Payachana Chareunsouk^1,2, Anantaporn Sena^1,3, Chaipat Chunharas^1,3; ¹Cognitive Clinical & Computational Neuroscience Lab, Faculty of Medicine, Chulalongkorn University, ²Medical Science, Faculty of Medicine, Chulalongkorn University, ³Chulalongkorn University Chula Neuroscience Center, King Chulalongkorn Memorial Hospital

When remembering multiple objects, people often group them into chunks to improve memory. Previous studies showed that people group objects by their basic features, such as location, orientation, or color. In the real world, objects often possess multiple basic features and belong to semantic categories. Whether and how semantic categories influence visual working memory remains debated. An answer might not simply be a yes-or-no as individuals may use categorical labels as a gist, resulting in more stable memory but a higher probability of within-category false alarms. This study investigates the role of semantic categories in real-world object visual working memory. Eleven participants performed a delay-match-to-sample task involving five to-be-remembered items, with a 5-second viewing of an object array, followed by a 2-second delay. The number of objects within a single category was manipulated across conditions: 5 (all same category), 4, 3, or 1 (all different categories). Participants were then shown a single object and asked to indicate whether it appeared in the original array. Test objects included items from the initial array (50%), new objects within the same category (25%), and new objects from different categories (25%). Participants had more accuracy when all objects belonged to the same category than from different categories (79±8%, 69±4%, 65±7% for 5 , 4 , 3 objects in the same category respectively). However, this improved accuracy was accompanied by a slightly higher false alarm rate, suggesting participants relied on semantic labels for memory grouping (11±6%, 8±6%, 9±6% false alarm rate of 5, 4, 3 objects in the same category respectively). Our findings indicate that semantic categories enhance visual working memory performance, albeit at the cost of increased within-category false alarms. This aligns with long-term memory research and highlights the potential interplay between semantic organization and working memory, offering insights into real-world memory processes.