Introduction

It has repeatedly been shown that facial appearance strongly affects various social interactions [1], [2]. Face perception is domain-specific, heritable, and independent of other cognitive abilities [3]. It should be highly adaptive in any social environment, especially for distinguishing a friend from a foe. In particular, primates show a highly developed capacity to recognize faces accurately even after the viewing position has been changed [4]. Humans are no exception.

Face perception provides information on trustworthiness, which in turn assists an individual’s social, economic, and reproductive success [5], [6]. During economic trust games, male participants with broader faces were more likely to exploit their counterparts’ trust than were male participants with narrower faces, and other players were more likely to entrust their money to males with longer and narrower faces [7]. In the human brain, the amygdala has been identified as the key structure for automatic evaluation of trustworthiness [5], [8]. Patients with bilateral amygdala damage cannot judge the trustworthiness of other persons appropriately [9]. Perceived trustworthiness also correlates with perceived happiness. Conversely, untrustworthy faces are considered to be angrier [10].

Most previous research has ignored the eyes as a facial feature that may influence personality assessment. Yet the eyes are highly conspicuous because the color of the iris contrasts with the whiteness of the sclera and can vary considerably [11]–[13]. There is some evidence that eye color correlates with various psychological and biosocial factors, such as temperament [14]. Blue-eyed infants are reportedly more inhibited, shy, and timid than brown-eyed infants [14]–[16]. Similarly, Coplan et al. [17] found that boys with blue eyes were socially warier than boys with brown eyes, although such differences were absent between blue- and brown-eyed girls. In a previous study, we showed that eye color affects male face shape, specifically the mouth, eye, and chin area [18]. Brown-eyed males tend to have a face shape that suggests happiness and, hence, higher perceived trustworthiness while blue-eyed males tend to have a face shape that indicates anger and, hence, lower perceived trustworthiness [5], [18], [19]. There is also evidence that people are felt to be more honest if they are baby-faced with shorter chins and lower positioned features [20], [21]. Here we hypothesize that faces of brown-eyed individuals should be perceived as more trustworthy-looking than blue-eyed individuals because of certain stable facial features that correlate with eye color. Further, we wish to identify the specific features that evoke a perception of trustworthiness by using geometric morphometrics to analyze face shape variation among male and female university students.

Methods

Results

Cronbach’s alpha values showed high inter-rater agreement for trustworthiness (0.970), dominance (0.835), and attractiveness (0.979). Eye color significantly influenced perceived trustworthiness (F_1,76 = 4.24, p = 0.043, η² = 0.053). The brown-eyed faces were perceived as more trustworthy than the blue-eyed ones. Sex of face had a significant effect on perceived trustworthiness (F_1,76 = 59.90, p = 0.001, η² = 0.441); female faces were rated as more trustworthy than male faces. We found a significant negative correlation between perceived trustworthiness and perceived dominance (r = −0.437, N = 80, p<0.01) and a positive correlation between perceived trustworthiness and perceived attractiveness (r = 0.829, N = 80, p<0.01). After we controlled for dominance and attractiveness, eye color still correlated with trustworthiness (F_1,74 = 18.54, p<0.001, η² = 0.2). Perceived dominance (F_1,74 = 38.49, p<0.001, η² = 0.342) and attractiveness (F_1,74 = 126.82, p<0.001, η² = 0.632) had a significant effect on perceived trustworthiness. Nevertheless, controlling for dominance and attractiveness did eliminate the effect of sex (F_1,73 = 0.024, p = 0.877, η² = 0.000), although the interactions “sex*dominance” (F_2,73 = 19.21, p<0.001, η² = 0.345) and “sex*attractiveness” (F_2,73 = 68.28, p<0.001, η² = 0.652) remained significant, indicating that male raters and female raters perceived attractiveness and dominance differently. For this reason we reran the analysis separately for male raters and female raters. Eye color had a significant effect on perception of trustworthiness by male raters (F_1,36 = 6.72, p = 0.014, η² = 0.157) and by female raters (F_1,36 = 11.01, p = 0.002, η² = 0.234). Both sexes considered the brown-eyed faces to be more trustworthy than the blue-eyed ones (Fig. 1).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 1. Perceived trustworthiness of blue-eyed and brown-eyed male faces and female faces.

