VOL. 14, NO. 4, JULY 2003 Copyright © 2003 American Psychological Society
Cultural Differences in Facial Expressions of Emotion
Abigail A. Marsh, Hillary Anger Elfenbein, and Nalini Ambady
We report evidence for nonverbal “accents,” subtle differ-
ences in the appearance of facial expressions of emotion across cul-
tures. Participants viewed photographs of Japanese nationals and
Japanese Americans in which posers’ muscle movements were stan-
dardized to eliminate differences in expressions, cultural or otherwise.
Participants guessed the nationality of posers displaying emotional
expressions at above-chance levels, and with greater accuracy than
they judged the nationality of the same posers displaying neutral ex-
pressions. These ﬁndings indicate that facial expressions of emotion
can contain nonverbal accents that identify the expresser’s nationality
or culture. Cultural differences are intensiﬁed during the act of ex-
pressing emotion, rather than residing only in facial features or other
static elements of appearance. This evidence suggests that extreme po-
sitions regarding the universality of emotional expressions are incom-
Do facial expressions of emotion constitute a universal “language,”
or are there unique ways of expressing emotions within different cul-
tures? This question has become a running debate among psycholo-
gists (e.g., Ekman, 1972, 1997; Elfenbein & Ambady, 2002a, 2002b;
Matsumoto, 2002; Mesquita, Frijda, & Scherer, 1997; Russell, 1994).
Frequently, when two psychological theories compete, the answer lies
somewhere in the middle. Seminal studies on facial expression of
emotion have supported the universality hypothesis, demonstrating
that people of different cultural backgrounds can display similar ex-
pressions in response to similar stimuli (e.g., Camras, Oster, Campos,
Miyake, & Bradshaw, 1997; Ekman, 1972; Ekman & Rosenberg,
1997). However, although there may exist basic commonalities, it
would be reasonable to expect local variations in emotional expression
across cultures. Emotional expression may function as a universal lan-
guage, but one with regional accents. In the present research, we in-
vestigated the existence of nonverbal “accents,” or variations in facial
expressions across cultures.
In linguistics, the word
denotes the characteristic differ-
ences in pronunciation used by subsets of speakers of a single lan-
guage. People come to speak in the accent of speakers around them
(Baron-Cohen & Staunton, 1994; Munro, Derwing, & Flege, 1999).
This may occur not only because similarities in accents help to iden-
tify members of a group, but also because similar spoken accents im-
prove the understanding of verbal communication. Listeners have
more difﬁculty in understanding speakers using unfamiliar accents
(Munro & Derwing, 1995).
Research on the judgment of emotions across cultures suggests
that people also have more difﬁculty understanding nonverbal com-
munication of individuals from foreign cultures. Classic judgment
studies demonstrated that facial expressions of the basic emotions of
anger, disgust, fear, happiness, sadness, and surprise can be accurately
recognized across cultural groups, which implies the existence of a
universal schema for the appearance of basic emotions (Ekman, 1972,
1997; Ekman et al., 1987; Ekman, Sorensen, & Friesen, 1969; Izard,
1971). However, when viewing American stimuli, non-Americans
generally do not recognize emotions as accurately as Americans do.
For example, in Izard’s (1971) large-scale study, Americans correctly
identiﬁed the emotions in 83% of the facial photographs of Ameri-
cans, and Europeans scored between 75% and 83%, whereas Japanese
scored 65% and Africans only 50%.
Along these lines, a recent meta-analysis of cross-cultural emotion
recognition has provided support for an in-group advantage (Elfenbein
& Ambady, 2002a, 2002b), whereby people more accurately recog-
nize emotional expressions of members of their own cultural in-group
than those of members of a cultural out-group. This effect was consis-
tent across a range of emotions, experimental methods, and nonverbal
channels of communication, and was smaller for cultural groups who
were physically closer to one another or had greater cross-cultural ex-
posure. Interestingly, this meta-analysis found a signiﬁcant in-group
advantage in many studies with designs in which the expressers’ back-
ground was not obvious from the stimulus materials. In Izard’s (1971)
study, Caucasian groups from European nations identiﬁed Caucasian
Americans’ expressions less accurately than did Americans.
