Specializations Of The Visual System Overview

Discussion of visual system evolution requires a very brief review of the anatomy of the visual system (for more details, see especially Kaas and Huerta, 1988; Kaas et al., 1978). Visual information reaches the brain by way of projections from the ganglion cells of the retina. Retinal ganglion cells are of several types, with different morphologies and physiological properties. Currently, three main classes of ganglion cells are distinguished, usually called M, P, and K cells in primates. This designation reflects the fact that these cells project to separate magnocellular, parvocellular, and koniocellular layers of the lateral geniculate nucleus. It is likely that nonprimate mammals have cells homologous to the M, P, and K cells (usually termed Y, X, and W cells, respectively, in nonprimates).

M cells have large cell bodies and dendritic fields; they have large visual receptive fields that integrate information from both rods and cones and respond well to light over a broad range of the visual spectrum. M cells have good temporal resolution, which is to say they respond well to flickering or moving stimuli, but have relatively coarse spatial resolution, so they are not well suited for fine visual discrimination. P cells are smaller than M cells, and have smaller receptive fields that integrate inputs from cones; they respond well to light from a more restricted part of the spectrum than M cells (i.e., they are relatively wavelength selective), have poor temporal resolution (they track stimulus onset and offset poorly), but have good spatial resolution. P cells and their efferent targets process color information in diurnal anthropoids, although they must do more than this, because P cells are present in nocturnal primates, and nocturnal primates lack the density of cones and diversity of cone types necessary to support color-opponent processing, which is the basis of fine color discrimination in diurnal anthropoids (Dkhissi-Benyahya et al., 2001). K cells are poorly characterized at present; they seem to be intermediate between M and P cells morphologically and physiologically. Their heterogeneity suggests that the K class is actually a composite of multiple cell classes. Interestingly, the short-wavelength photoreceptors (S cones), which respond maximally to light in the blue part of the spectrum, have recently been indicated (in anthropoids) to have a privileged anatomical relationship to K cells rather than to P cells (Dacey, 2000), which challenges the conventional idea that color processing is carried out exclusively in the P pathway. It is unclear at present whether all S cones are related to K cells and whether all K cells receive S-cone input.

Retinal ganglion cells project to many structures in the brainstem, the strongest projections targeting the lateral geniculate nucleus (LGN) of the thalamus and the superior colliculus (SC) in the midbrain (Figure 1). The LGN contains separate cell types linked to each of the different retinal ganglion cell types, and these cells are segregated into distinct strata or laminae. In primates, there are separate magnocellular and parvocellular layers, which receive inputs from the M and P retinal ganglion cells, respectively. These layers come in pairs, so that separate M and P layers represent each eye. The layers sometimes subdivide further, as in humans, in which it is conventional to recognize six main LGN layers: one pair of magnocellular layers and two pairs of parvocellular layers (Kaas et al., 1978). In strepsirhine primates, the konio-cellular layers also form pairs of layers that segregate input from the two

Konio Cortical Field Area

Retina LGN

Ancestral condition

Lgn Superior Colliculus

Retina LGN

Primate condition

Figure 1. Schematic representation of the distribution of visual information to the thalamus, superior colliculus, and cortex, illustrating ancestral organization and derived characteristics of primates. Visual information originating from the P, M, and K cells of the retina reaches the thalamic lateral geniculate nucleus (LGN) and the superior colliculus. The lateral geniculate nucleus (LGN) relays information from the retina and superior colliculus to the primary visual area (V1) and to extrastriate visual cortex. In primates, the K layers of the LGN target specialized compartments within area V1, the blobs (represented here as an array of dark spots). The LGN also projects to extrastriate cortex, although these projections are weak in primates and appear to arise primarily from the K layers of the LGN. Extrastriate cortex receives additional projections from inferior pulvinar of the thalamus in primates, and from its homologue in nonprimates, known as the pulvinar/lateral posterior (LP) nucleus. The inferior pulvinar receives major inputs from the superficial (visual) layers of the superior

(Continued)

Retina LGN

Primate condition

Figure 1. Schematic representation of the distribution of visual information to the thalamus, superior colliculus, and cortex, illustrating ancestral organization and derived characteristics of primates. Visual information originating from the P, M, and K cells of the retina reaches the thalamic lateral geniculate nucleus (LGN) and the superior colliculus. The lateral geniculate nucleus (LGN) relays information from the retina and superior colliculus to the primary visual area (V1) and to extrastriate visual cortex. In primates, the K layers of the LGN target specialized compartments within area V1, the blobs (represented here as an array of dark spots). The LGN also projects to extrastriate cortex, although these projections are weak in primates and appear to arise primarily from the K layers of the LGN. Extrastriate cortex receives additional projections from inferior pulvinar of the thalamus in primates, and from its homologue in nonprimates, known as the pulvinar/lateral posterior (LP) nucleus. The inferior pulvinar receives major inputs from the superficial (visual) layers of the superior

(Continued)

retinas; the situation is less clear in anthropoids, because their K layers are not all distinctly separate from the M and P layers (Kaas et al., 1978).

