Ocular pigmentation in humans, great apes, and gibbons is not suggestive of communicative functions

Pigmentation patterns of the visible part of the eyeball, encompassing the iris and portions of the sclera, have been discussed to be linked to social cognition in primates. The cooperative eye hypothesis suggests the white sclera of humans to be a derived adaptive trait that enhances eye-mediated communication. Here, we provide a comparative analysis of ocular pigmentation patterns in 15 species of hominoids (humans, great apes & gibbons) that show marked differences in social cognition and quantify scleral exposure at the genus level. Our data reveals a continuum of eye pigmentation traits in hominoids which does not align with the complexity of gaze-mediated communication in the studied taxa. Gibbons display darker eyes than great apes and expose less sclera. Iridoscleral contrasts in orangutans and gorillas approach the human condition but differ between congeneric species. Contrary to recent discussions, we found chimpanzee eyes to exhibit a cryptic coloration scheme that resembles gibbons more than other apes. We reevaluate the evidence for links between social cognition and eye pigmentation in primates, concluding that the cooperative eye hypothesis cannot explain the patterns observed. Differences in scleral pigmentation between great apes and humans are gradual and might have arisen via genetic drift and sexual selection.


Materials and methods
We conducted an extensive internet search for pictures of great apes and gibbons from wild as well as captive environments and pooled images found online with private digital photographs (see Suppl. Table 1). In case of data gathered online, information on the identity, sex and location of photographed apes was derived from the source websites to avoid repeated sampling of particular individuals. If available, international studbooks were used to check whether zoo-housed individuals were born in the wild or in captivity. Species level classification and nomenclature were adopted from Burgin et al. 25 . Our species selection was dictated by the availability of photographs, which forced us to exclude specific ape lineages, such as Tapanuli orangutans (P. tapanuliensis) and the northerly distributed species of crested gibbons (N. concolor, N. hainanus, and N. nasutus). Within hylobatids, we pooled data from recently diverging populations to reach larger sample sizes for the respective groups. As a consequence, Nomascus siki was grouped under N. leucogenys, N. annamensis under N. gabriellae, Hoolock tianxing under Ho. leuconedys, and Hylobates abbotti and Hy. funereus under Hy. muelleri (see Suppl. Tables 1 and 2 for taxonomic identities of each subject). Because ocular pigmentation and the shape of the eye contour can differ between juvenile and mature apes, we restricted our sampling to pictures of adults. Subadult animals were included for some gibbon species but only if they had already developed adult pelage traits (e.g., after entering the pale color phase in female gibbons of the genera Hoolock and Nomascus).
We followed the methodology of Mayhew and Gómez 7 , Perea García 9 , and Perea García et al. 8 in using ImageJ 26 to take quantitative measurements from the digital images we gathered.
To be included in the sample, picture resolution had to be high enough to unequivocally distinguish between pupil, iris, and sclera in at least one eye of the photographed subject. We noticed that all gibbon species, as well as orangutans and gorillas, display a thin gray line encircling the peripheral iris, which appears as a salient demarcation to the sclera (see Fig. 1). We were unable to find discussion of this trait in the literature. While this structure is superficially reminiscent of the arcus senilis 27 commonly found in aging humans, it probably has different physiological underpinnings. We only sampled pictures on which this demarcation line was visible, to clearly distinguish iridial and scleral portions of the visible eyeball. In case the aforementioned criteria were met, pictures capturing both direct and averted gaze were included.
We used the plot profile function in ImageJ ( Fig. 1) to retrieve luminance values from the images 9 . Only traits from one eye per subject were quantified. In doing so, we chose the better illuminated eye for measurements. In case both eyes appeared equally well visible (direct gaze conditions), we selected the one with higher contrast values to further mitigate shadow-induced biases.
