Unique human orbital morphology compared with that of apes

Humans’ and apes’ convergent (front-facing) orbits allow a large overlap of monocular visual fields but are considered to limit the lateral visual field extent. However, humans can greatly expand their lateral visual fields using eye motion. This study aimed to assess whether the human orbital morphology was unique compared with that of apes in avoiding lateral visual field obstruction. The orbits of 100 human skulls and 120 ape skulls (30 gibbons; 30 orangutans; 30 gorillas; 30 chimpanzees and bonobos) were analyzed. The orbital width/height ratio was calculated. Two orbital angles representing orbital convergence and rearward position of the orbital margin respectively were recorded using a protractor and laser levels. Humans have the largest orbital width/height ratio (1.19; p < 0.001). Humans and gibbons have orbits which are significantly less convergent than those of chimpanzees / bonobos, gorillas and orangutans (p < 0.001). These elements suggest a morphology favoring lateral vision in humans. More specifically, the human orbit has a uniquely rearward temporal orbital margin (107.1°; p < 0.001), suitable for avoiding visual obstruction and promoting lateral visual field expansion through eye motion. Such an orbital morphology may have evolved mainly as an adaptation to open-country habitat and bipedal locomotion.

The group comprised of Hylobatidae (CA = 99.2°) and humans (CA = 98.1°) had significantly (p < 0.001) less convergent orbits than the group comprised of Gorilla, Pan and Pongo (Fig. 3, Table 1). In humans, we measured a 107.1° OA and calculated a 9° OA-CA. Both of these values were significantly higher than those of the non-human apes (p < 0.001) (Fig. 3, Table 1). A zero OA-CA value would denote no discrepancy between OA and CA and would imply that the orbital margin is flat and therefore inscribed in a plane. The highest OA-CA value recorded in humans denoted a high discrepancy between CA and OA and showed that the human orbital margin fell short of being inscribed in a plane. As illustrated in Fig. 4, a flat lid cannot therefore close the human orbital margin but comes closer to closing the orbital margins of non-human apes who have a lower OA-CA.
The orbital W/H ratio and OA-CA were supposed to promote lateral visual field extension. We therefore wanted to know if these factors were correlated. A positive correlation was found between the orbital W/H ratio and orbital OA-CA values (0.64 [0.58; 0.69]: p < 0.001).
For the different human populations (Fig. 5), the OA of skulls from Europe was significantly greater (p values range: <2 × 10 −16 to 0.012) than that of any other population except Australians. The OA of skulls from China was significantly lower (p values range: <2 × 10 −16 to <9.2 × 10 −6 ) than that of any population. Moreover, the CA of the populations group comprised of skulls from the Republic of the Congo, Europe and North America was significantly higher (p values range: 1.9 × 10 −7 to 0.0012) than that of the populations group comprised of skulls from Australia and China. The OA-CA was significantly greater (p values range: 3.8 × 10 −14 to 1.1 × 10 −8 ) in skulls from Australia than in any other population.
Gender influence on evaluation of the measured parameters. No sexual dimorphism was noted for the CA. Female OAs were, on average, 0.4° larger than male OAs. This difference was less than our measurement error (see Methods) and was therefore considered to be insignificant. For the W/H ratio, in humans, male ratios were, on average, 0.04 larger than those of females. This difference was greater than our measurement error and was therefore considered to be significant (p < 0.001). Because the human sex ratio is 1.31 (Table 2), the higher W/H found in human males could potentially bias the result indicating that the human W/H ratio is significantly larger than that of non-human apes. However, the multi-way analysis of variance followed by Tukey's range test showed that the higher W/H ratio found in human males did not change the fact that the human W/H ratio was significantly larger (p < 0.001) than that of non-human apes.

