Overview of the visual system of tarsius



Tarsiers, which are currently considered to constitute the sister group of anthropoid primates, exhibit a number of morphological specializations such as remarkably large eyes, big ears, long hind legs, and a nearly naked tail. Here we provide an overview of the current state of knowledge on the tarsier visual system and describe recent anatomical observations from our laboratory. Its large eyes notwithstanding, the most remarkable feature of the tarsier brain is the large size and distinct lamination of area V1. Based on the need of tarsier for optimal scotopic vision and acuity to detect small prey in low lighting conditions, tarsiers may have preserved a high level of visual acuity by enlarging V1 at the expense of other areas. The other classically described visual regions are present in tarsier, albeit many borders are not clearly distinct on histochemical or immunohistochemical preparations. Tarsiers also have a large number and unusual distributions of cones in the retina, with high numbers of M/L-cones in the central retina and S-cones surprisingly at the periphery, which may be sensitive to UV light and may be useful for prey detection. These adaptive specializations may together account for the unique nocturnal predatory requirements of tarsiers. © 2005 Wiley-Liss, Inc.

The unusual appearance of tarsiers, with their large eyes, big ears, long hind legs, and nearly naked tail (Fig. 1), has held the interest of biologists since they were first described in 1769 (Polyak, 1957). They were originally misclassified as a rodent or an opossum, and even today, their phylogenetic position relative to other primates is under debate. They have been considered the most primitive living primate (Smith, 1924) and on the line leading to humans (Yoder, 2003). They were placed in the prosimian suborder with lemurs, lorises, and galagos (Gregory, 1915) and they have been widely recognized as such, but they have also been classified with monkeys and other anthropoid primates, with considerable support (Yoder, 2003). The confusion rests with the general conclusion that tarsiers arose from a branch very close to the separation of the prosimian and anthropoid lines, and they have traits that fail to place them clearly in either clade. Currently, there is a growing consensus that tarsiers are the sister group of anthropoids and join Anthropoidea in the semiorder Haplorhini (Schmitz et al., 2002; Meireles et al., 2003; Ross and Kay, 2004). Tarsier fossils, nearly identical to living species, date to 45–50 million years ago (MYA) (Beard et al., 1994), but the branching from Anthropoidea likely occurred 60 or more MYA (Martin, 2004). Today, four modern species are recognized, all confined to Southeast Asia.

Figure 1.

Front view of nocturnal primate Tarsius spectrum. Tarsiers range in size from 120 to 134 g. Their huge eyes, each approximately the same volume as the whole brain, lack a reflective tapetum. Their large lens and cornea are highly effective light collectors.

There have been a number of descriptions of aspects of the visual system and visual behavior of tarsiers. Most notably, Smith (1924) noted the large overlap of the visual fields of the two eyes and the large size of primary visual cortex (striate cortex or area 17), and Clark (1930) has described the visual thalamus (see also Chacko, 1954; Simmons, 1982). More recently, Polyak (1957) reviewed the visual system and visual behavior of tarsiers in 33 pages of text, Bonin (1951) described the architectonic subdivisions of neocortex, and Stephan (1984) described the brain with an emphasis on the visual system. Here we review these and other descriptions of visual system anatomy in tarsiers, while adding some of our own recent observations based on the histological examinations of three brains obtained by one of us (A.H.) from the Indonesian Primate Center.


The three brains available for histological study were adult Tarsius spectrum from northern Sulowesi. The tarsiers were euthanized by injection of sodium pentobarbital because they had injuries or untreatable health problems. One tarsier (01-119) was perfused through the heart with 4% paraformaldehyde in 0.1 M phosphate buffer and the brain was postfixed in 2% paraformaldehyde overnight. A second tarsier (01-92) was perfused with 4% paraformaldehyde/0.1% glutaraldehyde, followed by 2% paraformaldehyde. The third tarsier (00-58) was perfused with 4% paraformaldehyde in 0.1 M phosphate buffer and postfixed in 2% paraformaldehyde. Brains were stored for 2 days after fixation in 0.1 M phosphate-buffered saline (during transport to the United States), then transferred to a 30% sucrose solution for cryoprotection in preparation for sectioning.

One tarsier brain (01-119) was cut into 40 μm thick sections in the coronal plane on a sliding microtome with a frozen stage. Sections were collected into 0.1 M phosphate buffer and separated into 10 series. A second tarsier brain (01-92) was first bisected in the midsagittal plane. Both hemispheres were sectioned in the horizontal plane. Sections from each hemisphere were collected into 0.1 M phosphate buffer and separated into 10 series. The third tarsier brain (00-58) was bisected midsagitally. The right hemisphere was sectioned in the sagittal plane and the left in the coronal plane. Sections were collected into 0.1 M phosphate buffer and separated into 10 series.

One or more sets of sections were reacted for cytochrome oxidase (Wong-Riley, 1979) and one set was stained for myelinated fibers (Gallyas, 1979). One set of sections was stained for Nissl substance. Additional sets of sections were stained for intensified cytochrome oxidase and various immunocytochemical markers as detailed below.


