You\'re only as pretty as you feel: Facial expression as a determinant of physical attractiveness

Journal of Personality and Social Psychology 1984, Vol. 46, No. 2, 469-478

Copyright 1984 by the American Psychological Association, Inc.

You're Only as Pretty as You Feel: Facial Expression as a Determinant of Physical Attractiveness Kim T. Mueser, Barry W. Grau, Steve Sussman, and Alexander J. Rosen University of Illinois at Chicago Photos of IS target persons posing happy, neutral, and sad facial expressions were rated for facial attractiveness using paired comparisons and Likert scales. Half the raters were instructed to compensate for the effects of facial expression. Paired comparisons and Likert ratings were highly correlated. Target persons were less attractive when posing sad expressions than when posing neutral or happy expressions, which did not differ. No effect of compensation instructions was found. In addition, independent ratings of four dimensions of the target persons' facial expression were obtained: pleasantness, surprise, intensity, and naturalness. Changes in these dimensions from the neutral to the happy and sad expressions and the corresponding changes in attractiveness were consistently related only to pleasantness supporting the reinforcement-affect theory of attraction. A second study related overall attractiveness to facial and bodily attractiveness. Both facial and bodily attractiveness were predictive of overall attractiveness, but the face was a slightly more powerful predictor. Results are discussed with respect to the stability of physical attractiveness, and alternative explanations of the mental-illness/physicalunattractiveness relation were proposed.

Physical attractiveness is both popularly and changeable components, such as facial expresscientifically accepted as one of the most salient sion, grooming, body position. The static features involved in interpersonal attraction components of physical attractiveness have and interaction (Berscheid & Walster, 1974; been addressed by some aestheticists (e.g., Finck, 1891). In general, physically attractive Schopenhauer—see Durant, 1928) and psyindividuals appear to benefit considerably in chologists studying aesthetic preferences (e.g., social relationships. For example, physically Iliffe, 1960; Perrin, 1921), as well as by inattractive individuals are assumed to possess vestigators of heterosexual preferences for difpositive character traits, are more likable ferent somatotypes and body types (Beck, (Miller, 1970; Stroebe, Insko, Thompson, & Ward-Hull, & McLear, 1976; Kleinke & StaLayton, 1971), and have been found to be more neski, 1980;Lavrakas, 1975; Wiggins, Wiggins, socially skillful than unattractive people (Dion & Conger, 1968). The changing components of physical attractiveness have not been ex& Stein, 1978; Goldman & 'Lewis, 1977). Despite the predictive power of physical at^ tensively studied, although some researchers tractiveness, little is known about its deter- have found that use of makeup, grooming, minants. Physical attractiveness may be viewed and clothes affect physical attractiveness (Baras being composed of static components (rel- ocas & Karoly, 1972; Sigall & Landy, 1973). atively enduring physical properties, such as One important changing component of physshape of nose or roundness of face) and ical attractiveness involves nonverbal communication. This study focuses on one means of nonverbal communication, the communiThis article is based on a thesis by the first author under the supervision of the fourth author submitted in partial cation of affect via facial expression and how fulfillment of the requirements for the .roasters degree at it influences physical attractiveness. The face is the primary site of affective the University of Illinois at Chicago. We wish to thank Herbert Stenson and Leland Wilkinson expression, and people are remarkably adept for their comments on this study and Shirley Glynn for at recognizing the emotional meaning of difher help in preparing the manuscript. Requests for reprints should be to Kim T. Mueser, Psy- ferent facial expressions (Ekman, 1971; Izard, chology Department, University of Illinois, Box 4348, 1971; Tomkins, 1962) even at distances as great Chicago, Illinois 60680. as 45 m (Hager & Ekman, 1979). Facial 469

470

MUESER, GRAU, SUSSMAN, ROSEN

expressions have been found to be cross-culturally similar (i.e., environmentally stable; Eibl-Eibesfeldt, 1972; Ekman, 1971; Izard, 1971) as well as present in blind people (EiblEibesfeldt, 1973), although the presentation of the face may differ between cultures (Argyle, 1969). Numerous studies have assessed the dimensionality of facial expression, and all have found that the pleasantness/unpleasantness dimension is prominent (Abelson & Sermat, 1962; Frijda & Philipzoon, 1963; Osgood, 1966; Schlosberg, 1952; Stringer, 1967). In addition to pleasantness/unpleasantness, several other dimensions of facial expression have been proposed, including intensity, control, and naturalness. The intensity of a facial expression includes components such as tension/sleep, active/passive, attention/rejection, and clear/dull (Frijda & Philipzoon, 1963; Osgood, 1966; Schlosberg, 1952). The control dimension reflects how surprised a person appears to be (Frijda & Philipzoon, 1963; Osgood, 1966; Stringer, 1967). Naturalness is a term used by Frijda and Philipzoon (1963) and Frijda (1969) to describe a group of intercorrelated scales: natural/artificial, mild/ derisive, and admiring/despising. The above review suggests that when a person is perceived, the pleasant/unpleasant affective messages sent via facial expressions will be important. Tomkins (1962) has suggested that one learns to respond with negative affect to negative affect on the face of another and with positive affect to positive facial expressions. If this is the case, then according to the Byrne-Clore (1970) reinforcement-affect theory of attraction, pleasant facial expressions in a target person should induce attraction toward the target and unpleasant expressions should cause repulsion from the person. In four previous experiments, a rater's affective state was manipulated before rating the interpersonal attractiveness of a stranger, and in each experiment changes in positive affect resulted in increased attractiveness ratings, and conversely, negative affective changes lowered attractiveness ratings (Bell, 1978; Gouaux, 1971; Gouaux, Lamberth, & Friedrich, 1972; Veitch, 1976). The relation between the facial communication of affect and attractiveness, however, has never been explored experimentally.

