Payachana Victoria Chareunsouk^1,2,3, Yong Hoon Chung⁴, Viola Störmer⁴, Chaipat Chunharas^1,2,3; ¹Cognitive Clinical and Computational Neuroscience Center of Excellence, Faculty of Medicine, Chulalongkorn University, Bangkok, Thailand, ²Division of Neurology, Department of Medicine, Faculty of Medicine, King Chulalongkorn Memorial Hospital, Thai Red Cross Society, Bangkok, Thailand, ³Chula Neuroscience Center, King Chulalongkorn Memorial Hospital, Thai Red Cross Society, Bangkok, Thailand, ⁴Department of Psychological and Brain Sciences, Dartmouth College

When remembering multiple objects, people often group them into chunks to improve memory performance. Evidence shows that people group items by low-level visual features, but real-world objects also share semantic relationships that meaningfully shape working memory. Because forming a semantic gist may require time, we tested whether shortening encoding duration would weaken the semantic benefit or shift reliance toward high-level category labels rather than object-specific details. In an earlier study with a 5000 ms encoding time, arrays containing more items from the same category produced higher accuracy but also more within-category false alarms, indicating reliance on semantic gist. The present study examined whether this benefit persists under brief viewing. Twelve participants completed a delayed-match-to-sample task with five real-world objects under four semantic-composition conditions (five, four, three, or zero items from the same category). Arrays were shown for 500 ms followed by a 2000 ms delay. Semantic structure influenced memory under both encoding durations, but the strength of the effect differed. With long encoding, same-category arrays produced a robust semantic benefit (estimate = 1.075, t(14.6) = 3.442, p = .0037). Under brief encoding, semantic benefit was smaller and not reliable (estimate = 0.442, t(14.6) = 1.414, p = .178). Same-category false alarms also increased from 0.0719 at 5000 ms to 0.1387 at 500 ms. Increases in false alarms under brief encoding reflect either greater reliance on category-level similarity when item details are weak or more general errors when semantic structure is not extracted. Overall, semantic structure mattered at both durations, but the size of the effect varied with encoding time. Shorter exposure reduced semantic benefit and constrained the use of categorical information.