Anterior Inferotemporal Cortex Research Articles

Population Coding of Visual Space: Comparison of Spatial Representations in Dorsal and Ventral Pathways

Although the representation of space is as fundamental to visual processing as the representation of shape, it has received relatively little attention from neurophysiological investigations. In this study we characterize representations of space within visual cortex, and examine how they differ in a first direct comparison between dorsal and ventral subdivisions of the visual pathways. Neural activities were recorded in anterior inferotemporal cortex (AIT) and lateral intraparietal cortex (LIP) of awake behaving monkeys, structures associated with the ventral and dorsal visual pathways respectively, as a stimulus was presented at different locations within the visual field. In spatially selective cells, we find greater modulation of cell responses in LIP with changes in stimulus position. Further, using a novel population-based statistical approach (namely, multidimensional scaling), we recover the spatial map implicit within activities of neural populations, allowing us to quantitatively compare the geometry of neural space with physical space. We show that a population of spatially selective LIP neurons, despite having large receptive fields, is able to almost perfectly reconstruct stimulus locations within a low-dimensional representation. In contrast, a population of AIT neurons, despite each cell being spatially selective, provide less accurate low-dimensional reconstructions of stimulus locations. They produce instead only a topologically (categorically) correct rendition of space, which nevertheless might be critical for object and scene recognition. Furthermore, we found that the spatial representation recovered from population activity shows greater translation invariance in LIP than in AIT. We suggest that LIP spatial representations may be dimensionally isomorphic with 3D physical space, while in AIT spatial representations may reflect a more categorical representation of space (e.g., “next to” or “above”).

A Motion Direction Map in Macaque V2

In mammals, the perception of motion starts with direction-selective neurons in the visual cortex. Despite numerous studies in monkey primary and second visual cortex (V1 and V2), there has been no evidence of direction maps in these areas. In the present study, we used optical imaging methods to study the organization of motion response in macaque V1 and V2. In contrast to the findings in other mammals (e.g., cats and ferrets), we found no direction maps in macaque V1. Robust direction maps, however, were found in V2 thick/pale stripes and avoided thin stripes. In many cases direction maps were located within thick stripes and exhibited pinwheel or linear organizations. The presence of motion maps in V2 points to a newfound prominence of V2 in motion processing, for contributing to motion perception in the dorsal pathway and/or for motion cue-dependent form perception in the ventral pathway.

How do Task-dependent Attentional Demands Alter How Objects are Learned?

We learn to recognize objects in the world in environments whose attentional demands vary greatly. How does such learning depend upon task-dependent attentional demands? Object recognition needs to be tolerant, or invariant, with respect to position, size, and object view changes. In monkeys and humans, a key area for recognition is the anterior inferotemporal cortex (ITa). Recent neurophysiological data show that ITa cells with high object selectivity often have low position tolerance. We propose a neural model whose cells learn to simulate this tradeoff, as well as ITa responses to image morphs, while explaining how invariant recognition properties may arise gradually due to processes across multiple cortical areas, including the cortical magnification factor, multiple receptive field sizes, and top-down attentive matching and learning properties that may be tuned by task requirements to attend to either concrete or abstract visual features. The model predicts that data from the tradeoff and image morph tasks emerge from different task-dependent levels of attentive vigilance in the animals performing them. Computer simulations predict how receptive field properties would change under different task-sensitive vigilance levels. The model also predicts how vigilance may be controlled by mismatches between top-down learned expectations and bottom-up perceptual inputs, leading to acetylcholine release in neocortical circuits and an increase in vigilance. These results emphasize the importance of top-down attentional mechanisms in object learning and recognition, and of the need to carefully monitor task demands in studies of perceptual and cognitive processing.

An anterior temporal face patch in human cortex, predicted by macaque maps

Increasing evidence suggests that primate visual cortex has a specialized architecture for processing discrete object categories such as faces. Human fMRI studies have described a localized region in the fusiform gyrus [the fusiform face area (FFA)] that responds selectively to faces. In contrast, in nonhuman primates, electrophysiological and fMRI studies have instead revealed 2 apparently analogous regions of face representation: the posterior temporal face patch (PTFP) and the anterior temporal face patch (ATFP). An earlier study suggested that human FFA is homologous to the PTFP in macaque. However, in humans, no obvious homologue of the macaque ATFP has been demonstrated. Here, we used fMRI to map face-selective sites in both humans and macaques, based on equivalent stimuli in a quantitative topographic comparison. This fMRI evidence suggests that such a face-selective area exists in human anterior inferotemporal cortex, comprising the apparent homologue of the fMRI-defined ATFP in macaques.

