Long-term Memory Contributions to Working Memory: A Contralateral Delay Activity Study

Abstract How does long-term memory (LTM) support working memory (WM) performance? Proposed mechanisms include “boosting,” where LTM enhances WM representations through additional activation but does not reduce WM load; “chunking,” where WM holds references to LTM, reducing WM load; and “offloading,” where LTM substitutes active maintenance in WM. To distinguish between these mechanisms, we measured the contralateral delay activity (CDA) and behavioral performance during a three-alternative forced-choice WM task, in which some trials (varying in memory array set size) included a prelearned color–object binding. The CDA is a well-established neural correlate that indexes the amount of information actively stored in WM, allowing us to distinguish between the three accounts by examining whether including prelearned bindings in a memory set of a given size reduced its amplitude compared with maintaining entirely new bindings. Overall, our findings supported the boosting account. Prelearned bindings improved performance for matching items in WM, but the benefit did not extend to newly encoded items, indicating no reduction in WM load. CDA amplitude increased with set size irrespective of whether the items were prelearned or new, reflecting active maintenance of all the to-be-remembered items. Overall, these results show that prelearned associations in LTM can boost WM performance without substituting for WM storage, providing evidence against offloading or chunking accounts in the context of our paradigm.