creativity,” and these cues are removed duringthe making of the composites.

It is clear that our work and earlier work demonstrate that thefive–factor model may not be the best tool for assessing accuracy ofpersonality trait attribution to faces. Irrespective of accuracy, judg-ments of personality clearly do not resolve into five independent fac-tors, as indicated by the correlations between rated trait dimensions.We currently have little idea of how many dissociable “traits” or di-mensions individuals can perceive when attributing personalitycharacteristics to faces. When judging faces, raters tend to group de-sirable and less desirable personality traits into separate groups,probably as a result of halo effects of attractiveness (see Table 3). Al-though it is likely that there is some specificity in attributions to faces(faces that are rated high in extraversion are not identical to facesrated high in agreeableness, for example), there are also moderatecorrelations between ratings of socially desirable characteristics(e.g., faces rated high in agreeableness are also rated as being some-what high in extraversion, conscientiousness, emotional stability,and openness to experience; see Table 3). Socially desirable traits areascribed to attractive faces (Eagly, Ashmore, Makhijani, & Longo,1991) so some cross–talk between desirable traits is expected. Workinto the structure of personality, however, indicates that

628 PENTON-VOAK ET AL.

Page 23: from individuals high or low on the self–report dimensions ...from individuals high or low on the self–report dimensions of agreeableness, extraversion, and, for male faces only,

PERSONALITY JUDGMENTS FROM FACIAL IMAGES 629

TA

BL

E 5

.Wilk

’sL

ambd

a fo

rd

iscr

imin

antf

unct

ions

and

sta

ndar

diz

ed c

anon

ical

dis

crim

inan

tfun

ctio

n co

effi

cien

ts fo

r pe

rson

alit

y ra

ting

s of

mal

eco

mpo

site

s hi

gh a

nd lo

w o

n ag

reea

blen

ess,

extr

aver

sion

, and

em

otio

nal s

tabi

lity

Com

pos

ite

Dim

ensi

onW

ilk

’sL

amb

da

Cor

rect

lyC

lass

ifie

dR

atin

gD

imen

sion

Sta

nd

ard

ized

Can

onic

alD

iscr

imin

antF

un

ctio

n C

o -ef

fici

ents

Agr

eeab

lene

ss0.

733

p<

.005

71.1

%A

gree

able

ness

0.54

8

Con

scie

ntio

usne

ss0.

292

Ope

nnes

s0.

272

Em

otio

nal S

tabi

lity

0.22

6

Ext

rave

rsio

n0.

220

Att

ract

iven

ess

0.09

8

Ext

rave

rsio

n0.

705

p<

.001

71.1

%E

xtra

vers

ion

0.78

4

Agr

eeab

lene

ss0.

241

Em

otio

nal S

tabi

lity

0.23

0

Att

ract

iven

ess

0.10

0

Ope

nnes

s–0

.115

Con

scie

ntio

usne

ss–0

.174

Em

otio

nal S

tabi

lity

0.65

7p

<.0

001

78.9

%E

mot

iona

l Sta

bilit

y0.

640

Att

ract

iven

ess

0.34

7

Agr

eeab

lene

ss0.

217

Ext

rave

rsio

n0.

186

Ope

nnes

s0.

024

Con

scie

ntio

usne

ss–0

.005

Con

grue

nt r

atin

gs u

nder

lined

.

Page 24: from individuals high or low on the self–report dimensions ...from individuals high or low on the self–report dimensions of agreeableness, extraversion, and, for male faces only,

630 PENTON-VOAK ET AL.

TA

BL

E 6

.Wilk

’sL

ambd

a fo

rd

iscr

imin

antf

unct

ions

and

sta

ndar

diz

ed c

anon

ical

dis

crim

inan

tfun

ctio

n co

effi

cien

ts fo

r pe

rson

alit

y ra

ting

s of

fem

ale

com

posi

tes

high

and

low

on

agre

eabl

enes

s an

dex

trav

ersi

on.

Com

pos

ite

Dim

ensi

onW

ilk

’sL

amb

da

Cor

rect

lyC

lass

ifie

dR

atin

gD

imen

sion

Sta

nd

ard

ized

Can

onic

alD

iscr

imin

antF

un

ctio

nC

oeff

icie

nts

Agr

eeab

lene

ss0.

