Harini Sankar¹ (), Diane M. Beck¹, Stefan Uddenberg¹; ¹University of Illinois Urbana-Champaign

A central challenge for vision is not just to recover what is where, but to anticipate what is about to happen next, such as whether a tower of blocks will remain standing or collapse. Studies of intuitive physics have shown that observers can rapidly and accurately judge the stability of such towers from a single glance. But this work typically begins only after the scene has already been perceived. Here we asked whether the physical structure of a scene can shape this earliest stage of processing: how quickly the scene becomes visible at all. We first created random block towers that, as a whole, could either be stable or unstable, as determined by a physics engine. We also created images of scrambled blocks, where the blocks’ positions were allowed to vary freely, up to and including floating in free space and clipping into one another. On each trial of the experiment, observers were briefly shown a single image, which could either be an intact tower (stable/unstable), or scrambled blocks, followed by a mask. Observers were tasked with discriminating between the intact and scrambled stimuli. Stimulus duration varied adaptively using a one–up two–down staircase procedure. Prior work on visual detection suggests that images which are more statistically regular, and which better match the visual system’s expectations, reach awareness more readily than those that violate those expectations. Thus, if the visual system encodes physical stability and embodies a prior favoring stable configuration, then stable towers should be detectable at shorter durations than otherwise similar unstable towers. Preliminary results indicate that the stable towers were not detected at shorter presentation durations than unstable towers, indicating that a scene’s physical stability may not affect the very earliest stages of scene processing. However, more data using other masks is needed to verify this finding.