August 2012
Volume 12, Issue 9
Vision Sciences Society Annual Meeting Abstract  |   August 2012
Peripheral representation by summary statistics: An alternative to 3-D shape and lighting direction as basic features for search
Author Affiliations
  • Ruth Rosenholtz
    Dept. of Brain & Cognitive Sciences, MIT\nCSAIL, MIT
  • Xuetao Zhang
    Institute of Artificial Intelligence & Robotics, School of Electronic and Information Engineering, Xi'an Jiaotong University
  • Jie Huang
    Dept. of Brain & Cognitive Sciences, MIT
Journal of Vision August 2012, Vol.12, 267. doi:
  • Views
  • Share
  • Tools
    • Alerts
      This feature is available to authenticated users only.
      Sign In or Create an Account ×
    • Get Citation

      Ruth Rosenholtz, Xuetao Zhang, Jie Huang; Peripheral representation by summary statistics: An alternative to 3-D shape and lighting direction as basic features for search. Journal of Vision 2012;12(9):267.

      Download citation file:

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

  • Supplements

Searching for a side-lit shaded cube among top-lit is efficient (~8 ms/item), compared to search using "equivalent" 2-D items (>20 ms/item), or to search for top-lit shaded cubes among side-lit (21 ms/item) (Enns & Rensink, 1990). Arguably these results are puzzling, from the point of view of some theories of search, as they suggest that 3-D shape and direction of illumination might be available preattentively, whereas distinguishing between, say, a ‘T’ and an ‘L’ supposedly requires attention. We have argued that the information available in peripheral vision is a key determinant of search difficulty (Rosenholtz et al, in review), and that this information is strongly limited by early representation in terms of summary statistics (Balas et al, 2009). Here we revisit Enns & Rensink's results in light of this rethinking of search.

Experiment 1 tests the association between peripheral discriminability and search performance. Observers were asked whether a peripheral item, flanked by 1-4 distractors, was a target or distractor. Experiment 2 tests whether the information available in our hypothesized set of local summary statistics can predict search performance. We extracted a number of target-present and target-absent patches of varying set size (2-5) from search displays. For each one, we synthesized patches with approximately the same summary statistics as the original (Portilla & Simoncelli, 2000). Observers were asked whether each synthesized patch came from a target-present or target-absent original. Conditions corresponded to Enns & Rensink’s Experiments 2ABC and 3AB.

Statistical discriminability predicts peripheral discriminability (R2=0.77). Both measures predict search difficulty (R2= 0.65 & 0.55), including easier search for 3-D vs. 2-D items, and an asymmetry for lighting direction (though the latter may be an issue of criteria setting rather than a difference in available information). Peripheral representation in terms of summary statistics provides a parsimonious account of search with 3-D vs. 2-D items.

Meeting abstract presented at VSS 2012


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.