We quantified gray scale luminance values to quantify ocular contrasts and classified eyes into two phenotypic groups following Perea García et al. 8 . In type 1 eyes, the sclera is (at least partially) lighter than the iris (e.g., humans), while the opposite is true for type 2 eyes (e.g., most chimpanzees). Dependent on eye type, we chose either the highest or lowest grayscale luminance values from either portion of the eye to achieve the highest Table 1. Summary of data on ape ocular pigmentation patterns. Type I phenotype describes eyes in which the sclera is lighter than the iris. HC and RIL correspond to species means. *Data derive from Perea García et al. 8 . **n = 47. www.nature.com/scientificreports/ possible contrast difference. Subsequently, these values were used to calculate the absolute (highest ocular contrast = HC) and relative (relative iris luminance = RIL) differences between scleral and iridal luminance per eye. RIL reflects the percentage of grayscale luminance shown by the darker portion of the eye (sclera or iris) in comparison to the lighter one, which per definition is assumed to represent 100% luminance 28 . Higher HC values indicate greater conspicuousness, while the opposite is the case for RIL. Procedures were adopted from Perea García 9 and Perea García et al. 8 . Reflections mirrored in the eye as well as the demarcation line between iris and sclera were carefully avoided. We also quantified the gray value slope between iris and sclera 9 for each individual but did not incorporate these data into our analyses (Suppl. Table 1). Besides interspecific comparisons, we attempted to compare pigmentation patterns in wild and captivebred individuals. This way we could test for potential biases resulting from an animal's rearing background. It is known that the physiology and morphology of zoo animals can significantly differ from wild conspecifics in various ways 29 , and an effect on eye pigmentation should be considered possible. However, we only considered our samples sufficient for such a comparison in Gorilla gorilla (n captive-bred = 23; n wildborn = 17) and Hylobates lar (n captive-bred = 13; n wildborn = 9).
Quantifying ocular shape and sclera exposure in hominoids. Ocular shape was quantified at the genus level for all genera of extant hominoids except for Hoolock. The latter was omitted due to a lack of suitable photographs. A set of images from 20 individuals representing either direct (eyes oriented forwardly towards the camera; n = 10) or sideways averted glances (n = 10) was analyzed for each genus, respectively. Congeneric species were grouped because preliminary screenings did not reveal notable differences. This resulted in a total sample of 140 pictures (Suppl. Table 2). In all these photos, subjects consistently faced the camera with fully opened eyes to reduce effects of the angle of photography on the measurements. While pictures of non-human primates were collected as described beforehand, all photos of humans derived from the private archives of the authors. All persons pictured gave consent for the photos to be used in this study. The human subjects were of diverse Eurasian descent. For nonhuman species, a trained observer (KRC) assigned photos into the "direct glance" and "averted glance" categories but a naïve coder (SB) scored them into said categories as well, allowing us to compute Cohen's Kappa as a measure of inter-rater reliability.
The width-height ratio (WHR) and the exposed sclera size index (SSI) of eyes were quantified, following the procedures of Kobayashi and Kohshima 4 and Mayhew and Gómez 7 to allow for meaningful comparisons (Fig. 1). WHR is a measure to approximate eye shape, while SSI indicates the amount of exposed sclera. WHR was calculated from all images in the dataset for the respective genus, while SSI was calculated for direct and averted glance images independently to approximate changes in visible scleral surface during averted glancing. In case both eyes of an individual were clearly visible (n = 137), measurements from the left and right eye were averaged (Suppl. Table 2). Statistical analysis. All statistics were performed in R 30 . After log-transformation, normal distribution of data was assessed by applying the Shapiro-Wilk test and homogeneity of variances was checked by running the Bartlett test. Parametric data were compared via ANOVA, while non-parametric data were analyzed using Wilcoxon's rank sum test. Tukey's Honest Significant Differences was employed as a post-hoc test correcting for multiple comparisons subsequent to ANOVA, while a Bonferroni correction was applied to address this issue for Wilcoxon tests. Differences in RIL and HC between wildborn and captive-bred individuals were tested by  We visualized phylogenetic patterns, quantified phylogenetic signals (Pagel's λ, phylosig function) and computed maximum likelihood ancestral state estimates (fastAnc function) with the phytools package version 0.7-70 31 . The hominoid phylogeny used was derived from the 10kTrees website 32 .