Discussion
The human palpebral fissure width/height (W/H) ratio is highest among primates 16,17 . Besides the width of the palpebral fissure, the underlying orbital morphology could greatly influence the "eye-motion visual field" (EMVF 14,15 ). The convergent human orbit, assumed to provide limited lateral visual field   [4][5][6]13 , in fact allows for significant lateral visual field expansion through eye abduction 14,15 . The comparison of human and ape orbital morphology was undertaken to appreciate factors likely to promote or limit EMVF. To the best of our knowledge, our study is the largest comparative series analyzing human and non-human apes orbits 4,7 (Table 3) to date. However, one of the limitations of this study was the difficulty in obtaining a balanced sex ratio. Female skulls were especially difficult to find for gorillas.
As originally recommended by Broca himself 18 , we used the neuro-ocular plane (NOP) as a reference plane to orient skulls and take all the angular measurements in order to ensure effective and precise comparisons between modern humans and other primates 19,20 . We used the exact same anatomical landmarks as Broca to materialize the NOP i.e. on both sides the center of the optic foramina and a point equidistant from the superior and inferior orbital margins 18 . To quote Cartmill 1 "Visual convergence is not a simple univariate quantity. The orientation of the apparatus of vision can be estimated by several different procedures non two of which yield uniformly comparable results. " To measure the convergence angle (CA) we used purposedly the NOP as a reference plane and not, as originally recommended by Cartmill a reference plane perpendicular to both the mid-sagittal plane and to the orbital plane. In humans and apes, the visual pathways and the NOP are transverse, and perpendicular the body axis 19 . Hence, in the orbit, the NOP provides a meridional cut through the globe and through the horizontal recti muscles 19 . It can therefore be considered, as Strait and Ross 21 did for the horizontal recti muscle, that the NOP is approximately perpendicular to the orbital plane. Moreover, the NOP is perpendicular to the mid-sagittal plane. As a reference plane to record the CA, the NOP is therefore close to the reference plane (i.e. perpendicular to both the mid-sagittal plane and to the orbital plane) initially defined by Cartmill 1 . Besides, as discussed later, using the NOP as a reference plane allows meaningful discussion regarding eye motion influence on visual field eccentricity and regarding published anatomical about eyeball position into the orbit in humans and apes.
To record convergence angle (CA) values, which vary in the same direction as those of temporal visual field eccentricity, we measured the CA between the orbital plane (OP) and the part of the sagittal plane which is anterior rather than posterior to the orbital margin, as had been done in two previously published studies 4,7 . Our study yielded CA values close to previously published ones 4,7 (Table 3) with the advantage of having data from many more skulls. The opening angle (OA) was assessed to study how rearward the lateral orbital margin position was. In 1939 Winckler reported measurements of the "inclination of orbit entry relative to the frontal plane" in modern human skulls. This measurement, whose closest equivalent in our study is OA, indicated in his words "the degree to which the outer orbital rim is more rearward than the inner orbital rim" 22 . However he gave no indications about the skulls' orientation relative to the frontal plane whereas the skulls were oriented using the NOP as a reference plane in our study. The human orbital base has been described as flat with its margin inscribed in a plane 1,4,8 . The high discrepancy between OA and CA, which is unique to humans and denoted by OA-CA, proves that these notions are erroneous. Hence, far from being flat, the human orbital margin is warped so that its outer part is much more rearward than if it was simply a continuation of the inner part. In Fig. 4 (side view), this complex three-dimensional curvature accounts for the pronounced forward concavity of the human lateral orbital margin 23,24 . A black cardboard lid (representing the convergence angle) has been placed in the right orbital plane of each skull. In humans, the lateral orbital margin is far more rearward than the lid (a') because of the high discrepancy between convergence and opening angles. In non-human apes, this discrepancy is lower and the cap is much closer to the lateral orbital margin (b' ,c' ,d' ,e'). The specimens are from the Natural History National Museum (Paris, France) anthropological collections (a/a') and Comparative Anatomy Laboratory (b/b' to e/e').
