What and where pathways
|Leslie G. Ungerleider and Luiz Pessoa (2008), Scholarpedia, 3(11):5342.||doi:10.4249/scholarpedia.5342||revision #91940 [link to/cite this article]|
What and where pathways refer to a proposed organization of the visual system based on neuroanatomical, electrophysiological, and lesion studies. It describes two information processing streams originating in the occipital cortex, dorsal (which goes to parietal cortex) and ventral (which goes to temporal cortex), which exhibit relative specialization in object recognition (what) and spatial vision (where). This entry will present the general ideas of the what-and-where proposal, with an emphasis on the organization of the what processing stream.
Early studies of human patients pointed to the inferior temporal cortex as a site for object agnosia
The earliest clues about the neural bases of object perception and recognition came from the study of brain-damaged humans with visual agnosia. Visual agnosia refers to a set of disorders affecting object recognition, in which elementary visual capacities such as acuity and visual fields are preserved or grossly intact (for review, see Farah, 1990). Early attempts to localize the lesions in such cases appeared mainly in the German neurological literature of the early 1900s. Writers as early as Potzl in 1928 noted the importance of inferior regions of the temporal lobes, along with adjacent ventral occipital cortex, in most cases of visual agnosia. Subsequent studies upheld these general findings and confirmed the central role of the lingual and fusiform gyri ( Figure 1).
Object recognition in monkeys depends on the inferior temporal cortex
Ablation studies in monkeys have 1) confirmed the crucial role of the temporal lobes in visual object perception and recognition, and 2) clarified the nature of the visual representations used in the object recognition process. One of the earliest experimental studies on the role of the temporal cortex in visual recognition was that of Klüver and Bucy, who described in the 1930s what is now known as the Klüver–Bucy syndrome. When these researchers removed the temporal lobes of monkeys bilaterally, the monkeys demonstrated complex changes in visual, social, sexual, and eating behavior. These changes appeared to result from a combination of perceptual, memory, and motivational impairments. Subsequent research eventually identified the region now known as inferior temporal (IT) cortex as the critical area for producing the visual component of the syndrome. Monkeys with bilateral IT lesions are impaired both in learning to distinguish between different visual patterns or objects, and in retaining previously acquired visual discriminations. Figure 1 shows the location of this area in the macaque brain, relative to the areas implicated in agnosia in humans.
The visual cortex of monkeys can be divided into occipitotemporal and occipitoparietal pathways
Much of our detailed knowledge of visual cortical organization derives from studies of Old World monkeys of the genus Macaca. Similar regions and processing stages are now known to exist in both New World species and in humans, as described later. The original evidence that the visual cortex is divided into separate processing streams was based on the contrasting effects of inferior temporal and posterior parietal cortex lesions in monkeys. For example, inferior temporal lesions cause severe deficits in visual discrimination tasks, but they do not affect animals’ performance on visuospatial tasks (e.g., visually guided reaching and judging which of two objects lies closer to a visual landmark). In contrast, parietal lesions do not affect visual discrimination ability but instead cause severe deficits on visuospatial performance. On the basis of behavioral results such as these, Ungerleider and Mishkin (1982) proposed the existence of two processing streams: an occipitotemporal or ventral processing stream for mediating the visual recognition of objects (“what” an object is) and an occipitoparietal or dorsal processing stream for mediating the appreciation of the spatial relationships among objects and the visual guidance toward them (“where” an object is). Goodale and Milner (1992) have proposed a modification of this model, emphasizing “perception” vs. “action” for ventral and dorsal processing streams, respectively.
Physiological studies of cell properties also support the functional distinction between ventral and dorsal pathways. Neurons within the ventral stream (in distinct modules within V1 and V2, area V4, and inferior temporal areas TEO and TE) respond selectively to visual features relevant for object identification, such as color, shape and texture. Neurons within the dorsal stream (in other modules within with V1 and V2, areas V3, V3A, middle temporal area MT, medial superior temporal area MST, and additional areas in inferior parietal cortex) instead respond selectively to spatial aspects of stimuli, such as the direction and speed of stimulus motion. Such cells also respond when the animal visually tracks a moving target.
Although the two-stream idea has served as a useful scheme with which to understand the organization of the visual system, it is an oversimplification in several ways. For instance, some areas in the dorsal stream exhibit selectivity for simple, two-dimensional geometric shapes that are comparable to those observed in the ventral pathway (Sereno and Maunsell, 1998).
