Abstract
We empirically evaluated anecdotal reports that changes in the variance of image speed due to motion parallax produce changes in perceived global (mean) speed of a moving pattern (Atchley & Andersen, 1995, Vision Research, 35). Observers sequentially viewed pairs of 1–3 s computer-generated displays of moving dots. The mean image speed and variance in image speed (parallax) of the displays were varied independently by simulating observer movement relative to a set of three transparent planes of dots and varying the simulated distance between the planes. The display pairs consisted of a standard, in which the mean and variance (parallax) of image speed remained fixed throughout the experiment, and a comparison, for which the mean and variance of image speed each varied over five levels. Observers indicated which display appeared to move with greater global speed. For each level of variance (parallax), perceived speed and Weber fractions for speed discrimination were estimated by fitting 3-parameter sigmoid functions to the percentage of “faster” judgments as functions of mean speed. Results from a series of five experiments show that for displays of equal mean speed, perceived global speed is directly related to changes in the variance in image speed—the more parallax, the faster perceived global speed. Weber fractions for speed were unaffected by changes in variance, which is consistent with previous research indicating that motion parallax did not affect speed discrimination sensitivity (Watamaniuk and Duchon, 1992, Vision Research, 32). The parallax effect generalizes to both parallel and diverging optical flow patterns, as well as to a range of image speeds spanning more than two octaves. Further, the results rule out explanations based on oculomotor factors. Implications for models of speed perception will be discussed.