Prefrontal cortex in visual perception and recognition

Time/Room: Friday, May 17, 2019, 5:00 – 7:00 pm, Talk Room 2
Organizer(s): Biyu Jade He, NYU Langone Medical Center
Presenters: Diego Mendoza-Halliday, Vincent B. McGinty, Theofanis I Panagiotaropoulos, Hakwan Lau, Moshe Bar

< Back to 2019 Symposia

Symposium Description

To date, the role of prefrontal cortex (PFC) in visual perception and recognition remains mysterious. While it is well established that PFC neuronal activity reflects visual stimulus features in a wide range of dimensions (e.g., position, color, motion direction, faces, …), it is commonly thought that such feature encoding in PFC is only for the service of behaviorally relevant functions, such as working memory, attention, task rules, and report. However, recent emerging evidence is starting to challenge this notion, and instead suggests that contributions by the PFC may be integral for perceptual functions themselves. Currently, in the field of consciousness, an intense debate revolves around whether the PFC contributes to conscious visual perception. We believe that integrating insight from studies aiming to understand the neural basis of conscious visual perception with that from studies elucidating visual stimulus feature encoding will be valuable for both fields, and necessary for understanding the role of PFC in vision. This symposium brings together a group of leading scientists at different stages in their careers who all have made important contributions to this topic. The talks will address the role of the PFC in visual perception and recognition from a range of complementary angles, including neuronal tuning in nonhuman primates, neuroimaging and lesion studies in humans, recent development in artificial neural networks, and implications for psychiatric disorders. The first two talks by Mendoza-Halliday and McGinty will address neuronal coding of perceived visual stimulus features, such as motion direction and color, in the primate lateral PFC and orbitofrontal cortex, respectively. These two talks will also cover how neural codes for perceived visual stimulus features overlap or segregate from neural codes for stimulus features maintained in working memory and neural codes for object values, respectively. Next, the talk by Panagiotaropoulos will describe neuronal firing and oscillatory activity in the primate PFC that reflect the content of visual consciousness, including both complex objects such as faces and low-level stimulus properties such as motion direction. The talk by Lau will extend these findings and provide an updated synthesis of the literature on PFC’s role in conscious visual perception, including lesion studies and recent developments in artificial neural networks. Lastly, Bar will present a line of research that establishes the role that top-down input from PFC to the ventral visual stream plays in object recognition, touching upon topics of prediction and contextual facilitation. In sum, this symposium will present an updated view on what we know about the role of PFC in visual perception and recognition, synthesizing insight gained from studies on conscious visual perception and classic vision research, and across primate neurophysiology, human neuroimaging, patient studies and computational models. The symposium targets the general VSS audience, and will be accessible and of interest to both students and faculty.


Partially-segregated population activity patterns represent perceived and memorized visual features in the lateral prefrontal cortex

Speaker: Diego Mendoza-Halliday, McGovern Institute for Brain Research at MIT, Cambridge MA
Additional Authors: Julio Martinez-Trujillo, Robarts Research Institute, Western University, London, ON, Canada.

Numerous studies have shown that the lateral prefrontal cortex (LPFC) plays a major role in both visual perception and working memory. While neurons in LPFC have been shown to encode perceived and memorized visual stimulus attributes, it remains unclear whether these two functions are carried out by the same or different neurons and population activity patterns. To systematically address this, we recorded the activity of LPFC neurons in macaque monkeys performing two similar motion direction match-to-sample tasks: a perceptual task, in which the sample moving stimulus remained perceptually available during the entire trial, and a memory task, in which the sample disappeared and was memorized during a delay. We found neurons with a wide variety of combinations of coding strength for perceived and memorized directions: some neurons preferentially or exclusively encoded perceived or memorized directions, whereas others encoded directions invariant to the representational nature. Using population decoding analysis, we show that this form of mixed selectivity allows the population codes representing perceived and memorized directions to be both sufficiently distinct to determine whether a given direction was perceived or memorized, and sufficiently overlapping to generalize across tasks. We further show that such population codes represent visual feature space in a parametric manner, show more temporal dynamics for memorized than perceived features, and are more closely linked to behavioral performance in the memory than the perceptual task. Our results indicate that a functionally diverse population of LPFC neurons provides a substrate for discriminating between perceptual and mnemonic representations of visual features.