The pattern of LGN lamination varies considerably among mammals (reviewed by Kaas and Preuss, 1993; Kaas et al., 1972; Sanderson, 1986). Cell types that are mixed in primates can be segregated into separate laminae in other taxa. On current evidence, primates are the only mammals in which P cells are known to be completely segregated from M cells. The pattern of lamination in dermopterans and chiropterans is not well understood, however, and they may have a primate-like pattern, as indicated by Pettigrew et al. (1989).

Each geniculate layer receives projections from one hemiretina, which form a map of the contralateral half of the visual scene. So, for example, the left LGN contains a stack of maps of the right visual scene; half of these maps are formed by inputs from the left retina, half from the right retina. Since the right side of the visual scene is projected onto the left side of each retina, the left LGN receives its projections from the left (temporal) side of the left retina and from the left (nasal) side of the right retina. The fibers from the nasal hemiretinas cross the midline on their way to the thalamus, forming the optic chiasm; projections from the temporal hemiretinas are uncrossed.

In most nonprimate mammals, which have eyes set on the side of the head, the left eye sees mainly the left visual field and the right eye the right visual field; there is only a relatively small region of binocular vision. In these animals, the LGN on each side of the brain receives its major input from the contralateral retina; only a small projection arises from the ipsilateral retina, corresponding to the field of binocular overlap. Primates, by contrast, have forward facing eyes and consequently a large region of overlap between the images cast onto the left and right retinas. Correspondingly, the projections of the ipsilateral retina to the LGN are nearly as numerous as the projections from the contralateral LGN.

Figure 1. (Continued) colliculus. The deeper layers of the superior colliculus project to thalamic nuclei connected mainly with frontal cortex. These include the mediodorsal nucleus (MD) and the intralaminar and ventral thalamic nuclei, which project primarily to premotor cortex (PM) and dorsolateral prefrontal cortex (DLPFC), located on the lateral surface of the frontal lobe, and to the anterior cingulate (AC) cortex, located on the medial wall of the frontal lobe. In primates, the strongest tectal projections reach MD and the medial division of the ventral anterior nucleus; these nuclei project primarily to DLPFC. In nonprimates, the intralaminar nuclei are the main targets of tectal projections. Additional abbreviations: Ml = primary motor area.

The second major target of retinal projections is the superior colliculus (SC), a layered structure occupying the roof (or "tectum") of the midbrain, posterior to the thalamus (Huerta and Harting, 1984). Only the M and K ganglion cells project to the SC, where they terminate most densely in its superficial layers. Retinal terminations are arranged in an orderly, topographic manner. Inputs from each eye are at least partially segregated, although the patterning of segregation is more complex than the simple laminar pattern of the LGN, and unlike the LGN, some tectal cells have binocular inputs. Moreover, the SC is not solely a visual structure—the middle and deep layers receive inputs from the auditory and somatosensory systems—and the auditory and somatosensory maps are in spatial register with the visual map of the superficial layers.

The deep layers of the SC play an important role in organizing orienting movements, mediating the so-called "visual grasp reflex" by which animals rapidly shift their head and eyes so that the image of potentially significant objects fall onto the central retina, where resolving power is greatest. The role of SC as an eye-movement control center is emphasized especially by researchers who study anthropoid primates, animals that make unusually large-amplitude eye movements. In vertebrate neurobiology more generally, however, the SC is seen as a structure that, among other things, coordinates rapid orienting movements of the head, eyes, pinnae, and perioral face region (Dean et al., 1989). There is, furthermore, evidence for forelimb representation in the colliculi of nonprimate vertebrates, and it is significant that recent studies by Werner and colleagues in macaque monkeys implicate SC in the control of proximal forelimb movements associated with reaching (e.g., Werner et al., 1997).

The SC influences other brain regions by several routes (Huerta and Harting, 1984). Descending projections target the oculomotor nuclei and other motor structures in the brainstem and upper cervical spinal cord. Ascending projections target the thalamus and other diencephalic structures, as illustrated in Figure 1. The upper layers of the colliculus project to thala-mic nuclei that provide inputs to visual cortex—the LGN (specifically its K layers) and inferior portions of the pulvinar. Deep layers of the SC project to more anterior structures of the thalamus, particularly the intralaminar, ventral (VA), and mediodorsal (MD) nuclei. The projections to the MD and VA nuclei, which project in turn to frontal cortex, may be especially strong in primates (Huerta and Harting, 1984).

Retinotectal Organization

With this background, we can now consider the best-known specialization of primate brain structure, namely, the distinctive retinotopic organization of the primate superior colliculus (Figure 2). The term retinotopy refers to the way retinal information is mapped onto visual structures. In primates, the pattern of retinal projections to SC and the resulting visual field representation in SC are similar to the pattern exhibited by the LGN: each SC contains a complete representation of the contralateral visual field. This representation is supported by major projections from the nasal retina of the contralateral eye and from the temporal retina of the contralateral eye. There is no substantial representation of the ipsilateral visual field in the SC of primates: the representation of the vertical meridian is located at the anterior limit of the colliculus. This pattern of organization has been found in every primate that has been examined with microelectrode mapping methods or by tracing projections from the retina to the SC, a sample that includes numerous New World and Old World anthropoid species, as well as lorisid strepsirhines (see especially Kaas and Huerta, 1988, for a review of the published literature).