Scientific RepoRts | 5:11528 | DOi: 10.1038/srep11528 The term "frontal" does not describe properly orbital orientation because if often conflates orbit convergence and orbit frontation 4 . Frontation is, to quote Cartmill, "a measure of the extent to which the two orbits face forward toward the end of the snout (as in ape or man) rather than upward toward the skull roof (as in a crocodile)". Humans and non-human apes have therefore similarly high frontation 1 . Furthermore, frontation is not expected to be correlated with a horizontal parameter like the maximum horizontal extent of overlap of the two monocular visual fields 4 . Our study also deals with a horizontal parameter, i.e. with the influence of orbital morphology on horizontal visual field in humans and non-humans apes. We therefore believe, as previously recommended 4 , that reporting orbital convergence but not frontation is acceptable in our study.
The issue of orbital proportions has traditionally been approached with Broca's orbital index, that is "(Height of orbit × 100) / Width of orbit" [23][24][25] . In humans, this index yields figures lower than 100, which conveys the idea that the human orbit is shorter (less tall) than it is wide. In this study on the contrary, the orbital W/H ratio yielded figures higher than 1 in humans, which conveys the idea that the human orbit is wider than it is high. This approach makes it easy to parallel Kobayashi and Kohshima's finding that modern humans have the largest W/H palpebral fissure ratio of all primates 16,17 . The elongated human palpebral fissure, believed to allow visual field expansion through ample eye movement 16,17 appears to have an underlying orbital margin whose unique morphology serves the same purpose. Interestingly, we found that the orbital W/H ratio was significantly higher in Aboriginal Australians skulls than in all the other human sub-groups as reported by Winckler 22 .
In accordance with previously published reports 4,7 , we found that human and Hylobatidae orbits were significantly less frontal (higher CA) than those of the group comprised of Gorilla, Pongo and Pan (p < 0.001). This finding, combined with the fact that the human orbit has the largest W/H ratio compared with that of non-human apes, suggests a unique morphology favoring lateral vision. More specifically, while convergent, the human orbit has by far the largest OA and OA-CA values (p < 0.001), reflecting the fact that the lateral orbital margin is far more rearward in humans than in non-human apes       (Fig. 4). This finding is not obvious from frontal orbital views. This must account for our finding that human orbital morphology differs from that of non-human apes, whereas a general closeness of Pan and human orbits using Fourier analysis of the orbital margin based on frontal photographs was reported 25 . Using his aforementioned system, Winckler reported an "inclination of the orbit entry relative to the frontal plane", with values between 15 and 17.5°2 2 . These values, that would be 105° and 107.5° relative to the sagittal plane, are close to those recorded in our study (mean: 107.1° ± 0.245).
In the NOP, an average 65.7% of the human eyeball length is located in front of the line joining both lateral orbital margins (external bicanthal line) 26 . For this reason, plainly relating eye volume to orbit volume in humans (32% for Schultz 27 ) leads to an overestimation of the orbit volume fraction the eye really occupies (20% for Bron et al. 28 ). In non-human apes, the eyeball is much less protruding than in human. In one Hylobates lar (gibbon) and one Pan troglodytes (chimpanzee), the eyeball portion located in front of the external bicanthal line using computed tomography cross-sections in the NOP was 11.11% 29 and 34.78% 30 respectively. Schultz has reported that, in chimpanzees, the eye lies deep in the orbit which protects the eye amply on all sides and that in orangutans the eyeball reaches slightly beyond the orbital margin 27 . Any photograph of a living gibbon or gorilla [31][32][33] shows that their eyeball is deeply tucked into their orbits as well. The human eyeball position in the orbit, much further forward than that of non-human apes, must increase the efficiency of the human lateral orbital margin, which is much more rearward than that of non-human apes, in avoiding lateral visual obstruction, especially when the eyeball is moved outward 15 . This point is relevant, Land having reported that the maximum excursion of eyes in humans and monkeys is similar (+/−45°) 34 . It should be mentioned that a significant (p = 0.013) positive correlation between the forward position of the human eyeball in the orbit and the degree of visual field gained through eye motion in the temporal quadrant has been found 15 . In this study, overall, the orbital W/H ratio and OA-CA values were positively correlated which suggests that orbital horizontal elongation and double orientation serve the same purpose in increasing the extent of the lateral visual field. We have no valid explanations to account for the inter-continental variability of OA, CA and OA-CA observed in our study. We do not relate them simply to global knowledge of anthropometric morphology data, as an increase in nose breadth from north to south, or an increase in bi-zigomatic breadth from west to east 35 .
