The Semantic Level of a Testing Question Influences Subsequent Memory Reactivation

Poster Session C - Sunday, April 14, 2024, 5:00 – 7:00 pm EDT, Sheraton Hall ABC

Amy Qi¹, Marc Coutanche¹; ¹University of Pittsburgh

Our knowledge of a concept can be represented at multiple levels of semantic granularity - from item-specific perceptual features to contextual meaning. Each level has meaningful consequences for whether and how we access memories. In this study, we asked if the granularity through which we reactivate a concept impacts its later retrieval. During fMRI scans, participants learned and retrieved pairs of novel words and visual images of everyday items. After initial encoding, participants actively restudied pairings by answering questions drawing on one of three granularities: item, category, or theme. Finally, participants performed a recognition test of encountered images versus similar lures. To test whether each concept was represented in multiple semantic granularities, their neural representational similarity matrices (RSMs) were tested against model RSMs. Six regions-of-interest (ROIs) showed a significant correspondence between recorded and model RSMs for item and category patterns in the ventral temporal cortex (VT), the bilateral anterior temporal lobe (ATL), the perirhinal cortex, and the visual word form area (VWFA). Testing-induced retrieval through different semantic granularities significantly influenced VT and vlPFC pattern reactivation during subsequent recognition. A Gaussian Naïve Bayes (GNB) classifier was able to distinguish recognition patterns based on their testing history (item, category, theme questions) for remembered and non-remembered concepts in VT, and only remembered trials in VWFA. These findings suggest that the semantic level evoked during testing can influence subsequent memory reactivation.

Topic Area: LONG-TERM MEMORY: Semantic