The Skull of Tarsius: Functional Morphology, Eyeballs, and the Nonpursuit Predatory Lifestyle
- First Online:
- Cite this article as:
- Rosenberger, A.L. Int J Primatol (2010) 31: 1032. doi:10.1007/s10764-010-9447-x
- 346 Views
Little is known about the impact of enormous eyeballs on the tarsier's head, apart from facial morphology. I used a biomechanical analysis to compare the cranium of Tarsius with the Eocene fossil Necrolemur, a moderately large-eyed surrogate for ancestral tarsiid cranial morphology. Eyeball hypertrophy has radically influenced the neurocranium and basicranium, driving the evolution of such derived features as recession of orbital fossae, ectopically located eyeballs, uptilted brain and rounded braincase, anteroventrally shifted foramen magnum, enlarged and horizontally leveled nuchal plane, laterally displaced and narrowed tympanic cavities, and shortened external auditory tubes. The gestalt is an adaptation to efficient orthograde head carriage, balanced head-turning movements, and spatial packaging of cranial components, responses to an extreme loading regimen in which the eyes, with a mass approximating twice the bulk of the brain, profoundly eccentrically load the skull. Specializations of the retina and cortex suggest tarsiers have an acutely developed spatial sense, especially adept at detecting and mapping motion. Spanning several anatomical systems, this configuration represents an extreme form of vertical clinging and leaping (XVCL) geared for noiseless, nonpursuit predation, an energy-minimizing procurement strategy that may be a trade-off for relying on metabolically expensive, outsized eyeballs, maintained by a highly nutritious, super-specialized, animalivorous food source. A more varied galago-like locomotor profile and foraging habit was common among fossil tarsiiforms and preadaptive to this lifestyle, partly by canalizing the forward orientation of the tarsier's predatory gaze in VCL mode. Tarsier ecomorphology evolved to minimize the costs of being extraordinarily “top heavy,” carrying a heavy load that is roughly equivalent to 3 brains.
The approach used in the first part of the article builds on modeling biomechanics and anatomical transformations based on accurate lateral view illustrations of skulls as well as direct study of the morphology. Theoretically modeling cranial functional morphology is a time-tested research strategy that has yielded numerous insights into the evolution of the mammalian skull. Weishampel (1993) gives many examples, commenting (p. 338) that “….modeling approaches yield both accurate and heuristic information about the operation of vertebrate skulls.” Plotnick and Baumiller (2000) refer to the application of these approaches to fossils as “paleobiomechanics,” wherein an analysis strives to identify what a structural design is capable of doing efficiently, according to mechanical principles. One of the most influential studies of primates utilizing a comparable method was Cartmill’s (1974) comparative functional analysis of aye-aye facial morphology.
I used Necrolemur as an exemplar of a more primitive tarsiiform skull from which a tarsier-like pattern can be derived. Its orbits are clearly not as enlarged as in Tarsius (Martin 1990), and the only evident cranial feature that appears to be autapomophic relative to tarsiers is not material to the analysis, the hypertrophic petromastoid region (Szalay 1976). There is phylogenetic justification that makes Necrolemur appropriate for this purpose as well for, while its precise affinities are still unresolved, the fossil shares a host of homologous derived features with Tarsius in the skull and postcranium (Beard and MacPhee 1994; Beard et al.1991; Dagosto et al.1999; Rosenberger 1985, and references cited therein) which, I would argue (Rosenberger et al.2008), justifies placing it within a monophyletic tarsiid clade that includes tarsiers and a number of Eocene genera frequently allocated to Omomyidae (including Microchoerinae).
Materials and Methods
Measurements of the moment arms, dimensionless variables, used in each of the free body diagram models and of mass estimates, in cubic millimeters, of the brain and eyeball
Major features and source variables used in the models
Actual tarsiers, wet specimens
Actual tarsiers, osteology specimens
Actual tarsiers, wet, with minimal eye size
Actual tarsiers, wet, with maximal eye size
Necrolemur at 2/3 tarsier eye size
Necrolemur with minimal brain size
Necrolemur with minimal eye size
Necrolemur with minimal eye and brain size
Necrolemur with tarsier eye and brain size
Necrolemur with tarsier eye and average Necrolemur brain size
For each, the moment arms represent dimensionless units, calculated by taking the length of line segments, measured in Photoshop, and converting each to a ratio against maximum skull length × 100. The center of mass of the brain was calculated via Image J, by finding the geometric center of its outline. The center of mass of the eye was taken as the center of a sphere whose fit into the orbital fossae was approximated by eye. The center of rotation of the skull was assumed to be the midpoint of the plane of the foramen magnum, for convenience and because it was deemed a reliably fixable point on Radinsky’s Necrolemur illustration. The placement of the vector representing the nuchal muscles is simply at the midpoint of the nuchal plane, with a perpendicular drawn from the plane to approximate their line of action.
To compensate for the fact that our knowledge of Necrolemur remains incomplete and must be conjectural, and because the parameters describing eye size and brain size in Tarsius varies in the literature owing to differences in the measurement techniques used, to individual and ontogenetic variability, and to taxonomic differences among the samples, I provide several permutations of the models describing both genera. Several come directly from weights of fluid preserved specimens, others from osteological measurements. For comparability, mass values are presented in cubic centimeters, converted from diameter measurements in cases where eye size was determined by the span of the external orbital aperture. In addition, I assume the eyeballs of tarsiers and Necrolemur are spherical, even though in tarsiers they are anteroposteriorly elongate and quasi-tubular (Castenholtz 1984).