The object recognition pathway in monkeys consists of an interconnected set of cortical areas
In both monkeys and humans, cortical regions comprising the object recognition pathway lie directly adjacent to the primary visual cortex (V1) in the occipital lobe, extending progressively into more anterior and ventral portions of the temporal lobe.
Figure 2 illustrates the location of these areas and diagrams the forward flow of information within this occipitotemporal pathway in monkeys (see also Figure 5). The cortical analysis of objects begins in V1, where information about contour, orientation, color composition, and brightness is represented in subsets of neurons coding for each point in the visual field (note, however, that V1 is implicated in complex processes of segmentation and context-dependent effects; see Gilbert and Sigman, 2007). Information from V1 projects forward to neural subdivisions - inter-digitating thin, thick, and interstripe regions—within V2. From the thin and interstripe regions in V2 (specialized, respectively, for color and form information, roughy speaking), neural signals proceed forward to area V4 on the lateral and ventromedial surfaces of the hemisphere and to a posterior inferior temporal area just in front of V4, area TEO. From both V4 and TEO, signals related to object form, color, and texture proceed forward to area TE, the last exclusively visual area within the ventral stream for object recognition. Together, areas TEO and TE constitute the IT cortex.
Progressing forward along the occipitotemporal pathway from V1 to TE, there is a gradual shift in the nature of connectivity, which is significant for the eventual extraction of invariant object features. At successive stages, patterns of cortical projections become successively less topographic or “point to point.” At the same time, receptive field sizes increase as one advances through the occipitotemporal pathway. While connections from V1 to V2 maintain topographic order, the organization becomes looser between V2 and V4, and inputs to area TE from V4 and TEO retain no obvious retinotopic organization. Likewise, pathways interconnecting areas of the ventral stream on the two sides of the brain via the corpus callosum and/or the anterior commissure tend to be restricted to the representation of the midline of visual space early in the pathway, but their representation becomes more widespread at higher levels in the pathway. This loss of retinotopy within area TE means that single neurons respond to objects anywhere in the visual field. This evidence suggests that the initial explicit information about spatial position in the visual field (“retinotopy”) has been converted to information about object identity, in at least some parts of IT cortex.
However, neural processing in the object recognition pathway is not a simple elaboration of information from lower order to higher order areas. At all stages, the neural connections are reciprocal, such that an area receiving feedforward projections from an area earlier in the ventral stream also provides feedback projections to that area. Long-range nonreciprocal connections, such as from area TE back to V1, also exist. On the one hand, feedforward projections provide bottom-up, sensory-driven inputs to subsequent visual areas, and severing these projections disconnects subsequent areas from their visual input. On the other hand, the precise functions of reciprocal, feed-back projections are still unknown. Probably these feedback projections play a top-down role in vision (as in selective attention), by modulating activity in earlier areas of the ventral pathway. In addition to feedforward and feedback projections, there also appear to be “intermediate-type” projections that link areas at the same level of the visual hierarchy. These are seen most notably between areas of the dorsal and ventral streams. For example, area V4 is interconnected with motion-sensitive areas MT and MST. Thus, an anatomical substrate exists for interactions between the two processing streams, and these might serve to integrate what an object is with where an object is.
All areas within the ventral stream also have heavy interconnections with subcortical structures, notably the pulvinar, claustrum, and basal ganglia. In addition, each ventral stream area also receives subcortical modulatory inputs from ascending cholinergic projections from the basal forebrain, and ascending noradrenergic projections from the locus coeruleus. These projections are thought to play a role in the storage of information in the cortex, and the influence of arousal on information processing, respectively. Finally, visual information is sent from IT cortex to the most ventral and anterior reaches of the temporal lobe, notably the perirhinal cortex and parahippocampal areas TF and TH. These regions in turn project, via the entorhinal cortex, to medial temporal lobe structures, such as the hippocampus, which contribute to forming long-term memories of visual objects and their contexts. Information is also sent from IT cortex to the prefrontal cortex, which plays an important role in working memory, holding an object briefly in mind when it is no longer visible. Finally, there are direct projections from IT cortex to the amygdala, which is important for attaching emotional valence to a stimulus.
