Purchase this article with an account.
James Tse, Peter Gerhardstein; A higher-order mechanism beyond good continuation in contour integration. Journal of Vision 2005;5(8):348. doi: 10.1167/5.8.348.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
Concerns regarding a reported effect of closure in contour integration include: Whether closure alone is a sufficient condition for enhanced contour visibility; the small observed magnitude of the effect; and the need for positing a separate global mechanism beyond low-level local association or “good continuation”. We show that a higher-level mechanism responsible for closure does exist and that this mechanism affects the integration of orientation information into contours. In Exp. 1, we compared the visibility of oriented Gabor elements positioned and aligned to form two half-circle arcs, arranged an enclosed configuration “( )” versus the same two arcs arranged in a non-enclosed configuration “) (“. The two configurations were equal in terms of local curvature, but differed on a global level (i.e., one enclosed a common area and the other did not). The enclosure condition in a 2AFC task was more detectable, suggestive of an effect of closure in contour detection. There remained, however, the possibility that a good continuation effect bound the oriented elements on the ends of the arcs in the enclosed configuration through secondary or tertiary local association (i.e., a smooth line could be extended across the spacing between the arcs). In Exp. 2, we controlled for this possibility by using pairs of arcs that formed sharp corners on the ends in the enclosed configuration. A good continuation account would need to bind oriented elements on the arc ends across path curvatures of at least 75 degrees, which previous research has shown to be highly unlikely. The enclosed condition was still easier to detect than the non-enclosed configuration. We conclude that a closure effect beyond good continuation does exist in contour integration. The effect is most observable in situations where good continuation cannot assign edges to contours with sufficient statistical confidence (e.g., in cases where path curvature is high or where contour path is short).
This PDF is available to Subscribers Only