Purchase this article with an account.
David Kane, Peter Bex, Steven Dakin; Reverse correlation reveals the limits of observers' ability to solve the aperture problem in translating natural scenes. Journal of Vision 2009;9(8):644. doi: 10.1167/9.8.644.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
Natural scenes contain areas of high and low local orientation variance corresponding to texture and edges, respectively. While direction selective neurons in V1 are well suited to signalling the motion of edges, the motion signals arising from such regions will be ambiguous due to the aperture problem. In contrast, high variance textured regions will drive V1 neurons poorly, but can be used unambiguously to determine direction. Here we used a novel reverse correlation paradigm to determine whether contour- or texture-defined structure determines the perceived global direction of natural scenes. Subjects indicated the absolute direction of a translating natural scene, viewed through a series of randomly positioned static apertures, allowing us to relate, on a trial-by-trial basis, subjects‘ performance (bias and precision) to the statistical properties of the exposed image. We report that edges provide the most reliable direction signal when they are oriented parallel or orthogonal to the direction of motion (unbiased performance, ∼20% better precision than that for texture). In contrast edges oriented oblique to the direction of motion lead to precision ∼20% worse than for textures and typically bias observers towards reporting directions orthogonal to the edge-orientation. Differences in bias across subjects suggest that they pursue a stable but individual response strategy in the face of high uncertainty. Finally, all the effects noted (including the classic oblique effect), are invariant to random image rotations; thus effects are the result of the observers‘ interpretation of the scene, rather than being stimulus-dependant. In summary (compared to performance with texture) edges produce both the best and the worse performance, indicating vulnerability of our observers to the aperture problem under the conditions tested and that observers are equally able to use the orientations orthogonal or parallel to the direction of motion in judging the direction of complex natural scenes.
This PDF is available to Subscribers Only