Cells along the occipitotemporal pathway are sensitive to increasingly more complex physical features of objects
The different cortical areas in the occipitotemporal pathway share a number of physiological characteristics. Consistent with a role in object recognition, all areas in the pathway contain populations of cells sensitive to the shape, color, and/or texture of visual stimuli. However, at progressively higher levels, cells have larger receptive fields and their stimulus selectivity is more complex to characterize. Thus, higher order properties (consistent with invariant representation of objects) are usually attributed to cells in higher tier areas. For instance, many V1 neurons function as local spatial filters, signaling the presence of contours at particular positions and orientations in the visual field. In contrast, an increasingly higher proportion of cells in higher tier areas (e.g., V2 and beyond) apparently respond to illusory contours (i.e., contours implied by stimulus context and higher-order properties, not due to simple light-dark contrast), across increasingly larger regions of the visual field.
Farther forward in area TEO of the occipitotemporal pathway, and then into posterior and anterior portions of area TE, there are further increases in the complexity of the critical features needed to activate many neurons, and relative increases in the proportion of cells that are selectively driven by some kind of complex pattern or object (Gross et al., 1972). Thus at the highest levels, it can become very difficult to discover (or prove) the “optimal stimulus” for a given single cell, by simply testing various visual stimuli and intuiting the response constraints - because the number of possible test stimuli is literally infinite.
This experimental problem is related to the idea, developed sometime in the late 1960s, that every object would be coded by maximal firing in a single cell in IT cortex; this was termed the “grandmother cell” hypothesis because each “grandmother” was thought to cause maximal firing in a single cell (Gross, 2002). This idea, which was controversial from its inception, is considered by most researchers to be overly simplistic. Indeed, there simply are not enough neurons to represent all the different objects we can recognize. Instead, a “population code” is more likely, in which each specific object (and perhaps each viewpoint) causes a correspondingly unique pattern of firing in multiple cells that share connections and overlapping functional selectivity.
This latter idea is supported by evidence suggesting that IT cells are anatomically grouped into functionally similar patches, columns, or areas (see below). Based on findings from single-unit and optical recording, Tanaka (1996) has suggested that IT cells that respond to common visual features are grouped together into cortical columns that run perpendicular to the cortical surface. Newer evidence from functional magnetic resonance imaging in awake monkeys is consistent with the notion that cells with common functional properties are grouped together in IT cortex (see below) – although details of this organization remain incompletely understood.
Thus, a given object is apparently not represented by the activity of a single cell, but by the activity of many cells across different columns or regions. This unique-but-overlapping population code satisfies two otherwise conflicting requirements in visual recognition: robustness to subtle changes in input, and precision of representation. Although the image of an object projected on the retina changes in response to variations in illumination and viewing angle, the pattern of activity in IT cortex will be largely maintained, insofar as the clustering of cells with overlapping but slightly different selectivity serve as a buffer to absorb such changes.
The human brain also has ventral and dorsal processing streams
The differential visual impairments produced by focal lesions in clinical cases suggest that the human visual cortex, like that of the monkey, contains two anatomically distinct and functionally specialized pathways: the ventral and dorsal streams. For example, one study demonstrated a double dissociation of visual recognition (face perception camouflaged by shadows) and visuospatial performance (maze learning) in two men with lesions of the occipitotemporal and occipitoparietal cortex, respectively, confirmed by postmortem examination (Newcombe et al., 1987). The specific clinical syndromes produced by occipitotemporal lesions include visual object agnosia, as noted earlier, as well as prosopagnosia, an inability to recognize familiar faces, and achromatopsia, or cortical color blindness. In contrast, syndromes produced by occipitoparietal lesions include optic ataxia (misreaching), visuospatial neglect, constructional apraxia, gaze apraxia, akinetopsia (an inability to perceive movement), and disorders of spatial cognition. Interestingly, imagery disorders involving descriptions of either objects (especially faces, animals, and colors of objects) or spatial relations (geographic directions) are also dissociable following temporal and parietal lesions, respectively.
The development of functional brain imaging, including PET (positron emission tomography) and fMRI (functional magnetic resonance imaging), has made it possible to map the organization of the human visual cortex with far greater precision than is possible with human lesion studies, and without the confounding influence of compensatory responses to brain injury.