Although the primate condition was once thought to be typical of mammals, work in the early 1970s established that the primate condition is unusual. In most nonprimate mammals examined, the visual representation in SC crosses the vertical meridian to include a significant portion of the ipsilat-eral visual field, in addition to the contralateral visual field. Moreover, rather than receiving nearly equal projections from both retinas, SC projections in nonprimate mammals arise mainly from the contralateral retina, and span a territory that includes both the temporal and nasal hemiretinas, with a relatively small contribution from the ipsilateral retina. This type of organization has been found in a variety of mammalian taxa, including tree shrews, rodents, lagomorphs, artiodactyls, perissodactyls, carnivores, marsupials, and monotremes (see Allman, 1977, Kaas and Huerta, 1988, Pettigrew, 1986, and Rosa and Schmid, 1994, for citations to the extensive primary literature). This is presumably the ancestral mammalian SC organization.

The apparently clear dichotomy between primate SC organization and the ancestral mammalian condition makes SC organization potentially a useful character for sorting out phyletic relationships among the mammalian taxa that have been considered to be particularly closely related to primates (Figure 3).The taxa most commonly touted as close relatives of the Order Primates are

Figure 2. Schematic representation of retinotectal organization, illustrating ancestral organization and derived characteristics of primates. The sections are oriented approximately in the horizontal plane, passing through the eyes, optic chiasm, and superior colliculi. In primates, each colliculus contains a complete representation of the contralateral visual field; the representation of the vertical meridian (VM) lies at the rostral pole of the colliculus, and strong projections reach the colliculus from both eyes. In the ancestral, each colliculus represents the field of view of the contralateral eye; inputs arise mainly from the contralateral eye, with a small contribution from the ipsilateral eye. This schematic is based on Figure 1 of Pettigrew et al. (1989).

Figure 2. Schematic representation of retinotectal organization, illustrating ancestral organization and derived characteristics of primates. The sections are oriented approximately in the horizontal plane, passing through the eyes, optic chiasm, and superior colliculi. In primates, each colliculus contains a complete representation of the contralateral visual field; the representation of the vertical meridian (VM) lies at the rostral pole of the colliculus, and strong projections reach the colliculus from both eyes. In the ancestral, each colliculus represents the field of view of the contralateral eye; inputs arise mainly from the contralateral eye, with a small contribution from the ipsilateral eye. This schematic is based on Figure 1 of Pettigrew et al. (1989).

Figure 3. Alternative interpretations of the phyletic relationship of primates to other mammalian orders. Most modern workers accept that primates are closely related to tree shrews (order Scandentia), flying lemurs (order Dermoptera), and bats (order Chiroptera). Bats are comprised of two major subgroups, the Megachiroptera (fruit bats, flying foxes, and related taxa) and the Microchiroptera (echolocating bats). The status of bats is a matter of major disagreement in these interpretations. (A) Many accounts consider bats to be monophyletic and to belong within the Archonta, as in the interpretation of Novacek (1992). (B) In the "flying primate" hypothesis favored by Pettigrew et al. (1989), bats are considered to be diphyletic. Megachiropteran bats are included in Archonta, but microchiropterans are held to be distantly related to both megachiropterans and to primates. (C) The recent molecular results of Murphy et al. (2001) indicate that bats are monophyletic, but distantly related to primates. This analysis, like a number of others, places the superorder Archonta within a larger group, called Euarchontoglires, that includes rodents and lagomorphs.

Figure 3. Alternative interpretations of the phyletic relationship of primates to other mammalian orders. Most modern workers accept that primates are closely related to tree shrews (order Scandentia), flying lemurs (order Dermoptera), and bats (order Chiroptera). Bats are comprised of two major subgroups, the Megachiroptera (fruit bats, flying foxes, and related taxa) and the Microchiroptera (echolocating bats). The status of bats is a matter of major disagreement in these interpretations. (A) Many accounts consider bats to be monophyletic and to belong within the Archonta, as in the interpretation of Novacek (1992). (B) In the "flying primate" hypothesis favored by Pettigrew et al. (1989), bats are considered to be diphyletic. Megachiropteran bats are included in Archonta, but microchiropterans are held to be distantly related to both megachiropterans and to primates. (C) The recent molecular results of Murphy et al. (2001) indicate that bats are monophyletic, but distantly related to primates. This analysis, like a number of others, places the superorder Archonta within a larger group, called Euarchontoglires, that includes rodents and lagomorphs.

