Schedule of Events | Symposia

Category learning drives neural repulsion initially but integration at a delay

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

Marlie Tandoc¹ (tandoc@sas.upenn.edu), Sarah Solomon^1,2, Ashley Williams¹, Alex Gordienko¹, Jacob Parker¹, Anna Schapiro¹; ¹University of Pennsylvania, ²Binghamton University

There has been accumulating evidence in the memory literature of a striking neural repulsion effect, where learning associations between highly similar items can drive their representations apart. These findings stand in contrast with work suggesting that we represent the commonalities across related items, especially with consolidation. Here we investigate the tension between these phenomena in a category learning task, where item similarity and task structure could motivate both integration and differentiation, measuring representations initially and after a period of consolidation. Thirty two participants learned novel flower categories, where each flower had petal types shared with category members as well as a unique petal. In an fMRI scan immediately after learning, we found that the hippocampus represented flowers from the same category as more dissimilar than flowers from different categories, consistent with other remarkable findings of repulsion in the literature. In the precuneus, this repulsion manifested at the feature level, with unique petals exhibiting lower similarity to others in the same category. After a 1-2 week delay, the repulsion effects vanished and there was instead evidence of integration in other cortical regions, including the anterior temporal lobe. Behaviorally, the delay resulted in forgetting of individual items but persistence of category knowledge. The findings suggest that the initial ability to differentiate highly similar information may be supported by an exaggeration of their differences, but that neither the detailed memory nor the differentiated representations persist over time, giving way to integration that might support a more generalized understanding of the environment.

Topic Area: LONG-TERM MEMORY: Semantic