Rui Hu¹, Isabelle Boutet¹, Hanna Chainay²; ¹University of Ottawa, ²Université Lumière Lyon 2

False memory, the recall of events that never occurred or the misremembering of details, has been extensively studied using semantic stimuli within the Deese–Roediger–McDermott (DRM) paradigm. However, far less is known about false memory for non-semantic and perceptually meaningful stimuli such as faces. This project investigates how perceptual categorization based on social cues influences the retrieval of face representations in memory. Using a modified DRM paradigm, we created lists composed of faces sharing a perceptual category to test whether categorically related critical items (CIs) elicit stronger false recognition than unrelated distractors. Methods: Participants (target N= 34; aged 17–26) were tested in two blocks consisting of one Study and one Recognition phase. For the study phase of each block, participants were shown 25 identities from the FACES database, each morphed along two continua: emotion (happy-angry) and age (young-old). All faces belonged to one category. During recognition, participants were shown previously studied faces (old items), new faces from the opposite category (unrelated distractors), and new faces from the same category as the studied items (CIs). CIs were designed to elicit false recognition due to perceptual similarity and shared category membership. Participants judged via key press whether each face was “old” or “new.” Results: Preliminary analyses indicate that categorically related CIs elicit higher false recognition rates than unrelated distractors, as reflected by a greater percentage of “old” responses. This pattern suggests that participants were influenced by the overall category of the studied faces. CIs belonged to the same category as the studied items, they felt familiar, which increased the likelihood of false “old” responses. By extending this paradigm to socially meaningful facial cues, this project advances understanding of false recognition of faces and highlights conditions under which errors are most likely. These findings have implications for applied settings, including contexts involving eyewitness identification.

Acknowledgements: Funding: Supported by the Natural Sciences and Engineering Research Council of Canada Discovery Grant (2022-03998) to IB.