In everyday life, the eyes rarely remain in the primary position of gaze 14,15,36 and we have recently shown that in humans, the "base visual field" (BVF), i.e the visual field with the eyes pointing straight ahead, is much less extensive than the EMVF, i.e. the visual field including eye motion 15 . This is especially true in the temporal quadrant where the median EMVF surface area is 46% more than that of the median BVF. These results are in accordance with reports that humans readily use eyeball rather than head movements to scan their environment 6,[14][15][16][17]36,37 . These results also suggest that, as previously hypothesized 15 , humans may be unique among mammals in combining overlapping monocular visual fields and, through eye motion, large (enlarged) lateral visual fields. This issue is tackled in much greater depth in another report (submitted).
The human rearward lateral orbital margin (RLOM) could be a by-product of selection for other aspects of craniofacial anatomy; in other words, an exaptation. Loss of the snout with posterior facial retraction below the anterior cranial fossa in modern humans may be such an aspect 38 . This retraction may have positioned the zygomatic bone, and hence the lateral orbital margin, more posteriorly. However, compared with archaic Homo species and the non-human apes, modern humans have a very spherical (globular) cranial vault with a frontal squama that rises steeply above the orbits 39,40 and this may have driven the upper lateral orbital margin forward. Compared with non-human apes, modern humans eat soft, highly processed foods and do not spend much time chewing 38,41 . Accordingly, modern humans have masticatory muscles that are much less developed than those of non-human apes 38 . In anthropoid primates, the line of action of the anterior temporalis muscle is roughly vertical 42 . In humans, posterior facial retraction has resulted in a more posteriorly placed anterior temporalis muscle, with a line of action which is expected to have more of an antero-posterior component than is found in non-human apes. However, not much stress transfers to the upper face, including the postorbital septum, during chewing, in anthropoid primates including humans 38,43 . Furthermore, assuming that there is more antero-posterior strain on the postorbital septum in humans than in the non-human apes, the expected response would be to add bony mass 38,44 in the zone under strain, the result of which would not be expected to change the position of the lateral orbital margin. Finally, the human anterior temporalis muscle, which is proportionally thinner than that of non-human apes, may provide less support for the postorbital septum. However, the influence of this factor on the position of the human RLOM position is unsubstantiated.
To quote Lieberman 45 "heads defy many efforts to simplify because they are, by nature, complex and highly integrated systems". The RLOM likely represents a compromise among many factors including the demands of the temporal fossa content and the demands of the orbit. Apart from exaptation, there are good reasons to think that natural selection has driven the evolution of a RLOM in humans. The human RLOM does not offer much lateral eyeball protection, which may have had little negative selective pressure in humans. Indeed, humans live in a branchless environment with much less risk of branch-related eyeball trauma than the non-human apes who almost exclusively inhabit tropical forests [31][32][33]46,47 . Of course, many prosimians (e.g. lemurs) presumably exposed to as much risk of branch-related eyeball trauma as non-human apes, lead a successful arboreal existence without a deep positioning of their eyeballs in their orbits. The point here is not that a protective lateral orbital margin is mandatory for safe arboreal existence but that it may be for the non-human apes, may have been for early hominins and is no longer Scientific RepoRts | 5:11528 | DOi: 10.1038/srep11528 in the case of modern humans. Apart from eyeball trauma, the human RLOM and anterior eyeball position in the orbit may result in UVB-related eyeball conditions like pterygium or cataracts developed in the nasal aspect of the crystalline lens 48,49 . However, these conditions are usually non-blinding 48 . Hence, far from being a "design fault" in the human visual system 48 , the RLOM position and anterior eyeball position in the orbit may represent a trade-off between usually non-blinding UVB-related eyeball conditions and a large visual field 48 , enlarged through eye motion 15,16,37 , which may aid survival 48 . Humans are ground-dwellers 31 , live in more open spaces than tropical forests 41,50,51 and, being the only habitual mammalian bipeds 31,50,52,53 , have most of their visual targets at or parallel to ground level. Compared to knuckle-walking, human bipedal locomotion involves a higher head position and a more forward-facing orbital plane orientation relative to the frontal plane 21 . This overlooking, forward-facing orbital position comes in handy in humans whose large, heavy heads are -according to Galileo's principle of similitude 17,54,55 -much more difficult to move than those of smaller primates. In primates, the eye scales with greater negative allometry with respect to body mass than the orbit does 27,56,57 . The eyes of large primates (e.g. humans) therefore fill proportionally less orbital volume that the eyes of small primates 56,57 . Large primates therefore have proportionally more orbital space for oculomotor muscles 17 . Owing to Galileo's principle of similitude 17,54,55 , this favors swift and ample eyeball movements in large primates, especially humans 6,14,16,36,58,59 . Set in their overlooking anatomical position, the human eyes may thus efficiently scan their environment, mostly at or parallel to ground level. This process, very useful in challenging environments 37 , saves head movements, increases spatial awareness and vigilance through visual and visual field exploration, the RLOM avoiding obstruction of the EMVF 14,15 .

