August 2010
Volume 10, Issue 7
Vision Sciences Society Annual Meeting Abstract  |   August 2010
Tiny Memory: How many pixels are required for good recognition memory?
Author Affiliations
  • Yoana Kuzmova
    Brigham & Women's Hospital
  • Jeremy Wolfe
    Brigham & Women's Hospital
    Harvard Medical School
Journal of Vision August 2010, Vol.10, 1244. doi:
  • Views
  • Share
  • Tools
    • Alerts
      This feature is available to authenticated users only.
      Sign In or Create an Account ×
    • Get Citation

      Yoana Kuzmova, Jeremy Wolfe; Tiny Memory: How many pixels are required for good recognition memory?. Journal of Vision 2010;10(7):1244. doi:

      Download citation file:

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

  • Supplements

We can remember hundreds of pictures given only a few seconds of exposure (Standing,1973). How much stimulus resolution is necessary for successful picture memory? Torralba (2009) reported that 32x32 pixel photographs can be categorized with 80% accuracy, but can these thumbnails be effectively coded into memory? In Experiment 1, observers saw a sequence of natural scene images, and gave a new/old keypress response after each image. We varied picture resolution across blocks (16x16, 32x32, 64x64, or 256x256 pixels). Old and new pictures within a block had the same resolution. The second (old) presentation of an image could lag 2, 4, 8, 16, 32, 64 or 128 trials after the first. Higher resolution produced better performance, longer lags worse performance. However, performance was well above chance even at 16x16: 89% correct at lag 2, 52% at lag 128, d′ of 2.52 and 1.18, respectively (using the 16% overall false alarm rate). Similar performance was obtained whether lower resolution images were presented as smaller, thumbnail versions of 256x256 images or as highly blurred 256x256 images. Is resolution more important at encoding or at recall? In Experiment 2, the first presentation of a picture could be 32x32 or 256x256 pixels. The second (old) presentation of an image was always at a different resolution from the first. Results were strikingly asymmetric. Encoding at 256x256 produced good memory at 32x32 (d′=1.80). Encoding at 32x32 produced very poor memory at 256x256 (d′=0.15), far worse than encoding and testing at 32x32 (Exp 1: d′=2.04). We conclude that the representations of highly degraded images can support robust recognition memory. However, when observers see full resolution images they are unable match them to degraded representations of the same picture in memory.

Kuzmova, Y. Wolfe, J. (2010). Tiny Memory: How many pixels are required for good recognition memory? [Abstract]. Journal of Vision, 10(7):1244, 1244a,, doi:10.1167/10.7.1244. [CrossRef]

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.