Immunocytochemical markers for calbindin (CB; calbindin D-28k), parvalbumin (PV), and the lectin Wisteria floribunda agglutinin (WFA) were used to reveal subpopulations of neurons in the tarsier brains.

Sets of sections were processed for the calcium binding proteins calbindin D-28k (catalog no. 300) and parvalbumin (catalog no. 235; both Swant, Switzerland). On the first day of processing, sections were washed in 0.05 M tris-buffered saline (TBS) containing 0.25% triton X-100 (TBS/triton). Sections were incubated for 36–48 hr in primary antibody solution (1:5,000 PV and CB) containing 0.25% triton X-100, normal horse serum, and TBS. Sections were washed three times for 10 min each in TBS/triton, followed by incubation in secondary antibody solution (Vector kit 4002; 1:200 in 0.05 M TBS/triton) for 2 hr. Sections were again washed with TBS/triton, then transferred to ABC solution for 1 hr. Labeled cells and neuropil were visualized by DAB reaction.

Sets of sections were processed for the lectin WFA, which stains proteoglycans in the extracellular matrix, following the method of Preuss et al. (1998). Briefly, sections were rinsed in 0.05 M TBS, incubated in a blocking solution containing 0.05 M TBS, sheep serum, and 0.1% triton for 2–4 hr, then transferred to primary antibody solution (2 mg/ml WFA; Sigma L-1766) for 24 hr of room temperature incubation. Sections were washed in 0.05 M TBS, incubated in ABC solution (Vector kit PK-6100) for 1 hr, and visualized by DAB reaction.


Here we review previous studies to outline the prominent features of the visual system of tarsiers, while adding observations from our own histological material.

Eyes and Retina

Early primates are thought to have been nocturnal visual predators living on small prey and some fruit in the tropical rain forests (Ross, 1996; Ross and Wall, 2000). Later, the emerging anthropoid primates shifted to diurnal life and their visual system became more dependent on cones. As a branch of the anthropoid radiation, the ancestors of present-day tarsiers apparently had adopted a diurnal niche and then readapted to a nocturnal niche by evolving enormous eyes for gathering light in the absence of a reflecting tapetum and a rod dominated retina while retaining the diurnal adaptations of a fovea and two types of cones.

The eyes of tarsiers are huge, both in absolute size and in proportion to the size of the 120–134 g animal. Polyak (1957) concluded that their eye size relative to body size is unmatched by any other mammal. The diameter of the eye is 15–18 mm; roughly, the length of the forebrain and each eye has a volume that equals that of the brain (Castenholz, 1984). Other adaptations to dim light include a large (10 mm diameter, 6.5 mm thick) lens. The lens and the large cornea are highly effective light collectors, and the large iris allows the admitted light to be regulated from a lot at night to little in the day. As tarsiers change their view by rotating the head rather than the eyes, the extrinsic eye muscles are poorly developed. In viewing the eyecup, there is no indication of a tapetum, and the pigment epithelium is heavily pigmented (Hendrickson et al., 2000). There is a deep, avascular fovea (Fig. 2A) 2.7–3 mm temporal to the optic disk on the horizontal meridian (Polyak, 1957; Wolin and Massopust, 1967; Castenholz, 1984; Hendrickson et al., 2000; Ross, 2004).

Figure 2.

Fovea and peripheral retina. Layers for both pictures are labeled in the center. A: The tarsier fovea has a prominent pit, which reaches the bottom of the outer nuclear layer (ONL). Note that the ganglion cell layer (GCL) on the foveal slope is only 2–3 cells thick. B: This paraffin section from the mid-peripheral temporal retina is typical for most of the retina. The pigment epithelium (PE) has pigment granules only in its apical processes, which extend in between the rod outer segments. The arrowhead marks the boundary between the outer and inner segments of the photoreceptors. The ONL is thick and the cell bodies are not readily identified as rods or cones. The outer plexiform layer (OPL) is relatively thin. The inner nuclear layer (INL) is quite thick, as is the inner plexiform layer (IPL). Arrows mark the blood vessel plexus at the outer edge of the INL and within the GCL. C: An immunostained section of tarsier central retina. Antiserum to rod opsin labels the rod outer segments (OS) very heavily, the inner segments (IS) less heavily, and the rod cell bodies in the outer nuclear layer (ONL) lightly. Rod synapses in the outer plexiform layer (OPL) are not obviously labeled. Arrows mark the unstained cone cell bodies at the outermost border of the ONL. D: An immunostained section of tarsier central retina. Antiserum to medium- or long-wavelength-selective cone opsin labels many cone OS very heavily. The lightly labeled cone cell bodies fill the gap in the rod opsin staining. Cone terminals are not labeled in the OPL. E: Immunostained section of far peripheral tarsier retina, which is much thinner than central retina. Antiserum to short-wavelength-selective (S) cone opsin labels a few cones. The OS, IS, cell bodies are well stained and even the cone pedicles in the OPL can be identified. Scale bar in E for C–E. Bottom row: Photographs are from tarsier retina near the fovea (F,G) and for peripheral retina (H,I) immunolabeled as whole mounts using diaminobenzidine visualization. F: Retina near the fovea shows all cones immunolabeled for cone-specific α-transducin. At this eccentricity, cone density ranges between 11,600 and 14,200/mm2. G: This image on the right shows three cone outer segments labeled for S-opsin, illustrating the very low S-cone density found in central tarsier retina. H: Retina near the superior retinal edge. All cones immunolabeled for cone-specific α-transducin. Cone density is still quite high (4,200–7,000/mm2) even in the far periphery. I: This image shows the sharp increase in cones labeled for S-opsin from almost none centrally to 1,100–1,600/mm2 in the periphery. Scale bar in G for F–I.