The goals of this study were to assess whether an individual's facial expression affects how facially attractive he or she appears to others and to determine which dimensions of facial expression are related to changes in attractiveness within the same target persons. In addition, in order to assess the generalizability of this study and other studies on facial attractiveness to the body of research on overall physical attractiveness, the contribution of facial attractiveness to overall attractiveness was estimated. The hypotheses are: 1. Target persons will be rated more physically attractive when posing a happy facial expression than a neutral one and more attractive with a neutral expression than with a sad one (Experiment 1). 2. In line with the reinforcement-affect theory of attractiveness, within-target-person differences in physical attractiveness between the neutral and happy and neutral and sad facial expressions are primarily related to changes in pleasantness (Experiment 1). 3. Facial attractiveness accounts for a sizable proportion of overall physical attractiveness when photos of people are rated in their everyday appearance (Experiment 2). Experiment 1 Method Stimulus materials. Fifteen white female target persons with a mean age of 20.27 years were recruited on the grounds of the University of Illinois, Chicago, campus to pose for five photos to be used as stimuli in the subsequent experiments. The photographer stationed himself directly outside the student union during the daytime and asked every woman who passed by and who appeared to be between the ages of 17 and 25 years old if she "would you like to earn $3 by allowing me to take five photos of you?" Subjects who expressed an interest were informed how the photos would be used, and if they agreed to participate, they signed a consent form. All target persons were students at University of Illinois, Chicago. Photos were taken outside in natural light, and no attempts were made to modify the target person's appearance. First, the target person was instructed to "relax all of your facial muscles and allow your facial expression to be as neutral as possible." Following this instruction, two photos were taken, one of the whole person and one from the neck down only. Next, three facial photos were taken. The order in which facial expression photos were taken was balanced for the sad and happy expressions. The photo of the neutral facial expression was always taken between the photos of the happy and sad expressions. For the sad expression, target persons were instructed to "close your eyes and think of someone whom you are very close to.

FACIAL EXPRESSION

Imagine that he or she has just died. Show how that feels with a sad facial expression and open your eyes." One photo was taken after the target person opened her eyes, For the neutral expression, the target person was instructed: "Now close your eyes again. Imagine a blank wall. Let your facial muscles relax and try to appear as neutral as possible. Now open your eyes." The neutral facial expression photo was taken at this time. Finally, the target person was instructed to "close your eyes one last time. This time imagine you are with someone whom you like very much. Show how that feels with a happy facial expression and open your eyes." Apparatus. All photos were taken using Kodachrome film with a speed of ASA 400. A Konica Autoreflex T (a single-lens reflex) was used, with a Soligar 28-mm lens for the whole person and body photos and a Vivitar 135-mm lens for the facial photos. Subjects, Two hundred male University of Illinois, Chicago, students, enrolled in introductory psychology courses, served as raters for part one of the first experiment. Twelve male and 45 female students selected from introductory psychology courses were the raters for the second part of the first experiment. A diversity of racial and ethnic groups were included among the raters. Procedure for facial attractiveness ratings. In Part 1, subjects provided ratings of the facial attractiveness of the target persons. A Latin square design was used so that three sets of photos were formed, each containing 15 different target persons. Each set of photos consisted of five facial photos of target persons' displaying happy facial expressions, five of neutral expressions, and five of sad expressions, with each of the three sets including a different facial expression for any given target person. Thus, by showing a different set of photos to three different groups of raters, independent judgments of the facial attractiveness of the same target persons with different facial expressions were obtained. Subjects provided individual judgments in groups of 30 to 40 for photos of target persons' physical attractiveness. A total of six groups of subjects gave ratings. Within each group, half of the subjects were instructed to "compensate in your rating for the facial expressions of the people" and half were not given this instruction. All instructions were provided in writing. Photographic slides of the target persons were projected onto a 24 X 30 in. screen in the front of a large room in which the subjects were seated. Paired comparison judgments were obtained for the first rating procedure of facial attractiveness. Two photos were projected onto the screen simultaneously for IS s, and during this time subjects selected and recorded on their answer sheets which target person was more attractive. For each of the three sets of photos, all possible pairs of 15 photos were rated, resulting in a total of 10S comparisons. Photo comparisons were presented randomly under the constraint that each photo be shown at least once during the first IS comparisons, that no photo be shown twice consecutively, and that the number of times each photo appeared on the left equaled the number of times it appeared on the right. Following the paired comparisons procedure, each photo was projected onto the screen alone and subjects rated each target person's facial attractiveness on a 10-point Likert scale. The subjects who were instructed to compensate for facial expression in the paired-comparison ratings were also instructed to compensate in the Likert ratings.