Mixed selectivity for visual features and economic value in the primate orbitofrontal cortex

Speaker: Vincent B. McGinty, Rutgers University – Newark, Center for Molecular and Behavioral Neuroscience Rutgers University – Newark, Center for Molecular and Behavioral Neuroscience

Primates use their acute sense of vision not only to identify objects, but also to assess their value, that is, their potential for benefit or harm. How the brain transforms visual information into value information is still poorly understood, but recent findings suggest a key role for the orbitofrontal cortex (OFC). The OFC includes several cytoarchitectonic areas within the ventral frontal lobe, and has a long-recognized role in representing object value and organizing value-driven behavior. One of the OFC’s most striking anatomical features is the massive, direct input it receives from the inferotemporal cortex, a ventral temporal region implicated in object identification. A natural hypothesis, therefore, is that in addition to well-documented value coding properties, OFC neurons may also represent visual features in a manner similar to neurons in the ventral visual stream. To test this hypothesis, we recorded OFC neurons in macaque monkeys performing behavioral tasks in which the value of visible objects was manipulated independently from their visual features. Preliminary findings include a subset of OFC cells that were modulated by object value, but only in response to objects that shared a particular visual feature (e.g. the color red). This form of ‘mixed’ selectivity suggests that the OFC may be an intermediate computational stage between visual identification and value retrieval. Moreover, recent work showing similar mixed value-feature selectivity in inferotemporal cortex neurons suggests that neural mechanisms of object valuation may be distributed over a continuum of cortical regions, rather than compartmentalized in a strict hierarchy.

Mapping visual consciousness in the macaque prefrontal cortex

Speaker: Theofanis I Panagiotaropoulos, Neurospin, Paris, France

In multistable visual perception, the content of consciousness alternates spontaneously between mutually exclusive or mixed interpretations of competing representations. Identifying neural signals predictive of such intrinsically driven perceptual transitions is fundamental in resolving the mechanism and identifying the brain areas giving rise to visual consciousness. In a previous study, using a no-report paradigm of externally induced perceptual suppression, we have shown that functionally segregated neural populations in the macaque prefrontal cortex explicitly reflect the content of consciousness and encode task phase. Here I will present results from a no-report paradigm of binocular motion rivalry based on the optokinetic nystagmus (OKN) reflex read-out of spontaneous perceptual transitions coupled with multielectrode recordings of local field potentials and single neuron discharges in the macaque prefrontal cortex. An increase in the rate of oscillatory bursts in the delta-theta (1-9 Hz), and a decrease in the beta (20-40 Hz) bands, were predictive of spontaneous transitions in the content of visual consciousness that was also reliably reflected in single neuron discharges. Mapping these perceptually modulated neurons revealed stripes of competing populations, also observed in the absence of OKN. These results suggest that the balance of stochastic prefrontal fluctuations is critical in refreshing conscious perception, and prefrontal neural populations reflect the content of consciousness. Crucially, consciousness in the prefrontal cortex could be observed for faces and complex objects but also for low-level stimulus properties like direction of motion therefore suggesting a reconsideration of the view that prefrontal cortex is not critical for consciousness.

Persistent confusion on the role of the prefrontal cortex in conscious visual perception

Speaker: Hakwan Lau, UCLA, USA

Is the prefrontal cortex (PFC) critical for conscious perception? Here we address three common misconceptions: (1) PFC lesions do not affect subjective perception; (2) PFC activity does not reflect specific perceptual content; and (3) PFC involvement in studies of perceptual awareness is solely driven by the need to make reports required by the experimental tasks rather than subjective experience per se. These claims are often made in high-profile statements in the literature, but they are in fact grossly incompatible with empirical findings. The available evidence highlights PFC’s essential role in enabling the subjective experience in perception, contra the objective capacity to perform visual tasks; conflating the two can also be a source of confusion. Finally we will also discuss the role of PFC in perception in the light of current machine learning models. If the PFC is treated as somewhat akin to a randomly connected recurrent neural network, rather than early layers of a convolution network, the lack of prominent lesion effects may be easily understood.

What’s real? Prefrontal facilitations and distortions

Speaker: Moshe Bar, Bar-Ilan University, Israel
Additional Authors: Shira Baror, Bar-Ilan University, Israel

By now, we know that visual perception involves much more than bottom-up processing. Specifically, we have shown that object recognition is facilitated, sometimes even afforded, by top-down projections from the lateral and inferior prefrontal cortex. Next we have found that the medial prefrontal cortex, in synchrony with the para-hippocampal cortex and the retrosplenial cortex form the ‘contextual associations network’, a network that is sensitive to associative information in the environment and which utilizes contextual information to generate predictions about objects. By using various behavioral and imaging methods, we found that contextual processing facilitates object recognition very early in perception. Here, we go further to discuss the overlap of the contextual associations network with the default mode network and its implications to enhancing conscious experience, within and beyond the visual realm. We corroborate this framework with findings that imply that top-down predictions are not limited to visual information but are extracted from social or affective contexts as well. We present recent studies that suggest that although associative processes take place by default, they are nonetheless context dependent and may be inhibited according to goals. We will further discuss clinical implications, with recent findings that demonstrate how activity in the contextual associations network is altered in visual tasks performed by patients experiencing major depressive disorder. To conclude, contextual processing, sustained by the co-activation of frontal and memory-relate brain regions, is suggested to constitute a critical mechanism in perception, memory and thought in the healthy brain.

< Back to 2019 Symposia