The retina is large, with an area of about 400–500 mm2 and horizontal and vertical meridians of 26–27 mm (Hendrickson et al., 2000). The retina contains rods throughout at a rod density of > 300,000/mm2 (Fig. 2B and C). Because cones in tarsiers resemble rods morphologically, early investigators considered tarsiers to have an all rod retina (Polyak, 1957; Castenholz, 1984). However, as shown in Figure 2D–I, immunocytochemical staining using antisera that label monkey and human cones clearly labels cones in tarsiers (Hendrickson et al., 2000). Tarsiers have a dominant population of M- (medium-wavelength) and/or L- (long-wavelength) cones (Fig. 2D, F, and H) (Tan and Li, 1999) and a smaller population of S- (short-wavelength) and/or UV (ultraviolet-wavelength) cones (Fig. 2E, G, and I). The M/L-cones are concentrated in the central retina in a manner similar to that of diurnal primates, with M/L-cone densities dropping from 14,200/mm2 (Fig. 2F) around the fovea to 4,200/mm2 near the retinal edge (Fig. 2H). Foveal cone density has not been published. S-cones are rare in the central retina (< 300/mm2; Fig. 2G) but rise to 1,600/mm2 in a 3–4 mm wide ring encircling the peripheral retina (Fig. 2E and I), perhaps to assist the peripheral rods in detecting prey in dim light (Hendrickson et al., 2000). The lack of calcium-binding protein immunolabeling of tarsier cones resembles nocturnal New World monkeys more than diurnal Old World monkeys (Hendrickson et al., 2000). Tarsiers are more accurately considered crepuscular than nocturnal, as they are active before and after sunset and sunrise, rather than in the darkest midnight, and thus 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. However, tarsiers have central retinal ganglion cell densities more comparable to galagos than macaque monkeys (Tetreault et al., 2004).

Lateral Geniculate Nucleus (LGN)

Any description of the LGN of tarsiers is best understood in the context of LGN organization in other primates. In most mammals, the larger dorsal lateral geniculate is distinguished from the smaller ventral lateral geniculate nucleus, which does not project to cortex. In primates, the dorsal LGN is usually simply called the LGN, and the ventral LGN is often called the pregeniculate nucleus because of its anterior position. The LGN receives the bulk of the projections from the retina and about 80% of its neurons project to primary visual cortex (Kaas et al., 1972). The LGN in most primates has 4–6 layers numbered so that 1 is the most ventral layer. Each LGN represents the contralateral visual hemifield via inputs from the temporal retina of the ipsilateral eye and from the nasal retina of the contralateral eye. Layers are also distinguished by functional type (Kaas et al., 1978; Casagrande, 1994). In most anthropoid primates, the dorsal parvocellular mass is rather thick toward the caudal representation of central vision, and the two parvocellular layers subdivide and interdigitate to form four or more sublayers or leaflets that are usually not distinguished from full layers in counts of layers. Thus, anthropoid primates, such as macaque monkeys and humans, are usually described as having four parvocellular layers. Especially in diurnal primates, the larger dorsal four layers of the LGN receive inputs from the P or midget ganglion cells (GCs), which make up about 80% of the GCs. In nocturnal owl monkeys, this dorsal portion forms only two layers, a dorsal layer with inputs from the contralateral eye and a ventral layer with inputs from the ipsilateral eye. The dorsal parvocellular or P-layers function in detailed vision and color vision (Livingstone and Hubel, 1988).

Ventrally in the LGN, there are two magnocellular or M-layers, with the most ventral or external layer (layer I) getting input from the 10% or so of the M (or parasol GCs) of the contralateral eye, and the less ventral inner M-layer getting input from the ipsilateral eye. M-cells are not involved in color vision, have larger receptive fields, and are sensitive to low visual contrast in dim light.

Prosimian primates (lorises, lemurs, and galagos) have two additional layers wedged between the two parvocellular layers, which are called the koniocellular or K-layers, because of the small size of the neurons in these layers. The more external K-layer receives inputs from the contralateral eye, and the internal K-layer receives inputs from the ipsilateral eye. The K-layers receive inputs from a special subset of ganglion cells and project to supragranular layers of primary visual cortex (Casagrande, 1994). Their functions are not well understood. Recently, more attention has been directed to the existence of small cells in the interlaminar zones between the traditional layers of the monkey LGN. For instance, the nocturnal owl monkey has a broad interlaminar zone between adjoining M- and P-layers that is packed with small cells. The small cells in the K-layers of prosimians or in the interlaminar zones of all primates are now known to constitute a functional class, in part because they all express the calcium binding protein, calbindin, while M- and P-cells do not (Jones and Hendry, 1989; Hendry and Yoshioka, 1994; Johnson and Casagrande, 1995). Thus, K-cells exist in all primates, forming well-recognized K-layers in prosimians, and thinner K-layers in interlaminar zones of all primates (Casagrande, 1994; Hendry and Reid, 2000).

