Schedule of Events | Symposia

Neural representation of relational memory types

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

Catherine Saunders¹ (cgs47@duke.edu), Paul Bogdan¹, Roberto Cabeza¹, Simon Davis¹; ¹Duke University

Relational memory plays a fundamental role in our daily lives, helping us link names to faces, understand movie plots, and many other seemingly mundane tasks. Past research demonstrates that semantic relations between concepts are encoded based on distinctions between the semantic relationships themselves, not on the concepts involved. However, it remains unclear how the brain encodes specific types of semantic relations. We investigated this using fMRI. During encoding, participants saw word pairings corresponding to objects and compared each pair across one of four dimensions: color, price, sound, or touch. Retrieval included an old/new (identical/recombined) relational memory task wherein participants were asked whether a given pair was previously shown. Participants then completed a source retrieval task where they identified which relational type (color, price, sound, or touch) had been used to compare a previous word pairing. Univariate analyses showed strong effects of memory success in both the hippocampus and angular gyrus. Additionally, multivariate tests were conducted across the brain, searching for regions that encode the relation types. This was done by analyzing neural pattern similarity across trials of the same comparison condition. From this analysis, the angular gyrus emerged as representing multiple relation types. This supports previous research suggesting the inferior parietal lobe and the temporal lobe form a convergence zone enabling us to form abstract representations of perceptual experiences. Furthermore, identifying regions like the angular gyrus which are central to relational memory could prove helpful in designing future interventions to support relational memory with age.

Topic Area: LONG-TERM MEMORY: Episodic