Implementing manipulation in the face of distraction: Spatiotemporal and spectral dynamics in visual working memory

Visual working memory supports the transformation of goal-relevant information and the regulation of goal-irrelevant distraction. However, it remains unclear how the brain preserves and implements manipulation-related processing in the presence of competing irrelevant input. To address this question, we recorded electroencephalography while healthy adults performed a task in which manipulation and distraction were orthogonally crossed. Behaviorally, manipulation and distraction each increased response times and decreased accuracy, but their interaction was not significant. Neurally, the manipulation-related neural representation remained intact regardless of distraction, as evidenced by indistinguishable temporal decoding profiles, stable spatial activation patterns, and successful generalization of the neural decoder between distraction contexts. A follow-up full cross-temporal generalization analysis between manipulation and distraction indicated no reliable evidence for shared neural representations across time, supporting dynamically differentiable representational profiles. Direct contrasts between manipulation-related and distraction-related decoding further revealed that distraction decoding accuracy exceeded that of manipulation during encoding, whereas this relationship was reversed during the delay. During encoding, this dissociation was accompanied by different spatial activation-pattern distributions, with relatively stronger anterior contributions for distraction and posterior contributions for manipulation. During the delay, however, the dissociation was governed by a non‑topographic mechanism specifically within alpha-band activity, featuring a selective reduction of alpha power for manipulation and a selective elevation of alpha peak frequency for distraction. Together, these findings indicate that the brain preserves and implements manipulation in the face of distraction through dynamically differentiable representational patterns across time, space, and spectral dimensions.