August 2009
Volume 9, Issue 8
Vision Sciences Society Annual Meeting Abstract  |   August 2009
A crowded model of visual search
Author Affiliations
  • Ruth Rosenholtz
    Department of Brain & Cognitive Sciences, MIT
  • Stephanie Chan
    Department of Brain & Cognitive Sciences, MIT
  • Benjamin Balas
    Children's Hospital, Harvard Medical School
Journal of Vision August 2009, Vol.9, 1197. doi:
  • Views
  • Share
  • Tools
    • Alerts
      This feature is available to authenticated users only.
      Sign In or Create an Account ×
    • Get Citation

      Ruth Rosenholtz, Stephanie Chan, Benjamin Balas; A crowded model of visual search. Journal of Vision 2009;9(8):1197.

      Download citation file:

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

  • Supplements

Items within a crowded window defined by Bouma's Law — which states that the critical spacing between items for crowding to occur is approximately 0.5E — appear “jumbled,” such that target/distractor discrimination becomes intractable. In typical visual search displays where inter-item separation is small, crowding likely exerts a significant influence on performance. Presently, we examine the relationship between crowding and visual search in two ways.

First, we explored whether the “critical spacing” for crowded perception predicts search performance. In the context of visual search, Bouma's Law means that peripheral targets are more likely to be crowded and thus indistinguishable from distractors. Guided by this basic intuition, we constructed a model of visual search that can identify only “uncrowded” items and executes saccades as needed to identify previously obscured items. Can this model feasibly predict search reaction times (RT)? We show that for typical search displays the number of predicted fixations is an approximately linear function of set size. Thus it is trivial to fit the nearly linear RT vs. set size functions found for many search conditions. The resulting parameters are also quite plausible: approximately 400 ms decision time, and 200 ms /fixation.

Can the information available from viewing multiple crowded items predict search difficulty? We applied a model of visual crowding based upon a texture representation of stimuli by joint statistics defined over position, phase, orientation and scale. We have previously shown that this model predicts “crowded” identification across a range of conditions. This model allows the creation of visualizations of the information available under crowding, which we call Mongrels. These Mongrels successfully predict: Easy feature search for a tilted line among vertical, difficult conjunction search, search for T among L, and search for O among Q, and easy search for Q among O.

Rosenholtz, R. Chan, S. Balas, B. (2009). A crowded model of visual search [Abstract]. Journal of Vision, 9(8):1197, 1197a,, doi:10.1167/9.8.1197. [CrossRef]
 Supported by NSF grant BCS- 0518157 to RR.

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.