tree shrews (Scandentia), flying lemurs (Dermoptera), and bats (Chiroptera); collectively, these orders are held to constitute the superorder Archonta (Gregory, 1910; McKenna, 1975). A number of recent studies suggest that rodents and lagomorphs (superorder Glires) are also closely related to primates and tree shrews (e.g., Miyamoto, 1996; Murphy et al., 2001; Shoshani and McKenna, 1998). What is the condition of SC in these groups? Rodents, lagomorphs, and tree shrews all retain the ancestral character state, as noted in an earlier section. Pettigrew (1986), however, presented evidence indicating that some bats—specifically, the megachiropteran bats ("megabats")—possess the primate condition, while microchiropteran bats ("microbats") retain the ancestral SC organization. In addition, Pettigrew et al. (1989) maintained that flying lemurs also show a primate-like condition. Pettigrew and colleagues concluded that a clade comprised of Megachiroptera plus Dermoptera is the sister group of primates. This is the "flying primate" hypothesis. An important corollary of this view is that megachiropterans and microchiropterans are not sister taxa, and therefore Chiroptera is not a natural, monophyletic taxon.

Pettigrew's initial report (Pettigrew, 1986) was based on recordings and tracer injections in six individuals from three species of the megachiropteran genus Pteropus (flying foxes) and two individuals of Macroderma gigas, a microchiropteran species with a relatively well-developed visual system. In a subsequent monograph, Pettigrew et al. (1989) indicated that the megabat Rousettus aegyptiacus and the dermopteran Cynocephalus variegatus also have primate-like SCs. They adduced additional anatomical features (mostly features of the visual system) that unite primates, megachiropterans, and dermopter-ans. Anatomical data were not presented in detail in their monograph, however, and some of the character states they attributed to megachiropterans and dermopterans have been questioned (Kaas and Preuss, 1993).

The claim that megabats have a primate-like SC has also been challenged. Thiele et al. (1991) examined the megabat Rousettus aegyptiacus, using tracer injections of the eye and the SC, and recordings from the SC. In contrast to Pettigrew, they concluded that the SC visual representation in Rousettus is not restricted to the contralateral visual field, but rather extends at least 25° past the vertical meridian into the ipsilateral field. Furthermore, their tracing studies indicated that major projections to SC arise from both the nasal and temporal portions of the contralateral retina, there being only a small projection from the ipsilateral retina. They concluded, therefore, that Rousettus retains the ancestral mammalian condition. The results of Thiele et al. prompted Rosa and Schmid (1994) to reexamine SC organization in the megabat Pteropus using microelectrode recording, and they concluded that while visual receptive fields do indeed extend into the ipsilateral visual field, the ipsilateral representation is much less than reported in Rousettus by Thiele et al. Rosa and Schmid suggest that the differences between their results and those of Thiele et al. mainly reflect differences in methodology, differences that led Thiele et al. to systematically overestimate the extent of ipsilateral representation. Even if one accepts this, however, megabats would still seem to have more ipsilateral representation than is found in primates. While acknowledging that megabats retain some ancestral features of SC organization (including strong projections from the contralateral eye), Rosa and Schmid affirm the view that the pattern of visuotopic representation in Pteropus is primate-like.

At the same time that neuroscientists have been debating the condition of the megabat superior colliculus, phylogenetic studies have reduced its significance as an indicator of phyletic relationships. A growing body of evidence, particularly from comparative molecular investigations, indicates that tree shrews and flying lemurs, rather than bats, are the closest living relatives of primates (e.g., Adkins and Honeycutt, 1991; Ammerman and Hillis, 1992; Bailey et al., 1992; Murphy et al., 2001). Indeed, current evidence suggests that bats are very distantly related to primates and should not be included within Archonta at all (Murphy et al., 2001). In addition, an impressive array of anatomical and molecular data supports bat monophyly (see, for example, Allard et al., 1999; Honeycutt and Adkins, 1993; Murphy et al., 2001; Novacek, 1992; Shoshani and McKenna, 1998; Simmons, 1994). Thus, even if it were to be clearly demonstrated that the megabat SC is primate-like, that similarity would now have to be considered convergent rather than homologous in the context of the full range of comparative information currently available.

Blobs

In primates, and in the nonprimate taxa that have been studied, the SC sends projections to the LGN, which terminate specifically in the koniocellular (K) layers (Huerta and Harting, 1984). The K layers, along with the M and P layers, project to visual areas in the posterior part of cerebral isocortex, where the densest projections of the LGN terminate in the so-called primary visual area (V1). These projections have been studied in great detail in primates (for review, see Casagrande and Kaas, 1994). As shown in Figure 4, the P and M layers project to largely nonoverlapping levels within the middle cortical layer (layer 4) of V1, the P layers terminating in the deep part of layer 4 while the M projections terminate in a band just superficial to the main P projection. In most (but not all) anthropoid primates, an additional, thin band of P-layer projections terminates above the band of M projections. The K projections differ markedly from the M and P projections, terminating primarily in the superficial cortical layers, specifically in layers 1 and 3. Within layer 3, the K terminations are clustered into repeating, regularly spaced territories separated by tissue that lacks K inputs.

