Abstract
Background: Previously, we reported evidence that global and local effects in amodal completion reflect two different processes: a perceptual process that uses particular geometric relationships to interpolate edges under occlusion, and a global recognition process that suggests the likely shape of an object based on cues such as symmetry or familiarity but does not produce precise contour representations. These processes were distinguished experimentally using a dot localization procedure with shapes exhibiting potential familiarity or bilateral symmetry. Here, we attempt to distinguish these two processes in amodal completion of regular patterns. Method: Participants were assigned to either a local or global condition, which determined how they were instructed to complete shapes exhibiting potential radial or translational symmetry. On each trial, an occluded shape was presented, and a small red dot was flashed in front of the occluder. Subjects were asked to report whether the dot fell inside or outside of the line of sight of the interpolated boundary. Interleaved adaptive 2-up, 1-down staircases were used to estimate two points on the psychometric function: positions where the dot was .707 likely to be judged inside or .707 likely to be judged outside of the occluded object's boundaries. Imprecision (the distance between these two thresholds), location (their mean), and location error (the absolute value of deviation from predicted location) were measured. Results: Local completion produced markedly greater precision, more accurate location, and smaller location error. In addition, local subjects' responses for all measures were highly consistent with each other, while the global group demonstrated variances that were an order of magnitude larger. Conclusions: Local completion produces precise and accurate boundary representations. Global completion appears to be a cognitive registration of the possibility of a regular pattern behind an occluder; it does not, however, produce precise representations of occluded object boundaries.
Meeting abstract presented at VSS 2017