Neural Selectivity in Anterior Inferotemporal Cortex for Morphed Photographic Images During Behavioral Classification or Fixation

Anterior inferotemporal cortex (aIT) contributes to the ability to discriminate and classify complex images. To determine whether and what proportion of single neurons in aIT cortex can yield enough information to classify complex images, we recorded from aIT neurons during the presentation of morphed photographic images in sessions in which monkeys classified images in a two alternative forced-choice--delayed-match-to-sample (2AFC-DMS) task or in sessions in which they performed a fixation task. The sample stimuli were chosen from a sequence in which one image was gradually morphed into another in a pair, while the original pair of images served as choices. Responses of many individual neurons in aIT cortex during the behavioral classification of the images, decoded using an ideal observer analysis, were sufficiently selective to account for the observed behavioral classification of the images. The responses of a separate population of neurons in aIT cortex recorded in subsequent sessions while the monkeys viewed the same images, were less selective than neural responses measured during sessions in which the 2AFC-DMS task was performed. Our findings show that many neurons in aIT could provide sensory information sufficient for the classification of images when a 2AFC-DMS task was performed.

Individual faces elicit distinct response patterns in human anterior temporal cortex

Visual face identification requires distinguishing between thousands of faces we know. This computational feat involves a network of brain regions including the fusiform face area (FFA) and anterior inferotemporal cortex (aIT), whose roles in the process are not well understood. Here, we provide the first demonstration that it is possible to discriminate cortical response patterns elicited by individual face images with high-resolution functional magnetic resonance imaging (fMRI). Response patterns elicited by the face images were distinct in aIT but not in the FFA. Individual-level face information is likely to be present in both regions, but our data suggest that it is more pronounced in aIT. One interpretation is that the FFA detects faces and engages aIT for identification.

Action-Related Properties Shape Object Representations in the Ventral Stream

The principles driving the organization of the ventral object-processing stream remain unknown. Here, we show that stimulus-specific repetition suppression (RS) in one region of the ventral stream is biased according to motor-relevant properties of objects. Quantitative analysis confirmed that this result was not confounded with similarity in visual shape. A similar pattern of biases in RS according to motor-relevant properties of objects was observed in dorsal stream regions in the left hemisphere. These findings suggest that neural specificity for "tools" in the ventral stream is driven by similarity metrics computed over motor-relevant information represented in dorsal structures. Support for this view is provided by converging results from functional connectivity analyses of the fMRI data and a separate neuropsychological study. More generally, these data suggest that a basic organizing principle giving rise to "category specificity" in the ventral stream may involve similarity metrics computed over information represented elsewhere in the brain.

The Different Neural Correlates of Action and Functional Knowledge in Semantic Memory: An fMRI Study

Previous reports suggest that the internal organization of semantic memory is in terms of different "types of knowledge," including "sensory" (information about perceptual features), "action" (motor-based knowledge of object utilization), and "functional" (abstract properties, as function and context of use). Consistent with this view, a specific loss of action knowledge, with preserved functional knowledge, has been recently observed in patients with left frontoparietal lesions. The opposite pattern (impaired functional knowledge with preserved action knowledge) was reported in association with anterior inferotemporal lesions. In the present study, the cerebral representation of action and functional knowledge was investigated using event-related analysis of functional magnetic resonance imaging data. Fifteen subjects were presented with pictures showing pairs of manipulable objects and asked whether the objects within each pair were used with the same manipulation pattern ("action knowledge" condition) or in the same context ("functional knowledge" condition). Direct comparisons showed action knowledge, relative to functional knowledge, to activate a left frontoparietal network, comprising the intraparietal sulcus, the inferior parietal lobule, and the dorsal premotor cortex. The reverse comparison yielded activations in the retrosplenial and the lateral anterior inferotemporal cortex. These results confirm and extend previous neuropsychological data and support the hypothesis of the existence of different types of information processing in the internal organization of semantic memory.

