MESSAGES CONVEYED BY VARIOUS FORMS OF SMILING

Perceptual and Motor Skills, 1996,82, 1111-1121. O Perceptual and Motor Skills 1996

READING A SMILING FACE: MESSAGES CONVEYED BY VARIOUS FORMS O F SMILING 'L' EMMA OTTA, FABLANA FOLLADOR E ABROSIO, RACHEL LENEBERG HOSHINO Institute of Psychology University of Sao Poulo, Brazil Summary.-This study investigated the effect of various forms of smiltng (closed smile, upper smile, or broad smile) on person perception. Brazilian undergraduates ( N = 3 3 0 ) judged a photograph of a male or female stimulus person in three age ranges (young, middle-aged, and old) and smiling or not. 7-point scales were used to measure respondents' perception of the stimulus persons on various attributes (attractiveness, happiness, extroversion, sympathy, kmdness, submission, ambition, and intelligence). We found that a smile enhanced attractiveness and kindness ratings independently of its form, whereas the influence of the various forms on ratings of happiness was additive. As the neutral face changed to a closed smile and the closed smile became a broad smile, target stimuli were attributed greater rated happiness. We also found a contribution of perceivers' gender to the judgements of extroversion and symparhy, indicating a slightly greater discrimination of facial expressions among women than among men.

Nonverbal behaviors such as facial expression, gaze, spatial behavior, posture, gestures, and touch provide basic social information. According to Argyle (1988), nonverbal messages can be even more important than verbal ones. Their greater accuracy in communicating emotion may be due to their largely unconscious nature. S m h g is a particularly powerful nonverbal cue that is universally understood as a gesture of friendlrness (Eibl-Eibesfeldt, 1989). A smiling person generally creates a more favourable impression than a nonsmlhg one. Lau (1982), in a study conducted with Chinese college students, reported that a s m h g person was seen as happier, nicer, more attractive, and also more intelhgent than a n o n s r n h g one. Reis, Wilson, Monestere, Bernstein, Clark, Seidl, Franco, Gioioso, Freeman, and Radoane (1990) stated that American college students attributed s m h g persons greater sincerity, sociabhty, and competence but less independence and mascuhity. One major h i t a t i o n of these studies is that they have not distinguished 'This research was conducted with the support of Fundacao de Amparo i Pesquisa do Estado $e SPo Paulo and of Conselho Nacionol de Desenvolvime~toCientifico e Tecnolbgico. Special thanks are due to Vera Silvia Raad Bussab for her helpful comments on an earlier version of the manuscript. The authors also acknowledge the assistance provided by Dalton Francisco de Andrade in analysing the data. Please address correspondence and requests for reprints to Emma Otta, Departamento de Psicologia Experimental, Instituto de Psicologia, Universidade de Sgo Paulo. Av. Prof. Mello Moraes, 1721-CEP 05508-900-SPo Paulo, Brasil.

1112

E. OTTA, ET AL

dkferent topographical forms of smiling. The importance of this distinction has been suggested by the research of human ethologists who have been interested in the communicative function of s m h g and studied smiling as facial displays which regulate social interactions. Different topographical forms have been identified in different contexts, although there was some overlap. Brannigan and Humphries (1972) distinguished three forms: (a) the closed smile in which the mouth corners are drawn up and out while the teeth remain covered, (b) the upper smile in which thk mouth corners are drawn up and out, and the upper lip is raised showing part of the upper teeth while the lower teeth remain covered, and (c) the broad smile in which both upper and lower teeth are exposed. Jones, Raag, and C o h s (1990) looked for morphological dfferences between social (towards mother) and nonsocial smiles (towards toys) in toddlers. Upper smiles were typically displayed to an attentive mother, whereas smiles to toys were equally divided among the three smiling categories. Moreover, mothers' responses varied qualitatively with the infant's smlling mouth positions. They were most k e l y to laugh in response to a broad smile, whereas they were most l~kelyto nod and to vocalize in response to an upper smile and just to raise their brows in response to a closed smile. Brannigan and Humphries (19721, Cheyne (19761, and Otta and Sarra (19901, observing, respectively, English, Canadian, and Brazhan preschool children, found that the upper smile was the commonest expression. It was m a d y seen during friendly interactions. The closed smile was seen mainly in nonsocial situations when a child was engaged in solitary play or observing others. Sometimes it was also observed in social situations when a child was dl at ease, for example, after being called to order (Otta, 1994). The broad smile was usually seen during fast chasing play and during rough-and-tumble play. Otta and Sarra (1990) found that the broad smile was positively correlated with laughter, whereas closed and upper smiles were negatively correlated. Otta, Lira, Delevati, Cesar, and Pires (1994) investigated the influence of different topographical forms of smiling on person perception. For some traits such as happiness the influence of the various forms of s m h g was additive, whereas for others a more complex picture emerged. Adding a smile made a female stimulus person look more intelligent but did not influence the evaluation of a male stimulus person. The female stimulus person was perceived to be more intehgent with a closed or an upper smile than with a neutral face. No difference was found between the broad smile and the neutral face. The broad smile could be perceived as too expansive an expression, not appropriate for a female. One major h ~ t a t i o non the generalizability of the findings of Otta, Lira, Delevati, Cesar, and Pires (1994) is due to the fact that this study used

