A connectome-based prediction model of long-term memory

Abstract

Although many studies have investigated the neural basis of intra-individual fluctuations in long-term memory (LTM), few have explored the differences across individuals. Here, we characterize a whole-brain functional connectivity (FC) network based on fMRI data in an n-back task that robustly predicts individual differences in LTM. Critically, although FC during the n-back task also predicted working memory (WM) performance and the two networks had some shared components, they are also largely distinct from each other: the LTM model performance remained robust when we controlled for WM and vice versa. Furthermore, regions important for LTM such as the medial temporal lobe did contribute, but only partially, to predicting LTM. These results demonstrate that individual differences in LTM are dependent on the configuration of a whole-brain functional network including but not limited to regions traditionally associated with LTM during encoding and that such a network is separable from what supports WM.