Introduction
Most human observers can quickly evaluate facial attractiveness. In fact it takes only 100 ms to determine the attractiveness of a face (
Willis and Todorov, 2006). Faces that are more symmetrical, closer to the norm, or show greater evidence of estrogen or testosterone are generally perceived as more attractive (
Cunningham, 1986;
Grammer and Thornhill, 1994;
Perrett et al., 1998;
Perrett et al., 1999;
Ramachandran and Hirstein, 1999;
Rhodes, 2006;
Wong et al., 2008). The eyes receive greater attention than any other feature of the face (
Althoff and Cohen, 1999;
Davies and Hoffman, 2002) and provide important cues that influence attraction during social interactions. Dilated pupils indicate emotional arousal (
Bradley, Miccoli, Escrig, and Lang, 2008) and are rated more attractive by males and females (
Tombs and Silverman, 2004). Observers prefer a direct gaze when judging the attractiveness of a happy face, more than when judging the attractiveness of a disgusted face, and this difference in preference is greater when viewing a face of the opposite sex (
Conway, Jones, DeBruine, and Little, 2008). Here we show that a feature of the eye called the limbal ring also affects facial attractiveness.
The limbal ring appears as a dark annulus outlining the iris. This outline can be caused by pigmentation of the peripheral iris or by the internal optics of the eye (
Shyu and Wyatt, 2009). The visibility of the limbal ring can be decreased or completely removed by a variety of medical conditions that affect the transparency of the peripheral cornea. These include neovascularization and Pterygium related to unhealthy stem cells at the corneosclearal limbus (
Ang et al., 2010;
Mort, Ramaesh, Kleinjan, Morley, and West, 2009;
Sangwan, 2001;
Zheng and Xu, 2008) and cloudiness caused by corneal edema or corneal arcus (
Fernandez et al., 2009;
Khurana, 2007). The limbal ring also becomes less visible due to conditions more closely related to advanced age such as glaucoma and arcus sinilis (
Cavallotti and Cerulli, 2008). As demonstrated in
Figure 1, the thickness of the limbal ring is negatively correlated with age,
r(362) = -.303,
p < .001. As seen in this figure, the thickness of limbal rings begins to decline even in youth, before degenerative diseases usually begin. Thus a clearly visible limbal ring is a probabilistic indicator of both health and age.
Note: Ellipses were drawn to fit the inner and outer boundaries of the limbal ring on faces chosen randomly from
www.Facity.com. Thickness was computed as the mean difference of the major and minor axes of the two ellipses. The thickness of the limbal ring in this graph is reported as a proportion of the radius of the iris, to provide a scale invariant measure.
This raises the question: Are judgments of facial attractiveness shaped to use the information within the limbal ring in a potentially adaptive way? To answer this question, we showed observers, on each trial, two faces that were identical except that the eyes of one had dark limbal rings and the eyes of the other had no limbal rings. Observers chose which face was more attractive, and indicated the degree of their preference.
Results
Observers' ratings were assigned positive values if they moved the slider toward the dark limbal ring image, or negative values otherwise; mean values of zero indicate no preference. Right-tailed t-tests were used to determine if male and female observers preferred dark limbal rings in male and female stimuli. Mean ratings of all stimuli across male and female observers were significantly greater than zero (t(44) = 3.68, p < .001), as were mean ratings across observers when viewing only male (t(44) = 3.90, p < .001) or female (t(44) = 3.38, p = .001) faces. Male observers preferred the limbal ring in both male (t(19) = 2.30, p = .016) and female (t(19) = 2.02, p = .029) stimuli, as did female observers (t(24) = 3.13, p = .002 for male faces; t(24) = 2.70, p = .006 for female faces). Two-tailed independent-sample t-tests revealed no significant difference between male and female observers when viewing only male (t(43) = −0.323, p = .749) or female (t(43) = −0.174, p = .863) faces. Paired-samples t-tests showed that there was also no significant difference when viewing male and female stimuli for male observers (t(19) = −0.6558, p = .520) or female observers (t(24) = −1.621, p = .118).
Right-tailed t-tests demonstrate preference for the limbal ring in both upright faces (t(44) = 3.99, p < .001) and inverted faces (t(44) = 3.28, p = .001). This was the case for both male observers (upright: t(19) = 2.42, p = .0129; inverted: t(19) = 1.91, p = .0357) and female observers (upright: t(24) = 3.13, p = .00225; inverted: t(24) = 2.67, p = .00676). A paired samples t-test demonstrates that the preference for the limbal ring was greater in upright than inverted faces (t(44) = 2.76, p = .00845). A mixed design ANOVA confirmed these findings with significant main effects for face inversion (F(1,43) = 7.41, p = .009) and presence of the limbal ring (F(1,43) = 12.89, p = .001).