646

p<

0.00

0178

.9%

Agr

eeab

lene

ss0.

824

Con

scie

ntio

usne

ss0.

360

Ext

rave

rsio

n0.

203

Att

ract

iven

ess

0.17

3

Em

otio

nal S

tabi

lity

–0.1

23

Ope

nnes

s0.

115

Ext

rave

rsio

n0.

782

p<

0.01

72.4

%E

xtra

vers

ion

0.58

4

Agr

eeab

lene

ss0.

393

Att

ract

iven

ess

0.30

9

Ope

nnes

s0.

307

Em

otio

nal S

tabi

lity

–0.1

59

Con

scie

ntio

usne

ss–0

.060

Not

e. C

ongr

uent

rat

ings

und

erlin

ed.

Page 25: from individuals high or low on the self–report dimensions ...from individuals high or low on the self–report dimensions of agreeableness, extraversion, and, for male faces only,

self–reported personality traits are independent of each other (it ispossible to be both extravert and neurotic, for example).

Further research is needed to isolate how many dimensions of per-sonality people perceive in faces, how these dimensions differ be-tween sexes, and how they relate to the big five and/or other modelsof self–reported personality, and, finally, which of these dimensionsare perceived veridically.

Below, we consider three aspects of this work that merit furtherconsideration. First, we examine why composite faces should be per-ceived at least as accurately as, and possibly more so than, real facesof individuals. Second, we consider what facial characteristics mightmediate accurate social perception and what mechanisms could po-tentially link facial structure to self–reported personality. Finally, weexplore how these findings relate to other recent work on the socialperception of faces, such as recent developments in attractivenessresearch.

Why are composite faces perceived as accurately as, if not more so than,real faces?. The finding that composite faces appear to elicit personal-ity attributions as accurate (5/10 traits, across sexes) as individualfaces (4/10 traits, across sexes) is novel, and perhaps somewhat sur-prising. Indeed, this study shows more accurate perception of per-sonality than almost all of the existent literature (Zebrowitz &Collins, 1997). Although we predicted that composite faces shouldretain shared characteristics that are diagnostic of personality in in-dividual faces, we did not predict improved accuracy in thecomposite study.

There are several methodological differences between our studieswith individual and composites faces. While individual faces wereonly rated on a single trait by each participant, raters in the secondexperiment judged composite faces on all dimensions. This intro-duces the possibility that the trait judgments of the composite imagesmay be cross–contaminated. Given that such cross–contamination islikely to be in the form of halo effects, this methodological detailseems more likely to decrease, rather than increase, accuracy due tothe orthogonal nature of personality trait structure.

One possible explanation is that this apparent increase in accuracyis an artifact of examining only faces at the extremes of each trait di-mension distribution because the composite faces are derived fromthe highest scoring 10% and the lowest scoring 10% on each self–re-port personality dimension. The ratings of individual faces were

PERSONALITY JUDGMENTS FROM FACIAL IMAGES 631

Page 26: from individuals high or low on the self–report dimensions ...from individuals high or low on the self–report dimensions of agreeableness, extraversion, and, for male faces only,

made to nearly 300 targets, leaving the possibility that while theremight be accurate perception of faces at the extremes of each distri-bution, this could be “diluted” by the bulk of the individuals withless extreme personality types. Consequently, overall there might beno significant relationship between self–reported and rated person-ality. To test this possibility, we looked at the ratings given to each in-dividual face that went into each composite, and compared meanratings on the dimension of interest for the faces that formed eachcomposite image (Table 7). In general, faces that went into sociallydesirable composites were not rated higher on congruent personal-ity dimensions. There were, however, some significant differences.Individual females that went into the high self–report conscientious-ness composites were rated as appearing more conscientious thanthose who went into the low. Similarly, individual males who wentinto the high self–report extraversion and openness composites wererated as appearing significantly more extravert and open,respectively.

