Path Integration Reflects an Intermediate Representation in Contour Perception

Poster Presentation: Saturday, May 17, 2025, 8:30 am – 12:30 pm, Banyan Breezeway
Session: Perceptual Organization: Segmentation, grouping

Austin S. Phillips1, Philip J. Kellman1; 1University of California, Los Angeles

Path integration refers to perceptual grouping of oriented elements, often Gabor patches, embedded in a random array. Although the geometric relations that produce the pop-out of paths correspond to those that produce completion of illusory and occluded contours (Hess & Field, 1999; Kalar et al, 2010), path detection is not accompanied by perception of illusory contours, and the relation to contour interpolation has remained unclear. A two-process model of contour interpolation (Kellman & Fuchser, 2023) proposes that path detection depends on an intermediate representation of contours linked across gaps purely by their geometry, but that perception of continuous contours depends on additional constraints that act on this representation. The higher-level constraint appears to be that Gabors do not support perception of continuous contours because linked contours do not enclose a consistent surface color within and between elements. Altering Gabors so that their centers match the surround produces clear illusory contours connecting path elements (Kellman, Erlikhman & Carrigan, 2016). Here, we further tested path integration and illusory contour perception, using Gabors and modified elements with centers matching the surround. Participants searched for paths with 0-75 degree turning angles in a 2IFC paradigm (Field, Hayes & Hess, 1993). Consistent with the two-process model, we hypothesized that (1) elements supporting illusory contours would show no advantage in path detection, and (2) that illusory contours would be perceived only with elements satisfying both the geometry of relatability and the surface continuity constraint. The results confirmed both hypotheses. All element types showed similar path detection functions, and illusory contours were perceived only in displays supporting surface connections. For those displays, perceived illusory contours mirrored the path detection results. These results support the idea that path integration performance depends on an intermediate stage of contour-linking, necessary but not sufficient for perception of continuous contours across gaps.

Acknowledgements: We gratefully acknowledge support for this research from National Institutes of Health grant R01CA236791 to PK.