The y-axis shows residuals of perceived trustworthiness after statistical control for attractiveness and dominance (expressed by z-scores); whiskers denote standard deviations. Solid lines within the boxes indicate the group median.

https://doi.org/10.1371/journal.pone.0053285.g001

We tested the possible effect of rater’s eye color on perceived trustworthiness with linear mixed-effects models. After we added rater’s eye color to the model, the relationship between photo’s eye color and perception of trustworthiness was significant for male faces (F_1,36 = 6.882, p = 0.013) and for female faces (F_1,37 = 13.721, p = 0.001). For male faces, rater's eye color had no effect on ratings of trustworthiness (F_{2, 1327} = 0.606, p = 0.546). For female faces, the effect was significant (F_2,1745 = 3.928, p = 0.020). Blue-eyed female faces received lower ratings from brown-eyed raters than from blue- or green-eyed raters. Similarly, brown-eyed raters gave higher ratings to brown-eyed female faces than to blue-eyed female faces (F_1,74 = 20.04, p<0.001). Despite some differences due to rater’s eye color, this factor was not primarily driving the ratings. All raters, irrespective of their own eye color, perceived the brown-eyed faces as more trustworthy than the blue-eyed ones.

We used a one-sample t-test to measure the effect of artificially recoloring the eyes, i.e., by comparing the mean change in trustworthiness of individual faces with a theoretical value of 0. Recoloring the eyes from blue to brown or from brown to blue did not systematically influence perceived trustworthiness (male faces: p = 0.818, t₃₉ = 0.232, Mean±SE = 0.024±0.104; female faces: p = 0.748, t₃₉ = −0.324, Mean±SE = 0.024±0.075). Thus, despite the higher attributed trustworthiness of brown-eyed faces, eye color itself had no significant effect on perceived trustworthiness.

Facial WHR negatively correlated with perceived trustworthiness of male faces (F_1,38 = 5.197, p = 0.028, η² = 0.12). Eye color was uncorrelated with facial WHR (F_1,38 = 1.98, p = 0.168, η² = 0.049), so this ratio did not significantly differ between blue-eyed male faces (Mean±SD = 1.895±0.127) and brown-eyed ones (Mean±SD = 1.949±0.107).

Shape regression showed a significant correlation between perceived trustworthiness and male face shape (Goodall's F-test for 10,000 permutations: p = 0.0005, regression explained 8.6% of variance) but not for female face shape (p = 0. 17). Similarly, the correlation between eye color and face shape was significant only for male faces (Goodall's F-test for 10,000 permutations: p = 0.027, 5.5% of explained variance) but not for female faces (p = 0.058, 4.4% of explained variance) although the same trend was present. In contrast to blue-eyed male faces, brown-eyed male faces had a bigger mouth, a broader chin, a bigger nose, and more prominent eyebrows positioned closer to each other (Fig. 2d, 2f). A trustworthy male face was characterized by dilations of the TPS deformation grid in the mouth area and by constriction of the eyebrows (Fig. 2g, 2i). These characteristics, and the overall shape of a trustworthy face, matched TPS deformations in the direction of brown-eyed facial morphology. Female faces exhibited a similar pattern. Blue-eyed female faces tended to be oval with a less prominent and rounded chin, a mouth with corners parallel to the horizontal, and a longer distance between the eyes. In contrast, the narrower brown-eyed female faces tended to have a more prominent chin, a mouth with upward-pointing corners, and a shorter distance between the eyes (see Fig. 2a, 2c). The relationship between female face shape and perception of trustworthiness could not be reliably described because the correlation was not significant, nor did it show a trend.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 2. Shape changes associated with eye color and perceived trustworthiness.

Thin-plate spline visualizations of the way face shape correlates with eye color (a–f) and trustworthiness (g–i). Generated face shapes of blue-eyed woman (a) and brown-eyed woman (c) compared to average female face (b). Generated face shapes of blue-eyed man (d) and brown-eyed man (f) compared to average male face (e). Generated face shapes of untrustworthy-looking man (g) and trustworthy-looking (i) man compared to average male face (h). The TPS grids of perceived trustworthiness for women are not shown because shape analysis did not meet statistical significance. The generated facial images (a–f) were magnified 3x for better readability.

https://doi.org/10.1371/journal.pone.0053285.g002

Facial morphology was more variable in males than in females: p = 0.012, for 1000 permutations (male faces: MD = 0.00361, female faces: MD = 0.00285). Face shape variability did not significantly differ between brown-eyed male faces (MD = 0.00309) and blue-eyed ones (MD = 0.00373: p = 0.140, for 10,000 permutations). Body height did not significantly differ either between brown-eyed male faces and blue-eyed ones (t₃₈ = −0.711, p = 0.481), nor did it differ by eye color among female faces (t₃₈ = −0.312, p = 0.757). No difference in BMI was observed between blue-eyed and brown-eyed men (t₆₀ = 0.219, p = 0.827) or between blue-eyed and brown-eyed women (t₁₈₈ = 0.275, p = 0.783).

