Poster Abstracts


Unique and Overlapping Contributions of Posterior Medial Network Nodes to Predicting Recollection Outcomes

Kyle Kurkela, BA & Maureen Ritchey, PhD

    Recollection-based memory has been associated with activity in a set of medial temporal and posterior medial (PM) regions collectively referred to as the core recollection network. Although the activity and connectivity of PM regions has been shown to reliably distinguish between successful and unsuccessful episodic retrieval, it remains unknown how these regions complement and/or interact with another to predict memory success. In this study, we aim to disentangle the unique and overlapping contributions of PM regions to predicting memory outcomes during retrieval. To do so, we performed a univariate decoding analysis by iteratively building logistic regression models and testing the accuracy of their predictions using a leave-one-trial-out cross validation procedure. Importantly, we performed the decoding analysis using a combinatorial region of interest (ROI) approach, testing the predictive power of mean BOLD activation in each PM region both in isolation and in all possible combinations. The combinatorial approach allowed us to test for both redundancy and complementary effects on prediction for all subsets of nodes within the network. That is, for each combination of regions, we tested whether the nodes were informationally redundant with one another (i.e., they made the same contribution to predicting recollection outcomes) or whether they were complementary to one another (i.e., each region conferred unique prediction benefits). Results suggest that PM network regions play complementary roles in supporting memory recollection, such that mean activation across combinations of nodes supports the most accurate prediction of recollection success. 

episodic memory, fMRI, recollection

  • Twitter Clean Grey

Proudly created with