Not-so-working memory: Drift in fMRI pattern representations during maintenance predicts errors in a visual working memory task

Abstract

Working memory (WM) is critical to many aspects of cognition, but it frequently fails. Much WM research has focused on capacity limits, but even for single, simple features, the fidelity of individual representations is limited. Why is this? We used fMRI and a pattern-based index of \"representational drift\" to investigate how ongoing changes in brain activity patterns throughout the WM maintenance period predicted performance, using a delayed-match-to-sample task for a single item with a single critical feature: orientation. In trials where the target and probe stimuli matched, participants incorrectly reported more non-matches when their activity patterns drifted away from the target. In trials where the target and probe did not match, participants incorrectly reported more matches when their activity patterns drifted towards the probe. Our results suggest that WM errors are not simply due to unstructured noise, but also drift within representation space that can be indexed by neuroimaging.