Schedule of Events | Symposia

Learning and sleep reshape the representational geometry of visual experiences

Poster Session C - Sunday, March 30, 2025, 5:00 – 7:00 pm EDT, Back Bay Ballroom/Republic Ballroom
Also presenting in Data Blitz Session 1 - Saturday, March 29, 2025, 10:30 am – 12:00 pm EDT, Grand Ballroom.

Xianhui He¹ (xianhui.he@psy.ox.ac.uk), Philipp K. Büchel¹, Simon Faghel-Soubeyrand¹, Marcel S. Kehl¹, Bernhard P. Staresina¹; ¹University of Oxford

Neural and cognitive representations of the external world dynamically evolve over time. While research has suggested that new experiences undergo some transformation during memory consolidation, the mechanisms by which relationships among experiences are shaped remain poorly understood, including the role of sleep in this process. The present study used high-density electroencephalogram (EEG) recordings in 26 participants to elucidate potential changes in the representational geometry of visual stimuli as a function of learning and sleep. Prior to learning, participants performed a ‘localiser’ task (1-back) on 50 unique images from five categories (objects, faces, scenes, pseudowords, body parts). Participants then performed the ‘Memory Arena’ task in which the 50 images were presented in a specific spatial and temporal arrangement. Participants learned the sequence of the 50 images to criterion, followed by a ~2-hour nap. After the nap, they completed the localiser task again. We then used the first (pre-learning) localiser to train image-specific decoders based on multivariate EEG patterns and applied these decoders to the second (post-sleep) localiser. Interestingly, we found that during image processing in the second localiser, the image’s sequential successor from the learning task could be reliably decoded. This suggests that sequential representations persisted even when such information was not task-relevant. Importantly, our results further revealed that better learning performance and longer slow-wave (stage N3) sleep duration predicted the extent of incorporating successor representations across participants. Overall, our findings indicate that learning new inter-item relationships induces lasting changes in neural representational geometry, with N3 sleep strengthening these changes.

Topic Area: LONG-TERM MEMORY: Episodic