A Theory of Contour Interpolation in the Perception of Dynamically Occluded Objects
Evan M. Palmer and Philip J. Kellman University of California, Los Angeles
Thomas F. Shipley Temple University
under submission
Abstract
Humans readily perceive objects from visual information that is fragmented across space and accumulated sequentially over time, yet spatiotemporal object formation is poorly understood. We propose a new theory of contour interpolation and object formation for moving, partially occluded objects. The theory of spatiotemporal relatability (STR) extends the Kellman & Shipley (1991) theory of spatial relatability to predict visual unit formation for moving objects whose edges are revealed at different times throughout the visual field due to occlusion and object or observer motion. We report results of two studies using dynamically occluded objects that support the predictions of STR. A third study using dynamic illusory objects provides results consistent with the identity hypothesis in the spatiotemporal domain?that amodal and modal completion share common contour interpolation processes. Based on these results, we offer a formal account of STR that specifies the spatial and temporal relationships among visible regions that lead to visual unit formation. We propose that STR involves a particular kind of mental representation, the dynamic visual icon, which maintains the shape and current position of occluded regions of moving objects for a short time after their disappearance and allows for their perceptual unification with visible regions.
http://homepage.mac.com/evanpalmer/FileSharing2.html
Evan M. Palmer and Philip J. Kellman University of California, Los Angeles
Thomas F. Shipley Temple University
under submission
Abstract
Humans readily perceive objects from visual information that is fragmented across space and accumulated sequentially over time, yet spatiotemporal object formation is poorly understood. We propose a new theory of contour interpolation and object formation for moving, partially occluded objects. The theory of spatiotemporal relatability (STR) extends the Kellman & Shipley (1991) theory of spatial relatability to predict visual unit formation for moving objects whose edges are revealed at different times throughout the visual field due to occlusion and object or observer motion. We report results of two studies using dynamically occluded objects that support the predictions of STR. A third study using dynamic illusory objects provides results consistent with the identity hypothesis in the spatiotemporal domain?that amodal and modal completion share common contour interpolation processes. Based on these results, we offer a formal account of STR that specifies the spatial and temporal relationships among visible regions that lead to visual unit formation. We propose that STR involves a particular kind of mental representation, the dynamic visual icon, which maintains the shape and current position of occluded regions of moving objects for a short time after their disappearance and allows for their perceptual unification with visible regions.
http://homepage.mac.com/evanpalmer/FileSharing2.html