Poster F86, Tuesday, March 28, 8:00 – 10:00 am, Pacific Concourse
Mnemonic prediction errors modulate hippocampal connectivity patterns
Oded Bein1, Katherine Duncan2, Lila Davachi1; 1New York University, 2University of Toronto
Prediction errors (PEs) play a critical role in learning. Recent work showed that activation in hippocampal region CA1 linearly tracked how many current items were in conflict with a learned representation - or the number of mnemonic PEs. To the extent that PEs modulate memory updating and retrieval processes, we examined how PEs influenced CA1 connectivity. Over several sessions, participants learned a list of rooms; each room had a name (e.g., “John’s bedroom”) and an image depicting that room (Duncan et al, 2012). While in the scanner, participants were cued to recall the learned rooms followed by an image of the room. Crucially, this image could either be identical to the learned one (no PE), or had 1-4 changes (low-high PE). Focusing on the input to CA1, we asked how CA1-entorhinal functional connectivity was modulated by PEs. The results suggest that CA1-entorhinal functional connectivity was lower in the no-PE condition compared to all other PE conditions with no difference observed between different levels of PE. Interestingly, preliminary analyses suggest that CA1- CA3 connectivity decreased linearly as PE increased. CA1-subiculum connectivity, potentially the output of CA1, reflected that of a U-shape, in which connectivity was maximal for mid-level PE, and low for both no-PE and high-PE conditions. These results suggest dissociations between CA1-entorhinal and CA1-CA3 connectivity in the processing of mnemonic PEs.
Topic Area: LONG-TERM MEMORY: Episodic