CENTER for GENOME ENGINEERING

On January 13, the Center for Genome Engineering (CGE) at the Institute for Basic Science hosted Professor Soo Hyun Park of the KAIST Department of Brain and Cognitive Sciences for an internal seminar titled “Studying Social Vision in Non-Human Primates.”

Professor Park earned her B.A. in Psychology and Ph.D. in Neuroscience from Seoul National University before conducting postdoctoral research with David Leopold at the National Institute of Mental Health. She now leads the Visual Neurodynamics Lab at KAIST, where her team uses macaques and marmosets to study the neural circuits that support the perception of faces and behavior under natural viewing conditions.

Social vision refers to the visual processing that allows an individual to understand where another is looking and what its actions may signify. As primates adapted to diurnal life and complex social groups, they developed sophisticated color vision and specialized systems for processing gaze, faces, and behavior. Regions surrounding the superior temporal sulcus are particularly important for socially relevant visual information.

Different Neurons within the Same Face Patch

Many visual experiments use simplified stimuli, such as static face images, and measure the average response of a brain region. Real social scenes, however, are dynamic streams of gaze, expression, and action. Professor Park’s research recorded neurons in face-selective cortical patches while monkeys freely viewed natural movies.

Neighboring face-responsive neurons often responded differently to the same movie. The team combined single-neuron electrical recordings with whole-brain fMRI, generating a functional map for each recorded neuron based on activity covariation across the brain.

The 2017 Neuron study revealed distinct functional subpopulations within a single face patch. Neurons located in the same anatomical region could be associated with different brain-wide networks, demonstrating that a regional average cannot fully capture the computations occurring within that region.

Parallel Subnetworks across the Face-Processing System

In subsequent work, neurons with similar responses during movie viewing were grouped and their whole-brain maps were compared. Mixed populations within one face patch formed several distributed and parallel subnetworks. In some cases, a neuron was functionally more similar to a distant neuron in another region than to its immediate neighbors.

This finding reframes the face-processing system as a collection of parallel networks that process different features of a social scene. Professor Park also introduced calcium-imaging experiments in marmosets. By expressing a calcium indicator and monitoring individual cells during movie viewing, the researchers can identify groups with shared temporal response patterns and determine how those groups are distributed within the cortex.

Selecting the Right Primate Model for the Question

Macaques and marmosets offer different experimental advantages. Macaques are well suited for long, stable behavioral and neuroimaging tasks and possess a visual system closely related to that of humans. Marmosets are smaller, mature earlier, and often produce multiple offspring, making them attractive for the development of genetically modified primate models. Because viral vectors that work in rodents often perform differently in primates, primate-optimized genetic tools are also essential.

Professor Park emphasized that understanding social vision requires moving beyond simplified stimuli toward dynamic, naturalistic paradigms and measurements at the level of individual neurons. This work may help clarify circuit-level changes associated with disorders affecting social interaction and provide biological principles for artificial systems designed to interpret human behavior.

Extending Brain Maps to Natural Behavior

Natural movies avoid restricting neural responses to faces or gaze cues selected in advance by the experimenter. As a monkey watches freely, faces appear and move while other animals act and the background changes. Comparing the timing of individual-neuron responses to this complex stream allowed the team to distinguish cell groups within the same face patch that preferentially process different elements of a scene.

Combining single-unit recordings with fMRI connects cellular resolution to brain-wide networks within one experiment. When the response of a recorded neuron increases, correlated changes elsewhere in the brain reveal its functional partners. Anatomically neighboring cells do not necessarily belong to the same network, meaning that a regional fMRI average may blend several distinct neural populations.

Professor Park uses the detailed electrophysiological and imaging data available in macaques alongside the genetic accessibility of marmosets. Advances in calcium imaging and primate-optimized viral vectors may allow researchers to follow socially responsive cell populations over long periods and test the causal roles of specific circuits. Together, these approaches provide a foundation for connecting natural behavior with distributed brain networks.

References

Park, S. H., et al. (2017). Functional subpopulations of neurons in a macaque face patch revealed by single-unit fMRI mapping. Neuron, 95(4), 971–981.e5.

Leopold, D. A., & Park, S. H. (2020). Studying the visual brain in its natural rhythm. NeuroImage, 216, 116790.

Park, S. H., et al. (2022). Parallel functional subnetworks embedded in the macaque face patch system. Science Advances, 8(10), eabm2054.