Testing models of human visual motion perception

Cobo-Lewis, Alan

Abstract

The perception of motion is complicated enough to challenge our understanding of the human visual system, but is tractable enough to be amenable to analyzable models. At least as important, the neurophysiology of motion perception early in the visual pathway is sufficiently well known to offer substantial constraints on plausible models. We propose a series of behavioral experiments to evaluate the functional architecture of visual motion perception, and to examine the interaction of motion processing with the processing of other fundamental properties such as color and binocularity. In addition to using classical psychophysical methods, we will also exploit the missing-fundamental illusion and examine motion perception in the presence of high-amplitude counterphase grating. These approaches have the power to offer substantial insight into motion perception. We have two specific aims. One focuses primarily one-dimensional (1D) motion perception, and the other on two-dimensional (2D) motion perception. 1. We will ask how pervasive is the missing-fundamental illusion, and we will offer the first independent test of the canonical explanation of the missing-fundamental illusion (that it can be understood as evidence for motion-energy sensors selective for spatial frequency, such as have been found in Vi). 2. We will assess how many processes of 2D motion perception operate in the human visual system. Specifically, is it necessary to posit three mechanisms (intersection-of-constraints/non-Fourier, vector-sum/Fourier, and feature-tracking), or can fewer mechanisms suffice? Our long-term goals include extending our stimulus methodology to physiological paradigms such as fMRI so that a comprehensive view of motion processing can emerge. But, we propose to use the seed opportunity of an AREA grant to fund sharply focused behavioral experiments that are motivated by both physiological and computational concerns but which can be accomplished in the limited time and budget of an AREA grant. They promise both to create new knowledge and to validate the stimulus methodology we propose.