Memory's double take: Dissociating two forms of recollection in visual working memory

Visual working memory (VWM) supports both change detection and repetition detection, yet it remains unclear whether these operations can be functionally dissociated. We compared two VWM paradigms designed to isolate each process. The Any-Difference test presents an array of five colored squares, a brief retention interval, and then a test array in which either all squares retain their color or one square changes its shade; participants indicate whether any change occurred. The Any-Sameness test uses the same stimulus sequence but has participants report whether any square is a repeat of a previously shown item. In Experiment 1 (N = 24), participants provided six-point confidence ratings, and receiver operating characteristics (ROCs) were fit with a dual-process signal detection model, yielding estimates of recollecting a change, recollecting a repetition, and familiarity-driven discrimination (d'). The Any‑Difference test produced steep, curvilinear ROCs with higher recollection estimates for changes than repetitions, whereas the Any‑Sameness test produced shallower ROCs with higher recollection estimates for repetitions. Experiment 2 (N = 28) added subjective reports in which participants indicated whether they had actively maintained an item and "perceived" a change/repetition, or merely "sensed" it. ROC patterns replicated Experiment 1 and subjective reports confirmed that recollection estimates corresponded to trials involving active maintenance. Together, these findings show that the two tests provide complementary leverage on change- and repetition-based judgments and that the ability to recollect changes can be dissociated from recollecting repetitions. These methods offer a new approach for probing the underlying VWM processes in both healthy and clinical populations.