Abstract
The environment is replete with redundant visual information, which the visual system can compress into an efficient, ensemble representation. Ensemble perceptual systems have been shown to operate with remarkable flexibility across a host of visual domains. In the current set of experiments, we however, tested whether ensemble information may be represented at a conceptual level, that is, derived in the absence of physical stimuli. In an extension to Emmanouil and Ro's work on conceptual size averaging (VSS, 2015), we used illusory surfaces, Kanizsa triangles, to partially occlude sets of discs varying in size, instead of solid, visible bars. Our results revealed that observers could represent the average size of discs that were partially occluded by an illusory surface just as well as when the discs were fully visible. In a series of follow-up experiments, we tested whether observers implicitly represented the amodally completed surface. Observers judged which of two successively presented pacman sets, one of which gave rise to illusory contours (Kanisza configuration) and one of which did not, had the larger average size. To our surprise, there was no bias to perceive the average size of the Kanizsa configuration as larger, even though the pacman amodally completed behind the illusory surface. It seems that observers were unable to remove the missing 'pie wedge' in their estimation of mean size in either pacman configuration.
Meeting abstract presented at VSS 2016