Primates, Visual Perception and Memory in Nonhuman
PRIMATES, VISUAL PERCEPTION AND MEMORY IN NONHUMAN
Anatomical and physiological studies have revealed at least thirty separate areas with visual functions in the cortex of monkeys, and there could be even more visual areas in the cortex of humans. One view of the division of labor among the visual areas holds that there are two main processing systems: one devoted to identifying what an object is and the other devoted to identifying where an object is (Ungerleider and Mishkin, 1982). On this view, the occipitotemporal cortex is mainly the province of the former system, known as the ventral stream because of the idea that visual information flows, in some sense, from occipital cortex forward and ventrally toward the temporal cortex (see Figure 1). The other system, known as the dorsal stream because information is held to flow from occipital to parietal cortex, seems to play a role in either spatial perception, as noted above, or alternatively, in the guidance of movements to spatial targets (Milner and Goodale, 1996). This article will focus on the ventrally directed, occipitotemporal processing stream and its role in object perception and memory.
Anatomy of the Occipitotemporal Pathway for Object Identification
The occipitotemporal pathway begins with the projection from the striate cortex (the primary visual cortex, V1) to the second and third visual areas, V2 and V3, which in turn project to area V4 (see Figure 1). These prestriate visual areas are arranged in adjacent cortical belts that nearly surround the striate cortex. The major output of area V4 is to a widespread region within the inferior temporal cortex, including area TEO posteriorly and area TE anteriorly. The inferior temporal cortex is in receipt of highly processed visual information arising not only from lower-order visual areas but also from subcortical structures, including the thalamus and basal ganglia. Area TE is often considered to be the last, or highest-order, visual area in the cortical system for object identification because its principal cortical outputs are to areas in the temporal and frontal lobes that are not exclusively concerned with vision.
Murray and Bussey (1999) have argued that perirhinal cortex, a multimodal region that lies adjacent to area TE, should be considered a ventral extension of the ventral visual stream. The perirhinal cortex receives anatomical connections not only from area TE but also from brain areas that process auditory, somatic sensory, and visuospatial information. Although perirhinal cortex is not solely dedicated to the processing of visual sensory inputs, it may nevertheless carry out a higher-order level of visual processing than TE, perhaps by representing even more complex conjunctions of stimulus features than area TE, together with other kinds of sensory inputs. In this way, the perirhinal cortex may bring together disparate pieces of information about objects, including their associated attributes.
Neurons in the visual cortex see, or represent, pieces of our visual world. The amount of visual space that a neuron represents, or is responsive to, is the visual receptive field. In striate and prestriate areas the representation in the cortex is maplike or visuotopic ; that is, nearby parts of visual space are represented in nearby parts of a cortical area in a systematic mapping. The representations in these fields are also restricted to the contralateral visual field. In area TEO of inferior temporal cortex, this visuotopic organization seems to be coarser than that in earlier fields; in area TE it may be nonexistent. In area TE, in contrast to other visual fields, neurons have large receptive fields that often cross the midline. Thus, neurons in area TE can see an object regardless of its position in the visual field.
Also in the temporal cortex are at least two areas involved in processing motion vision: the middle temporal and middle-superior temporal areas, abbreviated as MT and MST, respectively (not illustrated). These areas are in the depths of the superior temporal sulcus, near the temporal-parietal-occipital junction, and are often considered a third visual stream, contrasting motion vision with the object vision of the ventral stream and the spatial vision of the dorsal stream.
Although much of the neural mechanism for object identification can be viewed as a bottom-up process, in which low-level inputs are transformed into a more useful representation through successive stages of processing, anatomical studies have shown that each of the feedforward projections between successive pairs of areas in the occipitotemporal pathway is reciprocated by a feedback projection. Such projections from higher-order processing stations back to lower-order ones could mediate some top-down aspects of visual processing, such as the influence of selective attention.
[Image not available for copyright reasons]
Neuronal Properties in the Visual Areas of the Occipitotemporal Pathway
Given the nearly sequential anatomical route to the inferior temporal cortex from V1, one would expect many types of visual information relevant to object identification to be processed within each area along the route. Indeed, physiological studies have shown that neurons in V1, V2, and V3 are sensitive to one or (usually) more stimulus qualities that we perceive, such as size, orientation, spatial frequency, texture, color, direction of motion, and binocular disparity. Neurons in the motion pathway reflect precisely the judgements reported by monkeys about the direction in which a field of light spots is moving on average. Furthermore, exciting cells by passing electrical current through an electrode near them systematically affects a monkey's perceptual judgement. For example, if a group of neurons represents upward motion, then stimulating these cells during a period when the light spots are tending to move to the left will induce the illusion that they are moving slightly up as well as left. Finally, in the perirhinal cortex and in the neighboring area TE, neurons respond less to familiar than to novel visual stimuli. Thus, these neurons carry signals about whether objects have been seen before (recognition memory). In addition, neurons in this region also reflect the relative recency of viewing of objects.