- Model 1
Fe and Fb, using the average volumes of 3 individuals representing 2 species (Spatz 1968), directly measured from wet specimens.
- Model 2
- Models 3
and 4 Fe, with volumes calculated from minimum and maximum diameters of the eyeball (Collins et al. 2005).
- Model 1
- Model 2
- Model 3
Fe, with eye volume calculated from the orbital diameters of 4 skulls (Kay and Kirk 2000).
- Model 5
Fe and Fb assume an eye and brain volume equivalent to a tarsier, using Tarsius Model 1 values.
- Model 6
Fe and Fb assume a tarsier eye volume and a Necrolemur brain volume using Tarsius Model 1 and Necrolemur Model 1 values.
Morphology and Cranial Biomechanics: Head Carriage
Anatomically, apart from sheer differences in orbit/eye size, the scaled lateral views of Tarsius and Necrolemur (Fig. 2) show important disparities in the position of the orbit relative to the braincase and the foramen magnum, the inclination of the craniocaudal axis of the brain relative to the long axis of the skull or toothrow, and the size and inclination of the nuchal plane. Anatomically, the common denominator here is the disposition of the brain. The uptilted set of the brain reduces the distance between its center of mass and the craniovertebral joint and enables the orbits to retreat posteriorly, and thus occupy a position partly below the forebrain. Resetting the position of the braincase also corresponds with a shift of the nuchal plane into a more horizontal orientation. The consequences of this redesign are a relatively shortened moment arm for the eyeballs and the brain and a relatively lengthened moment arm for the nuchal muscles. The theoretical advantage of this pattern is that in the neutral position less muscular force would be needed to maintain head balance in Tarsius by comparison with Necrolemur, all else being equal. The evolutionary explanation for the changes toward the derived tarsier condition is that the head of Tarsius has evolved accommodative adjustments for efficient balance as a response to selection for hypertrophic eyeballs.
The free body analysis provides estimates of the mechanical advantage of the morphological design under various loading conditions (Table I). I have used the summed weight of both eyeballs in the calculations, as this is more reflective of the true loading conditions. The moment arms of the eyeballs and brain are smaller in Tarsius than in Necrolemur, although the magnitude of the reduction in tarsiers appears less impressive because this is masked by the sheer enormity of the eyes, which displaces the center of mass rostrally. Without retracting the orbits posteriorly into a subcerebral position, Me would increase in proportion to the increase in the diameter of the globe. Anterior shift of the foramen magnum is chiefly responsible for the tarsier’s increase in the moment arm of the nuchal muscles. The models show that the amount of nuchal force (Fn) required to maintain equilibrium is basically the same in Tarsius if either wet measurements or osteological proxies of eyeball size are used, which helps justify applying these principles to the fossils. However, there is a notable range of differences in this measure if eye size is set to the minimum or maximum values presented in the literature. For illustrative purposes, I take the rounded average of all four measures, 14, to represent the genus.
Necrolemur presents a contrasting picture. In 3 of the 4 models that substitute different estimates of eye or orbit volume, Fn is substantially higher. Only in Model 4, which sets Necrolemur eyes and brains at their minimal values, does the fossil approximate the tarsier pattern. The highest estimates, in Models 5 and 6, where Fn is ca. 1.9–2.2 times the magnitude of tarsier’s average, occurs when Necrolemur eyeballs are set to be either 2/3 or the equivalent mass of a tarsier eye. These are anatomical impossibilities, but they highlight the dramatic biomechanical differences in skull design between these taxa. In other words, they illustrate how relatively inefficient the tarsier skull would be in controlling pitch of the head if the primitive cranial morphology had not been repackaged to accommodate the added weight of hypertrophic eyes.
Basicranial Morphology: Lateral Balance and Head Turning
As indicated, this shift of the foramen magnum corresponds with an increase in the moment arm of the nuchal muscles while also decreasing the moment arms of the eyeballs and brain. With the foramen magnum of Tarsius located anteroventrally, the nuchal plane is both enlarged and reoriented, having moved, more or less, from the coronal plane and into the transverse plane of the skull (Fig. 2). Thus the occipital condyles face ventrally rather than caudally, as in Necrolemur. The tarsier condition corresponds with the orthograde body posture of a habitual clinger, whereas the morphology of Necrolemur reflects a skull that is designed to be cantilevered at the end of a more pronograde vertebral column, although this does not obviate facultative vertical clinging and leaping (VCL) body and head postures. A fully horizontal disposition of the nuchal plane would tend to reorient the line of action of the nuchal and suboccipital muscles which, by analogy with the human, may increase their efficiency in a nonquadrupedal manner of head carriage. With the nuchal region oriented more nearly coplanar with the horizontal plane and optical axis, the tarsier condition may also advantage the suboccipital muscles in maintaining stability at the craniovertebral joint and efficient control of head turning movements.
