Abstract
Representations in working memory (WM) can have different levels of activation for representations within and beyond the focus of attention (FoA), according to some recent state models of WM. It is, however, still unclear whether the internal FoA in WM can maintain only a single representation at a time or multiple representations concurrently. To test the two competing hypotheses, we manipulated the number of retro-cues in a delayed estimation task in which participants were required to memorize three briefly presented colors over a short delay interval and then reproduced one of the remembered colors on a continuous colorwheel. The errors in these short-term recall responses were fit with Zhang & Luck (2008) mixture model, producing estimates of the probability of remembering and mnemonic precision of retained memory representations. In the middle of the delay interval, one, two, or three items in WM were cued simultaneously. The probability of remembering decreased significantly from the one-cue condition to the two-cue condition, but remained highly comparable between the 2-cue and 3-cue conditions, indicating a bottleneck for the FoA. In contrast, as the number of retro-cues increased, mnemonic precision decreased, reflecting the decrease in the amount of attentional resources that each cued item received. Similar pattern also manifested to reaction time (RT) based on ex-Gaussian model fits of RT distributions. Taken together, the results provide strong support for a highly-limited capacity for the FoA within WM.
Meeting abstract presented at VSS 2018