It is possible to view the areas along the occipitotemporal pathway as forming a processing hierarchy. As one moves along the pathway from V1 to perirhinal cortex, both neuronal response latencies and average receptive field size increase steadily, which is consistent with the notion that neuronal responses in later areas are built up from those in earlier areas. A hierarchical model further predicts that the product of visual processing will become more complex at each successive stage in the pathway. So far, the results of anatomical and physiological studies support this idea.
Although early findings led some theorists to propose that neurons in the anterior ventral temporal cortex code specific objects within their large, bilateral receptive fields, there is overwhelming evidence that neurons in this region, especially in area TE and perirhinal cortex, respond selectively to object features such as shape, color, and texture rather than to specific objects. Thus, the neural code for object recognition and identification must be a population code based on stimulus features. The only exception appears to be the small proportion of neurons selective for faces and, more rarely, hands. Why should the coding of faces be treated differently from the coding of other objects? One possibility is that faces are extremely important to primates not only for the recognition of individuals in the social group but also for social communication by facial expression. Thus, there may have been selective pressure to evolve specialized neural mechanisms for the analysis of faces and facial expression.
Effects of Lesions in the Visual Areas of the Occipitotemporal Pathway
Numerous studies have shown that lesions of V1 produce a scotoma, or blind spot, that corresponds to the part of the visual field represented in the damaged area. The effects of selective lesions of V2 and V3 on visual perception and memory have not yet been investigated. As described below, the impairments that follow lesions of areas V4, TEO, TE, and perirhinal cortex are consistent with a contribution from these areas to the object-identification process.
Lesions of Area V4
Several studies in monkeys have examined the contribution of V4 lesions to color and form vision. Although one reported that the lesion impaired color constancy but not hue or form discrimination, subsequent studies found impairments in both hue and form discrimination, which is consistent with the notion that V4 processes both color and form information and relays this information to the inferior temporal cortex.
Lesions of Areas TEO, TE, and Perirhinal Cortex
Damage to the inferior temporal cortex, including areas TEO, TE, and the perirhinal cortex, leads to deficits in the identification of objects through vision. Behavioral tasks designed to test visual abilities in monkeys indicate that inferior temporal cortical damage produces difficulties in discriminating and in matching objects on the basis of vision. If the damage is restricted to areas TE and TEO, objects can still be identified by touch. Because basic sensory capacities such as acuity and color vision are largely intact after inferior temporal cortical damage, the difficulty is considered to be one of higher-order vision. The nature of the impairment caused by inferior temporal cortex lesions is controversial. One idea is that damage to this region results in a loss of the ability to represent features (e.g. the color red), in such a way that the greater the damage, the fewer visual features that can be represented. Another idea is that it is not the ability to represent features that is lost, but, rather, the ability to represent conjunctions of features. Both these ideas emphasize the role of inferior temporal cortex in visual perception, that is, its role in representing visual features. On either view, damage to inferior temporal cortex will lead to disruption of stored representations of features, resulting in a loss of long-term visual memory. At the same time, this damage would result in an impaired ability to represent features, a perceptual deficit. Thus, it appears that neurons in the inferior temporal cortex contribute to both perception and memory and that perceptual and mnemonic functions are anatomically inseparable in this part of the brain.
Contributions of ventral stream areas to object identification typically are examined by presenting objects against a blank background. In a naturalistic visual scene, however, the visual system must identify and select objects from a multitude of objects, a process termed attention. Attention appears subject to at least two influences: the strength of the sensory signals processed by the ventral stream areas and the (learned) relevance of a particular object. Monkeys with combined lesions of V4 and area TEO are deficient in directing attention to objects in the visual field when there are multiple, distracting objects, and their difficulty increases when the distracting stimuli are made more salient by increasing their contrast (De Weerd, Peralto, Desimone, and Ungerleider, 1999).