Computer composites may remove some of the possibly spuriouscues to personality (e.g., spots, wrinkles) present in individual faces,as well as potentially misleading cues such as hair styles and cloth-ing, thereby facilitating perception of relevant predictive cues. An-other property of composite faces is that they tend to be moreattractive than the individual faces that comprise them. The compos-ite images are, then, more similar in attractiveness than the individ-ual faces that went into them. This would have the effect of levelingthe influence of halo effects across the composites, perhaps facilitat-ing perception of any residual diagnostic cues to personality present.In this sense, composites could prove better for judgment accuracythan individual faces.

What facial characteristics mediate accurate social perception, and whatprocesses link facial structure to personality?. The facial characteristicsthat trigger judgments of personality are not entirely clear. The fail-ure of feature–based approaches has led to theories suggesting thatpersonality judgments depend instead on the configural propertiesof faces (Alley, 1988). Halo effects and overgeneralization effects (ofage, emotion or health; see Zebrowitz & Collins, 1997, for review)seem likely to be among the mediating processes linking faces to at-tributions. Stereotypical attributions based on sex–typicality (essen-tially, a sex overgeneralization effect) may also be involved, withdominant/less prosocial characteristics being attributed to more

632 PENTON-VOAK ET AL.

Page 27: from individuals high or low on the self–report dimensions ...from individuals high or low on the self–report dimensions of agreeableness, extraversion, and, for male faces only,

masculine faces. While such processes can explain consensus in per-sonality judgments, the mechanisms leading to the correspondencebetween facial characteristics and self–reported personality remainfairly speculative.

Several mechanisms have been proposed to explain the apparentassociations between configural aspects of faces and self–reportedpersonality (Zebrowitz & Collins, 1997). The first category of expla-nation is that there are direct links between biology and personality.For example, there is recent evidence that facial appearance indicatescurrent endocrine status, with circulating testosterone levels (as as-sayed in saliva) being correlated with both ratings of masculinity(Penton–Voak & Chen, 2004) and expressive behaviors such as smil-ing (Dabbs, 1997) in male faces. Such evidence suggests that accuratesocial perception of some characteristics could have a biological ba-sis—particularly dominance (thought to represent some aspects ofextraversion) and agreeableness.

High circulating testosterone levels in males are associated withmarital dissatisfaction and poor father–child relationships (Julian &McKenry, 1989), low scores on a “spousal investment” measure(Gray, Kahlenberg, Barrett, Lipson, & Ellison, 2002), and are relatedto antisocial behavior (Dabbs, 1993). Moreover, male circulating tes-tosterone levels correlate negatively with self–reported prosocialpersonality characteristics (Harris, Rushton, Hampson, & Jackson,1996). Consequently, if certain facial characteristics are predictive ofcirculating testosterone levels, then they could provide reliable in-formation about behavioral dispositions, regardless of the directionof the causal associations between behavior and testosterone. Ofcourse, testosterone–dependent sexual dimorphisms in facial struc-ture provide exactly this sort of information in that they allow peopleto discriminate faces according to sex—and males and females ex-hibit significant and reliable behavioral differences. What remains tobe determined is the extent to which intrasexual variation in sexuallydimorphic facial features is associated with intrasexual variation inpersonality characteristics. Chronic exposure to elevated levels ofother hormones, such as cortisol, growth hormone, and perhapsestrogen, could also lead to changes in facial appearance that arecorrelated with behavioral dispositions (Cushing syndrome beingan extreme example).

Another category of explanation is that stereotypical responses tofaces may shape the social environment, leading to self–fulfilling

PERSONALITY JUDGMENTS FROM FACIAL IMAGES 633

Page 28: from individuals high or low on the self–report dimensions ...from individuals high or low on the self–report dimensions of agreeableness, extraversion, and, for male faces only,

634 PENTON-VOAK ET AL.

TA

BL

E 7

. Mea

n pe

rson

alit

y ra

ting

s (o

n a

7–po

int s

cale

[com

ma

here

] whe

re 7

= m

ost s

ocia

lly d

esir

able

) on

cong

ruen

t dim

ensi

ons

for

ind

ivid

ual

face

s th

at w

ere

incl

uded

in s

ocia

lly u

ndes

irab

le (l

ow) a

nd d

esir

able

(hig

h) c

ompo

site

s fo

r ea

ch s

elf–

repo

rt p

erso

nalit

y d

imen

sion

.

