Yiheng Qiu¹ (), Mowei Shen¹, Yingtao Fu¹, Hui Chen¹; ¹Zhejiang University, Hangzhou, China

How does long-term experience influence the representations formed from current sensory input? Evidence from expertise and long-term memory research suggests that extensive experience can yield representations that are more structured and place less weight on sensory detail. From this perspective, we predicted that familiarity would make working-memory (WM) representations less sensitive to transient, task-irrelevant sensory variation. To test this prediction, we conducted four behavioral experiments using faces as a controlled stimulus domain. In Experiments 1a–1c (two-session design), participants first learned name–face associations for six identities and then completed a change-detection WM task in which they judged identity changes while ignoring changes in facial expression. We systematically manipulated the degree of stimuli overlap between learning and test: Experiment 1a used the same six expressions at both stages; Experiment 1b used three expressions at learning and a non-overlapping set of three at test; Experiment 1c used neutral faces at learning and a full set of expressions at test. Across these variants, expression changes reliably slowed responses, indicating incidental encoding of expression. Crucially, the magnitude of the expression-induced interference was consistently reduced for learned (familiar) identities compared with unfamiliar faces. Experiment 2 replicated this attenuation using naturally familiar celebrity faces without laboratory training, demonstrating that real-world long-term knowledge produces the same pattern. Together, these findings converge to show that familiarity makes WM performance less affected by irrelevant sensory variation. The results provide behavioral evidence that long-term knowledge influences the degree to which WM relies on sensory-specific information, with implications for theories of representational abstraction and interactions between WM and long-term memory.