To visualize differences in the quantified ocular traits, a principal component analysis (PCA) was run on the species mean values for HC, RIL, and the species-specific proportion of type 1 eyes in the population (Type) as well as on the genus medians for SSI during averted glancing and WHR (see Tables 1, 2). Due to the lack of SSI and WHR measurements, Hoolock was omitted from the analysis. SSI and WHR data of congeneric species were assumed to be equal.

Results
Qualitative assessment of ocular pigmentation in great and small apes. In small apes, the sclera was found to be predominately darker than the iris (type 2 phenotype; n = 135 of 168; 80%), but notable interspecific variation was found (Table 1). In some species, the type 2 pattern was recovered for all individuals (Ho. leuconedys, Hy. moloch, Hy. muelleri, N. leucogenys). The highest prevalence of light sclerae (type 1 phenotype) was found in siamangs (S. syndactylus, n = 11 of 29; 38%) and white-handed gibbons (Hy. lar, n = 14 of 30, 46%). In most of these individuals, however, scleral depigmentation was moderate, leading to a medium to light brown sclera, which often was only minimally lighter than the iris. Of all sampled hylobatids, only four whitehanded gibbons displayed advanced bilateral depigmentation of the sclera, resulting in a mottled white appearance, vaguely resembling the human condition. The small ape sclera, if not depigmented, appears dark brown to almost black in all species. Iris color was found to be mostly dark to chestnut brown (see Fig. 2). However, in Hy. moloch and Hy. muelleri, the iris has a dark amber color, similar to that of chimpanzees (Fig. 2).
The sclerae of gorillas and orangutans tended to be, at least in parts, lighter than their irises, displaying various degrees of depigmentation. Eastern gorillas were an exception to this trend. Here the inverse pattern (G. beringei, n = 13 of 22, 59%) was found to be more abundant but both occured at high frequencies. In Western gorillas and orangutans, light sclerae were predominately found (G. gorilla, n = 34 of 40, 85%; P. abelii, n = 20 of 23, 87%; P. pygmaeus, n = 19 of 24, 79%). However, just as in many gibbon species, depigmentation in Bornean orangutans was often low, leading to predominately brownish instead of white sclerae in this species. In both gorillas and orangutans, the portions of the sclera immediately surrounding the iris typically (but not always) remained pigmented, creating a dark frame of variable thickness.
Gorillas deviated from both orangutans and gibbons in frequently displaying clearly asymmetric depigmentation patterns. In approximately one quarter of Eastern (n = 6 of 22; 27%) and Western gorillas (n = 10 of 40; 25%), conspicuous depigmentation was restricted to just one eye. This pattern was only noted for one Sumatran orangutan and was absent in the Bornean species. In gibbons, such asymmetries were also rare, occurring in white-handed gibbons (n = 4 of 30; 13%), pileated gibbons (n = 2 of 17; 12%) and siamangs (n = 2 of 29; 7%).
Quantitative assessment of ocular pigmentation in great and small apes. At the family level, we found a greater HC (Wilcoxon test: p < 0.001; Fig. 3) but lower RIL (Wilcoxon test: p < 0.001; Fig. 4) in hominids compared to hylobatids.
Humans, Sumatran orangutans and Western gorillas exhibited the highest mean HC values, followed by bonobos, Eastern gorillas, Bornean orangutans and chimpanzees ( Table 2). Chimpanzees were the only hominids for which the mean HC was recovered to lay within the range of variation of hylobatid species means. Among small apes, Bornean and Javan gibbons exhibited the highest HC values while the lowest were found among crested gibbons. Concerning RIL, species means for hominids fell within the range of hylobatid variation (Fig. 4). The lowest RIL in the sample was found in Javan gibbons, while the highest was recovered for Southern yellow-cheeked gibbons. Among hominids, humans displayed the highest RIL, while the lowest was found in Western gorillas (Table 2).
