Contour interpolation is a perceptual process that builds phenomenologically salient contour representations from spatially segregated edges. Interpolation not only groups visible fragments together, it also fills-in or represents boundaries that fail to project to the eye. A recently-discovered consequence is that seemingly irrelevant information appearing near an interpolated path will affect the path interpolated. In this talk, I characterize interpolation by exploring how, why, and under what circumstances filling-in regions become influential. In the first part of the talk, I use a classification image (CI) technique to show that filling-in proceeds between fragments separated in space and time, and that filling-in itself might best be considered a spatiotemporal process. Next, I offer a novel interpretation of CIs deriving from shape discrimination paradigms, and suggest that prominent CI features correspond to previously undocumented interactions between contour forming and lightness induction mechanisms. I conclude by presenting data that show that, in at least some cases, filling-in cannot be easily altered via top-down strategy. This last finding bears on neural models of object perception, and also the debate as to whether perception is encapsulated from the effects of cognitive expectation.
Event Details
Beyond phenomenological connectedness: Functional consequences of filling-in during contour interpolation
- Event Date: April 12, 2010
- Event Start Time: 12:00 PM
- Event End Time: 7:00 PM
- Event Location: University of Medicine and Dentistry of New Jersey and Rutgers University
- Event Type: Human and Computer Vision Series
- Event Semester: Spring 2010
- Event Contact: Dr. Brian Keane
- Event Extra info: <a href="http://www.briankeane.org/">Dr. Brian Keane</a>