Ambiguities in visual perception. The aperture problem and the role of eye movements in perceptual bistability

Abstract

The research reported in this thesis is devoted to visual ambiguities. First, we investigated the assigned binocular disparity for an 'infinitely' long line, which was occluded at the flanks (aperture problem). We showed that from a geometrical point of view, matching of the partially occluded line occurred in the horizontal direction when the occluders were well defined in terms of horizontal and vertical disparity. However, when matching of the occluders was ambiguous, the intersections between the occluded test line and the occluders played a role in the determination of the assigned horizontal disparity of the test line.In the next chapter we investigated the perception of a slanted plane when the depth cues of binocular disparity and linear perspective were manipulated independently. When these cues specify similar slants these cues are combined to provide one single percept. When the conflict between the two cues is large, then the percept alternates between a perspective-dominated percept and a disparity-dominated percept (slant rivalry). For this specific stimulus, we found that eye movements are not essential for perceptual alternations to occur and that on average there is no positive correlation between perceptual flips and both (micro)saccades and blinks that occurred prior to a perceptual flip. Further, we compared results for perceptual rivalry (in our study Necker cube rivalry and slant rivalry) and binocular rivalry (binocular grating rivalry and house-face rivalry). We found that, at the moments of perceptual alternations, there is high positive correlation between saccades and perceptual alternations for binocular rivalry, rather than for perceptual rivalry. For perceptual rivalry we found small differences in fixation positions at the moment of perceptual flips to the two different percepts.We investigated whether there was an interaction between voluntary control in percept dominance and eye movement schemes (for slant, necker cube and house-face rivalry). In separate experimental trials subjects were instructed to either view the stimulus in a natural way (without trying to influence the reversal rate) or to try to hold one specific percept. We found that the pattern of temporal correlation did not change with these different voluntary control conditions, indicating that subjects did not use different eye movement schemes in order to exert voluntary control (although the amount of correlation could be different for different voluntary control conditions). We did find small differences in fixation positions at the moments of perceptual alternations between the two different hold-percept conditions for each paradigm.Finally, we investigated whether the high correlation between saccades and perceptual alternations for binocular rivalry represented an intrinsic link between saccades and perceptual alternations or if this correlation is due to the retinal changes caused by the saccades. By using grating stimuli, which by definition are repetitive in space, we distinguished saccades that did produce local retinal image changes from saccades that did not. We found that there is a strong (causal) correlation between specific local retinal image changes and percept dominance (even for fixational saccades), indicating that retinal image changes are crucial for the correlation between saccades and perceptual alternations.