Comparison of Shape Encoding in Primate Dorsal and Ventral Visual Pathways

Ventral and dorsal visual pathways perform fundamentally different functions. The former is involved in object recognition, whereas the latter carries out spatial localization of stimuli and visual guidance of motor actions. Despite the association of the dorsal pathway with spatial vision, recent studies have reported shape selectivity in the dorsal stream. We compared shape encoding in anterior inferotemporal cortex (AIT), a high-level ventral area, with that in lateral intraparietal cortex (LIP), a high-level dorsal area, during a fixation task. We found shape selectivities of individual neurons to be greater in anterior inferotemporal cortex than in lateral intraparietal cortex. At the neural population level, responses to different shapes were more dissimilar in AIT than LIP. Both observations suggest a greater capacity in AIT for making finer shape distinctions. Multivariate analyses of AIT data grouped together similar shapes based on neural population responses, whereas such grouping was indistinct in LIP. Thus in a first comparison of shape response properties in late stages of the two visual pathways, we report that AIT exhibits greater capability than LIP for both object discrimination and generalization. These differences in the two visual pathways provide the first neurophysiological evidence that shape encoding in the dorsal pathway is distinct from and not a mere duplication of that formed in the ventral pathway. In addition to shape selectivity, we observed stimulus-driven cognitive effects in both areas. Stimulus repetition suppression in LIP was similar to the well-known repetition suppression in AIT and may be associated with the "inhibition of return" memory effect observed during reflexive attention.

Norm-based face encoding by single neurons in the monkey inferotemporal cortex

The rich and immediate perception of a familiar face, including its identity, expression and even intent, is one of the most impressive shared faculties of human and non-human primate brains. Many visually responsive neurons in the inferotemporal cortex of macaque monkeys respond selectively to faces, sometimes to only one or a few individuals, while showing little sensitivity to scale and other details of the retinal image. Here we show that face-responsive neurons in the macaque monkey anterior inferotemporal cortex are tuned to a fundamental dimension of face perception. Using a norm-based caricaturization framework previously developed for human psychophysics, we varied the identity information present in photo-realistic human faces, and found that neurons of the anterior inferotemporal cortex were most often tuned around the average, identity-ambiguous face. These observations are consistent with face-selective responses in this area being shaped by a figural comparison, reflecting structural differences between an incoming face and an internal reference or norm. As such, these findings link the tuning of neurons in the inferotemporal cortex to psychological models of face identity perception.

Using Neuronal Latency to Determine Sensory–Motor Processing Pathways in Reaction Time Tasks

We describe a new technique that uses the timing of neuronal and behavioral responses to explore the contributions of individual neurons to specific behaviors. The approach uses both the mean neuronal latency and the trial-by-trial covariance between neuronal latency and behavioral response. Reliable measurements of these values were obtained from single-unit recordings made from anterior inferotemporal (AIT) cortex and the frontal eye fields (FEF) in monkeys while they performed a choice reaction time task. These neurophysiological data show that the responses of AIT neurons and some FEF neurons have little covariance with behavioral response, consistent with a largely "sensory" response. The responses of another group of FEF neurons with longer mean latency covary tightly with behavioral response, consistent with a largely "motor" response. A very small fraction of FEF neurons had responses consistent with an intermediate position in the sensory-motor pathway. These results suggest that this technique is a valuable tool for exploring the functional organization of neuronal circuits that underlie specific behaviors.

Connections between the anterior inferotemporal cortex (area TE) and CA1 of the hippocampus in monkey.

In addition to the trisynaptic perforant pathway from entorhinal cortex to CA1, there are multiple direct parallel pathways between several cortical regions and CA1. These may be supposed to function cooperatively, in conjunction with the perforant pathway; but neither the functional nor anatomical organization of the extended network is well understood. In this report, we further investigate the connections between anterior inferotemporal cortex (area TE) and CA1. Injections of tracer substances demonstrate that part of the dorsal subdivision of TE sends projections to CA1, but does not receive reciprocating projections back. This contrasts with the bi-directional connections between the more ventral subdivision, TEav, and CA1, as reported by previous studies (and corroborated by tracer injections in this report). The corticohippocampal projections from dorsal TE are likely to be unimodal visual. They partially converge in the posterior portion of CA1 with connections from posterior TE and from the inferior parietal lobule, perhaps constituting a network related to visual or visuospatial processes.