How does the organization of the LGN of tarsiers compare to those of other primates? An understanding of the laminar organization of the LGN of tarsiers has only gradually emerged. Woollard (1922) provided an early description without recognition of the basic laminar organization. Clark (1930) soon followed with the assertion that the LGN of tarsiers “approximates closely to the anthropoid type,” while failing to distinguish more than three layers or recognize the magnocellular layers as such. Clark (1930) found no small-celled layers as in prosimians (the K-layers). Walls's (1953) interpretation of Clark's description of three layers, instead of the expected primate 6, was that they must all relate to the contralateral eye and that retinal projections must be totally, or nearly totally, crossed. Chacko (1954) followed with recognition of two large-celled or magnocellular layers, while also considering the parvocellular region as an “undivided mass.” In a broad comparative review, Hässler (1967) stated that the LGN of tarsiers has four layers, two magnocellular and two parvocellular, an observation repeated by Kaas et al. (1978). Simmons (1982) subsequently described the nucleus as consisting of two thin ventral magnocellular layers and two, possibly three, dorsal parvocellular layers. However, he concluded that it was not possible to distinguish clearly a separate layer V.

Perhaps the most informative report on the organization of the LGN in tarsiers is that of Rosa et al. (1996). These investigators were able to study retinal projections to the LGN layers in one tarsier (T. bancanus). In Nissl-stained sections from the same case, they recognized four layers, two ventrolateral magnocellular layers and two dorsomedial parvocellular layers. In addition, they described a koniocellular region of small cells between the internal parvocellular and magnocellular layers, much like the thick cell region between these layers in the nocturnal owl monkey (Norden and Kaas, 1978). While histologically the nucleus appears to conform to a basic four-layered anthropoid type (Kaas et al., 1978), with the addition of a broad K-cell interlaminar zone, the labeling of retinal terminations in the LGN revealed something rather startling. The external magnocellular layer (Me) appeared to receive inputs from the ipsilateral eye, and the internal magnocellular layer (Mi) from the contralateral eye, rather than the other way around as in other primates. Yet the external parvocellular layer received inputs from the contralateral eye, and the internal parvocellular from the ipsilateral eye, similar to other primates. Further evidence for this arrangement was based on the finding of a small discontinuity in the internal M-layer and in the external P-layer. Such a discontinuity had been previously shown in a range of species to correspond to the lack of projections from the site of the optic nerve head in the nasal retina, the hemiretina that projects contralaterally (Kaas et al., 1973). This apparent difference distinguishes tarsiers from all other primates, reinforcing the view that they arose in an early, independent line of primate evolution.

The histological organization of the LGN of tarsiers can be appreciated in the photomicrographs of prepared sections in Figure 3. Sections processed for calbindin clearly reveal the location of the broad region of small cells between the MI and PI layers of Rosa et al. (1996) and identify this interlaminar region as densely populated with K-cells. A thinner interlaminar region between the PI and PE layers also has a considerable population of K-cells, while a sparse population of K-cells exists on each side of layer ME. These interlaminar distributions of K-cells are characteristic of all studied primates. These K-cells have been previously described in the LGN of tarsiers by Hendry and Casagrande (1996). The calbindin preparations also help distinguish the four main layers by labeling the interlaminar zones.

Figure 3.

Sections from the lateral geniculate nucleus of three primate species, stained either for Nissl substance or calbindin. A: A Nissl-stained sagittal section from Tarsius illustrates a large parvocellular mass, wherein individual P-layers are not easily defined, and two ventral magnocellular layers. Calbindin (Cb) staining of the adjacent section reveals a thin interlaminar zone separating the P-layers, densely packed with koniocellular (K) cells. A wider interlaminar zone separates P- and M-layers ventrally with more sparsely distributed K-cells on either side of the M-layers. Asterisks mark points of alignment (blood vessels). B and C: Coronal and horizontal Nissl-stained sections from tarsiers show a clear laminar pattern. K-cells revealed by Cb staining occupy interlaminar zones. In B and C, asterisks mark a common blood vessel and arrows mark the location of the representation of the optic disk. D: Sagittal sections from a prosimian galago stained for Nissl and Cb illustrate the distinct laminar pattern typical of primates and the prominent K-layer typical of prosimians. E: Sagittal sections from a New World owl monkey illustrate a similar pattern, with K-cells distributed in interlaminar regions.

Another specialization of the LGN in tarsiers was described by McDonald et al. (1993). The intensity of acetylcholinesterase (AChE) staining was previously noted to be higher in the parvocellular layers of nocturnal prosimians and nocturnal owl monkeys (Fitzpatrick and Diamond, 1980), but in diurnal squirrel monkeys it was higher in the magnocellular layers (Hess and Rockland, 1983). McDonald et al. (1993) reported that AChE was higher in the parvocellular layers of tarsiers, suggesting a further specialization of the visual system for vision in dim light.

