Facial-attractiveness judgments influence many important social outcomes^1–3. For example, people prefer to date, mate with, associate with, hire, and vote for individuals with attractive faces^1–3. Given this link between facial attractiveness and social outcomes, many researchers have attempted to identify the physical characteristics that influence facial-attractiveness judgments. Because they are hypothesized to function as cues of immunity to infectious illnesses and/or increase the efficiency with which the perceptual system can encode and process faces (see Rhodes³ for a discussion of these two accounts of facial-attractiveness judgments), many studies have investigated the possible roles that symmetric, average (i.e., prototypical), and/or sexually dimorphic (i.e., masculine/feminine) shape characteristics play in facial-attractiveness judgments. To elaborate, some researchers have suggested that symmetry, averageness, and sexual dimorphism might influence facial attractiveness judgments because individuals possessing these characteristics also have strong immune systems and, consequently, are produce particularly healthy offspring⁴. However, other researchers have suggested that these characteristics might influence facial attractiveness, not because they advertise aspects of underlying physical condition, but simply because faces possessing these characteristics can be processed most efficiently by the perceptual system³.

Many studies have reported that more attractive faces have symmetric^5–8 and average^8–10 face shapes. Many other studies have reported that more attractive female faces possess feminine face shapes^11,12. By contrast with these results, evidence that more attractive male faces have masculine face shapes is rather mixed (see Little et al.² and Rhodes³ for reviews). For example, while some studies have reported that masculine male faces are perceived to be more attractive than feminine male faces¹¹, other studies have reported that feminine male faces are perceived to be more attractive than masculine male faces¹² or that masculinity–femininity does not have a significant effect on male facial attractiveness¹³. However, researchers have recently highlighted two potentially important problems with much of this research on the roles that symmetric, average, and sexually dimorphic shape characteristics play in facial-attractiveness judgments.

First, many studies investigating the roles that symmetric, average, and sexually dimorphic shape characteristics play in facial-attractiveness judgments have done so by experimentally manipulating these characteristics in face images using computer graphics methods^5,9,11,12. However, researchers have recently raised concerns about the ecological validity of this method^14–21. Indeed, several studies have now demonstrated that the large effects on social judgments of faces that are typically seen when this method is used are considerably smaller, and often not significant, when natural (i.e., unmanipulated) face images are rated and shape characteristics objectively measured from the face images^{14,15,17,22–24}. This has led some researchers to propose that experimentally manipulating shape characteristics in face images reveals what features participants can use to form impressions of others but does not necessarily reveal the features people actually do use under more naturalistic viewing condition where faces vary simultaneously on multiple dimensions (for a discussion of this issue, see Satchell et al.²⁰).

Second, most studies that have tested for possible relationships between attractiveness ratings of natural (i.e., unmanipulated) faces and measures of shape symmetry, averageness, or sexual dimorphism have done so by testing for bivariate (rather than partial) correlations only (see Rhodes³ for a review). This approach may be somewhat limited because it is not instructive about which of these face-shape characteristics independently predict attractiveness. Indeed, some studies have reported that averageness and symmetry^22,25 and symmetry and sexual dimorphism are positively correlated in face images^23,26–28.

Perhaps surprisingly, only a small number of studies have directly addressed these two potentially important issues. However, the specific patterns of results reported differ across these studies. Komori et al.²⁹ tested for independent relationships between attractiveness ratings of Japanese faces and both averageness and sexual dimorphism, finding that averageness and masculinity (in male faces) and averageness and femininity (in female faces) independently and positively predicted attractiveness ratings. In a follow-up study, Komori et al.³⁰ reported that averageness and symmetry independently and positively predicted attractiveness ratings of Japanese male faces and that averageness, but not symmetry, independently and positively predicted attractiveness ratings of Japanese female faces. By contrast with these results, Pavlovič et al.³¹ tested for independent relationships between attractiveness ratings of Vietnamese faces and symmetry, averageness, and sexual dimorphism, reporting a positive relationship between male attractiveness and averageness, a negative relationship between male attractiveness and masculinity, and a positive relationship between female attractiveness and averageness. By contrast, femininity was not significantly correlated with female attractiveness and symmetry was not significantly correlated with facial attractiveness in either sex. However, a different pattern of results was reported by Kleisner et al.³² in a study using a set of faces of diverse ethnicities. In this study, Kleisner et al.³² reported independent positive relationships between female attractiveness and averageness and femininity, but not symmetry, and independent positive relationships between male attractiveness and averageness, but not masculinity or symmetry. Thus, while each of these studies reported that averageness was positively correlated with facial attractiveness, findings for symmetry and both masculinity (in male faces) and femininity (in female faces) were mixed across studies.

