August 2014
Volume 14, Issue 10
Vision Sciences Society Annual Meeting Abstract  |   August 2014
Rethinking the aperture problem: a story of competing priors
Author Affiliations
  • Edgar Walker
    Department of Neuroscience, Baylor College of Medicine
  • Wei Ji Ma
    Center for Neural Science, New York University
Journal of Vision August 2014, Vol.14, 17. doi:
  • Views
  • Share
  • Tools
    • Alerts
      This feature is available to authenticated users only.
      Sign In or Create an Account ×
    • Get Citation

      Edgar Walker, Wei Ji Ma; Rethinking the aperture problem: a story of competing priors. Journal of Vision 2014;14(10):17. doi:

      Download citation file:

      © ARVO (1962-2015); The Authors (2016-present)

  • Supplements

How the visual system resolves ambiguity is a fundamental problem in vision science. A classic example is the aperture problem, in which a moving grating is viewed through an aperture. Although the stimulus is consistent with many motion directions, it typically produces a reliable motion direction percept. However, since very few studies have explored the effect of the shape of the aperture and of its orientation relative to the grating on perceived motion direction, the brains strategy for resolving ambiguity in this problem is not fully understood. We conducted an experiment in which subjects reported the perceived motion direction of a grating moving behind an elliptical or rectangular aperture with a variable aspect ratio and variable relative orientation. We found strong effects of relative orientation, aspect ratio, and shape on the perceived motion direction. These effects could not be captured by previous models one based on a prior favoring low speeds, and one based on line terminators. Instead, we reframed the observers decision process as Bayesian inference on the motion direction of an infinitely long patterned strip with fixed but unknown width viewed through the aperture. In the model, the observer a) computes for each candidate motion direction the speed and minimum strip width consistent with the scene, b) assigns posterior probability using both the low-speed prior and a prior we propose here, which favors narrower strips, and c) reports the posterior mean. The resulting model not only outperformed the other two models, but also captured the observed dependencies with high accuracy. One potential interpretation of the narrow-strip prior is as a "little-unseen stuff" prior, favoring scenes that require the fewest assumptions about unobserved regions of the scene. Perhaps our brain resolves ambiguity by performing a process analogous to model selection, where simpler models are favored over complex ones.

Meeting abstract presented at VSS 2014


This PDF is available to Subscribers Only

Sign in or purchase a subscription to access this content. ×

You must be signed into an individual account to use this feature.