Reflecting these findings, we found a significant but moderate phylogenetic signal for HC among the hominoid sample (Pagel's λ = 0.565, p = 0.026). RIL on the other hand was not found to correlate with phylogeny (Pagel's λ < 0.001, p = 1), which is mirrored by the inconsistent distribution of the trait among the studied taxa. Maximum likelihood ancestral state estimates of HC and RIL for each node within our hominoid phylogeny Table 2. Median values (± SD) of width-height ratio (WHR) and scleral size index (SSI) during direct and averted gaze in seven hominoid genera.     Table 3. For HC, only crested gibbons (genus Nomascus) exhibited values that significantly differed from other hylobatids, their eyes being notably dark. This genus also included the only gibbon species that significantly deviated from the patterns found in chimpanzees, bonobos and Bornean orangutans. Human HC differed significantly from all species in the sample except for Sumatran orangutans, bonobos, and gorillas, species which also tend to exhibit depigmented sclerae. Therefore, the pattern of contrasts found in the human eye is not unique. Comparisons of RIL did not produce similarly comprehensible patterns. While some species did not show significant differences to any others in the sample (Eastern hoolock, bonobo), Western gorillas did so in comparison to seven species, including humans, chimpanzees, and a range of small apes. There was a moderate but significant negative correlation between HC and RIL (Pearson's r =− 0.54, p = 0.04).
No significant differences in RIL and HC could be detected between captive-born and wild-born Western gorillas (Wilcoxon test: W ≥ 129; p ≥ 0.07) or white-handed gibbons (t-test: t ≥ -0.511; p ≥ 0.62).  www.nature.com/scientificreports/ Quantifying ocular shape and sclera exposure in hominoids. Inter-rater reliability for the scoring of glance direction was strong throughout, but agreement was slightly higher for the great ape (59/60; κ = 0.96, p < 0.001) than for the gibbon sample (56/60; κ = 0.86; p < 0.001). Mean SSI was found to be consistently smaller in hylobatids compared to those of hominids, with a particularly pronounced difference occurring in the averted gaze condition ( Principal component analysis of ocular traits. PCA grouped hominoids into three groups based on ocular traits (Fig. 6). The first two principal components of the PCA encompassed 89.8% of the total variance in the sample, exhibiting eigenvalues of 3.55 and 0.94, respectively. Variable contributions are visualized in Supplementary Fig. 1. The first group is constituted by hylobatids and chimpanzees, which clustered together in the PCA morphospace (Fig. 6). Bonobos grouped together with gorillas and orangutans, forming a second cluster. Finally, humans fell far outside of the range of variance of all the other hominoid genera. A second PCA omitting RIL, which is a problematic variable (see "Discussion" section), resulted in a similar pattern (Suppl. Figures 2 and  3), with the first and second principal component encompassing even more of the total variance in the sample (94.3%). In this PCA, gorillas were situated closer to humans than in the first one, diffusing the cluster constituted by bonobos, gorillas and orangutans.