In light of the mixed results described above, the current study tested for independent relationships between attractiveness ratings of white male and female faces and objective measures of asymmetry, distinctiveness (the converse of averageness), and sexual dimorphism (i.e., masculinity/femininity) of face shape. Given the mixed results from previous studies, we made no specific predictions about which particular features will or will not predict attractiveness significantly in this exploratory study.

All statistical analyses were carried out using R³⁶, with the packages lme4⁴², lmerTest 3.1-3⁴³, jtools 2.2.3⁴⁴, and robustHD 0.7.3⁴⁵. Data processing and display used kableExtra 1.3.4⁴⁶ and tidyverse 1.3.1⁴⁷. All data, full outputs, and analysis code are publicly available on the Open Science Framework (https://osf.io/xuz9r/).

Attractiveness ratings served as the dependent variable in a linear mixed effects model. The model included main effects of sexual dimorphism scores, distinctiveness scores, asymmetry scores, rater sex (effect coded so that − 0.5 corresponded to male raters and 0.5 corresponded to female raters), and face sex (effect coded so that − 0.5 corresponded to male faces and 0.5 corresponded to female faces) as predictors, as well as all possible two- and three-way interactions (excluding those involving multiple continuous predictors). The model also included by-rater and by-stimulus random intercepts, by-rater random slopes for sexual dimorphism, asymmetry, and distinctiveness (face sex varied between raters), and by-stimulus random slopes for rater sex. Sexual dimorphism, distinctiveness scores, and asymmetry scores were standardised prior to analyses by converting them to z scores (separately for each face sex). Full results for this model are summarised in Table 1. This analysis revealed significant negative main effects of both distinctiveness and sexual dimorphism. By contrast, the effect of asymmetry was not significant. The only significant interaction was the two-way interaction between sexual dimorphism and face sex. The significant effect of distinctiveness is shown in Fig. 2.Table 1

Results of our analysis of attractiveness ratings.

To interpret the significant interaction between sexual dimorphism and face sex, we repeated the analysis described above, this time analysing male and female faces in separate models. For female faces, there was a significant negative effect of sexual dimorphism (estimate = − 0.543, SE = 0.143, t = − 3.803, df = 50.802, p < 0.001). By contrast, the effect of sexual dimorphism was not significant for male faces (estimate = − 0.013, SE = 0.138, t = − 0.094, df = 50.145, p = 0.926). These results are shown in Fig. 3. Full results for both models are given on the Open Science Framework (https://osf.io/xuz9r/).Fig. 3

The significant effect of face-shape sexual dimorphism on female attractiveness and non-significant effect of face-shape sexual dimorphism on male attractiveness. The shaded areas show the 95% confidence intervals.

We also repeated the initial analysis described above, this time including all possible interactions among predictors in the model. This analysis revealed significant interactions among distinctiveness, rater sex, and asymmetry and among distinctiveness, face sex, rater sex, and sexual dimorphism. Results of these analyses are reported in full at https://osf.io/xuz9r/. Because these high-order interactions were unexpected, we do not discuss them further.

