July 2013
Volume 13, Issue 9
Free
Vision Sciences Society Annual Meeting Abstract  |   July 2013
Pruning of visual memories based on contextual prediction error
Author Affiliations
  • Ghootae Kim
    Department of Psychology, Princeton University
  • Jarrod A. Lewis-Peacock
    Department of Psychology, Princeton University\nPrinceton Neuroscience Institute, Princeton University
  • Kenneth A. Norman
    Department of Psychology, Princeton University\nPrinceton Neuroscience Institute, Princeton University
  • Nicholas B. Turk-Browne
    Department of Psychology, Princeton University\nPrinceton Neuroscience Institute, Princeton University
Journal of Vision July 2013, Vol.13, 930. doi:https://doi.org/10.1167/13.9.930
  • Views
  • Share
  • Tools
    • Alerts
      ×
      This feature is available to authenticated users only.
      Sign In or Create an Account ×
    • Get Citation

      Ghootae Kim, Jarrod A. Lewis-Peacock, Kenneth A. Norman, Nicholas B. Turk-Browne; Pruning of visual memories based on contextual prediction error. Journal of Vision 2013;13(9):930. https://doi.org/10.1167/13.9.930.

      Download citation file:


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

      ×
  • Supplements
Abstract

Visual experience leaves incidental traces in long-term memory. This encoding happens continuously and automatically, resulting in a potentially overwhelming number of memories. How does the brain decide whether to preserve or discard a visual memory over time? We propose a new ‘contextual unreliability’ principle for memory management based on visual statistical learning and prediction error: Memory for an object should be weakened if the object fails to appear in a visual context with which it has been associated. We tested this hypothesis in an fMRI study. Observers were exposed to a continuous sequence of faces and scenes while performing a cover task. Unbeknownst to them, this sequence was generated from triplets (e.g., faceA-faceB-sceneC). The first two ‘context’ stimuli in each triplet were later repeated, but to create contextual unreliability, a new stimulus appeared in the third position (e.g., faceA-faceB-faceD). We hypothesized that if the original third stimulus (e.g., sceneC) had been associated with the now-unreliable context, then this violation would lead to forgetting. To measure the contextual association for each third stimulus, we used multivariate pattern analysis to decode its category during the repeated context (e.g., how much scene information about C was present during faceA-faceB-faceD). We then related this ‘prediction’ strength to subsequent memory, and found a negative relationship: strongly (and incorrectly) predicted stimuli were recognized less well than weakly predicted stimuli and control stimuli not part of a context triplet. To verify that this category-level classifier reflected a specific prediction of the original third stimulus, we replicated all results with representational similarity analysis: Strong exemplar-level reinstatement of the third stimulus in the repeated context produced greater forgetting. These findings provide insights into how memory systems incorporate contextual information to manage preexisting representations. Moreover, they reveal dynamic and implicit interactions between memory mechanisms and the visual system during perception.

Meeting abstract presented at VSS 2013

×
×

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.

×