Discussion
General discussion of results. Although we found only few consistent differences in ocular traits separating small and great apes, some predictions of the cooperative eye hypothesis could be confirmed. Importantly, SSI was found to be consistently lower in gibbons compared to other hominoids and this difference became even more pronounced in averted gaze situations. This finding supports the conclusions of Kobayashi and Kohshima 4 that were drawn from a small-scale dataset and further demonstrates that gibbon eyes are indeed far less suited to convey glance signals than those of great apes and humans. Regarding the width-height ratio of the eyes, hylo- www.nature.com/scientificreports/ batids also displayed lower mean values than hominids. Nevertheless, siamangs approach the great ape genera Pan and Pongo in WHR and were found not to differ significantly from them in that regard. Still, this did not result in notably higher SSI values in siamangs compared to other gibbons, nor to equal SSI when compared to these great apes (Table 2). Our results on scleral exposure and eye outlines in human and great ape eyes match the data from previous studies 4, 11 . In particular, we replicated the finding of Mayhew and Gómez 7 that humans and gorillas do not differ in the degree of scleral exposure during averted glancing but do so in the direct glance condition. The reason for this disparity lies probably in the horizontally widened outline of the human eye. It causes rotations of the human eyeball to have less of an effect on scleral exposure when compared to other hominoids. Accordingly, the relative difference in the amount of visible sclera between direct and averted glance SSI shown by humans is exceeded by all great apes as well as by the gibbon genus Hylobates (Table 2). Pigmentation patterns followed a roughly similar pattern to SSI among hominoids, but they were not congruent. Hylobatids displayed less contrasted eyes than their large-bodied relatives, as indicated by values for highest ocular contrast (HC), which were found to moderately correlate with hominoid phylogeny. As with the comparatively small amounts of exposed sclera in the gibbon eye, this again demonstrates a greater signaling value of hominid compared to hylobatid eyes. At the species level however, this notion cannot be generalized as chimpanzees were found to exhibit mean HC values in the range of gibbons, rendering their eyes similarly inconspicuous. Chimpanzee HC differed significantly from that of hominids with strongly contrasted eyes (humans, Sumatran orangutans and Western gorillas) but not from the ones of siamangs, dwarf gibbons, and hoolocks. As exemplified by our results on Western gorillas and white-handed gibbons, wildborn and captive-bred apes do not differ in ocular contrasts, making biases through imbalanced sampling of natural and captive populations unlikely. Still, it should be pointed out that our methodology does not capture the full extent of scleral pigmentation patterning. For instance, pronounced local scleral depigmentation will yield similar results to a fully depigmented sclera, despite obvious phenotypic differences and related effects on glance direction signaling (compare Fig. 2C). This constitutes an important limitation of our method, which is also insensitive to asymmetric expressions of pigmentation. Merging quantitative analyses with a qualitative scoring of pigmentation patterns (compare 7 could constitute a way of overcoming this limitation in future studies. Another potential shortcoming of our, as well as of all previous approaches so far, is that only differences in ocular contrasts but not in hues were quantified (see 33 ) or scored to approximate salience. This could have led to an underestimation of conspicuousness, particular in species with dark sclerae.
Relative iris luminance (RIL) was recovered as a trait that varied independent of phylogeny. Importantly, we could not find support for the assumption that low RILs are reliable indicators of more conspicuously colored eyes 8 , despite a moderate negative correlation of the two traits within our sample. The lowest average RIL value found was that of the Javan gibbon (Hylobates moloch, mean RIL: 33.5) which shows an amber-colored iris and dark brown sclera, rendering its eyes obviously far less conspicuous than the ones of, for instance, humans (mean RIL: 48.7), which were found to have a significantly higher RIL. This fact points out a major issue in the usage of RIL as a meaningful measure of ocular pigmentation. If gray values for sclera and iris are both low and differ little from one another, resulting RIL values may still equal or range below those obtained from eyes that show salient contrasts between these regions. This insensitivity makes RIL an unsuitable proxy for the conspicuousness of ocular pigmentation patterns. However, it might still be used to analyze intraspecific pigmentation patterns in groups with uniformly colored sclerae such as humans, for which the measure has originally been established 28 . Given that HC is a more faithful proxy of general ocular conspicuousness, we propose to rely on HC rather than on RIL to quantify ocular pigmentation in future comparative studies. For this, large sample sizes are recommended to counteract effects of differing lighting regimes in the analyzed pictures. Perea García et al. 8 have argued based on RIL that chimpanzee, bonobo and human eyes are equally suited to convey gaze signals. However, for reasons just pointed out, this argument does not hold. In ocular pigmentation, chimpanzees' dark eyes resemble the ones of gibbons more than the human or even the average bonobo condition, arguing against a human-like social signaling function of chimpanzee eye coloration (compare Fig. 4). Above that, scleral exposure in Pan is the lowest among African apes, differing significantly from both humans and gorillas in averted and direct glancing situations. For these reasons, chimpanzee eyes are notable for being, on average, the least conspicuous of all hominids. Low RIL values alone fail to diagnose species that employ glance cueing or even sophisticated gaze cueing, as exemplified by several gibbon species in our sample.
