Abstract
Human observers are extremely sensitive to motion information. Under ideal circumstances, just a few milliseconds of motion is sufficient to generate a reliable percept of direction. However, we recently reported a surprising phenomenon in which observers consistently misperceive motion direction of brief, large, high-contrast gratings (Glasser & Tadin, VSS 2013). This perceptual reversal is seen across a wide range of temporal frequencies, but disappears when either size or contrast are reduced, which suggests that this effect is a consequence of psychophysical spatial suppression (Tadin, Lappin, Gilroy, & Blake, 2003). Here, we report a simple model, an extension of work by Derrington and Goddard (1989), which captures the key characteristics of the observed motion direction reversals. Specifically, the model compares responses of model neurons tuned to opposite directions of motion. At extremely brief durations, a moving stimulus contains a comparable amount of motion energy in both directions, so both neurons are activated by a stimulus that moves in only one direction. Critically, the responses of model neurons undergo supersaturating divisive normalization (Peirce, 2007), which yields non-monotonic responses as a function of stimulus strength (contrast, size, and duration). As a result, the veridical motion signal is more strongly suppressed than the opposing signal. This simple model produces motion direction reversals and, notably, yields size, contrast, and temporal frequency tuning that match our psychophysical findings. The current model does a good job of capturing our behavioral findings, but one limitation is that it ignores the temporal dynamics of responses to brief stimuli, which may yield insights into psychophysical spatial suppression, as well as motion processing in general. In ongoing work, we are exploring the dynamics of perceptual motion reversals to elucidate possible contributions of rapid adaptation, transients, motion opponency, or other mechanisms to reversed motion percepts.
Meeting abstract presented at VSS 2014