September 2011
Volume 11, Issue 11
Vision Sciences Society Annual Meeting Abstract  |   September 2011
A summary-statistic model of the visual periphery predicts the difficulty of visual curve tracing
Author Affiliations
  • Benjamin Balas
    Department of Brain and Cognitive Sciences, MIT
  • Alvin Raj
  • Ruth Rosenholtz
    Department of Brain and Cognitive Sciences, MIT
Journal of Vision September 2011, Vol.11, 1207. doi:
  • Views
  • Share
  • Tools
    • Alerts
      This feature is available to authenticated users only.
      Sign In or Create an Account ×
    • Get Citation

      Benjamin Balas, Alvin Raj, Ruth Rosenholtz; A summary-statistic model of the visual periphery predicts the difficulty of visual curve tracing. Journal of Vision 2011;11(11):1207.

      Download citation file:

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

  • Supplements

Observers can quickly and accurately report if two dots lie on the same contour, however task difficulty can vary substantially with the curvature and density of contours. Judging such spatial relationships is thought to require “visual routines” (Ullman, 1984), which are basic operations for marking items, diffusing activation within a region, or tracing a contour. Performance across a range of stimulus conditions has previously been modeled as an explicit tracing process in which a “zoom lens” rapidly moves along the contour between target dots. Though this model accounts well for human performance, its relationship to the human visual system is unclear. We propose that curve-tracing difficulty does not follow directly from constraints upon a high-level tracing “routine,” but may instead hinge on low-level information that is lost in peripheral vision (Balas et al., JOV 2009). Our model assumes that the visual system computes texture-like summary statistics in the periphery and estimates appearance from this lossy code. We assume that texture features are measured within peripheral pooling regions that scale with eccentricity in accordance with Bouma's Law. We first replicated previously-reported behavioral effects of contour density and curvature on “same-contour” tasks in human observers. We extend our previous methodology (Rosenholtz et al. VSS Symposium, 2010) to generate images that match the experimental stimuli in terms of local texture statistics. These images reflect what the stimuli “look like” to our model. We calculated d′ values for the model by observing how often the synthesized images preserved the original spatial relationship between the target dots. Our model's d′ scores capture the behavioral effects exhibited by human subjects, suggesting that performance in this task may be explicable by the lossy encoding of peripheral vision and consequent ambiguities imposed upon appearance.


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.