By comparison with Necrolemur, the larger foramen magnum of tarsiers corresponds with a wider transverse spread of the occipital condyles. This “wide wheelbase” arrangement increases the head’s lateral stability via the atlanto-occipital joints. Because the center of mass of an eyeball moves further from the midline as its volume increases, each imparts a moment that tends to roll the head laterally out of the horizontal plane, in proportion to eyeball diameter (Fig. 3). This lateral moment is minimized anteriorly by adjusting the location of the eyeballs: appression of the medial orbital walls places the centroid of the eyeballs as close to the midline as is possible. Posteriorly, the eccentricity of this load is minimized by the lateral shift of the condyles, which reduces the moment to roll the head. A “narrow wheelbase” would increase the magnitude of the moment to roll the head.
The lateral roll of the head to one side is also resisted by the tension of the contralateral suboccipital muscles acting on the opposite side. The mechanical advantage of this system can be enhanced by lengthening the transverse processes of the atlas, thus augmenting the muscles’ moment arms. Ankel-Simons (1999) pointed out that the atlas of Tarsius is relatively the largest of all primates. This corresponds with the enlarged f. magnum, laterally situated condyles, and an enhanced leverage of the suboccipitals inserting on the transverse processes.
When the craniovertebral pivot is posterior, as in Necrolemur, the component forces are relatively larger, tending to pitch and roll the skull downward, and especially to the side at larger excursion angles. A posterior f. magnum, therefore, places constraints on the evolution of eyeball hypertrophy. In the case of the cantilevered skull of Necrolemur, which has a vertical and frequently rugose nuchal plane, and a high and typically crested superior nuchal line, the nuchal muscles evidently play an important role in head balance during quadrupedal positional behaviors. Though this is generally the case in mammals, as a haplorhine Necrolemur has a rather foreshortened face, so less of an inherently forward, snouty mass to balance. But Necrolemur is also unusual among early haplorhines in having an enlarged petromastoid (Szalay 1976), which may have biomechanical significance. The latter presumably serves, in part, as an attachment site for the cleidomastoid muscles. Perhaps this development signals an enlargement of these prevertebral muscles as a complementary adaptation to help balance the head during VCL postures when the nuchal muscles have less leverage against the craniovertebral pivot, especially if the head is somewhat overbalanced anteriorly and laterally by the mass of moderately enlarged eyes and the primitively long moment arms of the brain and eyeballs.
Basicranial Morphology: The Auditory Region
The bullae of Tarsius and Necrolemur share several important resemblances, but tarsiers are widely acknowledged as being highly unusual morphologically (Fig. 3). Several features, all apparently interconnected, are of interest here: the transversely narrowed, toroidal shape of the tympanic cavity; medial orientation of the anteroposterior diameter of the tympanic cavity; transverse wasting at the junction between tympanic and hypotympanic sinus (the location of the intrabullar transverse septum); central, or anterolateral, as opposed to posteromedial, position of the posterior carotid foramen; inflation of the hypotympanic sinus; and narrowness of the ectotympanic tube. Much of this can be summarized in the form of a comparison. As in Necrolemur, the root of the tarsier ectotympanic emerges from a relatively flattened, vertical, lateral tympanic cavity sidewall, which represents the external, obverse bony surface that carries the crista tympani. But in Necrolemur and other tarsiiforms, the middle ear chamber is more evenly inflated anteromedially and ventrally opposite the tympanic cavity proper. In tarsiers, the inflation occurs anterior to the tympanum, which results in the latter appearing like a circular, mound-like rise. This constellation of unique features may be related to the spatial packaging of adjacent nonbullar structures.
Tarsiers have an unusually enlarged first cervical vertebra (Ankel-Simons 1999). Because of the forward position and wide span of the occipital condyles, which are now situated between the auditory bullae, the enlarged C1, including its anterior arch and transverse processes, encroaches the posteromedial face of the middle ear, including the medial aspect of the tympanic cavity posteriorly and the hypotympanic sinus anteriorly. This limits the mediolateral diameter of the middle ear compartment. Because Tarsius is evidently under selection to maintain a large inflated auditory bulla, it is possible that the unusually large size and irregular shape of the hypotympanic compensates for the limited space available for a medial or anteromedial dilation of the tympanic cavity proper, which is the condition of Necrolemur, where the tympanic and hypotympanic sinuses are continuous. Surely, the deep ventral inflation of the tarsier anterior accessory cavity is a function of the lack of expansion space medially, given the narrowness of the basioccipital, in turn related to mediolateral compression of the facial skeleton and choanae due to eyeball hypertrophy and reduction of the posterior nasal fossa.
An additional aspect of this pattern is that tarsiers have experienced a wholesale lateral displacement of the tympanic cavity as the foramen magnum shifted anteriorly, such that its lateral wall is much closer to the sidewall limit of the braincase. As a consequence, the transverse diameter of the external auditory tube is abbreviated, the orientation of the lateral wall of the tympanic cavity bearing the eardrum may have become more vertical from its primitively inclined setting, and the long axis of the tympanum is oblique to the sagittal plane.