In monkeys the delayed nonmatching-to-sample (DNMS) task has proved a reliable measure of visual recognition memory for objects. On each trial the animal is shown a sample object, which it displaces in order to find a food reward underneath; then, about ten seconds later, the animal is shown both the sample and a novel object for choice. The monkey can obtain another food reward by choosing the novel object. When the animal has learned the nonmatching rule, the task is made more difficult—either by increasing the delay between the sample presentation and the choice test or by increasing the number of items to be remembered—to tax the monkey's memory.
Lesions of area TE together with much of the ventrally adjacent perirhinal cortex lead to severe deficits on DNMS (Mishkin, 1982). More recent work suggests that area TE, at least dorsal portions of it, is not as important for visual-recognition memory as the perirhinal cortex. Indeed, lesions restricted to the perirhinal cortex lead to more severe impairments in recognition memory than do lesions to any other part of the inferior temporal cortex yet investigated.
The multimodal nature of the sensory inputs to this region is shown by the deficits in tactile and visual-recognition memory that result from lesions of the perirhinal cortex. Monkeys with lesions of perirhinal cortex can usually perform the nonmatching task if the delay between the sample and the choice is short, but their performance falls nearly to chance levels when the delays are as long as a minute or two; hence, the deficit seems to be one of memory. However, it is clear that the role of perirhinal cortex on visual processing extends beyond recognition memory and includes a role for representing conjunctions of features. Some accounts have questioned the memory interpretation and have argued instead for a unified theory of the contribution of the inferior temporal cortex to perception and memory.
Interactions of the Occipitotemporal Pathway with the Limbic System
In some instances, storage of visual memories must occur through the interaction of the visual cortex with medial temporal-lobe limbic structures (see Figure 1). It appears that the inferior temporal cortex is critical for storing knowledge about objects, the hippocampus for knowledge about places and events, and the amygdala for linking object, event, or place information with affective valence. Thus, the system might be organized in a hierarchical fashion with the inferior temporal cortex providing an initial stage of processing involving object recognition, association, and identification. In later stages, already-processed information about objects would be linked with events or affective valences by the hippocampus and amygdala, respectively. If so, we might expect that, to the extent that the later stages of processing require knowledge about objects, damage to the inferior temporal cortex, including the perirhinal cortex, would disrupt storage of information linking objects with events and affective valence.
See also:DISCRIMINATION AND GENERALIZATION
Dean, P. (1982). Visual behavior in monkeys with inferotemporal lesions. In D. J. Ingle, M. A. Goodale, and R. J. W. Mansfield, eds., Analysis of visual behavior. Cambridge, MA: MIT Press.
Desimone, R., and Ungerleider, L. G. (1989). Neural mechanisms of visual perception in monkeys. In F. Boller and J. Grafman, eds., Handbook of neuropsychology, Vol. 2. Amsterdam: Elsevier.
De Weerd P., Peralto, M. R. III, Desimone, R., and Ungerleider, L. G. (1999). Loss of attentional stimulus selection after ex trastriate cortical lesions in macaques. Nature Neuroscience 2, 753-758.
Gaffan, D., Harrison, S., and Gaffan, E. A. (1986). Visual identification following inferotemporal ablation in the monkey. Quarterly Journal of Experimental Psychology 38B, 5-30.
Milner, A. D., and Goodale, M. A. (1996). The visual brain in action. Oxford: Oxford University Press.
Mishkin, M. (1982). A memory system in the monkey. Philosophical Transactions of the Royal Society of London B298, 85-95.
Murray, E. A. (2000). Memory for objects in nonhuman primates. In M. Gazzaniga, ed., The Cognitive Neurosciences, 2nd edition. Cambridge, MA: MIT Press.
Murray, E. A., and Bussey, T. J. (1999). Perceptual-mnemonic functions of the perirhinal cortex. Trends in Cognitive Sciences 3, 142-151.
Newsome, W. T. (1997). The King Solomon lectures in neuroethology. Deciding about motion: Linking perception to action. Journal of Comparative Physiology A 181, 5-12.
Tanaka, K. (1996). Inferotemporal cortex and object vision. Annual Review of Neuroscience 19, 109-139.
Ungerleider, L. G., and Mishkin, M. (1982). Two cortical visual systems. In D. J. Ingle, M. A. Goodale, and R. J. W. Mansfield, eds., Analysis of visual behavior. Cambridge, MA: MIT Press.
Revised byElisabeth A.Murray