Macaca soo um P M K

Figure 4. The laminar and compartmental distribution of projections from the LGN to the primary visual cortex (area V1) in macaque monkeys. Left: A section through area V1 stained for cell bodies using a Nissl stain; cortical layers are numbered in conventional fashion. Right: The main targets of projections from the P, M, and K layers of the LGN indicated on a tissue section stained for cytochrome oxidase (CO). The main M and P projections are distributed in horizontal bands at different levels within layer 4. The K projections, however, are distributed mainly to patchy territories of dense CO staining within layer 3 called "blobs"; these are indicated here with asterisks. Additional projections from the M and P cells to layer 6 and from the K cells to layer 1 are not shown.

Macaca soo um P M K

Figure 4. The laminar and compartmental distribution of projections from the LGN to the primary visual cortex (area V1) in macaque monkeys. Left: A section through area V1 stained for cell bodies using a Nissl stain; cortical layers are numbered in conventional fashion. Right: The main targets of projections from the P, M, and K layers of the LGN indicated on a tissue section stained for cytochrome oxidase (CO). The main M and P projections are distributed in horizontal bands at different levels within layer 4. The K projections, however, are distributed mainly to patchy territories of dense CO staining within layer 3 called "blobs"; these are indicated here with asterisks. Additional projections from the M and P cells to layer 6 and from the K cells to layer 1 are not shown.

Conveniently, the punctate zones of K input are marked by high levels of cytochrome oxidase (CO) activity and can be revealed using a histochemical stain for CO. In fact, the discovery of CO-dense zones antedates the discovery of their relationship to the K pathway by Lachica and Casagrande (1992). The patchy distribution of CO staining in layer 3 of primate cortex was initially reported by Horton and Hubel (1981). These patches have come to be called "blobs" or "puffs" (Wong-Riley, 1994). The fact that blobs can be reliably revealed using a relatively simple histochemical stain, provided the tissue to be stained is fresh and not too strongly fixed, has made it possible to study blobs in a much wider array of species than usually receives the attention of neuroscientists. From the start, it has been appreciated that primates generally (and perhaps universally) have blobs, and nonprimates generally (and perhaps universally) do not (Horton and Hubel, 1981). Blobs have been observed in all catarrhine and platyrrhine species examined using suitable tissue, as well as in lorisoids and cheirogaleids (Horton, 1984; Horton and Hedley-Whyte, 1984; Horton and Hubel, 1981; Preuss and Kaas, 1996; Preuss et al., 1993; Wong-Riley, 1988), and evidently in lemurids (Jeo et al., 1997). Definitive evidence for or against the presence of blobs in tarsiers is presently lacking, because it has proven difficult to obtain suitable tissue. By contrast to primates, blobs are absent in tree shrews (Horton, 1984; Jain et al., 1994; Wong-Riley, 1988), murid and sciurid rodents (Horton, 1984; Wong-Riley, 1988), and rabbits (Horton, 1984). Blobs are absent in the one bat genus (Pteropus) that has been examined specifically in this regard (Ichida et al., 2000; see also the figures in Rosa, 1999; Rosa et al., 1993, 1994). Blobs appear to be absent in marsupials (L. H. Krubitzer, pers. comm.; see also the published photographs in Kahn et al., 2000; Martinich et al., 1990; Rosa et al., 1999).

Although carnivores were initially reported to lack blobs (Horton, 1984; Wong-Riley, 1988), more recent studies indicate that at least some carnivores—namely, cats and ferrets—possess alternating territories of dark and light CO staining in the upper layers of area V1 (Boyd and Matsubara, 1996; Cresho et al., 1992; Murphy et al., 1995). Like primate blobs, cat blobs appear to receive direct inputs from the K cells of the LGN (Boyd and Matsubara, 1996). Despite their similarities, the presence of blobs in primates and carnivores probably reflects homoplasy rather than homology, because animals considered to be more closely related to primates—tree shrews, rodents, lago-morphs, and bats—lack blobs (Preuss, 2000; Preuss and Kaas, 1996).

What is the function of blobs? Based mainly on microelectrode recordings of the visual properties of area V1 neurons, Livingstone and Hubel (1984, 1988) argued that blobs receive their major inputs from the P pathway and serve as specialized color-processing modules. While this idea has become a fixture of textbooks, it has struck some as problematic (Allman and Zucker, 1990; Casagrande, 1994; Preuss, 2000), if only because well-defined blobs are present in nocturnal primates, which have very limited capacity for color discrimination. Moreover, anatomical studies indicate that blobs receive direct inputs from the K layers of the LGN, as discussed above, but not from the P layers, which are usually thought to carry the color-opponent signals required for acute color discrimination. The blob story has taken a new twist with reports that at least some of the K retinal ganglion cells are specifically related to short-wavelength (S) photoreceptors (Dacey, 2000); these K cells could send information from the S-cone channel to the K layers of the LGN and thence to the blobs in area V1. There is at present, however, no evidence that blobs have a correspondingly strong relationship to the medium- and long-wavelength cones, and so a role for blobs in color vision (which requires the interaction of different cone types) has still not been established. It would also be premature to conclude that there is an exclusive relationship between S cones and blobs. In this regard, it is worth noting that blobs are present in two primate taxa, Otolemur and Aotus, in which mutations have inactivated the S-pigment gene (Jacobs et al., 1996).