471

Procedure for facial expression ratings. Forty-five female and 12 male subjects were instructed in writing to rate on a 10-point Likert scale the facial photos for four dimensions of facial expression: pleasantness, intensity, surprise, and naturalness. The subjects were assembled in one room and simultaneously made ratings on the photos of the target persons, which were projected one at a time for IS s each onto a screen in the front of the room. Subjects rated all three facial expressions for each of the 1S target persons, for a total of 45 photos. For each photo, subjects gave four consecutive ratings along the dimensions of facial expression. The order of the photos of target persons was randomized as was the order of facial expression photos within target persons, but all three facial expressions for each target person were shown consecutively. Results Facial attractiveness ratings. The pairedcomparison data were scaled according to Thurstone's method for Condition C, which assumes that the variances of the discriminal differences are constant for all pairs of stimuli (Torgerson, 1958). Because three groups of subjects rated photos of target persons using different facial expressions, and in each group half the subjects were instructed to compensate for facial expression and half were not, six independent sets of ratings were scaled. The derived scale values of the target persons were tested for goodness of fit for each of the six sets using Mosteller's test (Torgerson, 1958). No tests were significant at the .10 probability level. Chi-squares were 89.18, 105.89, and 63.96 for the three no-compensation groups and 50.24,69.24, and 77.46 for the same three respective compensation groups (djs = 91 for all six groups). The nonsignificance of these tests indicates that the scaled values adequately fit the obtained pair comparison data and that the physical attractiveness of the target persons can be measured on a single dimension. To compare the goodness of fit of the Thurstone scale values' of the groups that were instructed to compensate for facial expression and the groups not so instructed, the sum of the three chi-square statistics of the no-compensation groups was divided by the sum of the chi-square statistics of the compensation groups. Because the sum of independent chisquare variables is itself a chi-square variable with the degrees of freedom for the sum equal to the sum of the degrees of freedom, and because the ratio of two independent chisquare variables each divided by its degrees of freedom is an F statistic (Hays, 1973), and

472

MUESER, GRAU, SUSSMAN, ROSEN

finally, because the six groups are experimentally independent, this ratio should have an F distribution with 273 degrees of freedom for the numerator and 273 degrees of freedom for the denominator. This ratio was statistically significant, ^273, 273) = 1.31, p < .05, indicating that the compensation group fit the unidimensional model better than the nocompensation group. The Likert ratings of facial attractiveness were scaled by determining the mean ratings for each target person's happy, sad, and neutral photos independently for the no-compensation and compensation groups. Thus, as for the Thurstone data, a total of six Likert scale values were determined for each target person. In order to determine the similarity of the Thurstone and Likert scale values, Pearson product-moment correlation coefficients were computed for the scale values of the happy, neutral, and sad photos separately for the nocompensation and compensation groups. All of these correlations were highly significant: r = .95, .98, and .98 for the happy, neutral, and sad photos in the no-compensation group and .93, .97, and .96 for the respective photos in the compensation group (for each correlation, df= 13, p < .0001). These correlations indicate that the Likert scale values closely approximate the Thurstone scale values. The effects of facial expression and the compensation/no-compensation instruction on attractiveness ratings were examined with repeated measures analyses of variance- (ANOVAS) using REGM, a multivariate general linear model hypothesis program (Wilkinson, 1980). The target persons served as the units of observation. Facial expression (happy, neutral, sad) and compensation instructions (compensation, no compensation), each a repeated measure, were the independent variables, and attractiveness ratings were the dependent variables. The Likert and Thurstone scale data were analyzed separately. Thurstone scale values. The main effect of facial expression was significant, multivariate F(2, 13) = 4.11, p < .05. Specified contrasts indicated that the sad expression was less attractive than the neutral expression F(l, 14) = 8.29, p < .05, and that the happy expression did not differ from the neutral expression, F(l, 14) = 0.05, ns. The effect of compensation instructions was not significant, F(l, 14) =

Table 1 Mean Thurstone Ratings and Standard Deviations of Attractiveness Facial expression Rating instructions No compensation M SD Compensation M SD

Marginals

Marginals

Happy

Neutral

Sad

2.24 0.746

2.28 0.797

1.79 0.864

2.10

2.28 0.580

2.30 0.775

1.75 0.817

2.11

2.26

2.29

1.77

0.01, ns, nor was the Compensation X Facial Expression interaction, multivariate F(2,13) = 0.89, ns. Means and standard deviations for the Thurstone scale values are displayed in Table 1. Likert scale values. A similar pattern of results was found for the Likert data. The effect of facial expression was marginally significant, multivariate F\2,13) = 3.24, p