Mal

esFe

mal

es

Low

Hig

hL

owH

igh

Mea

nSD

Mea

nSD

Mea

nSD

Mea

nSD

Agr

eeab

lene

ss3.

80(0

.65)

3.93

(0.7

5)3.

97(0

.70)

4.21

(0.5

6)

Con

scie

ntio

usne

ss3.

79(0

.85)

4.13

(0.9

0)3.

94(0

.54)

4.57

*(0

.93)

Ext

rave

rsio

n3.

35(0

.88)

4.55

*(0

.79)

3.91

(1.1

2)4.

41(0

.73)

Em

otio

nal S

tabi

lity

3.47

(0.8

5)4.

15(0

.96)

4.01

(1.1

5)4.

37(0

.67)

Ope

nnes

s3.

82(0

.73)

4.62

*(0

.66)

3.64

(0.6

9)4.

01(1

.00)

Not

e. S

igni

fica

nt d

iffe

renc

es (p

<.0

5) b

etw

een

rati

ngs

for

face

s in

hig

h an

d lo

w c

ompo

site

s ar

e m

arke

d w

ith

an a

ster

isk.

Page 29: from individuals high or low on the self–report dimensions ...from individuals high or low on the self–report dimensions of agreeableness, extraversion, and, for male faces only,

prophecy effects. For example, people with facial features that elicitattributions of agreeableness may be treated as more trustworthyand may perhaps consequently develop more agreeable personalitycharacteristics. Although this sort of explanation begs the questionas to the origin of these stereotypes (which appear to be largelycross–culturally stable; see Albright et al., 1997; Keating, Mazur, &Segall, 1981), there is nevertheless some evidence that such a processoccurs (Zebrowitz & Collins, 1997).

Self–fulfilling prophecy mechanisms could also plausibly work inreverse. That is, individuals with particular personality attributescould, over time, develop facial features that elicit accurate personal-ity attributions. Repeated and frequent adoption of particular facialexpressions could lead to the development of creases, folds, wrin-kles, or even changes in tonic muscle tone that produce a facial ap-pearance that elicits certain personality attributions. For example,individuals who are dispositionally happy may develop “laughterlines” that provide an accurate indicator of this aspect of their per-sonality. It is also possible that the environment (either physical orsocial) may lead to congruence between personality and appearance.That is, external influences may simultaneously lead to the develop-ment of certain personality attributes and a facial appearance that isstereotypically associated with those attributes.

Accurate personality judgments and facial attractiveness. Compositesderived from individuals who self–reported socially desirable per-sonality characteristics were generally rated as more attractive thanthose derived from individuals who self–reported socially undesir-able personality characteristics.

The discriminant analyses, however, revealed that attractivenessratings were generally not as effective as ratings on personality di-mensions at discriminating between these composites. This suggeststhat, for example, while high-agreeableness composites were moreattractive than low-agreeableness composites, they differed more re-liably in their perceived agreeableness. Consequently, agreeablenessratings were better able to discriminate between them.

The finding that composites derived from individuals who self–re-ported socially desirable personality characteristics were generallyrated as more attractive is, in itself, quite interesting because it sug-gests that individuals with attractive faces possess socially desirablepersonality traits. Consequently, it would seem that the “what isbeautiful is good” physical attractiveness stereotype has some valid-

PERSONALITY JUDGMENTS FROM FACIAL IMAGES 635

Page 30: from individuals high or low on the self–report dimensions ...from individuals high or low on the self–report dimensions of agreeableness, extraversion, and, for male faces only,

ity. Such a positive association might arise through self–fulfillingmechanisms, with attractive individuals developing socially desir-able personality attributes as a result of being treated morefavorably.