In such studies, for participants’ accuracy to differ across cultural
groups implies the existence of cultural variants in the emotional ex-
pressions themselves. Such variants could serve to mark expressers’
nationality. If emotional communication is a universal language, the
coding for the basic muscle movements of certain expressions may be
inherited, but ﬁner-grained elements of expressions may be further
shaped by the expressive styles of the people in one’s environment.
Thus, speciﬁc appearances of facial expressions may differ among
cultures, as is the case for other forms of nonverbal behavior (Halber-
stadt, 1985). The existence of nonverbal accents could render the fa-
cial expressions of persons in cultures outside one’s own less easily
interpretable. Just as there is an in-group advantage for understanding
people’s speech, cultural differences could lead to an in-group advan-
tage in emotion recognition.
In the current study, we aimed to test for nonverbal accents in emo-
tional expressions. The strongest test for cultural variations in expression
would be to use stimuli already pretested for “stimulus equivalence”
(Matsumoto, 2002) in terms of identical facial muscle movement
across members of different cultural groups. For this reason, we used
photographs of Japanese and Japanese Americans in Matsumoto and
Ekman’s (1988) Japanese and Caucasian Facial Expressions of Emo-
tion (JACFEE) and Japanese and Caucasian Neutral Faces (JACNeuF)
sets. These photographs were coded according to the Facial Affect
Coding System (FACS; Ekman & Friesen, 1978), a version of the
same systems used by Ekman (1972) and Camras et al. (1997) in their
studies providing evidence for the universality of facial expressions.
“The muscles innervated in the expressions corresponded to the uni-
versal signals of emotion,” Matsumoto (2002) wrote of these photo-
Address correspondence to Abigail Marsh, William James Hall, Depart-
ment of Psychology, Harvard University, 33 Kirkland St., Cambridge, MA
02138; e-mail: firstname.lastname@example.org.
VOL. 14, NO. 4, JULY 2003
graphs, and they “are equivalent in emotion signaling properties
across encoder cultures” (p. 238). Given that exacting facial coding
systems have found these photographs to be identical in all measur-
able respects, use of these stimuli constitutes a conservative test of the
hypothesis that there are subtle differences in expressive style across
Because the purpose of our study was to test whether the style of
facial expressions differs across cultures, we wanted to exclude alter-
native explanations for participants’ ability to identify the nationality
of the individuals in the photographs. A crucial consideration was that
a person might be able to identify shared nationality from the physiog-
nomy of facial features, apparel, or other features. Thus, it was impor-
tant to demonstrate that nonverbal accents arise in the expression of
emotion itself, rather than only in features of the static face. We used
JACFEE and JACNeuF photographs for this reason also, because they
control for lighting, background color, and clothing. Furthermore, the
sets of nonexpressive and expressive photographs include the same in-
dividuals, allowing us to control for differences in static facial features
across individual posers. Thus, it would provide strong evidence for
nonverbal accents if participants were able to discern the nationality of
posers in these emotional photographs. It would also provide evidence
against alternative hypotheses concerning permanent facial differ-
ences if participants were substantially less able to discern the nation-
ality of the same posers with neutral expressions. Such a ﬁnding
would allow us to conclude that cross-cultural differences in the ap-
pearance of facial expressions emerge in the act of expressing emo-
Seventy-nine adults (40 females and 39 males) identifying them-
selves as native to the United States or Canada were recruited to par-
ticipate either for course credit or for $5 in compensation. Sixty-one
(77.2%) identiﬁed themselves as Caucasian, 6 (7.6%) as Black or Af-
rican American, and 5 (6.3%) as Asian American; 7 (8.9%) did not
identify their ethnic background.
The JACFEE and JACNeuF sets provided all the stimulus photo-
graphs (Matsumoto & Ekman, 1988). We used photographs of 18 pos-
ers, 9 Japanese nationals (i.e., individuals of Japanese citizenship and
ancestry) and 9 Japanese Americans (i.e., American citizens of Japa-
nese ancestry). For each individual, we selected both an emotional ex-
pression (anger, disgust, fear, sadness, or surprise) and a neutral
expression, for a total of 36 photos.
Happy expressions were omitted,
as no Japanese nationals pose happiness in the JACFEE set. Each par-
ticipant judged the nationality of all 18 posers, viewing one of two sets
of the 18 posers’ photographs. The assignment of stimulus set to each
participant was random. Each set contained some of the posers dis-
playing emotional expressions and some displaying neutral expres-
sions; expressions were balanced across nationality, and photographs
were presented in a different random order for each participant. All
participants thus judged a mix of neutral and emotional expressions
displayed by the 18 Japanese-national and Japanese American posers.
