Schedule of Events | Symposia

Timing is Everything: Temporal Community Structure is Sufficient for Categorical Inference

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

Alexander N. Lawriw¹ (alawri1@lsu.edu), Christopher R. Cox¹; ¹Louisiana State University

Sequences are parsed into events by tracking transitional probabilities. How well such structure has been learned correlates with activity in the hippocampus. Recent work has raised the possibility that the hippocampus has some capacity to learn categorical representations and may even support categorical inference. We critically evaluated this interpretation in two experiments where participants were exposed to sequences of stimuli that were, unbeknownst to the participant, generated by random walks on a 15-node graph with three 5-node “communities” and one path between each community. Learning this structure is understood to involve the hippocampus. In experiment 1, following sequence exposure, participants responded to a series of 2AFC trials consisting of a reference stimulus paired with an attribute that was either distinctive of its community or generic (shared by all stimuli). During these trials, participants were asked to choose which of two alternative stimuli was more likely to share that attribute with the reference stimulus. Behavior was similar regardless of whether the attribute was distinctive or generic. In experiment 2, participants were explicitly taught to associate a community-distinctive attribute with a single member from each community before being exposed to the structured sequence. Following sequence exposure, participants selected which of two alternatives were more likely to have a particular community-distinctive attribute on each 2AFC trial. Preliminary analysis of 35 of the planned 60 participants in experiment 2 indicates a significant category inference bias, t(34) = 2.12, p = .04, suggesting that hippocampus-mediated representation of temporal community structure is sufficient for categorical inference.

Topic Area: LONG-TERM MEMORY: Semantic