CAREER: Probabilistic inference in the primate visual system

University Of Texas At Austin, Austin TX

Investigators

Abstract

This project investigates the functional relation between sensory and prefrontal cortex during perceptual inference (i.e., inferring properties of the sensory environment) and perceptual introspection (i.e., evaluating the quality of this inference). Key to both tasks is that they require consideration of perceptual uncertainty. When perceptual uncertainty is high (i.e., when sensory measurements are ambiguous), perceptual inferences tend to be guided by prior experience, and confidence in perceptually guided decisions tends to be low. How neural circuits assess the reliability of sensory signals and use this assessment for perceptual inference and perceptual introspection is not well understood. The proposed research will advance our understanding in two ways. First, by simultaneously recording neural activity in early visual cortex and down-stream association cortex from animals performing perceptual inference and introspection tasks. Second, by using these empirical observations to evaluate theoretical proposals of how different brain regions work together to produce perception and cognition in the face of uncertainty. The data that will be collected and the data-analysis tools that will be developed will be made available to the neuroscience community. This project will also have a broader impact by providing students with a paid summer research internship in computational and systems neuroscience at UT Austin, and by engaging middle- and high-school students of central Texas in neuroscience. This project will study the neural implementation of probabilistic perceptual inference, using primate V1 and its downstream targets in the prefrontal cortex (area FEF) as the model system. The proposal employs behavioral tasks in which non-human primates communicate stimulus orientation judgements (and their confidence in this decision) with saccadic eye movements. The task paradigms distinguish the neural correlates of perceptual inference and perceptual confidence from those of action planning. Neural population activity will be recorded with multiple jointly inserted multi-electrode arrays. The proposal leverages functional models of neural coding to study the nature, strength, and dynamic evolution of the relation between V1 and FEF population activity. The experiments involve manipulations of stimulus features, sensory uncertainty, and prior probability. As such, the project aims to uncover canonical principles of neural coding that are instantiated in many brain regions and modalities. The data collected will be valuable for the development of models of large-scale distributed computation that seek to bridge circuit structure and function. This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.