READING A SMILING FACE

1113

only one stimulus person of each sex. Further, both were young people. The purpose of the present study was to overcome these limitations. We intended to examine the effects of various forms of s m h g on the perception of various traits, using 30 stimulus persons, half of each sex in three age ranges of young, middle-aged, and old. Two hypotheses were contrasted. The various forms of smiling could just express different intensities of a single emotion. The dosed, upper, and broad smiles could express, respectively, low, medum, and high intensity of positive affect. Alternatively, the various forms of s m h g could express different motivations. The problem is to know whether smiling is one signal or a set of discrete, although overlapping, signals (Hmde, 1974; Lockard, Fahrenbuch, Smith, & Morgan, 1977; Hooff, 1972; Whissell, 1982). Finally, we sought to assess whether the stimulus persons' age and gender and the perceivers' gender alter the reading of a smiling face. Women have been found to smile more than men in a number of social contexts (Bugental, Love, & Gianetto, 1971; Deutsch, 1990; Frances, 1979; Halberstadt, Hayes, & Pike, 1988; Mackey, 1976; Morse, 1982; Otta, 1994; Ragan, 1982). Women may have been socialized to present themselves as friendher and more sociable in interpersonal situations. In contrast, men may have been taught an internalizing mode of emotional expression (Berman & Smith, 1984; Buck, Savin, Miller, & Caul, 1972; Buck, Miller, & Caul, 1974). This gender difference in self-presentation may influence what a smile communicates to perceivers. In a study in which also no dstinction was made among forms of smiling, Deutsch, LeBaron, and Fryer (1987) reported that the absence of smiles had a greater influence on perceptions of women than on perceptions of men. When not s m h g , women were perceived as less happy, carefree, and relaxed than were men. Nonsmihg women were rated less happy, warm, relaxed, and carefree than the average woman, whereas smiling men were rated more favourably on those traits than the average man.

Subjects A total of three hundred and thirty college undergraduates in Sao Paulo, Brazil (166 men and 164 women) voluntarily consented to participate in a study on first impressions. Ages ranged from 18 to 29 years.

Material Seventeen men and eighteen women in three age ranges, young (of the same age as the subjects), middle-aged, and old, served as stimulus persons. Four colour photographs were taken of each one directly facing the camera, against a neutral background, from the upper trunk to the top of the head. In the smlling condition, the stimulus persons were asked to smile as natu-

1114

E. OTTA, ET AL.

rally as possible, exhibiting a closed, upper, or broad smile. In the neutral condtion, they were requested to maintain a neutral pose. Forty-four undergraduates (24 women and 20 men) who ranged in age from 19 to 23 years were asked to class& the stimulus set of 140 photographs by facial expression. The judges were wholly accurate in identdying all the types of smiles exhibited by 30 of the 35 stimulus persons. The 120 photographs of these persons were selected to compose the stimulus set used in the experiment. The remaining photographs weie discarded for lack of agreement among the judges concerning some of the s m h g forms.

Procedure The design of the experiment can be described as a 2 x 2 x 3 x 4 between-subjects factorial. The factors were perceivers' gender, stimulus persons' gender, age (young, middle-aged, or old), and facial expression (no smile, closed smile, upper smile, or broad smile). Randomly, each subject was presented only one photograph and was asked to evaluate the person on a response sheet which contained the attributes of intehgence, kindness, attractiveness, happiness, extroversion, sympathy, submission, and ambition. To measure respondents' perceptions of the stimulus person 7-point scales were used. RESULTS A multivariate analysis of variance was used to explore possible contributions of stimulus persons' facia1 expression, age, and gender, and of perceivers' gender, and interactions among these on the dependent measures (the eight characteristics the perceivers inferred from stimulus persons). After obtaining significant multivariate effects, we interpreted the corresponding univariate effects. Post hoc t ratios were performed to analyze further significant main effects and mteractions, using the Bonferroni correction for an over-all level of .05.

