Neural pattern classification reveals the temporal dynamics of competitive memory retrieval

Resolution of interference between competing memories is often critical for remembering. This study employed multivariate pattern analysis (MVPA) of electroencephalographic (EEG) data to investigate the temporal dynamics of retrieval competition and competition resolution. Competition between memories was created through the AB/AC interference paradigm. Memory retrieval of competitive AB and AC cue-associate word pairs was compared with non-competitive DE word pairs. Behavioural results showed worse memory performance for AC compared with DE word pairs, but comparable performance for AB and DE word pairs, revealing proactive without retroactive interference. Critically, the AB, AC and DE word pairs were encoded embedded in a movie with a distinct theme (first-person perspective of underwater, forest, and city environments). We trained classifiers to discriminate patterns of brain activity associated with the movies at encoding. The classifiers were later applied during memory retrieval to track memory reactivation. In an early cue time window, where only the word cue was presented (i.e. A or D), we observed classification performance for non-competitive retrieval, revealing target reactivation, but no classification for competitive retrieval, presumably due to the simultaneous reactivation of target and competitor memories. In a following probe time window, when participants were given a first-letter probe to retrieve the target word (i.e. B/C/E), we observed classification performance ~700 ms after probe onset for the competitive retrieval, revealing the time course of competition resolution. Importantly, classification accuracy in this late time window co-varied with memory performance, that is, with the resolution of proactive interference. This study offers novel insights into the time course of memory competition and competition resolution.