The Loss of Information from Visual Working Memory depends on Retro-Cue Reliability.

Retrospectively cueing an item retained in visual working memory during maintenance is known to improve its retention. However, literature has provided conflicting results regarding the costs of such retro-cues for non-cued items, which has led to a variety of theories on the role of cueing in visual working memory. We hypothesized that the differences in the reliability of retro-cues across studies might be a factor underlying conflicting results. We predicted that the more reliable the cues, the larger the costs for non-cued items. Participants performed a working memory task while electroencephalography (EEG) was recorded. Retro-cues indicated which of several items was most likely to be tested. We manipulated, between blocks, the ratio of trials on which the cue was valid vs. invalid. In addition to memory performance, we investigated the contralateral delay activity (CDA) in the EEG, which is claimed to index visual working memory maintenance. We also applied multivariate pattern analysis (MVPA) in the frequency domain to decode the location of the cued item. Reconciling previous contradictory findings, costs of invalid retro-cueing on recall performance (i.e. probability and precision estimates) were found only for highly reliable cues. Nevertheless, benefits of valid cueing were present for both reliabilities, though larger for highly reliable cues. Moreover, the CDA emerged only after a highly reliable cue. Finally, decoding accuracy was above chance level in the high alpha band (i.e. 10-12 Hz) following a retro-cue and was larger for highly reliable cues. Our results suggest that non-cued representations are removed from working memory only if the retro-cue reliability is high. Moreover, the presence of valid cueing benefits in the absence of invalid cueing costs suggests that an item in working memory can be prioritized without reallocation of memory resources. Meeting abstract presented at VSS 2015.