Remembering the past affects new learning: The temporal dynamics of integrative encoding

Abstract

Memories link elements across event boundaries, integrating overlapping content experienced at different times and places. Such memory representations are thought to rely on memory integration mechanisms. The present study used multivariate pattern analysis (MVPA) of electroencephalographic (EEG) data to elucidate the temporal dynamics of integrative encoding memory processes. Memory integration was investigated with the associative inference task. Participants learned overlapping paired associates, AB and BC pairs, comprising a word and a picture (a face or a bird) and were later tested on the AC inferred associations. MVPA was used to measure the online reactivation of previously learned AB memories while encoding overlapping but new BC events. Reliable AB reactivation was observed approximately 1500 ms after the onset of the BC event. Participants (n = 29) were divided into two groups according to their AC performance: high and low memory integration. Interestingly, while both groups showed comparable levels of AB reactivation, the reactivation effects were associated with different behavioral consequences. For participants in the high memory-integration group, reactivation was predictive of later AC performance, suggesting that an integrated ABC representation was formed during BC learning. Conversely, for participants in the low memory-integration group, the reactivation of AB was negatively correlated with BC performance, indicating that reactivation of overlapping memories impaired new learning. The present study extends previous literature by revealing the temporal dynamics of the integrative encoding mechanisms and by providing an account of inter-individual differences in the capacity to integrate memories across distinct episodes.