Passive but Accessible: Studied Information Is Not Actively Stored in Working Memory, Yet Attended Regardless of Anticipated Load

Abstract Working memory (WM) has a limited capacity of three to four items, yet people often encounter more information than this limit allows. A potential strategy, especially when overload is expected, is to hand off familiar content to long-term memory (LTM), freeing WM for new input. We tested whether anticipating additional memory load leads participants to rely less on active WM maintenance for familiar (studied) items by manipulating load expectation in an EEG experiment. Participants (n = 41) stored either novel or previously studied objects–color associations to be reported on a color wheel. Critically, we manipulated load expectation: In extra-load blocks, a second memory display followed the first on 80% of trials, whereas the second display followed in only 20% of trials in low-load blocks. Behaviorally, errors were higher in extra-load blocks, and studied items were recalled more accurately than novel ones, indicating a benefit of LTM availability. The contralateral delay activity, an EEG index of WM activation, was smaller for studied versus novel items, regardless of load expectation, indicating that participants relied less on active maintenance for studied items. In contrast, the N2pc and alpha-band (8–12 Hz) power lateralization were comparable across conditions, which indicates that sustained spatial attention toward the memorized items was deployed equally, even though WM recruitment was minimized. Together, these findings suggest that participants reduce WM activation when possible, keeping studied information accessible via attention, possibly through LTM, and that this mechanism is not modulated by anticipated memory load.