Pulvinar Complex

Traditionally, the pulvinar complex of primates has been divided into inferior, lateral, medial, and anterior nuclei (Stepniewska, 2003). The anterior or oral pulvinar has connections with somatosensory areas of cortex and is not part of the visual pulvinar. Likewise, the medial pulvinar has widespread connections with frontal, cingulate, and posterior parietal cortex and cannot be considered strictly visual. Thus, our concern here is with the lateral and inferior divisions of the pulvinar, which have connections with striate, extrastriate, and temporal visual areas. Fortunately, there has been considerable progress recently in understanding the organization of the lateral and especially the inferior pulvinar in monkeys and other anthropoid primates, so that the use of histochemical markers to identify functional subdivisions of the inferior pulvinar can be applied to tarsiers. Until now, the organization of the pulvinar complex has been almost ignored in tarsiers, and thus it has been poorly understood. Clark (1930) recognized a pulvinar in Tarsius at a time when its role in vision was uncertain, and he subsequently stated that the pulvinar was larger and more differentiated in tarsiers than in mouse lemurs. More recently, Simmons (1982) divided the pulvinar into oral, lateral, medial, inferior, and caudal regions, but these divisions were not obvious in the illustrated Nissl material.

The most significant finding of our present histochemical investigation of pulvinar organization in tarsiers is that subdivisions of the inferior pulvinar closely resemble those described in other primates. Based on differences in histochemistry, the inferior pulvinar of monkeys has been divided into posterior (PIp), medial (PIm), central medial (PIcm), and central lateral (PIcl) nuclei (Stepniewska and Kaas, 1997; Stepniewska et al., 2000). These nuclei differ in their connections with areas of visual cortex and in inputs from the superior colliculus. The organization of the inferior pulvinar in prosimian primates has been less certain because of a lack of histochemical studies that might reveal these four subdivisions. However, studies of cortical connections indicate that at least three nuclei exist within the inferior pulvinar (Symonds and Kaas, 1978), and our recent histochemical studies of the thalamus in galagos clearly reveal four subdivisions of the inferior pulvinar.

As shown in Figure 4, histochemical subdivisions of the inferior pulvinar are just as apparent in tarsiers as in owl monkeys. These results are consistent with the conclusion that the inferior pulvinar plays a similar functional role in tarsiers, prosimians, and monkeys. As the four subdivisions of the inferior pulvinar project to different areas of visual cortex in monkeys, the results suggest that tarsiers also have these cortical areas.

Figure 4.

The inferior pulvinars from a tarsier, a galago (Otolemur), and owl monkey (Aotus) appear to be similarly organized with four main subdivisions. The tarsier section is stained for calbindin, which densely stains neuropil and cell bodies in posterior (PIp) and central medial (PIcm) subdivisions and less densely stains medial (PIm) and central lateral (PIcL) subdivisions. In Aotus, the four subdivisions are defined with acetylcholinesterase (AChE) staining, which is densely distributed in PIcL and PIm and less densely in PIcm and PIp.

Architectonically, tarsiers also have a lateral pulvinar, PL. In monkeys and galagos, PL is interconnected with V1 and V2, and it contains a systematic representation of the contralateral visual hemifield (Stepniewska, 2003).

Superior Colliculus

Except for some difference in shape, the superior colliculus is not notably different in tarsiers than in galagos and owl monkeys (Fig. 5). The superficial gray is darkly stained in cytochrome oxidase preparations, and the laminar organization appears to be similar to that in other nocturnal primates. Tilney (1927) concluded that the superior colliculus of tarsiers is larger and better defined than in other primates. Polyak (1957) illustrated the superior colliculus in surface views and considered it well developed, but less so than in those mammals such as squirrels and tree shrews, where the superior colliculus dominates as a visual structure.

Figure 5.

Sections from the superior colliculus of Tarsius, Otolemur, and Aotus are shown stained for either Nissl substance (thionin) or cytochrome oxidase (CO). In coronal (A) and sagittal (B) sections from tarsiers, the large surface area of the superior colliculus is illustrated. The laminar pattern revealed by Nissl and the CO pattern appears to be similar to other nocturnal primates, such as the New World owl monkey (Aotus trivirgatus) shown in C and the prosimian galago (Otolemur garnetti) shown in D.

Primary Visual Cortex, V1 (Area 17)

As in other primates, primary visual cortex of tarsiers is located on the posterior pole of the cerebral hemispheres (Fig. 6). The locations of V1 in owl monkeys and galagos are shown for comparison. An obvious distinction of the tarsier brain in these photographs is the presence of few fissures in the neocortex. The lateral fissure appears as a groove on the cortical surface of the tarsier brain, but in sectioned material it clearly forms a shallow fissure (Fig. 9B). In contrast, owl monkeys and galagos have a deep lateral fissure. A clear superior temporal fissure is evident in owl monkeys. Galagos have an intraparietal fissure and two small fissures in the frontal lobe. All three primates, and all primates, have a calcarine fissure that indents the cortex of the medial wall of the occipital lobe (not shown).

