Abstract
Perceptual completion connects regions belonging to the same object but physically disconnected in the retinal image; a process of fundamental importance to visual scene analysis and object perception alike. We manipulate the distribution of contrast polarity reversals in inducing configurations to probe the mechanisms underlying different forms of perceptual completion. Novel configurations compatible with modal or amodal completion are equated in the structure of local features and the spatial distribution of local luminance relationships provides the only cue for the different depth relationships associated with modal and amodal completion. Use of these stimuli eliminates the confounding factors and interpretational difficulties related to differences in local structure in traditional configurations used to study modal and amodal completion.
Here, we directly compare the time course (50–300 ms) of spatial localisation of modally and amodally interpolated contours in two variants of a dot-probe localisation task. In a traditional version of this task, the dot always appears in a specified, fully predictable location (e.g., always close to the top contour in a Kanizsa square configuration). We developed a modified (“global”) version of this task, where the dot can appear close to any of the four contours. Positional certainty of a dot-probe's appearance in the traditional task affords reliance on the position of the nearest inducers whereas a global task depends more on the use of interpolated contours. While there was no difference in spatial localisation performance between modally and amodally completed contours with the traditional task, significant differences were observed with the global dot-localisation task. Modally completed contours were localised faster and more accurately than amodally completed contours. These results challenge the assumption that the boundary interpolation system depends solely on the geometrical spatial alignment of inducing fragments. Instead, they suggest that boundary interpolation depends on the spatial distribution of local luminance relationships.
Supported by Australian Research Council (ARCDPO66641).