Abstract
We understand in some detail how early visual mechanisms encode local oriented features, but comparatively little is known about the subsequent integration of this fragmentary information. One suggestion that has attracted some attention proposes the existence of a mechanism that extracts angle information in a manner that defeats the laws of statistically optimal combination of independent orientation estimates. However, the existing psychophysical evidence is conflicting and inconclusive. One major limitation of all previous studies on angle discrimination is their reliance on the direct comparison of angle sensitivity with orientation sensitivity for the angle-components presented in isolation. This is problematic because orientation-selective signals are likely to be modified by the presence of the second angle-component. We compared angle sensitivity with orientation sensitivity for the same stimuli, composed of two bisecting lines. Observers were asked to detect either a change in stimulus angle (the component lines are rotated in opposite directions) or a change in stimulus orientation (the component lines are rotated in the same direction). The summed absolute differences in line orientation between the reference and target stimulus is identical in these tasks. If perceptual judgments are based on the optimal combination of line orientation estimates, sensitivity will be identical. This is not what we find. Angle discrimination is consistently better than orientation discrimination. In a control experiment, we show that this superior performance cannot be explained in terms of correlated noise of orientation-selective signals. Together, these findings suggest that the visual system either contains mechanisms selective for angle size, or, alternatively, manages to extract information from orientation-selective mechanisms that is uniquely available in angle discrimination tasks.
Meeting abstract presented at VSS 2012