Figure 6.

Photographs of whole brains from Tarsius, Otolemur, and Aotus with estimates of rostral V1 boundaries indicated. Scale bars are 1 cm. The tarsier brain differs from other primates in its lack of deep fissures. Galagos and owl monkeys have a deep lateral fissure (LS). Owl monkeys also have a superior temporal sulcus (STS), while galagos have an intraparietal sulcus (IPS) and two other shallow fissures in the frontal lobe. V1 is large relative to whole brain size in tarsiers. Photographs of the whole tarsier brain and the whole owl monkey brain are reproduced with permission from images by W. Welker in the University of Wisconsin and Michigan State Comparative Mammalian Brain Collections. Preparation of images in the collection was funded by the National Science Foundation and the National Institutes of Health (http://www.brainmuseum.org/).

The primary visual area, V1, of tarsiers is unique among primates in two respects. First, the proportion of neocortex occupied by V1 (area 17) is the largest yet reported. In perhaps the most accurate measure, Stephan (1984) found that area 17 constitutes 21% of neocortex in tarsiers. Previous reports (Polyak, 1957) agreed on the large size, but estimates varied from as little as one-fifth to one-third to as much as one-half of the neocortex. At 21%, tarsiers have proportionately more primary visual cortex than any other primate, although the very small mouse lemur is close (Stephan, 1984). The proportionately large V1 in both of these small primates would be important in maintaining detailed vision (Kaas, 2000). Given the concentration of receptors in and near the fovea in tarsiers (Hendrickson et al., 2000), we would predict that much of V1 is devoted to central vision, further enhancing detailed vision, presumably for detecting small insect prey in dim light.

All previous investigations have noted the unique differentiation of six layers and additional sublayers in tarsiers (Fig. 7A). Here we use the terminology of Hässler (1967) for layers and sublayers, which places sublayers IVA and IVB of Brodmann (1909) in layer III (for a review of the issue, see Casagrande and Kaas, 1994). In Nissl-stained sections, layer III of tarsiers has a middle granular region (IIIb) that separates a less dense IIIa from a cell-sparse IIIc. Layer IV has two cell-dense bands (IVa and IVc), separated by a narrow cell-sparse sublayer (IVb). Layers V and VI were not subdivided by Hässler (1967), but both layers V and VI have an obvious division into an upper cell-dense layer Va and VIa and a lower cell-sparse layer Vb and VIb. The laminar pattern is clearly more distinct than in nocturnal galagos and owl monkeys (Fig. 6, top), but not markedly more than in area 17 of the small marmoset monkeys (Hässler, 1967). Other differences in lamination in V1 of tarsiers are revealed by the cytochrome oxidase (CO) preparation (Fig. 7B). Layer IV is very dark, corresponding to the common description in primates, and reflecting the high metabolic activity of layer IV neurons. Layer IVb is slightly less dark than layers IVa and IVc. Layer IIIb with its granular cells is somewhat darker than other parts of layer III, as this part of layer III receives some direct parvocellular geniculate inputs in most studied anthropoid primates (Casagrande and Kaas, 1994). This reflection of enhanced activity suggests a layer III geniculate projection in tarsiers as well, although such a projection has not been reported in galagos or in nocturnal owl monkeys. A narrow strip of higher CO activity is also seen along the junction of layers V and VI, reflecting the presence of the large Meynert cells that project to visual area MT (Spatz, 1975), as well as subcortically (Fries and Distel, 1983). Finally, close inspection reveals a sequence of patches of high CO activity across the length of layer III (Hendrickson, 1985). These patches are generally called blobs and have been variously thought to correspond to regions where neurons are color-selective and poorly selective to stimulus orientation (Casagrande and Kaas, 1994). In tarsiers, the blobs clearly extend into layers V and VI, but this is not apparent in all primates. McGuinness et al. (1986) previously failed to detect blobs in tarsiers, but it is difficult to get postmortem brains in good enough condition to reveal CO blobs (Preuss and Kaas, 1996). The present evidence supports the conclusion that the blob modules emerged early in primate evolution and have been retained in all major branches (Preuss and Kaas, 1996). The border of V1 with V2 is clearly revealed by changes in CO histochemistry (Fig. 7C).

Figure 7.