Modulation by context of a scene in monkey anterior inferotemporal cortex during a saccadic eye movement task.

We investigated the effect of a scene on the activity of cells in the anterior inferotemporal (AIT) cortex while the monkey performed a saccadic eye movement (SEM) task with and without the context of a scene (gray frame). Most neurons did not code for the presence of a scene when it appeared alone (monkey free viewing) or when the monkey was fixating. Nevertheless, when a peripheral target was turned on and the monkey had to make a SEM to it, some cells were capable of differentially coding the presence of the scene before and after the saccade.

Is the functional connectivity within temporal lobe influenced by saccadic eye movements?

Local evoked potentials (LEPs), recorded in response to electrical stimulation, were used to study functional connectivity between different sites of the temporal lobe. Permanent electrodes were implanted in anterior and posterior positions of both inferotemporal cortex (IT) and hippocampal formation (HF). In each experimental session, one of these four sites was stimulated and LEPs were recorded in the others. Clear LEPs were found in the anterior and posterior IT sites in response to stimulation of the anterior as well as posterior HF. Bidirectional connections (as judged by the potentials) were found between the anterior and posterior sites of the same structure (IT or HF). The timing of the LEPs indicates that much of the response was carried in multisynaptic circuits. Stimulation delivered just after the monkey made a saccade produced larger late components in the LEPs than the same stimulation delivered without a saccade. The influence was maximal when the delay between the end of the saccade and the electrical stimulation was in the range of 50-100 ms. This saccadic modulation of the functional connectivity was observed within IT (bidirectional) and between posterior HF and IT (unidirectional).

Dissociable Roles of Mid-Dorsolateral Prefrontal and Anterior Inferotemporal Cortex in Visual Working Memory

Functional neuroimaging in human subjects and studies of monkeys with lesions limited to the mid-dorsolateral (MDL) prefrontal cortex have shown that this specific region of the prefrontal cortex is involved in visual working memory, although its precise role remains a matter of debate. The present study compared the effect on visual working memory of lesions restricted to the mid-dorsolateral prefrontal cortex of the monkey with that of lesions to the anterior inferotemporal cortex, a region of the temporal cortex specialized for visual memory. Increasing the delay during which information had to be maintained in visual working memory impaired performance after lesions of the anterior inferotemporal cortex, but not after mid-dorsolateral prefrontal lesions. By contrast, increasing the number of stimuli that had to be monitored impaired the performance of animals with mid-dorsolateral prefrontal lesions, but not that of animals with anterior inferotemporal lesions. This demonstration of a double dissociation between the effects of these two lesions provides strong evidence that the role of the mid-dorsolateral prefrontal cortex in visual working memory does not lie in the maintenance of information per se, but rather in the executive process of monitoring this information. In addition, the present study demonstrated that lesions limited to area 9, which constitutes the superior part of the mid-dorsolateral prefrontal region, give rise to a mild impairment in the monitoring of information, whereas lesions of the complete mid-dorsolateral prefrontal region yield a very severe impairment.

Connections between Anterior Inferotemporal Cortex and Superior Temporal Sulcus Regions in the Macaque Monkey

We examined the connections between the anterior inferotemporal cortex and the superior temporal sulcus (STS) in the macaque monkey by injecting Phaseolus vulgaris leucoagglutinin (PHA-L) or wheat germ agglutinin conjugated to horseradish peroxidase (WGA-HRP) into the dorsoanterior and ventroanterior subdivisions of TE (TEad and TEav, respectively) and observing the labeled terminals and cell bodies in STS. We found a clear dichotomy in the connections of the rostral part of STS: the injections into TEad resulted in a dense distribution of labeled terminals and cell bodies in the upper bank of rostral STS, whereas labeling was confined to the lower bank and fundus of rostral STS after injections into TEav. The distribution of labeling in the rostral STS was discontinuous from the distribution of labeling surrounding the injection sites: the lower bank of the rostral STS was spared from labeling in the TEad injection cases, and TEad had only sparse distribution in the TEav injection cases. These results revise the classical view that the lower bank of rostral STS is connected with TE, whereas the upper bank of rostral STS is connected with the parietal, prefrontal, and superior temporal regions (Seltzer and Pandya, 1978, 1991, 1994). The upper bank of the rostral STS is called the superior temporal polysensory area (STP), because it was previously found that neurons there respond to auditory, somatosensory, and visual stimuli. The present results thus suggest that the polymodal representation in STP interacts more with information processing in TEad than TEav. It is also suggested that the information processing in the ventral bank of the rostral STS is distinct from that in TEad, and the former more directly interacts with TEav than TEad.

