Back

Semantic Memory Traces Reflect How They Were Last Retrieved

Qi, X.; Coutanche, M. N.

2025-06-01 neuroscience
10.1101/2025.05.30.657135 bioRxiv
Show abstract

Episodic memories are known to change with each act of retrieval. We hypothesize that semantic memories are altered in a similar way when retrieval draws on a core dimension of conceptual knowledge: semantic granularity. Each level--from accessing a concept via unique perceptual features to its thematic context--shapes the retrieval route used to access semantic memories. This study tests whether semantic granularities used to retrieve a concept influences its reactivation during later recognition. Human participants learned and retrieved novel word-image pairs while undergoing fMRI. Following encoding, items were retrieved via one of three semantic levels: item, category, or theme. Later, participants accessed the memories in recognition and cued recall tests. Although behavioral memory performance was matched across conditions, neural activity during recognition varied based on prior retrieval history. Recognition patterns could be classified according to prior retrieval granularity in ventral temporal cortex for both remembered and non-remembered concepts, and in the hippocampus for remembered concepts only. A whole-brain searchlight analysis revealed bilateral clusters along the ventral visual stream, from early visual cortex to fusiform gyrus, where retrieval history was decodable. Representational similarity analyses showed that category-level retrieval increased pattern consistency in early visual cortex and pattern reactivation in ventral temporal cortex, while item-level retrieval enhanced memory trace distinctiveness in visual word form area. Theme-level retrieval increased activity in left ventrolateral prefrontal cortex. These findings show that how we access a concept leaves a detectable trace on subsequent neural reactivation, subtly shaping how conceptual knowledge is organized in the brain.

Published in Journal of Cognitive Neuroscience (predicted rank #3) · training set

Matching journals

The top 3 journals account for 50% of the predicted probability mass.

50% of probability mass above

"Similar papers" are the closest papers from that journal in the model's embedding space. They show what the match is built on, but the ranking comes mostly from a classifier over the whole training set, not from these examples alone.