Limits on selecting multiple items from working memory: the role of context and item competition.

To support goal-directed behavior, working memory (WM) must allow flexible and efficient access to its contents. While much research has examined how individual items are selected from WM, less is known about the principles guiding the simultaneous selection of multiple contents. Prior work has shown that multi-item access is slower and more error-prone than single-item access. The present study aims to clarify the mechanisms underlying this cost. Specifically, we examine whether the performance decline identified in our prior work reflects increased competition among memory items, among the context representations that guide selection, or both. To distinguish between these possibilities, we used a spatial retrocuing paradigm in which we independently manipulated (1) the number of cued locations (i.e., spatial contexts) and (2) the number of to-be-selected memory items. Across five experiments, we consistently found that cueing two spatial locations-compared to one-substantially delayed selection, even when the number of relevant items was held constant. By contrast, relevant set size had a smaller and less consistent effect on selection speed. These results suggest that the bottleneck in multi-item WM selection arises from the need to use multiple contexts to retrieve the associated memory contents.