Recent work examining determinants of facial attractiveness sug-gests that while personality attribution is a critical factor in women’sjudgments of male attractiveness, it is apparently less important inmen’s judgments of female faces (Penton–Voak & Perrett, 2001;Perrett et al., 1998). Indeed, there is evidence that when women judgeattractiveness in male faces, they sometimes trade off putative cuesto biological “quality” (facial masculinity: dominance) against cuesto prosociality (facial femininity: warmth, agreeableness). For exam-ple, women prioritize cues indicative of biological quality whenjudging men’s attractiveness for short–term (sexual) relationships(Little, Jones, Penton–Voak, Burt, & Perrett, 2002) and when nearovulation and thus potentially fertile (Penton–Voak et al., 1999), butthey favor cues to prosociality (personality attributions made tofaces) for long–term relationships and in other phases of the men-strual cycle. Such shifts in preferences have been proposed to reflectthe functioning of evolved psychological mechanisms that increasedwomen’s reproductive success in ancestral environments. However,for such preferences to have been favored by selection, personalityattribution would have to be at least somewhat accurate. The evi-dence presented here suggests that, for individual faces, perceptionof male personality is more accurate than perception of female per-sonality, consistent with the importance of personality attributionsin women’s judgments of male attractiveness. Furthermore, thetraits perceived accurately in composites (especially agreeablenessand extraversion) seem to correspond well with the traits implicatedas important in attractiveness work (Perrett et al., 1998).

CONCLUSIONS

As outlined in the introduction, people believe strongly that it is pos-sible to make accurate judgments about an individual’s personalityfrom facial appearance alone. This is perhaps not surprising, becauseit has been demonstrated that people can make somewhat accuratejudgments after a fairly short period of interaction. Because facialphotographs alone are an impoverished stimulus, and seldom thesole basis for judgments in actual social interactions, our beliefs

636 PENTON-VOAK ET AL.

Page 31: from individuals high or low on the self–report dimensions ...from individuals high or low on the self–report dimensions of agreeableness, extraversion, and, for male faces only,

about accurate social perception may in fact reflect knowledge of ourability to make somewhat accurate judgments after fairly brief inter-actions. Our ability to combine cues from the face, verbal and non-verbal communication, and contextual information into accuratejudgments of individuals (Hassin & Trope, 2000) may lead us tooverestimate the accuracy of our judgments based on faces alone.Nonetheless, the data presented here indicate that social perceptionof faces may have a “kernel of truth,” and the cues that allow for ac-curate personality perception are visible in composite faces. Thecomputer graphic composite techniques that we use here have beenused extensively in attractiveness research. However, the use of suchtechniques here is novel, and to a certain extent exploratory. Find-ings of more consistent perception of personality across the targetsex when viewing composite rather than individual faces are intrigu-ing and suggest that these computer graphics techniques may, in thefuture, help provide valuable insights into mechanisms ofpersonality perception.

Our data indicate that the accuracy seen in personality perceptionis above chance, although still quite low. Even low levels of accuracy,however, may give individuals a slight edge in predicting how oth-ers will behave in social interactions over individuals who made nosuch judgments. Such a process may explain our readiness to attrib-ute personality on the basis of facial cues alone, and raises the possi-bility that we may underestimate the capabilities of human socialperception.

REFERENCES

Albright, L., Kenny, D. A., & Malloy, T. E. (1988). Consensus in personality judg-ments at zero acquaintance. Journal of Personality and Social Psychology,55(3), 387–395.

Albright, L., Malloy, T. E., Dong, Q., Kenny, D. A., Fang, X., Winquist, L., et al.(1997). Cross–cultural consensus in personality judgments. Journal of Per-sonality and Social Psychology, 72(3), 558–569.

Alley, T. R. (1988). Physiognomy and social perception. In T. R. Alley (Ed.), Socialand applied aspects of perceiving faces (pp. 167–186). Hillsdale, NJ: Erlbaum.

Ambady, N., Hallahan, M., & Rosenthal, R. (1995). On judging and being judgedaccurately in zero–acquaintance situations. Journal of Personality and SocialPsychology, 69(3), 518–529.

Barrett, L. F., & Pietromonaco, P. R. (1997). Accuracy of the five–factor model inpredicting perceptions of daily social interactions. Personality and SocialPsychology Bulletin, 23, 1173–1187.

