Study on intrinsic timescales of visual integration for active perception

Abstract

The human brain integrates time-varying sensory information to perceive the external world, and utilize perceived information to act. However, due to short presentation or high signal-to-noise ratio, sensory information may not be perfectly given for accurate perception. Under this condition, the brain often actively perceives the ambiguous information. Understanding how sensory stimulus is integrated and how integrated evidence is transformed into behavior was considered important in neuroscience, not only to understand the human cognitive process but it provides critical insight into cognitive diseases and ultimately provide key to resolve cognitive dysfunction. To understand the underlying process of active visual perception, we focused on two behavior characteristics, 1) the interpretation of information periodically alternates over time, and 2) the period of interpretation alternation largely varies from individual to individual – during active visual perception. Especially, the relationship between the distinct perceptual behavior and time course of information integration was focused, and neurological factors that determine the information integration time was explored that may eventually control individual perceptual decision. First, the time of information integration was quantitatively measured, by analyzing rotational perception while random dot stimulus was given to a human. We found that the information integration time largely varies between 1 second to 4 seconds across individuals, and the time was robustly maintained regardless of external stimulus type. The information integration time was positively correlated to the period of bistable alternation behavior, and a computational model could accurately suggest how individual perceptual accuracy of a given stimulus can be optimized. Next, to find a specific neurological factor that determines the information integration time, we hypothesized that eye movements may induce intrinsic time course of perceptual behavior. Using various bistable perception experiments, we found that specific eye movement patterns always precede the different perceptual states, which could accurately predict the perceptual behavior of an ambiguous stimulus. We also found that slow rhythmic eye gaze oscillation during bistable perception, which period is positively correlated to the period of perceptual behavior change. Moreover, the frequency of saccadic (and micro-saccadic) eye movement during free viewing was positively correlated to the period of perceptual behavior, which implies that intrinsic temporal dynamics of the eye movement circuit is closely related to active perception. However, we found that manipulating eye movement without intentional control does not induce altered behavior, suggesting eye movement may not induce the perceptual decision, but only reflects the intrinsic temporal profile. Taken together, I have shown that active perception of ambiguous visual stimulus can be accurately predicted by information integration time and rhythmic eye gaze patterns, and how two factors contribute to individual perceptual behavior change