Our results highlight differences in scleral depigmentation rates in hominids compared to hylobatids. The dataset of Kobayashi and Kohshima 4 would suggest the dark-eyed gibbon pattern to be the plesiomorphic one. Therefore, more contrastingly colored eyes would be expected to have evolved in hominids after their split from hylobatids. However, this hypothesis can be challenged, given the small species sample sizes in that study together with the fact that its assumptions were later shown not to hold for most hominids [7][8][9] . From just our own experience, we can anecdotally report the presence of light sclerae in multiple species of Old World and New World monkeys (Suppl. Figure 4). Studies on the frequency and phylogenetic distribution of this trait in primates other than apes are necessary to sufficiently characterize the ancestral state for hominoid eye pigmentation.
It is difficult to discern what underlies patterns of ocular appearance among the two hominoid families. Considering the cooperative eye hypothesis, it might be tempting to suggest that brighter and more exposed eyes in great apes reflect the more sophisticated gaze following behavior in this group when compared to gibbons 23 . However, the hypothesis fails to explain the derived chimpanzee phenotype and cannot account for the great variability of hominid ocular contrasts. Furthermore, differences in SSI and WHR between hylobatids and hominids might be more parsimoniously explained by scaling effects deriving from differences in body size instead of by communicative demands 4 .
It is notable that each great ape genus encompasses species that markedly differ in their ocular pigmentation patterns (compare 8,9 ). Perea García et al. 8 suggested that scleral depigmentation in apes might be an evolutionary byproduct of greater social tolerance, induced by pleiotropic genes controlling neural crest development and www.nature.com/scientificreports/ mirroring patterns found in domesticated mammals. Indeed, the dark-eyed Bornean orangutans and Eastern gorillas are less tolerant towards unfamiliar conspecifics than their congeners in the wild (van Schaik 1999; Cooksey et al. 2020). Nevertheless, it remains to be clarified whether this is a consequence of intrinsic behavioral predispositions rather than extrinsic factors relating to habitat characteristics. Between bonobos and chimpanzees, such intrinsic predispositions are far better characterized than in other ape taxa and point to marked physiological differences underlying behavioral disparities within the genus Pan 34,35 . Comparisons of socially tolerant and despotic monkey species could further test for a correlation between scleral pigmentation and aggressiveness. Yet, specific pleiotropic effects affecting scleral but not general skin or fur coloration appear to be yet undescribed in mammals (compare e.g., 15 ), making the link a speculative one at the moment. It is also unclear how the social tolerance hypothesis might apply to the hominid family in general when compared to hylobatids or specifically to white-handed gibbons. Although the latter show the highest scleral depigmentation rates among small apes, there is no evidence to suggest them to exhibit decreased levels of aggression towards unfamiliar conspecifics when compared to other species. To sum up, the cooperative eye hypothesis might fit the family-level patterns we describe but loses its explanatory power at the species level. It is further important to point out that the human eye is on average not more saliently contrasted than that of gorillas, bonobos and Sumatran orangutans, further challenging its validity. The occurrence of at least locally depigmented sclerae in varying portions of the total population likely is an ancient hominid trait but its biological significance remains obscure. The view that extensive scleral depigmentation is exclusive to the human lineage 4 or that it might even be diagnostic for the species Homo sapiens 6 cannot be maintained. Finally, reliance on RIL as an indicator of ocular conspicuousness may give rise to misleading results.

Is ocular pigmentation linked to specific socio-cognitive traits?
