Associating Everything with Everything Else, All at Once: Semantic Associations Facilitate Visual Working Memory Formation for Real-World Objects

Integrating prior semantic knowledge with environmental information is essential for everyday cognition, yet how this process affects ongoing perception and memory remains unclear. We investigate this by studying how associative semantic knowledge interacts with perceptual constraints induced by brief encoding times, thereby supporting visual working memory (VWM) for real-world objects. Study 1 reanalyzed data from Quirk et al. (2020), involving 75 participants across 13,750 trials of a VWM task with randomly chosen objects and verbal distraction. We found that objects' semantic associations, estimated by a natural language processing model, predicted trial-level VWM accuracy under brief but not prolonged encoding times (0.2s vs. 1-2s). These results, unaffected by image similarity from computer vision models, were replicated in Study 2 with 50 participants across 11,880 trials. Combined, these findings suggest that semantic associations facilitate effective grouping among VWM items to mitigate perceptual constraints, highlighting the role of semantic knowledge in VWM formation.