Casagrande (1994) has considered alternatives to the idea that blobs function as color modules, based on the recognition that blobs may have a privileged relationship to the superior colliculus, by virtue of the strong SC projection to the K layers of the LGN. She suggests that the functions of blobs are related to the attentional and eye movement functions of the colliculus. This possibility will be considered in more detail in a subsequent section.

The Critical Role of V1 in Primate Vision

In primates, lesions of the primary visual area have a devastating effect on visual detection and discrimination. This has been demonstrated both in anthropoids and strepsirhines (Galago) (Atencio et al., 1975). Humans with V1 lesions report they are blind in the affected parts of the visual field, and indicate they are unaware of the occurrence of stimuli presented therein. In electrophysio-logical experiments, lesion or deactivation of area V1 (also known as the "striate" area, owing to its possession of a conspicuous, horizontal band of myelinated fibers) results in marked suppression of stimulus-driven activity in regions of higher-order, "extrastriate" visual areas that represent the lesioned part of the visual field (reviewed by Rodman and Moore, 1997). Collectively, these results suggest that much of the visual information that reaches higherorder cortical centers traverses area V1. Remarkably, however, when patients are instructed to guess the location of stimuli presented in the lesioned visual field, or to identify the characteristics of those stimuli, they do better than chance, indicating that some visual processing capacity is retained in the lesioned part of the visual field representation—even though subjects insist they are unaware that stimuli have been presented (Poppel et al., 1973). This phenomenon, known as "blindsight," can be demonstrated in nonhuman primates as well as in humans (Cowey and Stoerig, 1995; Weiskrantz, 1996).

To understand why lesions of area V1 have such a destructive effect on visual processing in primates, it is necessary to consider the routes through the thalamus by which visual information reaches the cortex. The most numerous visual projections to the cortex arise from the M, P, and K layers of the LGN and terminate in area V1; in turn, area V1 projects to the second visual area (V2) and a variety of other extrastriate visual areas. There are also, however, direct LGN projections to area V2 and other extrastriate areas, although these are very much weaker than the LGN projections to area V1 (Rodman and Moore,

1997). In addition to these geniculocortical projections, the inferior pulvinar nucleus (which receives visual inputs from the superficial layers of the SC) sends projections to V1 as well as to V2 and other extrastriate visual areas. The projections to extrastriate cortex from the LGN and pulvinar presumably provide the anatomical substrates for residual visual capacity following V1 lesions.

Lesions of V1 in nonprimate mammals have much less dramatic effects than in primates. For example, tree shrews with lesions of area V1 retain considerable visual discriminative capacity (Killackey et al., 1971, 1972). In addition, V1 lesions in rats, cats, and bats typically do not produce decrements of visual responsiveness in extrastriate visual areas comparable to those observed in primates (see Funk and Rosa, 1998, and references therein). The reason V1 lesions in nonprimate taxa have relatively modest effects on extrastriate function than in primates may reflect differences in thalamocortical organization. Specifically, the LGN projections to extrastriate visual areas are probably less numerous in primates than in nonprimates, as suggested by the fact that these projections have only been recognized in primates quite recently, while they have long been recognized in nonprimates. Furthermore, the extrastriate projection may arise primarily from K cells in primates (Hendry and Reid, 2000; Rodman et al., 2001), whereas there appear to be more substantial projections from the M and/or P cells in nonprimates, especially to area V2 (e.g., Kawano,

1998). Alternatively, the LGN projections to extrastriate cortex might be less potent physiologically in primates than in nonprimates, and the projections from V1 to extrastriate visual areas more so (in this regard, see also Funk and Rosa, 1998). It is also possible that the strong influence of V1 on extrastriate areas in primates results from an increased potency of V1 projections to the inferior pulvinar, which projects in turn to extrastriate cortex (Cusick, 2002). By whatever mechanism, V1 exerts a much stronger influence on extrastriate areas in primates than in other mammals that have been examined.

Dorsal and Ventral Visual Processing Streams and Their Termini in Higher-Order Parietal and Temporal Cortex

In primates, area V1 projects to multiple extrastriate areas (principally V2, V3, V3A, V4, and MT) which serve in turn as the major sources of visual information to higher-order visual areas (Figure 5). Estimates of the total number of visual areas in anthropoid primates range from at least 15 (Kaas, 1989) to more than 30 (Felleman and Van Essen, 1991). Although strep-sirhines have not been investigated as exhaustively as anthropoids, they possess many of the same areas (Collins et al., 2001; Krubitzer and Kaas, 1990; Preuss and Goldman-Rakic, 1991c; Preuss and Kaas, 1996; Preuss et al., 1993; Rosa et al., 1997). By contrast to primates, comparative studies suggest that ancestral eutherians possessed only a few visual areas, which included V1, V2, perhaps two or three additional areas on the lateral surface rostral to V2, and a medial area (Rosa, 1999; Rosa and Krubitzer, 1999). Certain taxa that have large regions of posterior cortex devoted to vision, such as tree shrews and megabats, nonetheless appear not to possess many visual areas in addition to those that were present in ancestral eutherians (Lyon et al., 1998; Rosa, 1999). Thus, the large number of areas present in primates represents a derived condition, and many of the visual areas present in primates must lack homologues in other mammals, and are therefore neo-morphic (Allman, 1977; Allman and McGuiness, 1988; Kaas, 1987, 1989; Rosa, 1999).