PERSONALITY JUDGMENTS FROM FACIAL IMAGES 637

Page 32: from individuals high or low on the self–report dimensions ...from individuals high or low on the self–report dimensions of agreeableness, extraversion, and, for male faces only,

Berry, D.S. (1990). Taking people at face value: Evidence for the kernel of truthhypothesis. Social Cognition, 8, 343–361.

Berry, D. S., & Brownlow, S. (1989). Were the physiognomists right? Personalitycorrelates of facial babyishness. Personality and Social Psychology Bulletin,15, 266–279.

Bond, C. F., Berry, D. S., & Omar, A. (1994). The kernel of truth in judgments ofdeceptiveness. Basic and Applied Social Psychology, 15, 523–534.

Borkenau, P., & Liebler, A. (1992). Trait inferences: Sources of validity at zero ac-quaintance. Journal of Personality and Social Psychology, 62(4), 645–657.

Botwin, M. D., Buss, D. M., & Shackleford, T. K. (1997). Personality and matepreferences: Five factors in mate selection and marital satisfaction. Journalof Personality, 65(1), 107–136.

Brunswik, E. (1956). Perception and the representative design of Psychological experi-ments. Berkeley, CA: University of California Press.

Cleeton, G. U., & Knight, F. B. (1924). Validity of character judgment based on ex-ternal criteria. Journal of Applied Psychology, 8, 215–231.

Cloutier, J., Mason, M. F., & Macrae, C. N. (2005). The perceptual determinants ofperson construal: Reopening the social–cognitive toolbox. Journal of Per-sonality and Social Psychology, 88(6), 885–894.

Costa, P. T., Terracciano, A., & McCrae, R. R. (2001). Gender differences in per-sonality traits across cultures: Robust and surprising findings. Journal ofPersonality and Social Psvchology, 81(2), 322–331.

Dabbs, J. M. (1993). Salivary testosterone measurements in behavioral–studies.Annals of the New York Academy of Sciences, 694, 177–183.

Dabbs, J. M. (1997). Testosterone, smiling, and facial appearance. Journal of Non-verbal Behavior, 21(1), 45–55.

Eagly, A. H., Ashmore, R. D., Makhijani, M. G., & Longo, L. C. (1991). What isbeautiful is good, but . . . : A meta–analytic review of research on the physi-cal attractiveness stereotype. Psychological Bulletin, 110(1), 109–128.

Galton, F. J. (1878). Composite portraits. Anthropological Institute of Great Britainand Ireland, 8, 132–142.

Gray, P. B., Kahlenberg, S. M., Barrett, E. S., Lipson, S. F., & Ellison, P. T. (2002).Marriage and fatherhood are associated with lower testosterone in males.Evolution and Human Behavior, 23, 193–201.

Harris, J. A., Rushton, J. P., Hampson, E., & Jackson, D. N. (1996). Salivary testos-terone and self–report aggressive and pro–social personality characteris-tics in men and women. Aggressive Behavior, 22, 321–331.

Hassin, R., & Trope, Y. (2000). Facing faces: Studies on the cognitive aspects ofphysiognomy. Journal of Personality and Social Psychology, 78(5), 837–852.

Julian, T., & McKenry, P. C. (1989). Relationship of testosterone to men’s familyfunctioning at mid–life: A research note. Aggressive Behavior, 15, 281–289.

Kahneman, D. (2003). A perspective on judgment and choice—Mappingbounded rationality. American Psychologist, 58(9), 697–720.

Keating, C. F., Mazur, A., & Segall, M. H. (1981). A cross–cultural exploration of

638 PENTON-VOAK ET AL.

Page 33: from individuals high or low on the self–report dimensions ...from individuals high or low on the self–report dimensions of agreeableness, extraversion, and, for male faces only,

physiognomic traits of dominance and happiness. Ethology andSociobiology, 2, 41–48.

Kenny, D. A., Albright, L., Malloy, T. E., & Kashy, D. A. (1994). Consensus in in-terpersonal perception: Acquaintance and the big five. Psychological Bulle-tin, 116, 245–258.

Kenny, D. A., Horner, C., Kashy, D. A., & Chu, L. (1992). Consensus at zero ac-quaintance: Replication, behavioral cues and stability. Journal of Personal-ity and Social Psychology, 62, 88–97.