Study material.
A total of 220 skulls (440 orbits) were studied at the Natural History National Museum and its Comparative Anatomy Laboratory (Paris, France), at the Museum of Natural Sciences (Brussels, Belgium) and the Royal Museum for Central Africa (Tervuren, Belgium). These included 30 Hylobatidae (gibbons) skulls, 30 Pongo (orangutan) skulls, 30 Gorilla skulls, 30 Pan (chimpanzees and bonobos) skulls and 100 Homo sapiens (human) skulls (Table 2).
Orbital width, height and width/height ratio. For both orbits of each skull, maximum orbit width and height were recorded using a thin, metal ruler (model "1051.02", Facom, New Britain, CT, USA) or calipers (model "815A", Facom, New Britain, CT, USA) when, owing to rounded orbital margins, orbital points were out of reach of the flat ruler. The orbital width/height (W/H) ratio was then calculated.
Orbital angles definitions. The Neuro-Ocular Plane (NOP) is the plane that, in the primary gaze position (looking straight ahead into the distance), contains the center of both crystalline lenses, optic discs, and optic foramina 19,20 . In the primary gaze position, the pupil is equidistant from the superior and inferior orbital margins 18,28 . We therefore defined the NOP, as Paul Broca did in 1873 18 , as the plane which runs symmetrically through both optic foramina and through a point located mid-way between the highest and lowest points of the orbital margin.
Orbital convergence was evaluated with the convergence angle (CA), which is commonly used for that purpose 1,4,5,7,9,13 . The more or less rearward position of the temporal orbital margin was evaluated by the opening angle (OA). The CA and OA were measured in the NOP, which allows reproducible head orientation in space, and therefore can be used for reliable comparison between humans and other species 20 .
CA measurement requires the placement of 3 orbital margin points which together define the orbital plane (OP): the highest, lowest and foremost points 1 , respectively denoted by s, i and a. The CA is then defined as the angle -measured in the NOP -between the OP and the midsagittal plane (Fig. 6).
OA measurement requires the placement of point a' defined as the intersection of the NOP and the inner orbital margin. The OA is then defined as the widest angle allowing a view of a' when looking postero-laterally at the orbit in the NOP (Fig. 6).
Orbital angles measurement method. A cross-laser (reference "253003", Magnusson, Longpont sur Orge, France) mounted on a flexible camera tripod (GorillaPod SLR zoom, Joby, San Francisco, USA) was used to materialize the sagittal plane and the NOP. A wooden protractor prototype was mounted on a flexible camera tripod (GorillaPod SLR, Joby, San Francisco, USA), allowing proper positioning of the protractor in the NOP.