Participants completed the study in a private, sound-attenuated
room, using the software program SuperLab™ on an IBM-compatible
desktop computer. Photographs appeared on the monitor in gray scale
289 pixels. As is commonly done in such judgment studies
(e.g., Matsumoto, 1993), we ﬁrst familiarized participants with the set
of photographs while asking them to make a different set of judg-
ments, unrelated to the nationality of the posers. During the familiar-
ization trials, participants judged six personality traits (efﬁcient,
independent, ignorant, open, likable, and rude) for the 18 faces. After
the familiarization trials, the instructions presented on the computer
informed participants that half of the faces they had seen were Japa-
nese and the other half were Japanese American. This information en-
sured that participants understood that each individual shown had an
equal chance of being Japanese or Japanese American, which encour-
aged similar base rates in responses and limited test bias. Participants
then viewed the 18 photographs again and judged the nationality of
each individual poser.
The analyses we report use Wagner’s (1993) unbiased hit rate, cal-
culated as arcsin[hit rate * (1
false alarm rate)]. Although we mini-
mized response biases by informing participants that half of the
photos depicted Japanese and half depicted Japanese Americans, the
unbiased hit rate corrects for any lingering response biases across par-
ticipants, such as bias in the level of false alarms. Wagner (1993) also
provided a procedure to calculate the accuracy level expected due to
chance guessing, which is analogous to calculating expected values
for a chi-square analysis. All scores reported refer to the degree of ac-
curacy above that expected due to chance. Thus, a score of 0 would
indicate the participant was unable to distinguish Japanese from Japa-
nese American faces.
Participants could identify the nationality of emotional expressions
at above-chance levels, single-sample
.51. Participants could also identify the nationality of neu-
tral expressions at above-chance levels, single-sample
.26. A comparison of the effect sizes for accu-
racy judging nationality from emotional and neutral expressions, how-
ever, indicated signiﬁcantly greater accuracy for photographs showing
.05. A 2 (expression type: emo-
2 (participant’s sex: male, female) repeated measures
analysis of variance (ANOVA) with stimulus set as a covariate simi-
larly showed a main effect for expression type. Nationality judgments
were signiﬁcantly more accurate for emotional expressions than neu-
.34. The main effect
of participants’ gender and the interaction of expression type and par-
ticipants’ gender were not signiﬁcant,
The results of a one-way repeated measures ANOVA by expression
(fear, disgust, sadness, surprise) indicated that accuracy in judgments
of nationality differed among the four expressions,
1. From the JACFEE set, we used Photographs 05, 21, 23, 30, 31, 45, 47,
53, and 56 (Japanese nationals) and Photographs 06, 22, 24, 29, 32, 46, 48, 54,
and 55 (Japanese Americans). From the JACNeuF set, we used Photographs
30, 31, 32, 33, 38, 49, 50, 54, and 56 (Japanese nationals) and Photographs 29,
34, 35, 36, 39, 43, 46, 48, and 51 (Japanese Americans).
Abigail A. Marsh, Hillary Anger Elfenbein, and Nalini Ambady
VOL. 14, NO. 4, JULY 2003
.27. Note that only two anger exemplars from the
JACFEE could be used, so that each subject saw only the neutral ex-
pressions or the anger expressions of the anger expressers, and thus
anger was excluded from these analyses. Accuracy in judging nation-
ality was highest for sad expressions (
by surprised expressions (
0.05), then fear expressions
0.05) and then disgust expressions (
0.04). Post hoc paired
tests showed that the nationality of individuals
expressing sadness was recognized more accurately than that of indi-
viduals expressing the three other emotions:
.01, for fear; and
for surprise (all two-tailed). No other differences among emotions
were signiﬁcant. The evidence suggests that expressive information
carries information about individuals’ nationality that is detectable to
naive observers. By contrast, observers are less able to determine the
nationality of the same individuals’ neutral expressions.
