The Flexibility of Working Memory in Drawing on Episodic Long-Term Memory Representations in Serial Recall

TL;DR

This study explores how working memory uses prior long-term memories to improve recall, showing that benefits come from retrieving full episodic memories rather than just item activation.

Contribution

The study demonstrates that working memory benefits from retrieving full episodic representations, not just item activation, during serial recall.

Findings

Pre-learnt word pairs improved memory for matching words in serial recall tasks.

Benefits for new words depended on encoding strategies like chunking and offloading.

Position within lists influenced the effectiveness of pre-learnt and novel words.

Abstract

Prior episodic long-term memory (LTM) can enhance working memory (WM) by improving recall of WM representations that match pre-learnt information and by freeing up capacity for new information. In this study, we investigated the flexibility of WM in doing so. Specifically, we tested whether WM can make use of pre-learnt item-item associations in a serial recall task, which typically requires the formation of item-positional bindings. We examined whether any benefits arise from accessing full episodic representations or from item activation, and assess whether the observed benefits are best explained by compression accounts during encoding (e.g., chunking, offloading) or by redintegration at test. Furthermore, we tested whether the benefits for pre-learnt and novel words depended on the position within the lists. Across three experiments, we consistently found that incorporating pre-learnt word pairs into a serial recall task facilitated immediate memory for words that matched pre-learnt representations – speaking against an item activation account. However, the benefit on new words within lists that included pre-learnt pairs depended on whether the words could be easily submitted to encoding strategies, such as chunking or offloading, which was facilitated by providing matching grouping structures during encoding. Overall, our results expand our understanding of how prior experiences can benefit WM processes, demonstrating that such benefits mainly result from the retrieval of prior episodes, rather than enhanced item activation in episodic memory.

Figures