Retro-cue benefits across time

Abstract

Human observers can maintain multiple items in working memory. Unequal allocation of attention during memory encoding may lead to differences in how susceptible these memoranda are to visual interference. Even after the visual input is no longer available, observers can shift attention within the contents of working memory. This reallocation of attentional resources within working memory is often studied by means of a retro-cue, presented after offset of the memory array and indicates which item will be tested in a subsequent memory task. The retro-cue allows observers to prioritize this item, and shield it from visual interference, thus increasing memory performance compared to a post-cue condition, in which the test item is cued at the onset of the memory task. Such a retro-cue benefit, however, is not observed in all participants or in all studies. It often requires many trials before retro-cue benefits emerge, and some subjects seem to never demonstrate a retro-cue benefit. In this experiment, we investigated how much training is required for a reliable retro-cue benefit to emerge. Observers performed a change detection task for 10 hours over 7 days. Retro-cue benefits surpassed 5% in 9 out of 10 observers, emerged after about 250 trials on average and continued to increase until about 500 trials. However, we also observed large differences between subjects in required training time and final magnitude of the retro-cue benefit. We found that the retro-cue benefit typically reached a plateau after some time, while accuracy and associated measures of memory capacity (surprisingly) continued to increase in both the post-cue and the retro-cue condition even after 10 hours of task performance. Our results cast doubt on the idea of working memory capacity as a stable property of an observer and suggest that studies employing a retro-cue should always be prefaced by a training session.