September 2019
Volume 19, Issue 10
Open Access
Vision Sciences Society Annual Meeting Abstract  |   September 2019
Classification of load in visual working memory using single-trial EEG data
Author Affiliations & Notes
  • Kirsten Adam
    Department of Psychology, University of California San Diego
  • Edward Awh
    Department of Psychology, University of Chicago
  • Edward K. Vogel
    Department of Psychology, University of Chicago
Journal of Vision September 2019, Vol.19, 247a. doi:
  • Views
  • Share
  • Tools
    • Alerts
      This feature is available to authenticated users only.
      Sign In or Create an Account ×
    • Get Citation

      Kirsten Adam, Edward Awh, Edward K. Vogel; Classification of load in visual working memory using single-trial EEG data. Journal of Vision 2019;19(10):247a.

      Download citation file:

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

  • Supplements

Here, we used data acquired from 4 published studies (combined n = 340 participants) to test whether working memory load can be predicted from the broadband EEG signal across all electrodes, even at the single trial level. In Experiment 1 (Unsworth, Fukuda, Awh, and Vogel, 2015), we first demonstrate that we can classify memory load (2 versus 6 items) using a multivariate classifier across all electrodes. Importantly, the multivariate load signal was independent of the hemifield in which items were stored, such that combining trials across left and right lateralized displays did not impede accurate classification; thus the load signal has a global character. In Experiment 2 (Hakim, Adam, Gunseli, Awh & Vogel 2018), we demonstrate that the multivariate classification signal is specific to working memory task demands, as opposed to a physical stimulus confound or attention demands; we could classify load during the retention interval when participants were remembering the position of 2 versus 4 items, but not when they directed attention to those positions without storing the items themselves. Finally, in Experiments 3 and 4 (Fukuda, Mance & Vogel, 2015; Fukuda, Woodman & Vogel, 2015) we pushed this method further by classifying load across a finer parametric manipulation of load (set sizes 1–8). Intriguingly, the confusion matrix for these analyses revealed higher discriminability between lower loads (1–3) than between higher loads (6–8), suggesting that this multivariate signal respects hypothesized capacity limits and is not simply a measure of overall effort. Combined, these analyses demonstrate that multivariate classification of broadband EEG data can be used to predict visual working memory load in a manner that is (1) independent of where the items were encoded (2) specific to object-based storage demands and (3) precise enough to differentiate item-by-item increments in the number of stored items.

Acknowledgement: NIH grant 5T32-MH020002 (KA). NIH grant 5R01-MH087214 (EV and EA). ONR grant N00014-12-1-0972 (EV) 

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.