Abstract
Purpose: Perceived direction in moving Type 2 plaids is biased away from the veridical intersection-of-constraints (IOC) and toward the vector-sum direction. An analysis of a Bayesian model of motion extraction predicts that when one of a plaid's components gives a more reliable motion signal than the other, perceived direction should be biased toward the component whose motion signal is more reliable. A sufficiently large bias toward the slower grating could yield percepts beyond vector-sum. At VSS 02 we confirmed this prediction in Type 2 plaids whose components had unequal spatial frequency (SF), verifying that perceived direction was biased toward the direction of the higher-SF grating, sometimes yielding percepts beyond vector-sum. Now we offer a convergent test of this prediction by presenting Type 2 plaids whose components are of equal SF but unequal contrast. Methods: Naïve subjects judged direction of Type 2 plaids composed of 1-cpd gratings drifting in directions separated by 15 deg and at respective speeds of 2.21 and 1.80 deg/sec. Aperture diameter was 1.5 or 5.0625 deg, and duration was 0.5 sec. The geometric mean of grating contrast was 0.05. We varied the ratio of faster:slower grating contrast logarithmically from 0.54 to 1.84. Results: For gratings of equal contrast, perceived direction was close to vector-sum. When the faster grating had higher contrast, perceived direction was between IOC and vector-sum. When the slower grating had sufficiently higher contrast, perceived direction was beyond vector-sum, as predicted. Conclusions: A Bayesian model of motion extraction suggests that the vector-sum rule is a special case of an endpoint prediction whose general case is a weighted vector sum, with weights given by reliability of gratings' motion signal. Because reliability should increase with contrast, the model predicts bias toward the higher-contrast grating, sometimes entailing bias beyond the vector-sum direction. This prediction was confirmed.
Supported by NIH grant EY 013362