Abstract
Recent research suggests the existence of two neurophysiologically distinct working memory (WM) subsystems. One stores information actively by means of sustained activation maintained by recurrent feedback loops (WM-active), and the other stores information passively via rapidly decaying changes in synaptic efficacy (WM-passive). The present study tests the hypothesis that WM-active represents precise metric properties of visual stimuli and is needed to perform any non-automatic cognitive computations, whereas WM-passive represents information categorically and can be maintained during ongoing cognitive operations. To test these hypotheses, we used a dual-task paradigm in which participants performed a non-automatic letter discrimination during the retention interval of an orientation delayed estimation task (Attend-Letter condition). Participants ignored the intervening letter stimulus in a control condition (Ignore-Letter condition). We assessed the presence of WM-active by recording the EEG during the maintenance interval of the WM task and decoding the sample orientation (because EEG decoding is possible only with active representations). We predicted that the orientation would be displaced from WM-active by the intervening letter stimulus when this letter was being discriminated, eliminating the orientation information from the EEG, but the orientation would still be maintained in WM-passive. As predicted, we could decode the sample orientation during the retention interval in the Ignore-Letter condition, but the presentation of the letter eliminated this orientation information in the Attend-Letter condition. Analyses of behavioral responses showed that the orientation memory was less precise and more categorically biased in the Attend-Letter condition, in which the orientation was presumably stored in WM-passive but was no longer present in WM-active. Together, the present results indicate that (1) WM-active is needed for even simple non-automated tasks such as letter discrimination, and (2) WM-passive maintains visual information in a less precise and more categorical manner than WM-active.
Meeting abstract presented at VSS 2016