Purchase this article with an account.
Christian Kempgens, Tim Schade, Gunter Loffler, Harry S. Orbach; The influence of local orientation on shape discrimination. Journal of Vision 2011;11(11):1091. doi: https://doi.org/10.1167/11.11.1091.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
Purpose: Sensitivity of human observers for discrimination of continuous or sampled radial frequency (RF) contours from a circle is in the hyperacuity range. Recent research has focused on whether the visual system predominantly relies on orientation or position information in this task. For coarsely sampled contours, orientation has been shown to be more important than position, but both are necessary for optimum performance. We investigated how modifications of local orientation information in sampled RF contours influence RF discrimination performance. Methods: Stimuli were composed of either 9 or 40 Gabor patches (7 cpd), centred on (invisible) circular or RF 4 contours. We studied the effects of Gabor patch orientation and set-size (9 vs. 40 elements). Gabor patch orientations were either tangential (aligned with the contours), orthogonal (perpendicular to the contours), radial, parallel, randomly oriented or without orientation (7 cpd bull's eyes). Discrimination thresholds for four subjects were measured in a temporal 2AFC procedure (presentation time = 200 ms) as the RF amplitude yielding 75% correct responses. Results: Thresholds for all variations on element orientation were statistically significantly (p < 0.05) elevated relative to tangential patterns. Random and parallel orientations resulted in the poorest performance. Discrimination performance was always better with 40 elements than with 9 elements (p < 0.05). Conclusions: With identical sample position information, orientation information consistent with the shape contour (tangential) is necessary for optimum RF discrimination performance. When Gabor orientations are parallel or random, local orientation information masks the imaginary contour defined by their positions, with worse discrimination performance than where orientation information is absent (bull's eyes). Increasing set-size results in better sampling of the contour and in performance equivalent to published data for continuous contours. In summary, both position and orientation information are required for peak performance in shape discrimination but if orientation is inconsistent with the contour, orientation can mask the contour.
This PDF is available to Subscribers Only