IS FACE RECOGNITION NOT SO UNIQUE AFTER ALL?

In monkeys, a number of different neocortical as well as limbic structures have cell populations that respond preferentially to face stimuli. Face selectivity is also differentiated within itself: Cells in the inferior temporal and prefrontal cortex tend to respond to facial identity, others in the upper bank of the superior temporal sulcus to gaze directions, and yet another population in the amygdala to facial expression. The great majority of these cells are sensitive to the entire configuration of a face. Changing the spatial arrangement of the facial features greatly diminishes the neurons' response. It would appear, then, that an entire neural network for faces exists which contains units highly selective to complex configurations and that respond to different aspects of the object “face.” Given the vital importance of face recognition in primates, this may not come as a surprise. But are faces the only objects represented in this way? Behavioural work in humans suggests that nonface objects may be processed like faces if subjects are required to discriminate between visually similar exemplars and acquire sufficient expertise in doing so. Recent neuroimaging studies in humans indicate that level of categorisation and expertise interact to produce the specialisation for faces in the middle fusiform gyrus. Here we discuss some new evidence in the monkey suggesting that any arbitrary homogeneous class of artificial objects—which the animal has to individually learn, remember, and recognise again and again from among a large number of distractors sharing a number of common features with the target—can induce configurational selectivity in the response of neurons in the visual system. For all of the animals tested, the neurons from which we recorded were located in the anterior inferotemporal cortex. However, as we have only recorded from the posterior and anterior ventrolateral temporal lobe, other cells with a similar selectivity for the same objects may also exist in areas of the medial temporal lobe or in the limbic structures of the same “expert” monkeys. It seems that the encoding scheme used for faces may also be employed for other classes with similar properties. Thus, regarding their neural encoding, faces are not “special” but rather the “default special” class in the primate recognition system.

Are cortical models really bound by the "binding problem"?

Are cortical models really bound by the "binding problem"?

Inter-trial neuronal activity in inferior temporal cortex: a putative vehicle to generate long-term visual associations.

When monkeys perform a delayed match-to-sample task, some neurons in the anterior inferotemporal cortex show sustained activity following the presentation of specific visual stimuli, typically only those that are shown repeatedly. When sample stimuli are shown in a fixed temporal order, the few images that evoke delay activity in a given neuron are often neighboring stimuli in the sequence, suggesting that this delay activity may be the neural correlate of associative long-term memory. Here we report that stimulus-selective sustained activity is also evident following the presentation of the test stimulus in the same task. We use a neural network model to demonstrate that persistent stimulus-selective activity across the intertrial interval can lead to similar mnemonic representations (distributions of delay activity across the neural population) for neighboring visual stimuli. Thus, inferotemporal cortex may contain neural machinery for generating long-term stimulus-stimulus associations.

Activity of primate inferotemporal neurons related to a sought target in pair-association task.

Visual long-term memory in primates has been assessed by using the pair-association (PA) task, in which a subject retrieves and chooses the paired associate of a cue picture. Our previous studies on single neurons in the anterior inferotemporal (AIT) cortex suggested their roles in representing paired associates in the mind. To test the possibility that the delay activity of AIT neurons is related to a particular picture as a sought target, we devised the PA with color switch (PACS) task. In the PACS task, the necessity for memory retrieval and its initiation time were controlled by a color switch in the middle of the delay period. A control task, in which there is no color switch, corresponds to the conventional delayed matching-to-sample (DMS) task where the monkey chooses the same picture as a cue. We found that AIT neurons started to respond just after the color switch in the PACS task, when the cue-optimal picture's associate was presented as a cue. In contrast, they showed no response change in the DMS task. We confirmed that this effect is not due to the visual response to colors. Furthermore, when the cue-optimal picture was presented as a cue, these neurons showed suppression after the color switch in the PACS task. These results suggest that the activity of AIT neurons mediates gating mechanisms that preferentially pass information about a sought target, even when the sought target is retrieved from long-term memory.