Leslie Ungerleider - US grants

Disorders of perception, attention, and memory frequently accompany the major mental diseases. To understand the neural mechanisms of these mental processes, we are recording the activity of neurons in the extrastriate and prefrontal cortex of monkeys engaged in tasks requiring stimulus discrimination, attention, and memory. When confronted by multiple stimuli, cells in extrastriate cortex appear to engage in competitive interactions, and top-down signals bias these interactions in favor of cells representing the stimulus that is relevant to the task at hand. This bias appears to be accomplished, in part, through high-frequency (gamma) synchronization of the cells carrying the relevant information. Some of the sources of the top-down signals appear to be located in prefrontal and posterior parietal cortex.

Lesions of inferior temporal (IT) cortex in humans can result in the syndrome termed prosopagnosia, an inability to recognize familiar faces. Singel-cell recordings from IT cortex of monkeys have revealed the existence of neurons that are selectively activated by visual images of faces. Additionally, brain imaging studies in both humans and monkeys have demonstrated face-selective regions, in which the fMRI signal evoked by faces is greater compared to that evoked by non-face objects. Although many groups have reported face-selective regions in temporal and prefrontal cortex, the circuitry underlying face selectivity remains unclear. Here, we studied the functional connectivity among these face-selective regions in monkeys in the resting state. First, we mapped the regions by contrasting fMRI activation to images of monkey faces vs. non-face objects. Two face-selective regions in IT cortex were found in each hemisphere: the anterior and posterior face patches. The animals then underwent ten minutes of resting-state scans. Resting-state averaged time courses from the face patches of each hemisphere were the seeds for functional connectivity analyses. We found that a seed placed in the posterior face patch of one hemisphere correlated with activity in the posterior face patch of the other hemisphere and in the anterior patch, prefrontal face-selective areas and the amygdala of both hemispheres. A seed placed in the anterior face patch showed similar functional connectivity. These results demonstrate that face-selective regions form a network, which can be detected in the intrinsic, spontaneous fMRI signal fluctuations. We have also been exploring whether the face-processing network, revealed in healthy individuals at rest, differs in those with congenital prosopagnosia (CP), a lifelong deficit in face recognition that occurs despite normal intelligence and experience. We localized key regions of the face network using a localizer task and used these regions as seeds for a functional connectivity analysis. This revealed in the controls a set of both posterior and anterior cortical areas whose activity was significantly correlated during rest, reflecting the presence of a face-selective resting state network. However, in CP individuals, the network was compromised, with correlated activity in more anterior regions markedly lower. The results thus indicate that impaired connectivity within the face network may underlie CP. 



 We previously showed that facial expressions modulate fMRI activity in the monkey's amygdala and visual cortex: expressions with emotion yield greater activation than neutral faces, which we term the valence effect. We next tested whether amygdala lesions would eliminate emotional modulatory feedback to the visual cortex, thus disrupting the valence effects seen there. Activation to four different facial expressions was tested: neutral, aggressive (open mouth threat), fearful (fear grimace) and appeasing (lip smack). In control monkeys, as expected, faces with emotional expressions relative to neutral faces produced enhanced responses in face-selective regions. In monkeys with amygdala lesions, although face-selective patches were found in IT cortex, their activity was not modulated by facial expressions, demonstrating that the amygdala is the source of the valence effects seen in visual cortex but is not necessary for face processing per se. In related work in monkeys, we found that oxytocin (OT, nasally inhaled) reduced fMRI responses to both fearful and aggressive faces in face-responsive regions and the amygdala, while leaving responses to appeasing and neutral faces unchanged. We also found that OT reduced functional coupling between the amygdala and areas in the occipital and inferior temporal cortex when viewing fearful and aggressive faces, but not when viewing neutral or appeasing faces. Our results indicate that the monkey may be an ideal animal model to explore the development of novel OT-based pharmacological strategies for treating patients with dysfunctional social behavior, such as autism spectrum disorder. Additional projects are currently in progress: 
 1. We have been studying anatomical connectivity in monkeys by electrically stimulating a targeted structure and measuring the resultant neuronal activation in functionally connected brain areas with fMRI. When we targeted the basal nucleus of the amygdala, we found preferential activation of face-selective patches compared to object-selective patches within IT cortex. Amygdala inputs to these face patches are likely important for deciphering the emotional state of others from their facial expression. 
 2. Arguments for face-selective homologues between humans and macaques assume common processing strategies. Here, we trained monkeys on a face inversion task where, like humans, they responded slower to inverted than to upright faces. However, better efficiency scores for recognizing upright stimuli than inverted ones were only found for macaque and chimpanzee faces (not for human or sheep faces or non-face objects). These results reveal that macaques process macaque and chimpanzee faces holistically and support the idea that the inversion effect is specific for stimuli for which the individual has developed expertise. 3. In behavioral work in humans, we found that face recognition is disrupted by masks consisting of noisy curved stimuli, while object recognition (chairs) is disrupted by masks consisting of noisy straight stimuli, suggesting that separate functional streams in the brain might process curved and rectilinear shapes. We confirmed this idea in fMRI studies in monkeys and humans. Further, fMRI curvature-biased patches in both species partially overlapped face-selective patches, suggesting that curvature processing may contribute to face perception. 4. One model of face recognition posits that separate networks process the invariant aspects of a face, such as identity, and the changeable aspects of a face, such as expression. In support of this idea, we found (using multivoxel pattern analysis) that human face-selective regions responsive to visual motion accurately decoded facial expressions, whereas face-selective regions unresponsive to motion did not, and instead decoded identity. We are currently exploring whether homologous pathways exist in monkeys. 5. Monkey neuranatomy has demonstrated a pathway, specialized for biological motion, projecting down the superior temporal sulcus (STS) into the amygdala. We used theta-burst TMS (TBS) combined with fMRI to determine whether such a pathway exists in humans. TBS delivered over the right posterior STS (rpSTS) reduced activation to dynamic faces and bodies (but not objects) in the rrpSTS itself and activation to faces in the amygdala relative to TBS delivered over the vertex control site. These results demonstrate that the rpSTS is functionally connected to the amygdala for the perception of dynamic (moving) faces. 6. Human face recognition is often attributed to configural processing; namely, processing the spatial relationships among facial features. If so, do visuospatial mechanisms within the posterior parietal cortex (PPC) contribute to this process? We explored this question in humans using fMRI and TMS in a same-different face detection task. Within localized regions of the PPC, configural face differences led to stronger activation relative to featural face differences, and the magnitude of this activation correlated with behavioral performance. Critically, TMS centered on the PPC impaired performance on configural but not featural difference detections. We conclude that spatial mechanisms within the PPC are necessary for configural face processing and, more broadly, that the PPC may be necessary for the veridical face perception.