For tarsiers, the coupling of a forwardly migrated foramen magnum and an enlarged C1 needs to be considered as a possible driving factor behind the novel location of the posterior carotid foramen. Rather than entering the bulla medially and almost horizontally, as in Necrolemur and other early haplorhines like Rooneyia, in Tarsius the foramen in shifted laterally and anteriorly away from the condyles. It enters into the bulla from a ventral and more anterior spot, well forward on the promontorium of the tympanic cavity. I consider this combination of local factors more likely than MacPhee and Cartmill’s (1986) suggestion that migration of the posterior carotid foramen was influenced by the extreme narrowing of the basioccipital.
Cranioskeletal Adaptations and Evolutionary History
As is well known, the forward chamber of the auditory bullae of tarsiers is very large relative to skull size, and it is tempting to seek a unique acoustic explanation for this phenomenon and the abbreviated external auditory meatus, another tarsier novelty. However, some nuances of bullar morphology may be secondary consequences of the complex spatial packaging of the basicranium, especially constraints on the volume of its posterior compartment, a tympanic cavity that is laterally shifted to accommodate the spread of the occipital condyles, and transverse processes of the atlas as the foramen magnum moved forward to a position between the bullae. In other words, the hypotympanic sinus may have increased in volume to compensate for the reduced volume of the tympanic cavity proper, which is spatially constrained. Similarly, the inflated petromastoids of Necrolemur, which has a differently shaped hypotympanic sinus, may reflect an alternative way to add intrabullar volume to the middle ear as well as enlarged attachment sites for prevertebral muscles to counterbalance the heavily loaded rostrum. Another consequence of laterally shifting the tympanic cavity toward the sidewall of the skull in tarsiers is that the length of the external auditory meatus is correspondingly reduced; Necrolemur still retains a lengthy tube.
Clearly, the enormous eyes profoundly influence cranial packaging and localized epigenetic processes, like the development of circumorbital flanges. But it is preferable to recognize the historical interplay of 2 major domains interacting with selection to produce the tarsier’s specialized cranial morphology, the visual and positional behavior systems. As indicated, head carriage and control are probably fundamentally behind various novelties that contribute to the extensively modified neurocranium and basicranium. But these are specifically designed to function with the neutral position of the head set in an upright posture, as in VCL mode, not in a quadrupedal position. The precocious evolution of eyeball hypertrophy was thus first made possible by a VCL mode of head carriage and locomotion in early tarsiiforms. Once accomplished, the coordination of hypertrophic eyes and efficient head balance become core features of a new adaptive zone in primates occupied only by tarsiers, based on a set of radically new sensory capabilities that extend the VCL pattern.
This is not to say that facultative, pronograde head and body postures would be anomalous in Tarsius: quadrupedalism is an empirically documented aspect of its repertoire. Only that once the eyeballs hypertrophied in the tarsier lineage, certain biomechanical efficiencies would be virtually required and others could be gained while maintaining the head and trunk in orthogrady, obligatorily. Nor does this hypothesis suppose that small VCL strepsirhines like galagos, who frequently engage in quadrupedalism, are unable to hold their heads with eyes facing forward while clinging because they lack the same mechanism. But no form other than Tarsius has modified the head so extensively, or so radically altered the postcranial skeleton throughout to sustain a fully new manner of foraging. To emphasize this commitment, its uniqueness and adaptive pervasiveness across anatomical systems, I distinguish the tarsier mode of posture and locomotion as extreme VCL, or XVCL. To be clear, XVCL is not meant to emphasize either the leaping or the clinging, but to allude to an extensive suite of adaptations within and outside the locomotor skeleton that enables tarsiers to be successful at finding food without hunting actively and extensively for it: the nonpursuit predatory strategy.
The comparative evidence from modern primates suggests that a fundamentally orthograde style of head carriage based on intrinsic biomechanical adaptations of the head and postcranium is the distinctive skeletal basis of XVCL. I am aware of no reports that state that other VCL primates, such as galagos and indriids, have modified the skull to affect a comparable head carriage. Though there is no reason why the smaller-bodied VCL galagos could not be found to parallel tarsiers in this regard, galagos are also more persistent quadrupedal foragers. Nor have those strepsirhines with the largest eyes next to Tarsius, lorises, adopted a similar pattern; to the contrary, loris skulls are modified to enable them to stalk quadrupedally, with uptilted large eyes directed horizontally as they walk nose-to-the-branch. Thus the best explanation of the tarsier is that its skull is modified to direct the eyes horizontally while the head is perched vertically, and to swivel the head, instead of translating the whole body through space, as its scans for prey and initiates liftoff.
Although tarsiers have extensive, dramatic adaptations for saltation, they actually move little as part of the XVCL paradigm. The frequencies of positional behavior or support use have been documented for four species (Crompton and Andau 1986; Dagosto et al.2001; MacKinnon and MacKinnon 1980; Niemitz 1984; Tremble et al.1993). VCL and climbing, and usage of vertical and oblique supports, dominate the profiles in all studies. Quadrupedalism and horizontal support use is rare in the wild studies, with the highest frequencies of observations amounting to 11% and 21% in Tarsius syrichta, respectively, and locomotor bouts averaging only ca. 0.5 m in length (Dagosto et al.2001). Clearly, living largely in the understory, the tarsiers eschew the branch networks, travel pathways, and behaviors that make roaming effective. The distances covered in leaps are small, averaging 1.3 m among Tarsius bancanus, T. spectrum, T. syrichta, and, according to Dagosto et al., bouts of climbing up and down averaged only ca. 0.4 m.