Our analyses of facial-attractiveness ratings showed a significant positive relationship between averageness of face shape and attractiveness ratings of male and female faces (i.e., negative relationships between distinctiveness and facial attractiveness) and a significant positive effect of femininity of face shape and attractiveness ratings of female faces. By contrast, masculinity of face shape was not significantly correlated with male facial attractiveness and symmetry was not significantly correlated with either male or female attractiveness. This pattern of results (significant effects of averageness and female femininity, but not symmetry or male masculinity) is identical to those reported previously by Kleisner et al.³². Although other studies testing for independent effects of averageness, symmetry, and sexual dimorphism of face shape on attractiveness ratings also reported positive effects of averageness^29–31 and female femininity²⁹, we did not observe the significant effects of symmetry³⁰ or male masculinity^29,31 that these studies reported. Thus, our findings add to a growing body of research suggesting that averageness is a key shape characteristic for facial attractiveness judgments and that there is little evidence for a significant role of symmetry or male masculinity (see also Kleisner et al.³²).

There has been considerable debate in the facial-attractiveness literature regarding the extent to which preferences for faces with average shapes are due to a hypothesized positive effect of symmetry on attractiveness ratings^48–50. However, results from studies in which averageness was experimentally manipulated independently of symmetry suggest that symmetry contributes little to the positive effect of averageness on facial-attractiveness judgments^48–50. Our finding that averageness, but not symmetry, has an independent positive effect on attractiveness ratings provides further evidence that averageness is attractive independent of the putative effects of symmetry and suggests that symmetry may, in fact, contribute little to the positive effect of averageness on facial attractiveness (see also Kleisner et al.³²). Although some previous studies have reported that symmetry influences facial attractiveness judgments (see Rhodes³ for a meta-analytic review), those studies did not typically control for the positive effects of averageness on attractiveness judgments. This issue is potentially noteworthy, since averageness and symmetry in faces are correlated^22,25. Indeed, recent work has suggested that effects of averageness mediate effects of symmetry on social judgments of faces²³. Whether this mediation is due to symmetry and averageness having similar effects on the ease with which faces can be processed by the perceptual system or that both characteristics are associated with similar aspects of physical condition remains an open question. Similarly, whether averageness influences facial attractiveness because average faces can be processed more efficiently or because it advertises physical condition (the two most common explanations for effects of averageness on facial attractiveness, Rhodes³) is likely to be an important topic for future research. This issue is also an important direction for future research on the effects of femininity on women’s facial attractiveness.

Our null result for masculinity and male facial attractiveness contrasts with findings from some previous research that found male facial attractiveness was influenced by masculine shape characteristics^11,12,51. However, studies showing this latter pattern of results typically employed stimuli in which shape characteristics were experimentally manipulated. Several lines of evidence suggest that findings obtained using this type of manipulated stimuli generalise poorly to studies investigating judgments of natural (i.e., unmanipulated) stimuli, as were employed in the current work (see Dong et al.¹⁴ for a recent review of work on this issue). Nonetheless, while we suggest that differences between the results of the current study and those of previous research on preferences for male facial masculinity may be due to this methodological issue, we also acknowledge that putative effects of contextual factors (e.g., sociocultural factors) on preferences for male facial masculinity⁵² might also contribute to differences in findings across studies.

Although we observed significant relationships between facial attractiveness and face-shape averageness and between shape femininity and female facial attractiveness, these effects were quite weak (see also Kleisner et al.³²). This issue raises the question of what additional factors might contribute to facial-attractiveness judgments. While most research on facial attractiveness has focussed on investigating possible roles of averageness, sexual dimorphism, and symmetry of face shape, other studies have suggested that surface information (e.g., color and texture information in faces) and cues of adiposity (i.e., facial correlates of body mass index) both may influence facial-attractiveness judgments (see Holzleitner et al.³⁸ for a review). Considering these (and other) additional characteristics may improve the predictive power of models of facial attractiveness.

To summarise, much of the research on predictors of facial-attractiveness judgments has focussed on the possible roles played by symmetry, averageness, and sexual dimorphism in shape characteristics. However, relatively few studies have investigated the extent to which these face-shape characteristics independently predict attractiveness and those studies that have done so have reported somewhat mixed results^29–32. Our finding that facial attractiveness is significantly correlated with averageness in both male and female faces and femininity in female faces, but not masculinity in male faces or symmetry directly replicates the pattern of results reported by Kleisner et al.³². Thus, our results present further evidence that averageness and femininity, rather than symmetry and masculinity, predict facial-attractiveness judgments.