For instance Haxby and colleagues (Haxby et al., 1994) were able to confirm a functional difference between ventral (object) and dorsal (spatial) pathways in humans, by measuring PET activity while subjects performed object identity and spatial location tasks ( Figure 3). The results identified occipitotemporal and occipitoparietal regions associated with face and location matching, respectively. In addition, regions in the ventral and dorsal frontal cortex were also selectively activated by face and location matching, respectively. These findings thus indicate the existence in humans, as in monkeys, of two functionally specialized and anatomically segregated visual processing pathways. The findings further suggest the extensions of each into the frontal lobe, which plays a role in working memory for objects and their spatial locations.
The “building blocks” of object recognition in human cortex include early visual areas and the lateral occipital complex (LOC)
Functional neuroimaging studies have also revealed a number of distinct visual areas within the ventral and dorsal streams in humans. Many of these areas appear to be equivalent (and perhaps homologous) to specific monkey visual areas, including V1, V2, V3, V3A, V4v, and the middle temporal area (MT) (see also Figure 5). As in monkeys, most of these areas have been defined on the basis of retinotopic maps (Sereno et al., 1995). Area MT, a small area with somewhat disorganized retinotopy, has instead been defined on the basis of its distinctive functional properties (strong selectivity for visual motion and direction) in both human neuroimaging and single-unit recordings in macaque.As described above, receptive field size increases, and stimulus requirements become more “stringent”, as one progresses from V1 through various stages of the ventral stream.
For instance, single units in macaque IT cortex do not respond to “scrambled” versions of object images. To test for analogous regions in human visual cortex, Malach and colleagues (Malach et al., 1995) compared responses evoked by a range of images to activations evoked by “scrambled” versions of those same images, while attempting to equate several image properties (such as size and contrast). Regions responding more to objects than to scrambled stimuli were interpreted as being involved in object-related processing; these regions were collectively named the Lateral Occipital Complex (LOC) because of their anatomical location ( Figure 4). Later neuroimaging studies revealed that the cortical regions activated by such stimuli include classical IT cortex in macaques (Tsao et al., 2003; Denys et al., 2004); see Figure 5 – note, however, that the relationship between LOC in humans and areas V4 and IT in monkeys is unclear and is the object of current studies.
A number of fMRI findings suggest that the LOC is an important stage in human object processing. First, while early visual areas V1–V3 respond similarly to intact, recognizable stimuli and highly scrambled, unrecognizable pictures (scrambled pictures tend to elicit even stronger responses), the LOC exhibits a high sensitivity to image scrambling (in fact, the LOC is often defined by contrasting objects to scrambled objects). Second, a four-fold change in visual object size does not affect the LOC activation, although it changes some stimulus properties greatly (e.g., local contrast); responses in the LOC also exhibit some invariance to changes in image position. Third, the LOC responds to multiple visual cues. For example, the shape of an object can be defined by luminance, by texture, or by motion cues alone; objects defined by any one or more of these cues reliably evoke responses in the LOC. Fourth, objects that vary widely in their recognizability (e.g., famous faces and unfamiliar three-dimensional abstract sculptures) produce similar activations in the LOC. Together, these findings indicate that the LOC is an intermediate link in the processing of objects, i.e., a stage following low-level processing and preceding object processing stages that involve memory.
While the processing of general object shape observed in the LOC appears to be important for object perception and recognition, the LOC probably is not responsible for object recognition per se. Whereas the LOC responds to essentially any three-dimensional shape, areas in the ventral stream anterior to the LOC, most notably on the fusiform gyrus in the ventral temporal cortex, respond preferentially to recognizable objects. It is therefore thought that object recognition is critically dependent on the ventral temporal cortex. Indeed, it is likely that the storage of object representations takes place in the ventral temporal cortex as well.
Early studies of brain damage in humans and lesions in monkeys pointed to the crucial role of tissue in the ventral parts of the temporal cortex for object perception and recognition. After many decades of intensive research, one can now say that the ventral portions of the temporal cortex form the last stations of an occipitotemporal ventral pathway, beginning in the primary visual cortex and progressing through multiple visual areas beyond it. In macaque monkeys, important regions for object perception and recognition include areas TEO and TE of the inferior temporal cortex. In humans, homologous circuits appear to exist in both occipital and temporal regions, including the lateral occipital complex, the fusiform gyrus, and nearby regions in the ventral temporal cortex.
Single-unit recording studies in monkeys show a progressive increase in the complexity of object features needed to “trigger” cells at progressively more anterior stations in the occipitotemporal pathway. Many TE neurons show preservation of their selectivity over transformations that change the physical stimulus, such as retinal position, size, distance in depth, and degree of ambient illumination. At the highest levels of the pathway, within area TE, some neurons are preferentially responsive to faces. More recently, neuroimaging and evoked potential studies in humans with implanted electrodes (Allison et al., 1999) have also shed light on the mechanisms by which the primate brain represents objects.