A typical scene contains many different objects that compete for neural representation due to the limited processing capacity of the visual system. At the neural level, competition among multiple stimuli is evidenced by the mutual suppression of their visually evoked responses. The competition among multiple objects can be biased by both bottom-up sensory-driven mechanisms (exogenous attention), such as stimulus salience, and top-down, goal-directed influences, such as selective (endogenous) attention. Although the competition among stimuli for representation is ultimately resolved within visual cortex, the source of top-down biasing signals likely derives from a distributed network of areas in frontal and parietal cortex. We previously reported that monkeys with lesions of prefrontal cortex (PFC) are impaired in their ability to switch top-down control. We then asked whether monkeys with lesions of posterior parietal cortex (PPC) would show similar or different behavioral effects. Our results showed that, unlike monkeys with PFC lesions, those with PPC lesions are not selectively impaired in their ability to switch top-down control. Rather, they have a selective impairment in spatially locating targets they are required to discriminate. Thus, the PFC plays a critical role in the ability to switch attentional control on the basis of changing task demands, whereas the PPC plays a critical role in allocating attentional resources to behaviorally relevant spatial locations. Another major goal has been to better characterize the nature of distractibility in ADHD by testing hypotheses about whether distractibility arises from increased sensory-driven interference or from inefficient top-down control. We employed an attentional filtering paradigm in which discrimination difficulty and distractor salience were parametrically manipulated. Increased discrimination difficulty should add to the load of top-down processes, whereas increased distractor salience should result in stronger sensory interference. We found a striking interaction of discrimination difficulty and distractor salience: For difficult discriminations, ADHD children filtered distractors as efficiently as healthy children and adults, and all groups were slower to respond with high vs. low salience distractors. In contrast, for easy discriminations, ADHD children were much slower and made more errors than healthy children and adults. For easy discriminations, healthy children and adults filtered out high salience distractors as easily as low salience distractors, but ADHD children were slower to respond on trials with low salience distractors than on trials with high salience distractors. The fact that ADHD children exhibit efficient attentional filtering when task demands are high, but show deficient and atypical distractor filtering under low task demands suggests that filtering mechanisms remain intact in these children but the trigger for activating attention is selectively impaired. Previous research has suggested that the right middle frontal gyrus (rMFG) may serve as a node of interaction between neural networks for top-down goal-directed endogenous attention and bottom-up, stimulus-driven exogenous attention. We tested this hypothesis by comparing the performance on an orientation discrimination task of a patient with a rMFG resection (to remove a brain tumor) and healthy controls. On endogenous attention trials, a valid central cue predicted with 90% accuracy the location of a perithreshold Gabor patch. On the 10% invalid trials, the Gabor patch appeared in the opposite location to the cue. On exogenous attention trials, a cue appeared briefly at one of two peripheral locations, followed, after a variable interstimulus interval (ISI; range 0 to 700 ms), by a Gabor patch in either the same (valid) or opposite location (invalid). Analysis of behavioral data showed that for both patient and controls, valid cues facilitated faster reaction times compared to invalid cues, on endogenous and short ISI exogenous trials. However, at longer ISI exogenous trials, the patient was unable to withhold his responses, resulting in reduced performance compared to controls. This may be related to the patient's inability to reorient attention in a top-down fashion after the effect of the exogenous cue has dissipated, and suggests a putative role of the rMFG in switching between exogenous and endogenous modes of attention. We are continuing to test this hypothesis.