This study provides evidence for the existence of nonverbal ac-
cents—cultural variations in the appearance of basic facial expres-
sions of emotion. These ﬁndings suggest that extreme positions
regarding the universality of emotion are incomplete. That is, across
cultures, emotions may be expressed in a manner largely consistent
with universal prototypes, but there can still exist subtle cultural dif-
ferences in the appearance of these universal emotions. Participants
less accurately distinguished the cultural background of posers when
they judged neutral rather than emotional expressions of the same pos-
ers. This result indicates that emotional expressions carry information
about nationality or culture beyond the information conveyed by per-
manent differences between cultural groups in their members’ physi-
ognomy of facial features, apparel, or other static features. Differences
may exist in the extent to which various emotional expressions convey
evidence of cultural origins. However, because expresser and emotion
were confounded in the photo set we used, the results of this analysis
are difﬁcult to interpret.
It is possible that nonverbal accents differentially shape even the
appearance of static facial features of members of different cultures.
Zajonc, Adelmann, Murphy, and Niedenthal (1987) showed that the
physical appearances of people in prolonged social contact converge
over time. They explained this ﬁnding in terms of permanent changes
in the structure of facial musculature, as a result of people who are in
contact with one another tending to develop common habits in using
certain muscles. This phenomenon might similarly entail permanent
changes in the facial appearance of members of an ethnic or national
group, and would make it possible to identify members of a particular
national or ethnic group on the basis of their neutral facial expres-
Our test for nonverbal accents in this study was conservative and
stringent. Participants accurately identiﬁed posers’ cultural back-
grounds in emotional expressions expressly designed for stimulus
equivalence. That is, the photograph set we used was designed with
the explicit purpose of eliminating all possible cultural differences in
the appearance of emotional expressions. Using the FACS (Camras et
al., 1997; Ekman, 1972; Matsumoto, 2002), Matsumoto and Ekman
(1988) validated these emotional expressions by identifying muscle
movement patterns that corresponded to their prototypes for universal
emotions. Thus, the current ﬁndings suggest that cultural differences
in the appearance of facial expressions are so robust that they are
nearly impossible to eliminate. These cultural differences also seem
not to be explicitly detectable using the FACS. They likely consist of
very slight variants in the muscle movements, or in the relative inten-
sity of these movements. This assumption is supported by the varia-
tion in the accuracy of nationality judgments across emotional
expressions. The expressions are, by deﬁnition, composed of different
sets of muscle movements. It therefore would be expected that differ-
ent emotional expressions would convey cultural differences to differ-
ent extents. More detailed analysis of the speciﬁc components of
expression carrying the implicitly detectable nonverbal accents is
The current study demonstrated people’s sensitivity to subtle dif-
ferences in the appearance of facial expressions across cultures. It
would be worthwhile for further work to examine how people use this
information in naturalistic situations. Nonverbal accents may provide
a mechanism underlying well-known social phenomena such as attrac-
tion to that which is familiar (Byrne, 1997) and xenophobia, the fear
or dislike of foreigners or strangers (Hall, 1959; Warnecke, Masters, &
Kempter, 1992). Attractiveness, familiarity, and similarity to self may
be mutually reinforcing qualities (Moreland & Zajonc, 1982). If peo-
ple from one’s own culture look familiar when expressing emotion,
this may help one to form favorable attitudes about individuals with
similar cultural backgrounds. Our ﬁndings suggest that people can
judge cultural background through nonverbal accents, just as they can
judge the geographic backgrounds of people speaking a common lan-
guage—for example, a Texan versus a Scot—through verbal accents.
Baron-Cohen, S., & Staunton, R. (1994). Do children with autism acquire the phonology
of their peers? An examination of group identiﬁcation through the window of bilin-
Byrne, D. (1997). An overview (and underview) of research and theory within the attrac-
Journal of Social & Personal Relationships
Camras, L.A., Oster, H., Campos, J.J., Miyake, K., & Bradshaw, D. (1997). Japanese and
American infants’ responses to arm restraint. In P. Ekman & E.L. Rosenberg (Eds.),
What the face reveals: Basic and applied studies of spontaneous expression using
the Facial Action Coding System (FACS)
(pp. 289–301). New York: Oxford Univer-
Ekman, P. (1972). Universals and cultural differences in facial expressions of emotion. In
J. Cole (Ed.),
Nebraska symposium on motivation, 1971
(pp. 207–282). Lincoln:
University of Nebraska Press.