Previous research on scleral pigmentation in primates has highlighted a potential connection between cognitive traits such as social glance cueing and conspicuous ocular pigmentations 3,4,6,8 . Whether these characters are correlated is not yet clear, however, because humans are the only primate species that evidently combines them (see below). Deducing glance cueing and other cognitive abilities from eye coloration alone may quickly lead into an adaptationist pitfall. Even if primate species converge in ocular pigmentation, the ways in which these species perceive conspecifics' eyes may differ. For example, Western gorillas and Sumatran orangutans exhibit ocular contrasts resembling the human condition. Yet, their viewing patterns of conspecific's faces and particularly eyes, is more reminiscent of chimpanzees than humans, they exhibit pronounced gaze avoidance in diverse social contexts and evidence of conspecific glance cue exploitation is, to our knowledge, absent 11,36 .
The hypothesis that dark eyes evolved to mask glance direction in competitive social environments must be critically reevaluated as well. Kobayashi and Kohshima 4 proposed that all non-human primates would exhibit dark sclerae to benefit from "gaze camouflage", but this idea remains hypothetical rather than empirical. This assumption is not based on any experimental evidence and does not address how this trait would be advantageous to primates with low SSI and across the extreme diversity of social, activity, and foraging regimes that the 91 species they studied encompass. In line with this, the additional notion that dark sclerae would lower predation risk via gaze concealment 1,4 is purely speculative as well. Kobayashi and Kohshima 4 also ignore the relevance of other facial ornaments for gaze/glance communication (compare 11,37 and the great variety of cooperative behaviors in non-human primates that are often linked to gaze following (compare 38 ). Dark sclerae appear to be a plesiomorphic primate trait 4 , and disentangling its evolutionary roots will require accounting for other mammals or even more distantly related vertebrate groups. We therefore discard both the glance cueing and cryptic gaze hypotheses of eye coloration as too simplistic to be helpful in interpreting the evolution of ocular pigmentation in primates. Our dataset could be expanded to test, whether ocular morphology indeed correlates with specific cognitive traits across primate groups. For this, however, there must be agreement on the definitions of the cognitive characteristics studied. Currently, this is not the case for primate gaze or glance cueing.
So far, most information on gaze and glance cue understanding in apes derive from studies in which human experimenters signal to an ape subject in a captive setting 3,14,21,22 . It is important to note that responsiveness to human eye orientation by habituated animals does not equate with the usage of glance cues among conspecifics, as for instance the successful exploitation of human glances by Californian sea lions (Zalophus californianus) demonstrates 39 . The evidence that apes utilize conspecifics' glancing independent from head orientation to inform their actions is meager. Many studies assume that apes' head direction and eye orientation align with each other for the most part, despite evidence of the contrary 12 . Equating head and eye direction in interpretations of glancing behavior has at times led to confusion. For example, Perea García et al. 8 cite studies that either did not differentiate between the two 40,41 or that approximated gaze by head orientation alone 42 to support the notion that glance cues are relevant to chimpanzee communication. We are not aware of studies that unequivocally show exploitation of conspecific glance cues in any ape species within a social context. An effect of ocular contrast on such behaviors would need to be demonstrated further. Because humans can reliably deduce the glance direction of chimpanzees, it can be hypothesized that from a perceptual perspective, conspecifics might do so as well, despite their cryptically colored eyes 12 . The additional effect of an increased iridoscleral contrast on gaze salience might well be comparatively minor. On a different note, it should be discussed how exactly the inclusion of glance cues could enhance apes' communicative repertoire in the wild. What referential information could glance cues convey in a naturalistic setting that head orientation cannot? In the absence of evidence for conspecific glance cueing in great and small apes or an unambiguous link between ocular pigmentation and cognitive traits relating to gaze/glance following, as well as for reasons of parsimony, we assume that these characters evolve independently from one another. www.nature.com/scientificreports/ Which factors underly the pigmentation of the human eye? The evolutionary trend of scleral depigmentation in hominids finds its strongest expression in humans. Although rudimentary pigmentation of the conjunctiva and inconspicuous scleral spots can frequently be found in humans 43 , apparent