Abstract
Working memory (WM) performance can be facilitated by an informative cue presented during storage. This effect – termed a retrocue benefit – is contingent on the reliability of the cue, with larger retrocue benefits observed for highly reliable cues (e.g., 80%) relative to less reliable cues (e.g., 50%). In principle, these differences could reflect cue-driven changes in the strength of a cued memory representation, changes in the strengths of uncued memory representations, or a mixture of both. Here we report two experiments designed to test these possibilities. We recorded EEG while participants performed variants of a retrospectively cued spatial WM task. On each trial, participants were asked to remember the spatial positions of two discs over a short delay. On critical trials a retrospective cue indicated which of the two discs was most likely to be probed at the end of the trial, and across blocks we varied the reliability of the retrospective cue between 50%, 75% and 100%. Next, we reconstructed and quantified time-resolved representations of the cued and uncued disc locations by applying an inverted encoding model to patterns of induced alpha band (8–12 Hz) activity measured over occipitoparietal electrode sites during WM storage. This allowed us to examine whether changes in retrocue benefits with cue validity are better explained by changes in the strength of the cued memory representation, changes in the strength of the uncued memory representation, or both. In both experiments, reconstructed representations of the cued disc were unaffected by retrocue reliability, while reconstructed representations of the uncued disc were reliably stronger when retrospective cues were less reliable. Our findings suggest that changing the reliability of a retrospective cue influences memory performance primarily by modulating the strength of uncued memory representations.