Nissl- and CO-stained sections from primary visual cortex (V1) of tarsier, owl monkey and galago. A: A section from V1 of a tarsier reveals the extremely distinct laminar pattern of this nocturnal species compared to owl monkey and galago. Three subdivisions of layers III and IV are easily defined, while both layers V and VI have two clear sublayers, a and b. A section from V1 of a New World owl monkey shows a fairly clear laminar pattern, but fewer sublaminar divisions than tarsier. In owl monkey, both layers III and IV have three sublayers, but neither layer V nor VI are subdivided. V1 from a prosimian galago is not as well differentiated. Layer III has no sublamination. Layer IV is divided into two sublayers and layers V and VI have no sublamination. B: CO-stained sections from tarsier, owl monkey, and galago help to define laminar patterns further in V1. Tarsiers, like some diurnal species of monkey, have a band of CO density in layer IIIb, representing a direct parvocellular input to this layer. Neither owl monkeys nor galagos have this CO band in layer III. Tarsiers have a typical dense CO staining pattern in layer IV, resulting from the direct thalamic terminations to layer IV. This CO-dense layer IV is visible in both owl monkeys and galagos. Tarsiers also appear to have CO-dense blobs in V1 (indicated by arrows), similar to other primates, including owl monkeys and galagos. Tarsier blobs appear to extend throughout the depth of cortex to layer VI. This is not apparent in owl monkeys or galagos. C: Lower-magnification views of the cytochrome oxidase-stained sections from tarsier, owl monkey, and galago illustrate the sharp boundary between V1 and V2 in all three species (marked by arrows) and the regular distribution of blobs in V1. Blobs are less apparent in the low-magnification tarsier section due to heavy fixation of the tissue, which resulted in suboptimal CO staining.

Other stains and histochemical evidence add to the evidence that the layers and sublayers of V1 are functionally distinct. Immunocytochemistry for the calcium binding protein, calbindin, revealed that the cell-sparse sublayers IIIc and IVb have high levels of calbindin, layer Vb has a moderate level, and IIIa and II have moderate levels (Fig. 8A). The high levels in IIIc and IVb are specific to V1 as the label ends precisely at the V1/V2 border. Parvalbumin and calbindin typically form complimentary distributions, and V1 of tarsiers is no exception. Thus, parvalbumin is highly expressed in sublayers IVa and IVc (Fig. 8B). A row of solitary giant cells of Meynert at the layer V–VI junction is also highly parvalbumin-positive. The lectin WFA labels a more varied collection of neurons in V1 (Fig. 8C). Finally, myelin stains indicate that V1 is characterized by both a dense array of radial fibers and horizontal fibers (Fig. 8D). The radial fibers are more dense in V2, and the horizontal pattern is not apparent. The horizontal band of fibers in the middle layers of V1, the band of Gennari of other primates, is hardly evident.

Figure 8.

Features of V1 and V2 revealed by staining with calcium binding proteins, CB and PV, WFA, nissl, myelin, and cytochrome oxidase. A: CB staining defines the V1/V2 border sharply and allows close approximation of the rostral V2 boundary. In V2, cell bodies and neuropil are densely stained in layers II and more sparsely in layer III. Stained neurons are sparsely distributed in lower layer III and layer IV. Neuropil staining is slightly more dense in layer V and layer VI contains a population of CB-positive cells, but neuropil is CB-light. B: PV staining of V2 results in the opposite pattern with PV-immunoreactive (IR) cells throughout the cortical depth, but more sparse in upper layers and reduced neuropil PV-IR. The rostral V2 border is easily defined, as most of V2 has reduced neuropil PV-IR. C: V2 staining with the lectin WFA defines both caudal and rostral V2 boundaries. Most WFA-IR is confined to layer V (neuropil and cells) and VI (cells), but some sparse immunoreactivity is seen in layer IV. D: Myelin staining is more dense in V2 compared to V1 and the V1/V2 border is easily defined. The rostral V2 boundary is less clear. E: Nissl staining crisply defines the V1/V2 border and layer IV of V2 appears to be densely packed with small cells relative to layers III and V. No sublamination is evident in V2, but the six cortical layers are well defined. At the rostral V2 boundary, cellular density appears to increase throughout the cortical depth. F: CO staining of tarsier cortex shows an abrupt reduction in CO density in layer IV at the V1/V2 border. Upper layers of V2 are more CO-dense, while layers V and VI are CO-light relative to V1. The rostral V2 border is evident, but not as clear as the caudal boundary.

Second Visual Area, V2

While the border of V2 with V1 is exceptionally distinct in tarsiers, the rostral border of V2 is not. The histological characteristics of V2 are very similar to those of adjoining extrastriate visual areas, presumably V3 and perhaps DL (Kaas and Lyon, 2001). Yet in Nissl and CO preparations (Fig. 8E and F), we detect a subtle change in features about 1.5 mm from the V1/V2 border. Thus, we estimate V2 to be about 1.5 mm in width. However, this architectonic evidence is questionable, and more certain evidence for V2 would depend on visualization of the CO banding pattern in brain sections cut parallel to the cortical surface, as has been described mainly in anthropoid primates (Krubitzer and Kaas, 1989), but also some prosimians (Preuss et al., 1993). In Nissl-stained sections, Hässler (1967) identified a V2 (area 18) in tarsiers of about the same width, and Bonin (1951) delimited a similar area as parakoniocortex.