While measurements vary among the species and studies (Gursky 2007a), the home ranges of tarsiers, which usually amount to a small few hectares, compare well with the most predaceous galagines and lorisines (Nekaris and Bearder 2007). In some studies, tarsier home ranges are smaller, approaching that of the smallest primate gumivore, Cebuella pygmaea (Digby et al.2007), whose life is tied to a small number of sessile trees. With regard to finding food, stealth, in this case cryptic non-motion, is the tarsier modus operandi. While being coy has obvious benefits for a predator as a form of concealment, its energetic returns would be vast if a semisedentary prey detection system could substitute for mobile prey foraging. As tarsiers have evolved a generally conservative energy expenditure strategy (Niemitz 1984), the prodigious leaping abilities of tarsiers are better understood as benefiting prey-capture accuracy and efficiency rather than a method for traveling large distances. In a small primate such as the tarsier, prone via ancestry to some form of VCL habit, the price of dragging around a massive head quadrupedally may be a costly limiting factor, as would lifting off and maneuvering a body with a massive ocular bulk added to the head end. Gursky (2007b, p. 121) argues similarly that tarsier mothers switch their locomotor preferences to quadrupedalism when carrying infants in their mouths because “…it may not be biomechanically possible to utilize vertical clinging and leaping when transporting a large load anteriorly.”
There are other reasons why a lifestyle selected for overall energetic efficiency would be under strong selection in tarsiers for, as a derivative of neural tissue, maintaining the 2 enormous eyes metabolically is also likely to be very expensive. As noted by Wangsa-Wirwan and Linsenmeier (2003, p. 547; see also Nivens and Laughin 2004), “The retina is one of the most metabolically active tissues, consuming O2 more rapidly than many other tissues, including the brain.” This large combined caloric overhead can be offset by moving little to forage and by having small home ranges, i.e., forsaking plant foods that need to be collected by shopping and, alternatively, assuming a lamppost scanning and ambush strategy, where the prey comes to you. Roberts and Kohn (1993) emphasized the efficiency of this manner of prey selectivity in their study of a colony of captive Tarsius bancanus, where the capture success rate of adults was 88%.
Obviously, this would be best accomplished if prey detection methods are finely tuned. Though the specialized eyes of tarsiers have received much attention, hearing is likely to be equally important (Stephan 1984), and certainly the 2 senses are powerfully integrated (Niemitz 1985). The wide head and laterally displaced eardrums of tarsiers may augment prey detection by making the ears more adept at localizing sound though the binaural time delay principle. Thus it would be shortsighted to presume that the only advantage of tarsiers being able to swivel the head in a large arc is to adjust the gaze because the ocular muscles are weakly developed, as tradition has it. Head movements likewise position the ears for hearing. Perhaps freezing the eyeballs and moving the head instead is a way of accurately synchronizing the localizing functions of eyes and ears in high-resolution prey detection, as an alternative to hunting noisily and expensively through the forest via locomotion. An added benefit to nonpursuit predation is that tarsier mothers can be watchfully close to their parked infants (Gursky 2007a,b), whose eyes, spatial sense and limb coordination, one imagines, would be awkwardly slow in developing.
Eyeball Hypertrophy: Why?
Recent contributions to the basic form and function of the tarsier eye and visual cortex have been given by Castenholtz (1984), Hendrickson et al. (2000), Collins et al. (2005), and others (Martin and Ross 2005, and reference cited therein) have provided an evolutionary and ecological context for assessing tarsier vision in broad surveys of the primate visual system. These articles discuss various novelties of Tarsius, provide updates on fuzzy anatomical issues, and demonstrate newly found characteristics covering a range of topics, including among other features a finer characterization of the retinal fovea; demonstration and quantification of both rods and cones in the retina; descriptions of cornea size and pupil dilation; affirmation that the eyeballs are intrinsically immoveable; and characterization of the cell types in the lateral geniculate nucleus, the processing center and relay station joining the retina with the cortex. The emphasis arising from the physiological studies in particular is that tarsier eyes, in spite of being nocturnally adapted, are capable of high visual acuity and the perception of color (Collins et al.2005), in keeping with its phylogenetic roots as a haplorhine. However, it is evident that much still needs to be learned about the functionality of tarsier eyes.
Of course, eyeball hypertrophy is almost always part of the discussion. The enormity of tarsier eyes has been explained as compensation for the lack of a tapetum lucidum, which is typical of nocturnal strepsirhine primates, in an animal that has reentered a nighttime niche but without the original, necessary optic equipment (Cartmill 1980; Martin 1973; Martin and Ross 2005). Martin (1990, p. 664) provided a concise statement of the hypothesis, saying that “…in comparison to the eyes of strepsirhines primates, those of tarsiers seem to have become adapted in a difference [sic] way for nocturnal vision—namely, through a radical increase in the size of the retina (and hence in the number if photoreceptors) well beyond the level of nocturnal strepsirhines.” To explain why the convergent, secondarily nocturnal Aotus has experienced less eyeball hypertrophy, he notes that tarsiers have had a longer time to evolve their big eyes than owl monkeys did.
