Early or late distractions hurt working memory differently depending on how long you look

Visual Working Memory (VWM) is essential for temporarily retaining goal-relevant visual information, yet its limited capacity renders it vulnerable to distraction. While previous studies have examined the effects of distractors occurring during encoding or maintenance, it remains unclear how stimulus presentation duration modulates these effects and whether different types or quantities of distractors exert similar costs. Across three experiments, we systematically investigated how distraction timing, distractor content, and distraction load interact with encoding duration to influence VWM performance. In Experiment 1, participants performed continuous recall and change detection tasks under four distraction conditions (no-, encoding-, delay-, and full-distraction) and two encoding durations (short: 200 ms; long: 1000 ms). Encoding-stage distractions impaired performance only in the short-duration, high-precision task, whereas delay-stage distractions consistently disrupted memory regardless of duration or task type. Experiment 2 manipulated distractor-target similarity (same-category vs. different-category distractors) and revealed that homogeneous distractors exerted stronger disruption, particularly when presented during the delay period. In contrast, heterogeneous distractors could be effectively suppressed when sufficient encoding time allowed for robust consolidation. Experiment 3 examined perceptual load (low vs. high) and showed that increasing distractor quantity did not amplify interference, suggesting that once memory consolidation is complete, delay-stage distractions disrupt VWM representations regardless of distractor load. Together, these findings reveal an asymmetry in how VWM handles distractions at different processing stages. While extended encoding supports resistance to early distraction, maintenance-stage distractions exert persistent effects-especially when distractors are similar to targets. Our results highlight the importance of presentation duration and distractor similarity in shaping VWM robustness, and suggest that distinct cognitive mechanisms may underlie suppression at encoding and maintenance stages.