Studies of the connections between primate visual areas reveal that extras-triate cortex is organized into at least two, partly independent, processing streams, that have been termed the dorsal and ventral pathways (Boussaoud et al., 1990; Felleman and Van Essen, 1991; Livingstone and Hubel, 1988; Ungerleider and Mishkin, 1982; Young, 1992). The dorsal pathway includes the middle temporal area (MT) and areas downstream from MT, which contain neurons that are sensitive to object motion. The ventral pathway includes area V4 and areas downstream from V4; these areas contain neurons sensitive to features of object form and (in diurnal anthropoids, at least) to color. The dorsal and ventral streams terminate in the posterior parietal (PP) cortex and inferior temporal (IT) cortex, respectively, two regions classically regarded as higher-order association cortex. In humans and nonhuman primates, lesions of PP and IT have been found to produce very different kinds of visual deficits. Damage to posterior parietal cortex results in sensory neglect (inattention), errors in the spatial localization of objects, and misdirected reaching. Damage to IT cortex produces an inability to recognize familiar objects (visual agnosia). Individual neurons in portions of IT cortex respond to the sight of specific classes of objects such as faces. In view of these differences, the dorsal/parietal and ventral/temporal pathways have been characterized as

Ventral Dorsal Color Motion

Figure 5. Visual areas of tree shrews and primates. (A) The visual cortical territory of tree shrews (Tupaia), denoted by gray shading, as described by Lyon et al. (1998). Tree shrew visual cortex includes a primary visual area (V1), second visual area (V2), and several extrastriate areas (TA, TD, TP). The connections of these visual areas extend forward into the region posterior to the primary somatosensory area (S1), a region that may contain additional visual areas or polysensory cortex. (B) Primates have a very large number of visual areas, which collectively occupy the occipital lobe and large portions of the parietal and temporal lobes. (C) Because visual cortex covers such a large territory, including tissue that in some primates is buried within deep sulci, it is convenient to represent the visual areas as if flattened, as in this schematic drawing. Some primate areas are known by different designations; alternative names are indicated in italics. Included among the primate divisions are three sets of higher-order areas, located in the posterior parietal lobe (PP), superior temporal sulcus (STS), and inferotemporal region (IT), respectively, which are highlighted here with dark shading. Within the visual cortex of

Figure 5. Visual areas of tree shrews and primates. (A) The visual cortical territory of tree shrews (Tupaia), denoted by gray shading, as described by Lyon et al. (1998). Tree shrew visual cortex includes a primary visual area (V1), second visual area (V2), and several extrastriate areas (TA, TD, TP). The connections of these visual areas extend forward into the region posterior to the primary somatosensory area (S1), a region that may contain additional visual areas or polysensory cortex. (B) Primates have a very large number of visual areas, which collectively occupy the occipital lobe and large portions of the parietal and temporal lobes. (C) Because visual cortex covers such a large territory, including tissue that in some primates is buried within deep sulci, it is convenient to represent the visual areas as if flattened, as in this schematic drawing. Some primate areas are known by different designations; alternative names are indicated in italics. Included among the primate divisions are three sets of higher-order areas, located in the posterior parietal lobe (PP), superior temporal sulcus (STS), and inferotemporal region (IT), respectively, which are highlighted here with dark shading. Within the visual cortex of the "where is it" and "what is it" systems, respectively (Ungerleider and Mishkin, 1982).

However, recent evidence suggests there is more to parietal function than spatial localization. Posterior parietal cortex has been found to receive not only visual inputs, but also somatosensory inputs from the forelimb as well as information about eye and head position. The different subdivisions of PP combine these inputs in particular ways to perform very specific types of sensorimotor computations (Andersen et al., 1997; Colby and Duhamel, 1996). For example, some posterior parietal areas have visual receptive fields that shift to accommodate eye movements. This shift allows the receptive fields to be transformed from eye-centered coordinates to head-centered or body-centered coordinates, which presumably are the reference frames in which forelimb movements are programmed. In other posterior parietal areas, neurons respond to tactile stimuli on a particular set of fingertips, say, or on the lips, and also to the sight of an object when it is near the fingertips, or near the lips. Some parietal neurons respond well only to visual stimuli that are within reaching distance and many parietal neurons are responsive during active looking or reaching. For these and other reasons, the functional role of posterior parietal cortex has been characterized as "vision for action" (Goodale and Milner, 1992). (For recent reviews of the structure and function of higher-order parietal and temporal cortex, see Farah, 2000; Jeannerod, 1997; Milner and Goodale, 1995.)