Though generally acknowledging that tarsier eyes are adapted to predation, these studies do not address how generalized interpretations such as “visual acuity” may actually confer selective benefits. The tarsier eye may be hypertrophied and photoreceptor cell count may be augmented and sight may be better than previously expected of a nocturnal haplorhine primate, but to what end? For what purpose is the receptor rich retina of a tarsier optimized: huge eyes and good nighttime vision to do what, exactly? Though the core idea has merit, the full hypothesis falls short as an explanation for the grotesquely large size of tarsier eyes, obviously the starting point for understanding what a tarsier is in essence. It does little to integrate eye size with other unique, and probably contingent features, such as the immobile eyeballs —is this really an anomalous consequence of supersized eyeballs?— or, with our improved understanding of the physiological capacities of the retina. In addition, these notions seem to exaggerate the negative, anticipated optical difficulties encountered by the tapetum-less, foveate visual system operating in low light, for the central retina of haplorhines is typically stocked with cones that rely on high levels of luminance to focus an image onto a large field of these color-sensitive cells. While such a “deficit” may be real in terms of a tarsier’s ability to discriminate the shape and orientation of an object-image, which is exacerbated by the loss of color information when light is insufficient and constrained by a relatively small field of cone receptors, it need not be true for the other major function of the visual system, to detect, localize, and guide movement. This involves a different cortical pathway but is nonetheless predicated on the same fundamental purpose of the visual system, to map and interpret object information onto the retina and cortex (Morgan 2003).
I suggest the unique specializations of the tarsier eye are the basis of an adaptive complex that maximizes the visual system’s spatial mapping properties, as opposed to a system designed for object identification. In other words, tarsiers may be more adept at finding prey and plotting a course to capture than knowing at first what it is. Visual acuity to a tarsier means being an excellent cartographer. It is not equivalent to “acute eyesight” in the human sense of being able to spot and encode the identifying features of objects, which to us seems more important cognitively than finding it in the first place. As an early branch of the haplorhine stock, tarsiers are not even expected to have either the same anatomical configuration of the central retina as anthropoids or the same kind of keen eyesight.
Regarding the extreme size, it is likely that tarsiers benefit in several ways from the geometry of outsized eyes such that selection eventually maximized this quantity until it became limited by some basic physiological threshold, or by simply relaxing selection. A key to the beginnings of hypertrophy may have been the development of an ectopic eye, for tarsiers have eyeballs that have largely escaped the osseous confines of the orbital fossa. Schultz (1940) pointed out long ago that more than half of the globe is located outside the margins of the orbital aperture, and Rosenberger et al. (2008) discussed anatomical evidence that may relate to this phenomenon among several Eocene fossil tarsiiforms in the form of supraorbital flanges, among other features. Castenholtz (1984) provided the following ratios of eyeball volume: orbital volume to provide perspective on how much the eye of Tarsius bulges out beyond the eyecup. The values are: tarsiers, 1.79; nocturnal strepsirhines, ca. 1.0; chimpanzees, 0.26; orangutans, 0.17; humans, 0.32.
Whereas the anthropoid eye is an organ whose core ecological adaptations may be understood to be concentrated near the center of the retina where the fovea is located and the color-sensitive cone receptors are densely distributed, tarsier retinas show evidence of specialization in the retinal periphery. This may be critical to explaining eyeball hyperetrophy. Given the fundamental topographic organization of the retina about a center, while cell populations near the central fovea could grow in number and density as tarsier eyes evolved to enormous size, the largest potential for adaptations relating to surface area increase can be expected to occur in the periphery, off center, where surface is more plentiful. The periphery is where rods are concentrated. Collins et al. (2005) reported that tarsiers have ca. 135,000,000 rods in the periphery of the retina, distributed in a density of ca. 300,000/mm2. Humans, by comparison, have ca. 120,000,000 peripheral rod cells, at a density of 95,000/mm2 (Curcio et al.1991). Dichromatic ground squirrels have ca. 1,270,000 rods, with a peak density of 13,000/mm2 (Kryger et al.1998). This comparison suggests the possibility of capturing a larger and more finely grained image in the retinal periphery of tarsiers, all else being equal. Because in humans, and presumably in tarsiers, different populations of the rods are optimized for the luminance levels of starlight, dawn and twilight (Sterling 2004), this advantage in cell count is likely to be a specifically nocturnal adaptation, perhaps even tuned to different phases in the light-dark continuum.
Collins et al. (2005) also suggest another unusual feature in a peripheral retina of tarsiers, a wide band of cones at the rim of the photoreceptor array that is sensitive to ultraviolet light. They suggest these S-cones are useful for detecting prey that fluoresces in UV light. Their location contrasts with the M- and L-cones of tarsiers, which are not UV sensitive and occur in the central retina. If Collins and colleagues are correct, this means a specialized encoding feature of the tarsier night vision system is located in the retinal periphery, not in the central retina where the diurnal haplorhines appear to have specialized with their large populations of cones. Why?