- Allison T, Puce A, Spencer DD, McCarthy G. (1999). Electrophysiological studies of human face perception. I: Potentials generated in occipitotemporal cortex by face and non-face stimuli. Cereb. Cortex 9, 415-430.
- Denys, K., Vanduffel.,W., Fize, D., Nelissen, K., Peuskens, H., Van Essen, D., and Orban, G. A. (2004) The processing of visual shape in the cerebral cortex of human and nonhuman primate: a functional magnetic resonance imaging study. J. Neurosci. 24, 2551-2565.
- Farah, M. J. (1990). “Visual Agnosia: Disorders of Object Recognition and What They Tell Us About Normal Vision.” MIT Press, Cambridge, MA.
- Gilbert CD, Sigman M. (2007) Brain states: top-down influences in sensory processing. Neuron 54, 677-696.
- Goodale, M. A. and Milner, A. D. (1992) Separate visual pathways for perception and action. Trends Neurosci. 15, 20-25.
- Gross, C. G. (2002) Geneology of the “grandmother cell. Neuroscientist 8, 512-518.
- Gross, C. G., Rocha-Miranda, C. E., and Bender, D. B. (1972). Visual properties of neurons in inferotemporal cortex of the Macaque. J. Neurophysiol. 35, 96–111.
- Haxby, J. V., Horwitz, B., Ungerleider, L. C., Maisog, J. M., Pietrini, P., and Grady, C. L. (1994). The functional organization of human extrastriate cortex: A PET-rCBF study of selective attention to faces and locations. J. Neurosci. 14, 6336–6353.
- Malach, R., Reppas, J. B., Benson, R. R., Kwong, K. K., Jiang. H., Kennedy, W. A., Leden, P. J., Brady, Rosen, B. R., and Tootell, R. B. (1995). Object-related activity revealed by functional magnetic resonance imaging in human occipital cortex. Proc. Natl. Acad. Sci. USA 92, 8135-8139.
- Newcombe, F., Ratcliff, G., and Damasio, H. (1987). Dissociable visual and spatial impairments following right posterior cerebral lesions: Clinical, neuropsychological and anatomical evidence. Neuropsychologia 25, 149–161.
- Sereno AB, Maunsell JH. (1998). Shape selectivity in primate lateral intraparietal cortex. Nature 395, 500-503.
- Sereno, M. I.,Dale, A. M., Reppas. J. B., Kwong, K. K., Belliveau, J. W., Brady, T. J. Rosen, B. R., and Tootell, R. B. (1995) Borders of multiple visual areas in humans revealed by functional magnetic resonance imaging. Science 268, 803-804.
- Tanaka, K. (1996). Inferotemporal cortex and object vision. Annu. Rev. Neurosci. 19, 109–139.
- Tsao, D. Y., Freiwald, W. A., Knutsen, T. A., Mandeville, J. B., and Tootell, R. B. H. (2003) Faces and objects in macaque cerebral cortex. Nature Neurosci. 6, 989-995.
- Ungerleider, L. G. and Mishkin, M. (1982). Two cortical visual systems. In “Analysis of Visual Behavior” (D. J. Ingle, M. A. Goodale, and R. J. W. Mansfield, eds.), pp. 549–586. MIT Press: Cambridge, MA.
- Joseph E. LeDoux (2008) Amygdala. Scholarpedia, 3(4):2698.
- Lawrence M. Ward (2008) Attention. Scholarpedia, 3(10):1538.
- Peter Redgrave (2007) Basal ganglia. Scholarpedia, 2(6):1825.
- Valentino Braitenberg (2007) Brain. Scholarpedia, 2(11):2918.
- Olaf Sporns (2007) Complexity. Scholarpedia, 2(10):1623.
- William D. Penny and Karl J. Friston (2007) Functional imaging. Scholarpedia, 2(5):1478.
- Seiji Ogawa and Yul-Wan Sung (2007) Functional magnetic resonance imaging. Scholarpedia, 2(10):3105.
- Howard Eichenbaum (2008) Memory. Scholarpedia, 3(3):1747.
- Almut Schüz (2008) Neuroanatomy. Scholarpedia, 3(3):3158.
