Anticipated distractor processing relies on distinct representational states before and after distractor onset in visual working memory
Successful goal-directed behavior requires limiting the impact of salient information that is irrelevant to current goals. When distractors can be anticipated, task-irrelevant information may be encoded before it appears, yet the functional role of such anticipatory coding remains unclear. Does advance coding of irrelevant content itself support protection of visual working memory (VWM), or does effective control depend on how distractor representations evolve after entering processing? Here, using human EEG and multivariate decoding, we addressed this question by tracking distractor representations across temporal, spectral, and spatial dimensions. The distractor category was cued in advance and reliably decoded before physical onset, demonstrting content-specific anticipatory coding. However, the strength of this coding was not associated with behavioral protection. Instead, distractor processing showed a structured transition from relatively stable and temporally generalizable pre-distractor coding to more temporally unstable and time-specific post-distractor coding. Feature-level analyses further showed that distractor representations were shifted into a different representational configuration relative to representations of the same features when they served as task-relevant targets, indicating post-distractor reformatting of irrelevant content. Spectral and spatial analyses further showed that pre- and post-distractor coding relied on distinct frequency and scalp-topographic signatures. Strong anticipatory coding was associated with reduced subsequent post-distractor instability, whereas behavioral protection was associated with post-distractor instability even when pre-distractor metrics were taken into account. Together, these findings suggest a temporally coordinated control mechanism in which anticipatory coding may specify the upcoming irrelevant content, whereas post-onset reformatting may implement control by placing that content into a less stable representational state to reduce behavioral interference.