Effects of Stimulus Persons' Facial Expression The multivariate analysis of variance yielded significance for stimulus persons' facial expression (Wdks lambda = .55; F,,,,,,= 7.70, p < ,001). Significant main effects were obtained in the corresponding univariate analyses for the following dependent measures: happiness (F,,,,=51.52, p < .001), attractiveness (F,,,,= 3.45, p < .05),and kindness (F,,2,,2= 8.39, p < ,001). Means and standard deviations are presented in Table I . Post hoc tests indicated that s m h g increased rated happiness, lundness, and attractiveness more than a neutral expression. Further, the various forms of smiling expressed different intensities of happiness but not of the other traits. Stimulus persons were ~erceivedto be happier with a broad

READING A SMILING FACE

TABLE 1 MWNSA N D STANDARD DEVIATIONS ON DEPENDENT MEASURES as A FUNCTION OF STIMULUS PERSONS' FACIAL EXPRESSION Measure

No Smile

M

Closed Smile

M

SD

SD

Broad Smile

Upper Smile

M

M

SD

SD

1.49 5.63' 1.21 Happiness 2.94V.37 4.29b 1 . 5 8 . 4.73b Attractiveness 2.749.43 3.24b 173 3.32b 1.83 3.5 l b 1.83 Kindness 4.403 1.64 5 . 2 0 ~ 1.49 5.35b 1.40 5 . 3 ~ ~1.32 Note.-Means for a given dimension that do not share a common superscript are si nificandy dkferent according to port hoc r tests using the Bonferroni correction for an over-a% level of .05.

smile than with an upper smile or with a closed smile, whereas no differences were found among the various forms of s r n h g on luridness and attractiveness.

EfJects of Stimulus Persons' Age The multivariate analysis of variance ~ieldedsignificance for the main effect of stimulus persons' age ( Wilks lambda = .77, F,,,,,, = 4.84, p < ,001). Corresponding significant main effects were obtained in the univariate analyses for the following dependent measures of happiness (F,,,, = 5.29, p < .01), extroversion (F,,,, = 10.64, p = .OO I ) , ambition ( F,,,, = 8.48, p < .001), and h d ness (F,,2,,2= 14.15, p < .001). Means and standard deviations are presented in Table 2. TABLE 2 MEANS A N D STANDARD DEVIATIONS FORDEPENDENT MEASURES AS A FUNCTION O F STIMULUS PERSONS' AGECATEGORY Measure

Young

M

Middle-aged

SD

M

SD

Old

M

SD

Happiness 4.69" 4.41nb 1.72 4.06~ 1.69 1.70 Extroversion 4.7a3 1.79 4.08~ 1.86 3.92b 1.78 Am bition 3.63" 3.01b 1.67 2.75b 1.48 1.77 lndness 4.64' 4.96a 1.47 5.61b 1.32 1.59 Note.-Means for a given dimension that do not share a common superscript are si nificantly diFferent according to post hoc t tests using the Bonferroni correction for an over-a% level of .05.

Post hoc tests showed that both old and middle-aged persons were considered less extroverted and ambitious than young persons. In addtion, old persons were considered kmder than both young and middle-aged persons. Finally, old persons were attributed less happiness than young persons. The middle-aged were differentiated from neither the young nor the old on this dimension. In addition to the main effect of stimulus persons' age, the multivariate

1116

E. OTTA. ET AL

analysis of variance yielded significance for the interaction of stimulus persons' age x gender (Wdks lambda = .88, F,,,, = 2.22, p < .01). Corresponding significant univariate effects were found for attractiveness (F,,,, = 9.36, p < .001) and intelhgence (F2,2a2 =3.26, p < .05). A marginally significant univariate effect was also found for submissiveness (F,,,, =2.87, p < .06). Means and standard deviations are presented in Table 3 . TABLE 3 A ~ I Z A C T I V EAN ND S SSUBMISSIVENESS ~ RATINGSOF YOUNG, A N D O L DFACES BY MALEAND FEMALE ]UDGES MIDDLE-AGED, Measure

Sex of Judge Male Female Male Female

-

Youne

M

Middle-aeed "

SD

M

SD

Old

M

SD

2.85" 1.80 3163 1.85 3.14' 1.57 4.17" 1.65 3 08" 1.54 2.6gb 1.59 Submissiveness 3.299.00 376' 1.69 3.54V.49 2.95' 1.72 3.46'.b 1.78 4.13~ 1.87 Note.-Means for a given dimension that do not share a common superscript are significantly different according to port hoc t tests using the Bonferroni correction for an over-all level of .05. Attractiveness

Post hoc tests indicated that both middle-aged and old women were rated as less attractive than young women, whereas age did not affect attractiveness ratings of men. Moreover, old women were seen as more submissive than young women. Age also did not affect the submissiveness ratings of men. The multivariate analysis of variance did not yield significance for the interaction of stimulus persons' facial expression x age (Wdks lambda= .83, F,,,,,,,= 1.12, p>.O5) or for the interaction of the stimulus persons' facial expression x age x gender (Wdks lambda= .94, F,,,,,,= 1.14, p>.O5). Effects of Perceivers' Gender The multivariate analysis of variance yielded a significant multivariate perceivers' effect for gender (Wilks lambda = .90, Fa,,, = 3.79, p < .OO 1). Follow-up univariate tests indicated that women gave higher ratings than did = 11.47, p < .001), attractiveness (F,,*,,= 6.66, p < .01), men on sympathy (F,,2,2 h d n e s s (F,,2a2= 10.84, p