Ekman, P. (1997). What we have learned by measuring facial behavior. In P. Ekman &
E.L. Rosenberg (Eds.),
What the face reveals: Basic and applied studies of sponta-
neous expression using the Facial Action Coding System (FACS)
New York: Oxford University Press.
Ekman, P., & Friesen, W.V. (1978).
Facial Affect Coding System: A technique for the mea-
surement of facial movement
. Palo Alto, CA: Consulting Psychologists Press.
Ekman, P., Friesen, W.V., O’Sullivan, M., Chan, A., Diacoyanni-Tarlatzis, I., Heider, K.,
Krause, R., LeCompte, W.A., Pitcairn, T., Ricci-Bitti, P.E., Scherer, K., Tomita, M.,
& Tzavaras, A. (1987). Universals and cultural differences in the judgments of facial
expressions of emotion.
Journal of Personality and Social Psychology
Ekman, P., & Rosenberg, E.L. (Eds.). (1997).
What the face reveals: Basic and applied
studies of spontaneous expression using the Facial Action Coding System (FACS)
New York: Oxford University Press.
Ekman, P., Sorensen, E.R., & Friesen, W.V. (1969). Pancultural elements in facial displays
Elfenbein, H.A., & Ambady, N. (2002a). Is there an ingroup advantage in emotion recog-
Acknowledgments—This study was supported in part by two National
Science Foundation Graduate Fellowships to the ﬁrst and second authors, a
Rand grant, and a Presidential Early Career Award for Scientists and Engi-
neers (Grant BCS-9733706) to the third author. We thank Dana Gavrieli,
Ryan Jimenez, and Lauren Sheerr for their assistance in this research.
VOL. 14, NO. 4, JULY 2003
Elfenbein, H.A., & Ambady, N. (2002b). On the universality and cultural speciﬁcity of
emotion recognition: A meta-analysis.
Halberstadt, A.G. (1985). Race, socioeconomic status, and nonverbal behavior. In A.W.
Siegman & S. Feldstein (Eds.),
Multichannel integrations of nonverbal behavior
(pp. 227–266). Hillsdale, NJ: Erlbaum.
Hall, E.T. (1959).
The silent language
. Garden City, NY: Doubleday.
Izard, C.E. (1971).
The face of emotion
. New York: Appleton-Century-Crofts.
Matsumoto, D. (1993). Ethnic differences in affect intensity, emotion judgments, display
rule attitudes, and self-reported emotional expression in an American sample.
vation and Emotion
Matsumoto, D. (2002). Methodological requirements to test a possible ingroup advantage
in judging emotions across cultures: Comments on Elfenbein and Ambady and evi-
Matsumoto, D., & Ekman, P. (1988).
Japanese and Caucasian facial expressions of emo-
[Slides]. San Francisco, CA: San Francisco State University, Depart-
ment of Psychology, Intercultural and Emotion Research Laboratory.
Mesquita, B., Frijda, N.H., & Scherer, K.R. (1997). Culture and emotion. In J.W. Berry,
P.R. Dasen, T.S. Saraswathi, Y.H. Poortinga, & J. Pandey (Eds.),
Handbook of cross-
cultural psychology: Vol. 2. Basic processes and human development
(2nd ed., pp.
255–297). Boston: Allyn & Bacon.
Moreland, R.L., & Zajonc, R.B. (1982). Exposure effects in person perception: Familiarity,
similarity, and attraction.
Journal of Experimental Social Psychology
Munro, M.J., & Derwing, T.M. (1995). Foreign accent, comprehensibility, and intelligibil-
ity in the speech of second language learners.
Munro, M.J., Derwing, T.M., & Flege, J.E. (1999). Canadians in Alabama: A perceptual
study of dialect acquisition in adults.
Journal of Phonetics
Russell, J.A. (1994). Is there universal recognition of emotion from facial expression? A
review of the cross-cultural studies.
Wagner, H.L. (1993). On measuring performance in category judgment studies of nonver-
Journal of Nonverbal Behavior
Warnecke, A.M., Masters, R.D., & Kempter, G. (1992). The roots of nationalism: Nonver-
bal behavior and xenophobia.
Ethology & Sociobiology
Zajonc, R.B., Adelmann, P.K., Murphy, S.T., & Niedenthal, P.M. (1987). Convergence in
the physical appearance of spouses.
Motivation & Emotion