Cite as:
Ashida H, Kitaoka A, 2003, "A gradient-based model of the peripheral drift illusion" Perception 32 ECVP Abstract Supplement

A gradient-based model of the peripheral drift illusion

H Ashida, A Kitaoka

In the periphery, we see illusory rotation on a stationary radial grating with a sawtooth luminance profile (Fraser and Wilcox, 1979 Nature 281 565 - 566). Eye movement causes actual retinal image motion (Naor-Raz and Sekuler, 2000 Perception 29 325 - 335), but the percept of unidirectional motion is not fully explained. Faubert and Herbert (1999 Perception 28 617 - 621) argued that motion detectors are activated because of the latency difference in the light and dark parts. Here we report new findings and propose a gradient-based account. First, an optimised stimulus can induce motion aftereffect on a stationary sinusoidal grating, which provides evidence for the activity of the visual motion mechanism. Second, perceived direction can be altered by background luminance. The local mean luminance seems crucial. On the basis of these findings, we propose that asymmetric responses of motion detectors to the jittering luminance gradient are the source of this illusion. A gradient model of motion detectors, which takes the image velocity as local temporal gradient over local spatial gradient, can predict such asymmetric responses if we assume imbalance in the positive and negative parts of temporal derivative filters. The asymmetry then depends on the DC offset of the stimulus, which explains the effect of the background luminance. The sawtooth stimulus is not optimal in this respect. Reported 'polymorphism' might reflect variability in subtle viewing conditions that could change the effective local mean luminance. We created a better stimulus by reversing the gradient direction in the positive-contrast and negative-contrast parts of the sawtooth grating so that the directional bias is always in one direction when the spatial and temporal derivatives are coupled together.

[Supported by the 21st Century COE Program (D-2 to Kyoto University), MEXT, Japan.]

These web-based abstracts are provided for ease of seaching and access, but certain aspects (such as as mathematics) may not appear in their optimum form. For the final published version of this abstract, please see
ECVP 2003 Abstract Supplement (complete) size: 1712 Kb