A key benefit may derive from spatial factors that are a consequence of eyeball size. The enormity of tarsier eyes exaggerates point separations within and between the eyes, of the nasal (medial) and temporal (lateral) aspects within each retina and also in the position of the eyes with respect to one another: Their optical axes are spread apart as the diameters of the two eyeballs increase. Tarsiers are small bodied and would normally thus have absolutely small eyes. For a variety of structural and phylogenetic reasons (Rosenberger et al.2008) probably linked with a diminution of olfaction and the nasal fossa in ancestral haplorhines, they also have orbits more closely spaced near the midline than strepsirhines or plesiadapiforms, whose orbital fossae are widely separated by the interorbital pillar. So, there are inherent hard tissue constraints that would otherwise keep tarsier eyes positioned very close together. Separating the eyes, or more properly their optical axes, while maintaining a small body size can be accomplished by simply enlarging the spheres beyond the eyecups formed by the orbital fossae.
Separation of the eyes is the central anatomical feature behind binocular vision, and increasing the separation of forward facing eyes enhances depth perception (Morgan 2003). The mechanism involved is stereopsis, which is a distinct property of binocular vision. Stereopsis is based on the differences in what each eye “sees” in space and time as a function of their different viewpoints, i.e., their spatial separation. It is exactly comparable to the interaural time delay between the 2 ears hearing a sound coming from some direction. This horizontal disparity means that complementary areas on each retina are mapping different versions of the target image in real time. The differences are resolved and integrated by the retina and the brain to produce the sensation of spatial depth. Greater disparity of the retinal maps (up to a point) can enable more precision in sensing the target with forward facing eyes, a calculation that is done by measuring the differentials between the amount and distribution of light falling on the complementary receptive fields of the 2 eyes. For tarsiers, the peripheral low-luminance rods and the UV-sensitive S-cone bands that are farthest out on the retinal periphery would be most affected functionally as the eyeball expands, not the cones of the central retina. These peripheral photoreceptors are not designed to assess color or edges, properties useful for classifying the nature of an object. The information they convey is simply the cartographic information derived from mapping the object’s position and movement in space relative to the viewer.
Within the cortex, the tarsier visual system also suggests a priority for dynamic cartography, i.e., selection of motion detection functions. According to Collins et al. (2005), the lateral geniculate nucleus is dominated by magnocellular cells rather than parvocellular cells. The M-cells have a quicker response time, are best suited for spatial assessment and motion detection, and not for discriminating color. P-cells, in contrast, are slow to process and designed to discern color.
The square in the random-dot stereogram is totally invisible to either eye’s image alone…It exists only by virtue of the relationships between the two eyes…binocular vision allows us to see it. This idea of camouflage gives us a new evolutionary perspective on binocular vision. A moth might make its colour and texture match that of the bark on which it sits, but it is not able to make itself perfectly flat. Viewed with two eyes, the moth will leap out from the bark like the square from the random-dot stereogram.
With color information already reduced or absent in the tarsier’s low light domain, an even greater selective premium would be placed on a parallax spatial and motion detector, as described, able to see the topography of a moth and its movement.
An additional advantage of having a wide transverse diameter of the eyeball consonant with hypertrophy is that it allows for a larger cornea and a pupil that is capable of opening very widely. Both of these features are adaptations to maximize the amount of light entering the tarsier eye (Castenholtz 1984), i.e., to increase the size of the mapped image. The benefit of a larger image is not that the eye “sees better.” It is that more photoreceptors are excited, making it easier and faster to calculate and compare where the object is in space from instant to instant based on horizontal disparity (Morgan 2003). The panoramic effect of the large, maximally dilated pupil of a tarsier may also help stabilize the mapped object-image on the retina to permit calculations as the target moves infinitesimally small distances across the photoreceptor field from instant to instant.
In the long axis, the larger eyeball effectively becomes a telescopic lens, also increasing the size of the image mapped onto the retina, although the optics of tarsiers eyes, which also involve a large anterior chamber associated with the cornea (Castenholtz 1984), are likely to be complex. Ideally, the image should be as sharp as possible, i.e., focused, for the sake of accuracy. According to Collins et al. (2005, p. 1016), tarsiers are probably capable of projecting a focused image, as “…they are likely to have use of color vision, as well as the high visual acuity implied by the high receptor counts of rods and cones in the central retina and fovea.” Though it is not understood how tarsiers may use the light-intensive central retina and fovea in darkness, it seems that the retina is potentially very efficient in mapping an image because it is capable of sensing a nodal point within the mapped field that can be used in calculations. Photosensors in the periphery may be mapped as vectors against the fovea, as in anthropoids (Morgan 2003).