Although not as well known as the posterior parietal and inferotemporal cortex, there is a third region of higher-order cortex located in the depth and upper bank of the superior temporal sulcus (STS). Although this has been dubbed the superior temporal "polysensory" area (Bruce et al., 1981), the main sensory inputs to STS cortex are visual. These inputs arise particularly from the dorsal stream, although some divisions of STS cortex appear to integrate inputs primates, there are two main "streams" of visual processing extending forward from V1: a dorsal stream, which includes area MT and extends into PP, and a ventral stream, which includes area DL and extends into IT. The STS cortex receives inputs from both PP and IT. Patterns of cortical connectivity are represented in a highly simplified fashion here; for more detailed treatment, see Boussaoud et al. (1990) and Felleman and Van Essen (1991). Note that projections between cortical areas tend to be reciprocal, although "forward" and "backward" projections are not functionally equivalent. Additional abbreviations: Aud = auditory cortex; M1 = primary motor cortex.

from the dorsal and ventral streams (Boussaoud et al., 1990; Cusick, 1997). Electrophysiological studies in macaques have revealed that STS neurons are responsive to biological motion: they respond to the sight of animals, or parts of animals (especially faces and hands), moving in particular ways; some neurons are responsive to specific classes of hand-object interactions (e.g., Oram and Perrett, 1994; Perrett et al., 1990). In humans, functional imaging has revealed areas sensitive to biological motion in temporal cortex (Grossman et al., 2000); posterior parietal cortex and ventral premotor cortex exhibit similar properties as well (Buccino et al., 2001).

PP, IT, and STS have been studied most extensively in catarrhine primates (in particular, macaque monkeys and humans), but there is evidence that homologous regions exist in other primates. The posterior parietal region has been little studied in New World monkeys, but strepsirhines (galagos) possess multiple, histologically distinguishable divisions of PP; and the connections of this regions with frontal cortical areas (Preuss and Goldman-Rakic, 1991b) and with the pulvinar nucleus (Glendenning et al., 1975; Preuss and Goldman-Rakic; unpublished observations; Raczkowski and Diamond, 1981), resemble the connections of macaque posterior parietal cortex. As for IT, there is good connectional and architectonic evidence that homologous cortex exists in New World monkeys (Weller and Kaas, 1987). This region has received little attention in strepsirhines, although the histology of the IT region in galagos has been described as resembling that of catarrhines, and what little is known about the connections and functions of this region is consistent with the hypothesis that strepsirhines possess IT cortex (Preuss and Goldman-Rakic, 1991b; Raczkowski and Diamond, 1980). The STS cortex has not been studied in New World monkeys, but a region homologous to STS cortex has been attributed to galagos by Preuss and Goldman-Rakic (1991b, 1991c), based on similarities to macaque STS in histology and frontal lobe connectivity. However, Preuss and Goldman-Rakic also noted that whereas multiple architectonic divisions of the STS cortex could be distinguished in macaques, only a single division could be recognized in galagos. This suggests that the STS region underwent major changes during the evolution of anthropoids or catarrhines.

It is not surprising that a system of partly separate dorsal and ventral visual processing systems like that found in primates is not readily discerned in most other mammals, as most other mammals have relatively few visual areas. It is widely believed that at least some nonprimates possess cortex homologous to the primate parietal cortex. In rodents, for example, there is a very small zone located between primary somatosensory and extrastriate visual cortex that receives inputs from both sensory modalities, that could be homologous to one or more divisions of primate PP (Kolb, 1990; Reep et al., 1994). Others have considered the same region a division of somatosensory cortex, however (e.g., Krubitzer et al., 1986), and the connections of this region do not specifically resemble those of any one of the PP divisions of primates. There is no evidence, moreover, that this region consists of multiple areas in rodents, each with different connectional and functional characteristics, as it does in primates. Furthermore, there is no strong evidence for the existence of homologues of STS or IT cortex in nonprimate mammals (Preuss and Goldman-Rakic, 1991b).

The visual cortex of carnivores provides perhaps the most interesting comparison to primates, because like primates, carnivores possess a large number of visual areas. Payne (1993) has offered a detailed comparison of cat and primate visual cortex, and concludes that most cat areas have homologues in primates. Given that tree shrews—animals generally thought to be more closely related to primates than are cats—have few visual areas, it seems likely that many of the similarities between the extrastriate visual areas of primates and carnivores cited by Payne as evidence of homology are actually the result of convergent evolution. Moreover, cats do not exhibit the connectional segregation of visual areas into dorsal and ventral processing streams found in primates (compare Scannell et al., 1995, and Young, 1992), and even Payne notes that cats lack a strong candidate for homology with primate IT cortex. Interestingly, there is at least one group of mammals in addition to primates— sheep—that have neurons in visual cortex that respond specifically to faces; some of these neurons are responsive to sheep faces, while others are selective for the faces of humans or sheepdogs (Kendrick, 1991). Based on current comparative evidence, this is best interpreted as an instance of convergence rather than homology.

0 0

Post a comment