- Rodolfo Llinas (2008) Neuron. Scholarpedia, 3(8):1490.
- John Dowling (2007) Retina. Scholarpedia, 2(12):3487.
- Nicholas V. Swindale (2008) Visual map. Scholarpedia, 3(6):4607.
- Bar, M. (2004). Visual objects in context. Nature Rev. Neurosci. 5, 617-629.
- Barlow, H. B. (1972) Single units and sensation: A neuron doctrine for perceptual psychology? Perception 1, 371–394.
- Biederman, I. (1990). Higher-level vision. In “Visual Cognition and Action” (D. H. Osherson, S. M. Kosslyn, and J. M. Hollerbach, eds.), Vol. 2, pp. 41–72. MIT Press, Cambridge, MA.
- Bruce, C., Desimone, R., and Gross, C. G. (1981). Visual properties of neurons in a polysensory area in superior temporal sulcus of the macaque. J. Neurophysiol. 46, 369–384.
- Damasio, A. R. (1990). Category-related recognition deficits as a clue to the neural substrates of knowledge. Trends Neurosci. 13, 95–98.
- Desimone, R. (1991). Face-selective cells in the temporal cortex of monkeys. J. Cogn. Neurosci. 3, 1–8.
- Desimone, R., and Ungerleider, L. G. (1989). Neural mechanisms of visual perception in monkeys. In “Handbook of Neuropsychology” (F. Boiler and J. Grafman, eds.), Vol. 2, pp. 267–299. Elsevier; Amsterdam, The Netherlands.
- Grill-Spector, K., Kushnir, T., Hendler, T., Edelman, S., Itzchank, Y., and Malach, R. (1998). A sequence of early object processing stages revealed by fMRI in human occipital lobe. Hum. Brain Mapp. 6, 316–328.
- Gross, C. G., Rodman, H. R., Gochin, P. M., and Colombo, M. W. (1993). Inferior temporal cortex as a pattern recognition device. In “Computational Learning and Cognition” (E. Baum, ed.), pp. 44–73. Society for Industrial and Applied Mathematics, Philadelphia.
- Haxby JV, Hoffman EA, Gobbini MI (2000) The distributed human neural system for face perception. Trends in Cognitive Sciences 4:223-233.
- Kanwisher, N., McDermott, J., and Chun, M. M. (1997). The fusiform face area: a module in human extrastriate cortex specialized for face perception. J. Neurosci. 17, 4302-4311.
- Logothetis, N. K., and Sheinberg, D. L. (1996). Visual object recognition. Annu. Rev. Neurosci. 19, 577–621.
- Logothetis, N. K., Pauls, J., and Poggio, T. (1995). Shape representation in the inferior temporal cortex of monkeys. Curr. Biol. 5, 552–563.
- Mishkin, M., Ungerleider, L. G., and Macko, K. A. (1983). Object vision and spatial vision: Two cortical pathways. Trends Neurosci. 6, 415–417.
- Pessoa, L., Tootell, R. & Ungerleider, L.G (2008). Visual perception of objects. In Squire, L.R., Bloom, F.E., McConnel, S.K., Roberts, J.L., Spitzer, N.C., Zigmond, M.J. (Eds.), Fundamental neuroscience(3rdEdition). Academic Press.
- Rubens, A. B., and Benson, D. F. (1971). Associative visual agnosia. Arch. Neural. 24, 305–316.
- Sereno, M. I. and Tootell, B. H. (2005) From monkeys to humans: what do we now know about brain homologies? Curr. Opin. Neurobiol. 15, 135-144.
- Tanaka, K. (1993). Neuronal mechanisms of object recognition. Science 262, 685–488.
- Tootell, R. B., Tsao, D., and Vanduffel, W. (1993) Neuroimaging weighs in: humans meet macques in “primate” visual cortex. J. Neurosci. 23, 3981-3989.
- Ullman S. (1996). “High-Level Vision.” MIT Press, Cambridge, MA.
- Ungerleider, L. G., and Haxby, J. V (1994). “What” and “where” in the human brain. Curr. Opin. Neurobiol. 4, 157–165.
- Ungerleider, L. G. and Pasternak, T. (2004). Ventral and dorsal cortical processing streams. In “The Visual Neurosciences” (L.M. Chalupa and J.S. Werner, eds.) pp. 541-562. MIT Press, Cambridge, MA.