Discussion

We found significant sex-dependent correlations between eye color, face shape, and perceived trustworthiness. Brown-eyed faces were perceived as more trustworthy than blue-eyed ones (See Fig. 1). There was also a relationship between eye color and face shape. Although the results were statistically significant only for male faces, the trend for female faces was in the same direction. The non-significant results for female faces might be due to lower phenotypic variability among women in general [34]. Blue-eyed male faces were characterized by a more angular and prominent lower face, a longer chin, a narrower mouth with downward-pointing corners, relatively smaller eyes, and rather distant eyebrows. This was also the pattern of an untrustworthy face. In contrast, brown-eyed faces tended to have a rounder and broader chin, a broader mouth with upward-pointing corners, relatively bigger eyes, and eyebrows closer to each other. This was also the pattern of a trustworthy face. Our findings are consistent with those of Todorov [5], i.e., when facial features associated with perceptions of higher (or lower) trustworthiness are exaggerated, the result is an expression of happiness (or anger) (See Fig. 2g–i). Further, the longer chin of blue-eyed individuals makes them look less baby-faced. It has been shown that higher ratings for honesty are given to more baby-faced people with shorter chins and lower-positioned facial features [20].

Given the negative correlation between perceived dominance and trustworthiness, our present results seemingly contradict the finding by Kleisner et al. [18] that brown-eyed men are perceived as being more dominant. It may be that eye color transmits two different signals. The perceived dominance of brown-eyed men may weaken but not totally eliminate the relationship between brown-eyed face shape and perceived trustworthiness. In support of the two-signal hypothesis, eye color had a stronger effect on trustworthiness when perceived dominance was controlled.

Our results lead to three questions. First, why would a man with a less robust face, characterized by a smaller nose, chin, and mouth, be perceived as less trustworthy? Second, why would these less robust faces be associated with blue eyes? And third, why would this association be stronger for male faces than for female faces?

Men with smaller facial features are perceived as less masculine, and therefore as less dependable and trustworthy. For instance, when Japanese women judge Japanese men, they tend to associate darker skin (itself a male characteristic) not only with greater masculinity, strength, and assertiveness, but also with manly sincerity and dependability [35]. Nevertheless, there is also evidence that more masculine faces are generally thought to be less cooperative, less honest, and less parenting-oriented [36]. Men with wider faces are perceived as less attractive and less trustworthy [7]. Keep in mind that facial WHR does correlate with various facial features that in turn correlate with perceived trustworthiness but not with eye color. The GMM analysis also showed that differences in perceived trustworthiness are not triggered solely by facial WHR. Therefore, other facial traits than facial WHR, such as the shape of the nose, the mouth, the chin, and the eye area, may primarily contribute to a perception of trustworthiness or untrustworthiness. The same facial features that make one seem more trustworthy, namely bigger eyes, larger eyebrows, a mouth with upward-pointing corners, and a generally extended and narrower face shape, are also perceived as being more attractive (see [37] for attractiveness visualization). To date, it seems that relatively smaller eyes and a smaller mouth with downward-pointing corners are the facial traits that cause the lower perceived trustworthiness of blue-eyed males.

Adaptation may also affect how one perceives the trustworthiness and attractiveness of a face [38]. The extent to which male (or female) faces are judged trustworthy (or untrustworthy) would thus depend on visual experience with a particular face shape. The first question might be answered conclusively by cross-cultural comparisons of attributes of masculinity/femininity and judgments of trustworthiness. To answer the second and third questions we propose two explanations.

Acknowledgments

We are grateful to all the students of the Faculty of Science who participated in this research. We thank Jakub Kreisinger, Vojtěch Jarošík, and two anonymous reviewers for valuable comments on earlier drafts of the manuscript.