complete scleral depigmentation approaches 100% in our species 4 . Why do humans, but not other apes, exhibit such a uniformly white scleral phenotype? The assumption that this trait evolved to facilitate glance or gaze cueing is problematic. First, as already pointed out, its occurrence among mammals that do or do not show sophisticated gaze following or glance cueing has not been sufficiently investigated. Above that, available evidence suggests that humans can reliably assess the glance direction of chimpanzees from a distance of 2-10 m 12 as well as that of human models with artificially modified scleral colors 44 . Although it took naïve participants significantly longer to assess the glance direction of the latter compared to natural human eyes, the time differences only encompassed fractions of a second and no differences in the accuracy of deducing glance direction from normal human models and those with matched iris and scleral colors were found 44 . The relevance of the depigmented sclera for human communication might therefore be overstated, especially when compared to other morphological (e.g., the widened horizontal outline of the visible eyeball) or physiological ocular traits of our species (e.g., emotional tearing, particularly in infants), which are not shared by other hominoids 7,45 ). What requires an explanation is perhaps not that the human sclera simply exhibits depigmentation but that its expression is uniformly extreme across individuals and populations. The marked variability of scleral color in other hominid genera, which includes complete scleral depigmentation 7 , makes a similar phenotypic diversity in human ancestors appear likely. Following that, the human pattern does not necessarily require an adaptive explanation but may simply result from genetic drift acting on ancestral trait variability. It is also reasonable to assume that sexual selection has contributed to the evolution of human eye pigmentation, not excluding but possibly complementing effects of genetic drift. As we have shown, other apes do not only exhibit strong interindividual differences in scleral pigmentation but at times also asymmetric scleral coloration, particularly gorillas (compare also 7 ). Both may point to a relaxed evolutionary pressure on ocular appearance in great apes compared to humans.
It has been demonstrated that scleral brightness strongly affects the attractiveness of human eyes and that it can act as an indicator of individual age 46,47 . A negative effect of age on scleral brightness has also been shown for chimpanzees and bonobos 8 . Thereby, uniformly light, salient sclerae should contribute to a juvenilized appearance of the face, which complies to general sexual selection pressures for neotenic facial traits in humans compared to other hominids, particularly in females 6,[48][49][50] . Similarly, the symmetrical scleral depigmentation pattern in our species is in line with that human facial attractiveness is generally enhanced by increased symmetry 51 . Light sclerae in humans could take over signaling functions that are absent in other apes, for example because they pay less attention to or even prefer to avoid gazing into each other's eyes, e.g., gorillas and orangutans 11,36 ; or because correlates of scleral brightness such as youthfulness do not increase sexual attractiveness in these species, e.g., chimpanzees 52 . Distinct selection pressure on ocular appearance in humans is also indicated by iris color variability. Although there is at least one other primate genus, spider monkeys, in which two distinct iris color morphs co-occur in the adults of some species (Ateles fusciceps and A. hybridus 53 , A. paniscus-personal observation), humans appear to be unique among non-domesticated mammals in the diversity of iridal hues found in the global population, particularly Western Asia and Europe 54 . Other apes display uniform, speciesspecific iridal coloration. Given that humans exhibit notable eye color preferences in the context of mate choice 55 , it can be assumed that iris color represents a sexually selected ocular trait that differentiates our species from non-human apes, just as it might be the case with the plain white sclera.

Conclusion
Our data add to growing evidence suggesting a graded evolution of hominoid ocular coloration instead of a clear dichotomy between human and non-human primate eyes. Still, the evolutionary drivers behind the recovered trend of scleral depigmentation in hominids, which peaked in humans and reversed in chimpanzees, remain unidentified. However, the great intra-and interspecific variability of hominoid eye pigmentation opens possibilities for further comparative research which should include both the great apes and gibbons as well as a range of outgroup taxa. The classic cooperative eye hypothesis that proposes an evolutionary link between primate ocular morphology and social signaling, needs to be experimentally revisited and scrutinized. In order to uphold it, a clear relevance to ocular pigmentation traits for communication among conspecifics in nonhuman primates must be demonstrated.