Purchase this article with an account.
Deborah Apthorp, David Alais, Peter Wenderoth; Motion streaks improve fine direction discrimination. Journal of Vision 2007;7(9):33. doi: 10.1167/7.9.33.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
Due to temporal integration, fast-moving objects leave neural “streaks”. Because orientation perception is very finely tuned, Geisler (1999) suggested signals from these streaks could combine with those from more broadly-tuned motion-sensitive neurons to improve direction perception. Here, in a direction discrimination experiment, we directly manipulate streak length while holding other factors constant. In a 2IFC procedure, subjects judged the direction of motion of RDKs (120ms) composed of Gaussian blobs 12 min wide, moving horizontally at narrow directional separations. Experiment 1: High contrast dots were either fixed-walk or random-walk, with coherence held constant at 50%. Streak length was manipulated by varying the number of fixed-walk frames. Five speeds were tested (10, 5, 2.5, 1.25, 0.625[deg]/sec) and three streak lengths (8, 4, 2 frames). At the faster speeds, direction discrimination thresholds were significantly lower for long streaks, but not below a critical speed of around 2[deg]/sec, similar to Geisler's estimated “streak threshold” of one dot width per 100ms. Experiment 2: Contrast thresholds for detecting coherent motion were measured at each speed and streak length. We found no differences between streak lengths (and a U-shaped function of speed). Experiment 3: Direction discrimination was performed at 2*contrast threshold. The enhanced discrimination provided by long streaks was lost. We conclude that direction discrimination is improved by the presence of oriented “motion streaks”. The loss of this advantage at low contrast probably reflects a weakened a streak signal, due to the shallow contrast response functions for P-pathway units compared to M-pathway. Different results were found using hard-edged dots, suggesting it is important to control spatial frequency spectra by using Gaussian or DOG elements. It is probable that natural scenes contain enough high spatial frequency information for streaks to be useful even at low speeds. The similar contrast thresholds across streak lengths suggest that streak information is not used in coherent motion detection.
This PDF is available to Subscribers Only