Schedule of Events | Symposia

Natural Scene Representations in Parietal Cortex Predict Fine-Grained Representational Structure of Verbal Recall

Poster Session C - Sunday, March 30, 2025, 5:00 – 7:00 pm EDT, Back Bay Ballroom/Republic Ballroom

Anisha Babu¹ (ababu@uoregon.edu), Julian Gamez¹, Zhifang Ye¹, Brice Kuhl¹; ¹University of Oregon

Human fMRI studies frequently use representational similarity analyses to test whether brain regions reflect objective properties of visual stimuli during memory encoding or retrieval. While these fMRI measures have been linked to memory success (e.g., remembered or forgotten), there is less evidence linking representational similarity to subjective expressions of memory. Here, N=38 human participants learned associations between faces and naturalistic scene images during fMRI. Scene images consisted of six exemplars from each of three categories (e.g., ‘libraries’). After scanning, participants completed a verbal recall task in which each face was presented as a cue for participants to type a description of the corresponding scene from memory. Recall descriptions were then transformed into embeddings using a Natural Language model. Of critical interest was whether neural representations predicted the representational structure of recall for images within each category (e.g., which libraries were associated with the most similar verbal recall). To test this, pairs of images from each category were sorted according to the similarity of their recall embeddings. Then, fMRI pattern similarity was computed for these pairs, separately for regions of interest in the hippocampus, visual cortex, and parietal cortex. Across these regions, parietal cortex similarity (angular gyrus) was uniquely, and positively, related to verbal recall similarity (p<0.001). These findings indicate that angular gyrus representations contained fine-grained information that predicted subtle differences in verbal recall of naturalistic scenes.

Topic Area: LONG-TERM MEMORY: Episodic