The position of point a was determined using a line laser (reference "GPLL5", Bosch, USA) mounted on a camera tripod (reference "AK-2", Vanguard, Canton, China). Using a set square, the line laser was placed perpendicular to both the NOP and the sagittal plane. Keeping this orientation constant, the line laser was displaced antero-posteriorly until it crossed the inner orbital margin. The crossing point was denoted by a (Fig. 6). The OP was materialized using the line laser mounted on its camera tripod and crossing points i, s and a. In some skulls, bony interference with the line laser made proper alignment of these points impossible. In such cases, the OP position was determined by trial and error, aligning two of the three points and using the thin metal ruler placed in the plane of the line laser to reach the third point. In Fig. 6, points a and a' share the same location. When this was not the case, the protractor was centered on point a, in a plane parallel to the NOP, to record the CA. The lower the CA value, the more convergent the orbital margin orientation.
OA measurement required the use of the line laser mounted on its camera tripod and placed perpendicular to both the NOP and the sagittal plane. Keeping this orientation constant, the line laser, whose source was placed in the NOP, was displaced gradually in a postero-lateral position. The most postero-lateral position allowing point a' illumination defined the opening angle. The higher the OA, the more rearward the temporal orbital margin position. Statistical analysis. Statistical analysis was performed using R (http://www.r-project.org/). Orbital W/H ratio, OA, CA and OA-CA values were compared using a Student test and with correction of the type I error using the Holm-Bonferroni method. Pairwise comparisons between each pair of taxa (humans, Gorilla, Hylobatidae, Pan, Pongo) and between each of the different human populations (from Africa, Australia, China, Europe and North America) were conducted. To be considered significant, the observed values differences had to be strictly superior to the measurement error (see below) and have a p-value < 0.05.
Pearson's product-moment correlation was used to assess whether orbital W/H and OA-CA values were positively correlated, for all genera. This correlation yields a coefficient with a value varying between − 1 and 1. A zero coefficient value indicates no correlation. The higher or lower the coefficient value, the higher the positive or negative correlation.
Differences between right and left measurements on OA, CA and W/H ratio were measured using a Student test for paired values. For CA, the difference (mean [95% confidence interval]) was 0.47° [0.04 : 0.9]; For OA, the difference was 0.35° [0.05 : 0.76]; For orbital width the difference was 0.04 mm [0 : 0.16]; For orbital height, the difference was 0.14 mm [0.02 : 0.26]. For all these parameters, the upper limit of the 95% confidence interval was below our measurement error. We therefore chose to consider that the right and left measurements were the equivalent of one value measured twice.
Gender influences on OA, CA and the W/H ratio were measured using multi-way analysis of variance. Tukey's range test was used as a post hoc test to compare Gorilla, Homo, Hylobatidae, Pan and Pongo with one another (using pairwise comparisons between each pair of taxa).

Measurement error evaluation.
Twenty skulls (2 females skulls and 2 males skulls, i.e. a total of 4 skulls for Hylobatidae, Pongo, Gorilla, Pan and Homo) were chosen at random. For each skull, the CA and OA of both orbits that had already been measured were measured a second time 11 to 25 months later. We called m1 the value measured the first time; m2 the value measured the second time and m3 the average of m1 and m2. We calculated the absolute value of the difference between m3 and m1, for each of the 40 measurements. The measurement error was estimated by taking the upper value of the 95% confidence interval. The same principle was applied to angle difference (AO-AC), orbital width, orbital height Figure 6. Overview of the convergence angle and opening angle. The sagittal plane and the neuro-ocular plane (NOP: yellow) are represented by a cross-laser. The protractor is positioned in the NOP. The most anterior point of the orbital margin is denoted by a. The point where the NOP crosses the inner orbital margin is denoted by a'. In this skull, points a and a' coincide. The highest and lowest orbital points are denoted by s and i respectively. The orbital plane (OP) crosses points i, s and a. Its position is shown using a red line for laser crossing points i, s and a. The para-sagittal plane (PSP) is parallel to the sagittal plane and crosses points a and a'. The CA is the angle between OP and PSP, measured using the protractor. The OA is the maximum angle allowing point a' to be illuminated by the line laser placed postero-laterally in the NOP. The difference between the CA and the OA is denoted by "OA-CA".