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 hu-man female preferences for sexual dimorphism in male face shape. Proc RSoc Lond B Biol Sci, 269(1496), 1095–1100.

McCrae, R. R., & Costa, P. T. (1987). Validation of the five factor model of person-ality across instruments and observers. Journal of Personality and Social Psy-chology, 52(1), 81–90.

Norman, W. T. (1963). Toward an adequate taxonomy of personality attributes:Replicated factor structure in peer nominated personality ratings. Journalof Abnormal and Social Psychology, 66, 574–583.

Passini, F. T., & Norman, W. T. (1966). A universal conception of personalitystructure? Journal of Personality and Social Psychology, 4, 44–49.

Paunonen, S. V. (1991). On the accuracy of ratings of personality by strangers.Journal of Personality and Social Psychology, 61(3), 471–477.

Penton–Voak, I. S., & Chen, J. Y. (2004). High salivary testosterone is linked tomasculine male facial appearance in humans. Evolution and Human Behav-ior, 25(4), 229–241.

Penton–Voak, I. S., Jones, B. C., Little, A. C., Baker, S., Tiddeman, B., Burt, D. M.,et al. (2001). Symmetry, sexual dimorphism in facial proportions, andmale facial attractiveness. Proceedings of the Royal Society, B., 268,1617–1623.

Penton–Voak, I. S., & Perrett, D. I. (2001). Male facial attractiveness: Perceivedpersonality and shifting female preferences for male traits across the men-strual cycle. Advances in the Study of Behavior, 30, 219–260.

Penton–Voak, I. S., Perrett, D., Castles, D., Burt, M., Koyabashi, T., & Murray, L.K. (1999). Menstrual cycle alters face preference. Nature, 399, 741–742.

Perrett, D. I., Lee, K. J., Penton–Voak, I., Rowland, D., Yoshikawa, S., Burt, D. M.,et al. (1998). Effects of sexual dimorphism on facial attractiveness. Nature,394, 884–887.

Perrett, D. I., May, K.A., & Yoshikawa, S. (1994). Facial shape and judgments offemale attractiveness. Nature, 368, 239–242.

Rowland, D. A., & Perrett, D.I. (1995). Manipulating facial appearance throughshape and color. IEEE Computer Graphics and Applications, 15(5), 70–76.

Skowronski, J. J., Carlston, D. E., Mae, L., & Crawford, M. T. (1998). Spontaneoustrait transference: Communicators take on the qualities they describe inothers. Journal of Personality and Social Psychology, 74(4), 837–848.

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

PERSONALITY JUDGMENTS FROM FACIAL IMAGES 639

Page 34: from individuals high or low on the self–report dimensions ...from individuals high or low on the self–report dimensions of agreeableness, extraversion, and, for male faces only,

terly Journal of Experimental Psychology Section A—Human Experimental Psy-chology, 46(2), 225–245.

Tiddeman, B. P., Burt, M., & Perrett, D. I. (2001). Prototyping and transformingfacial textures for perception research. IEEE Computer Graphics and Appli-cations, 21(5), 42–50.

Todorov, A., Mandisodza, A. N., Goren, A., & Hall, C. C. (2005). Inferences ofcompetence from faces predict election outcomes. Science, 308, 1623–1626.

Todorov, A., & Uleman, J. S. (2002). Spontaneous trait inferences are bound to ac-tors’ faces: Evidence from a false recognition paradigm. Journal of Personal-ity and Social Psychology, 83(5), 1051–1065.

Todorov, A., & Uleman, J. S. (2004). The person reference process in spontaneoustrait inferences. Journal of Personality and Social Psychology, 87(4), 482–493.

Watson, D. (1989). Strangers’ ratings of the five robust personality factors: Evi-dence of a surpising convergence with self–report. Journal of Personalityand Social Psychology, 57, 120–128.

Zebrowitz, L. A., & Collins, M. A. (1997). Accurate social perception at zero ac-quaintance: The affordances of a Gibsonian approach. Personality and So-cial Psychology Review, 1(3), 204–223.

Zebrowitz, L. (1998). Reading Faces: Window to the soul? Boulder, CO: Westview.

640 PENTON-VOAK ET AL.