Abstract
Working memory (WM) has a limited representational capacity, and mechanisms of selective attention are needed to prioritize content in WM for behavioral output. Attentional prioritization in WM can be studied using retrospective cues. For example, memory performance can be facilitated by an informative cue presented during the delay period of a WM task. This effect – termed a retro-cue benefit – occurs frequently when participants are cued to prioritize a single WM representation, but it is unclear whether retro-cue benefits can also extend to multiple memory representations. Here, we report results from six experiments (N = 292) designed to test this possibility. In each experiment, participants were asked to remember displays containing four simple objects (oriented “clock-face” stimuli or colored squares) and later cued to prioritize zero, one, or two of these items. Participants then adjusted a probe to match one of the original memory items (i.e., continuous report) or were required to select the identity of a probed item from a set of discrete options (i.e., forced choice). In each experiment, we quantified how frequently participants reported items that were neither cued-nor-probed, items that were cued-but-unprobed, or items that were cued-and-probed. Unsurprisingly, participants reported cued-and-probed items with the greatest frequency in all conditions. However, during the critical cue-two condition, participants reported the cued-but-unprobed item more frequently than either of the uncued-and-unprobed items, suggesting that the cued-but-unprobed item was afforded a greater level of attentional priority. Based on these results, we conclude that participants can retrospectively prioritize multiple WM representations, but that doing so leads to increased competition between cued memory representations and an increase in “swap” errors.