Note: Positive values indicate preference for faces with a dark limbal ring.
Discussion
The limbal ring is caused by anatomical and optical components and becomes less visible with age or poor health. Male and female participants rated faces with clearly visible limbal rings as more attractive, suggesting that both sexes have been shaped by natural selection to use the limbal ring as a probabilistic indicator of reproductive fitness.
Inverted stimuli were included in the study to test if limbal rings are processed as local features or configural features of the face. When viewing upright faces, we exploit spatial relationships among facial features (i.e., their configuration), but when a face is presented upside-down we rely primarily on local features (e.g., the shape or color of the eye;
Diamond and Carey, 1986;
Maurer, Le Grand, and Mondloch, 2002;
Murray, Yong, and Rhodes, 2000;
Rossion and Gauthier, 2002). Our finding that inverted faces are rated more attractive when they have a dark limbal ring suggests that the limbal ring is processed as a local feature in the analysis of facial beauty. Our further finding that a dark limbal ring enhances attractiveness of upright faces more than of inverted faces could be interpreted as indicating that the limbal ring also has some role in the configural analysis of facial beauty (see, e.g.,
Bäuml, 1994). However, a more likely explanation is simply that the eyes, and therefore the limbal rings, receive fewer fixations and less attention when faces are inverted (
Barton, Radcliffe, Cherkasova, Edelman, and Intrilligator, 2006;
Davies and Hoffman, 2002).
When rating attractiveness of the opposite sex, men typically place more emphasis on signals of youth than do women (
Buss, 2007;
Buss and Barnes, 1986;
Etcoff, 1999). In principle, the limbal ring could serve as a signal of youth: Thickness of the limbal ring declines with increasing age, as shown in Figure 1. Our data show no evidence that the limbal ring affects male ratings of attractiveness differently than it affects female ratings. This might indicate that the limbal ring is used as a signal of health but not of youth. Or it might indicate that our experiment did not detect real differences in male and female ratings. Our stimuli have only a dark limbal ring or no limbal ring; perhaps adding more subtle variations to the limbal rings would lead to more nuanced results. It would be interesting to test directly whether the limbal ring affects judgments of age. One could, for instance, have observers estimate the ages of faces shown one at a time, and have each face appear twice at random in the sequence, once with a dark limbal ring and once with a minimal limbal ring. If the versions with dark rings are estimated to be significantly younger, this would be good evidence that the limbal ring can affect judgments of age.
One might argue that the preference for a dark limbal ring in our experiment is due to an artifact in our stimuli: One face in each pair has no limbal ring at all, and one might argue that this is so unnatural that it makes the face unattractive. However, eyes with faint or nonexistent limbal rings are common, and there are many attractive faces with faint limbal rings.
It may be that the preference for a dark limbal ring in our experiment is due not simply to its signaling of corneal clarity and iris pigmentation, but also to enhancing of the apparent whiteness and brightness of the sclera by increased local contrast. The relative contributions of these two factors to attractiveness can be tested in future experiments by using stimuli in which the scleras have been adjusted to have equal apparent brightness and whiteness. Even after such an adjustment, the luminance contrast between the sclera and iris plays a vital role in social interactions by indicating gaze direction (
Ando, 2002;
Sinha, 2000). The direction of gaze is interpreted more quickly for opposite sex faces with exaggerated sex-typical features, suggesting that detecting the gaze of potential high-quality mates was an important selection pressure (
Jones, Main, DeBruine, Little, and Welling, 2010). A dark limbal ring provides an additional contrast boundary and may further enhance gaze-dependent social interactions.
In principle, the limbal ring could be a rich source of probabilistic information about youth and health. Our experiment shows that a dark limbal ring is more attractive than no limbal ring; however, it is an interesting and open question whether human vision has been shaped to exploit subtle differences in limbal rings, or whether we have been shaped to use simple heuristics such as “dark is better” or “thick is better.” The results could have practical application: Colored contact lenses are now commercially available that mimic the appearance of a limbal ring.
The faces in our experiment had lightly colored irises; however, the limbal ring is more difficult to detect in darker irises. Therefore the limbal ring might have less influence on attractiveness in faces with dark irises. This is an interesting topic for further experiments.
Acknowledgements
We thank Naveen Hothi and Shane Johnston for help creating the stimuli, and David Buss, Shannon Cuykendall, Nancy Etcoff and Marjan Farid for helpful comments on an earlier draft of this paper. We also thank Claudia Gavrilescu, Alysha Kassam, Samantha Kiel, Faizan Memon, Ian Merion, Matt Richardson and Jessica Wong for making the limbal ring measurements displayed in Figure 1. This research was supported in part by a grant from Procter & Gamble.