Neural Code for Dynamic Visual Experience in the Primate Brain

Author: Lu, Jialiang

Year: 2026

Degree: Dissertation (Ph.D.)

Advisor: Tsao, Doris Y.

Committee Members: Meister, Markus; Adolphs, Ralph; Tsao, Doris Y.; Rutishauser, Ueli

Option: Computation and Neural Systems

Abstract

Neuroscience increasingly needs experimental tools that can study vision in naturalistic, continuous, and interactive scenes. This thesis presents a methodological and scientific framework for studying naturalistic vision and social scene representation in the primate brain. I developed MINOS, a Unity-based platform that integrates complex stimulus generation, behavioral control, hardware communication, synchronization, and analysis-ready data recording in one modular environment. I also developed a compact piezo-actuated Neuropixels insertion system for acute large-scale electrophysiology in awake macaques, designed to support precise, reproducible, and flexible multi-probe recordings in limited chamber space. Using these tools, I recorded neural populations from inferotemporal cortex and ventrolateral prefrontal cortex while a monkey freely viewed a continuous social scene in which one or two characters entered, waited, interacted, and left. This design kept precise event labels while reducing the strong visual transients of discrete image presentation. The results showed a clear functional difference between the two regions. Inferotemporal cortex mainly represented the currently fixated face and updated rapidly after fixation changes. Ventrolateral prefrontal cortex integrated information across time and scene state, represented current and recently relevant identities, encoded entering and leaving events, and maintained identity information in different low-dimensional formats depending on whether a person was fixated or not. It also carried a stable and largely identity-independent code for action category and action role across sessions. Together, these findings suggest that prefrontal cortex helps organize dynamic social scenes by combining context-dependent identity representations with a more general relational code. This thesis therefore contributes new experimental tools for primate systems neuroscience and opens a route for studying how the brain represents structured events during continuous visual experience.