Middle Temporal Area (MT)

MT is a visual area that has been identified in a wide range of primates, and thus it is thought to be common to all primates (Kaas, 1997). Hence, it would be surprising if evidence for MT were not found in tarsiers. Ideally, such evidence would consist of identifying histological characteristics, a demonstration of direct inputs from V1, and the presence of a retinotopic representation of the contralateral visual hemifield that mirrors that of V1. Here we can only consider histological features. Fortunately, one of the most notable features of MT is its dense myelination, and that feature alone has often been used to identify MT. In our myelin-stained sections from tarsier brains (Fig. 9, left), we do see a myelin-dense region in the expected location of MT, and thus MT appears to be present. However, further study of this material is needed to see how closely the myelin pattern, the dimensions and location of the field, and other histological features conform to those expected of MT. The presumptive MT of tarsiers is the correct size for MT (3–4 mm rostrocaudally), but this MT is a little close to V2, leaving little room for the dorsolateral visual area (DL or V4). Possibly DL is reduced in size in tarsiers. Note also in Figure 9 the histological evidence for primary auditory cortex, A1.

Figure 9.

Features of the MT revealed in horizontal sections stained for (left) myelinated fibers and (right) cell bodies in a Nissl preparation (tarsier 01-92). The sections show the V1/V2 border clearly, our estimate for the rostral border of V2, and our estimate of the MT boundaries. MT is most apparent as a myelin-dense area in the left panel. In Nissl preparations, MT is identifiable by the lower cell density in layers IV and VI. Primary auditory cortex, A1, has the well-developed layer IV of primary sensory areas. The small, more lightly myelinated area rostral to MT may be the middle superior temporal area (MST). Scale is 2 mm.


Overall, the most remarkable feature of the tarsier brain, to us and all previous investigators, is the large size and distinct lamination of V1. There is no obvious explanation for this specialization, but we suggest that it is related to the small size of the brain and the unique behavioral specialization of tarsiers as a predator of insects and small vertebrates that eats no vegetable food (Ross, 1996). Thus, even in dim light, tarsiers would require good visual acuity to detect small prey, and possibly good stereopsis so that camouflaged prey would stand out from their background. Hässler (1967) presented a related reason why V1 is so large in tarsiers. As leapers from branch to branch in its exclusively arboreal niche, Hässler (1967) proposed that tarsiers would need very precise stereoscopic vision. Visual acuity depends not only on the receptor array, but on the size of primary visual cortex. Cortical areas that are very small do not have enough neurons to mediate the locations of objects in space with a high degree of precision (Kaas, 2000), and mammals with very small brains may devote larger parts of cortex to behaviorally relevant primary sensory areas while reducing the proportional sizes of other areas or even eliminating some of them (Catania et al., 1999). Thus, tarsiers may have preserved a high level of visual acuity by enlarging V1 at the expense of other areas. Compensations for the loss of function in other visual areas might be the reason for the extreme differentiation of layers and sublayers in V1. It may be no coincidence that the lamination of V1 in tarsiers resembles superficially the high degree of lamination of the optic tectum of predatory birds. In both structures, the specialized layers may allow more functions to be stacked over each other for differential processing within the same fine-grain visuotopic map.

Another remarkable feature of the visual system of tarsiers is the large number and unusual distributions of cones in the retina (Hendrickson et al., 2000). The longer-wavelength cones (M or L) are concentrated in the central retina as one might expect for a fruit-eating diurnal anthropoid primate, but tarsiers capture small prey in dim light. One possible explanation might be an evolutionary retention from the diurnal ancestors of tarsiers, but this seems questionable in view of the great modifications that have occurred in the central visual system. The observation that other nocturnal primates, including galagos and owl monkeys, also have many cones in the central retina suggests that L- or M-cones in the central retina mediate useful functions, such as foveal acuity and color vision, when the light is not too dim. The unique distribution of S-cones in the peripheral rather than the central retina is another puzzle. Hendrickson et al. (2000) suggest that the S-cones “add value” to nocturnal vision, perhaps by aiding prey detection in peripheral vision. Preliminary sequencing of the S-opsin protein from tarsiers suggests that these cones are sensitive to UV, not blue light. If confirmed, this would make tarsiers the only primate with UV detection capability. Because field observations show that tarsier prey insects have cuticles and/or blood that fluoresces in the UV, a ring of peripheral UV-sensitive cones would be ideal for detecting prey in the dense Indonesian forest.

The lateral geniculate nucleus of tarsiers clearly conforms to the anthropoid rather than the strepsirrhine pattern in that the two koniocellular layers are absent. However, the nucleus also has specializations for nocturnal vision, in that the proportion of the parvocellular layers to the magnocellular layers is reduced, as in other nocturnal primates (Stephan, 1984), and the thickness of the interlaminar zone, between the M- and P-layers, is increased and packed with K-cells. The significance of K-cells for tarsier vision is uncertain, but nocturnal primates do have more K-cells (Johnson and Casagrande, 1995), suggesting they are especially important for vision in dim light.

The inferior pulvinar also shows the typical pattern of four distinct nuclei of other primates. Two of these nuclei are interconnected with V1 and V2 in other primates, and these areas are present in tarsiers. The other two nuclei largely interconnect with cortex in and around visual area MT in other primates, adding to the present architectonic evidence that tarsiers have an MT, and possibly the MT satellite areas (FST, MST, MTc).


The authors would like to acknowledge the assistance of the staff of the Indonesian Primate Center at Bogor Agricultural University, Bogor, Java.