Abstract
The processes underlying bistable rivalry (e.g., binocular, depth, and figure-ground rivalry) have long been believed to be stochastic. In other words, the length of a particular dominance phase cannot be predicted from the dynamics of the preceding dominance alternations. Bistable rivalry, however, has limits for examining sequential effects in spontaneous visual rivalry because transition probabilities are uninformative (transitions are always between two percepts). Using quadra-stable rivalry, we have demonstrated clear evidence of path dependence. Four psychophysical observers viewed a pair of overlapped shapes (diamond & hourglass, left & right-pointing chevrons, upright & inverted triangles, or left & right-skewed parallelograms) presented dichoptically. Each fused image was quadra-stable, perceptually alternating among two pairs of shapes which shared minimal contours and were “opponent” in that they produced opposite shape aftereffects (e.g., Suzuki & Cavanagh, 1998, JEPHPP; Suzuki, 2001, Vision Research). Observers viewed each fused image for 60–70 sec. while reporting changes in the dominant percept; the sequence and duration of the four dominant percepts were recorded for each image. We found evidence of “perceptual trapping” in that spontaneous shifting of percepts was substantially more likely within each opponent pair than across the two pairs. Furthermore, the visual system gradually adapted to opponent shapes while perception was trapped between them, increasing the probability of shifting to the other pair. All observers showed the same effects. These results demonstrate that changes in the percept of multistable stimuli can be highly structured. We will argue that these results suggest that opponent shape coding and high-level neural adaptation play a significant role in determining the dynamics of conscious perception in multistable rivalry.
Supported by NSF SBR-9817643.