Ironically, the immobile eyes may be another asset of the tarsier visual system that benefits motion detection. In the absence of localized motor control, one may presume that tarsiers suffer by not being able to efficiently manage their gaze, or track a target with minimal energy expenditure by using the eyes alone. On the other hand, fixing the eyes in the head means that the visual system is yoked to the vestibular-cochlear complex. It suggests a high level of cortical integration between these sensory systems. Thus, while a tarsier’s eyes may not be moving independently of the head, its gaze may be precisely registered against the vestibular-cochlear mapping system, giving the a high-fidelity spatial recognition and tracking apparatus. Stephan (1984) noted that the vestibular system was highly developed even though cortical motor systems were not, by comparison to strepsirhines. He also indicated that, in the midbrain, the superior colliculus of Tarsius was very well developed. The latter is essentially a mirror of the retinal mapping field, from which it receives input. Its functions have been well studied in macaques (Walton et al.2007). There the superior colliculus controls eye shifts during gaze and head movement, but Walton et al. theorized there is a large population of cells that may control head movement alone independent of eye shifts. How, precisely, these structures might work in tarsiers cannot be determined. But they imply a profound emphasis and coordination of detailed spatial mapping functions. If tarsier eyes are truly motionless as reported, it is likely this is an advantage for the tracking system rather than an anomaly. If anatomical studies confirm that frozen eyeballs relate both to eyeball size and their ectopic position, we may be able to trace the implied contingent behaviors in the fossil record.
Comparative behavioral information suggests other reasons why mapping space and movement may be more relevant to explaining tarsier eyeball specializations than feature detection. Predaceous lorises and galagos, both armed with a tapetum that scatters light, and both lacking a fovea where incoming light could be concentrated, also have impressively large eyes for their body and brain sizes, and by comparison with their more frugivorous relatives (Rosenberger pers. obs.). Yet their optics suggests an inability to produce a focused image. The coincidence of biological roles between lorisiforms and tarsiers does not negate the potential general benefits of increasing photoreceptor count by enlarging the retina, along the same lines originally proposed to explain tarsier hypertrophy (Cartmill 1980; Martin 1973), but it suggests the need for a more specific interpretation of the underlying functions that were selected. Increasing photoreceptor count in lorisiforms would not increase visual acuity in the haplorhine sense because the retinal image would not be a focused image. The form-function common denominator behind their shared specialization, eyeball enlargement, may be an enhanced spatial mapping and motion detection system that comes with increasing binocular disparity between peripheral retinal fields.
The tarsier cranium has been extensively redesigned to accommodate huge eyeballs, not only by making space for them but also by arranging cranial components to produce a biomechanical system capable of balancing and accurately moving a head that is massively eccentrically loaded. Its compact, roundish skull, with the nuchal region rotated into a ventral position, is organized to balance the head in the neutral position of the vertical clinging body posture while keeping the skull light in weight and structurally strong. The wide brain case and forward position of the foramen magnum set the ears and tympanic membranes widely spaced apart, which may enhance the localizing abilities of the auditory system, also an asset to predation as hearing and vision may be exceptionally integrated and equally important to the tarsier lifestyle. Many specializations of the tarsier eye, including eyeball hypertrophy and the eye’s cortical representation, may also serve to optimize the visual system’s ability to map space and detect motion, rather than to discriminate patterns for object identification, an aspect that may be under less selective pressure given the low light conditions of the animal’s nocturnal manner. Seeing movement and navigating the body through space in order to intercept prey are functions critical to the tarsier’s unique style of non-pursuit predation which, overall, is an energy-minimizing ecological strategy. The selective imperative to conserve locomotor energy may counterbalance the massively hypertrophic eyes, which are metabolically expensive. The exclusively animalivorous diet of Tarsius may also be a prerequisite to evolving such a costly visual system.
The derived functional morphology of the tarsier visual system and cranium probably evolved as the lineage shifted the emphasis of posture and locomotion from a pattern that involved a rather varied profile of biological roles, supporting a mobile hunting lifestyle that almost surely involved a significant amount of quadrupedalism, to one that became highly efficient in supporting a relatively sedentary, sit-and-wait mode of prey detection, punctuated by the prey capture pounce. The integration of specialized optic, auditory, cranial, and postcranial elements is the XVCL habit, evident only in tarsiers. None of the fossil tarsiiforms for which we have relevant evidence have hypertrophic eyeballs on the order of a tarsier’s or the advanced biomechanical features related to head carriage and eccentrically loaded skulls, although several appear to have at least precociously developed ectopic eyeballs and some, such as Necrolemur, exhibit extensive tarsal elongation and tibio-fibular fusion, in the absence of greatly enlarged orbits. This indicates that VCL locomotion and insect foraging (as well as frugivory and/or soft-prey preferences, viz. Necrolemur) were central to a branch of the tarsiiform radiation, but none thus far known are likely to have practiced the nonpursuit, XVCL foraging method of modern Tarsius, which relies so heavily on sight, and probably sound. Taxonomically then, it remains to be seen if the fossil record reveals this unique lifestyle to be anything more than a genus-specific phenomenon, which would perhaps be another cause to draw similarities between the evolutionary histories of 2 of the odder primates, Tarsius and Homo.
The research was partially supported by grants from PSC-CUNY and the Tow Travel Research Fund, Brooklyn College. For access to collections I thank authorities and colleagues at the American Museum of Natural History; the United States National Museum; the San Diego Museum of Man; Yale Peabody Museum; Museum of Comparative Zoology, University of California, Berkeley; Museum National d’Histoire Naturelle, Paris; British Museum, London. I express special thanks to Emily Rosenberger for scientific assistance, Julia Zichello for help with illustration, Mark Dow for developing and producing Fig. 1, and to Carsten Neimitz. I thank Siobhan Cooke and anonymous reviewers with deep expertise in tarsier biology for thoughtful comments that improved the manuscript.