Keynotes | K&Ts | GACs | Talks | Posters | Search

Poster E in Poster Session E: Thursday, August 6, 10:30 am – 12:15 pm, Kimmel Center, Shorin & Rosenthal Rooms

Replay Supports Both Future and Immediate Decisions by Updating a Successor Representation

Shuheng Guo¹, Alexa Booras¹, Michael Kahana², Itzhak Fried³, John J. Sakon³, Aaron M Bornstein¹; ¹University of California, Irvine, ²University of Pennsylvania, ³University of California, Los Angeles

Presenter: Shuheng Guo

Neural replay has been proposed to play two distinct roles: 1) Consolidating memories to support abstraction and 2) guiding planning for decision-making. Here, we examine if a single underlying principle—the successor representation (SR)—explains both. Analyzing single neuron recordings from the hippocampus of neurosurgical patients while they performed a memory-guided maze navigation task probing both spatial and sequential order memory, we find that rapid replay of neuronal pairs during separate task phases differentially influences subsequent behavior. Replay during rest prior to the start of each trial relates to the accuracy of spatial retrieval on future trials, while replay during the sequential order test relates to accuracy on this test for current and future trials–but critically, not vice versa. Modeling these replay events as random and directed walks, respectively, of trajectories along the learned maze structure, we find an SR captures the qualitative effects of replay on future trials in both task phases. Our results suggest neuronal replay supports memory-guided decisions to learn task structure by updating an SR of the environment at distinct timepoints, bringing seemingly disparate roles under the same theoretical framework.

Topic Area: Memory, Learning & Knowledge Structures