Working memory capacity differences in working memory offloading

Abstract

Effective memory performance relies on the dynamic exchange of information between working memory (WM) and long-term memory (LTM). Past research has shown that both semantic and episodic LTM can enhance WM performance, as prior knowledge facilitates recall through mechanisms such as chunking, boosting, and offloading. Recent work by Bartsch and Shepherdson (2023) suggests that offloading—storing previously learned information in LTM to reduce WM demands—may be a key strategy for minimizing the impact of distractions on information maintained in WM. However, the extent to which individual differences in WM capacity (WMC) influence offloading remains unclear. This dissertation had three main goals: to (1) replicate prior findings showing that LTM enhances WM performance, (2) replicate prior findings supporting the mechanism of offloading, and (3) examine whether offloading depends on individual differences in WMC. We hypothesized that (H1a) WM recall performance would be higher for lists containing pre-learned items in LTM, demonstrating that episodic LTM benefits WM. We expected (H2a) that WM performance would be poorer following a distractor task, but more so for lists with only new items, as previously learned items should be offloaded to LTM and thus less affected. Alternatively, (H2b) if pre-learned items are not offloaded to LTM but instead remain in WM, we expected that performance would decline across all lists, regardless of whether they contained only new items or a mix of new and pre-learned items supporting a boosting account. Finally, we hypothesized (H3a) that higher WMC would be associated with more offloading. To test these hypotheses, participants completed a multi-phase memory study. In the LTM learning phase, they first studied word pairs. Next, during the WM phase, they encoded and retrieved lists of word pairs with varying amounts of pre-learned information (0%, 25%, 50%, or 75% Old). WM trials were followed by either a distractor task designed to disrupt WM maintenance or a blank screen, which allowed us to assess the extent to which distractions disrupt memory performance. Finally, participants completed an LTM retrieval phase to test their memory for previously learned items and reported which strategies they used to remember each word pair. Results indicated that pre-existing episodic LTM representations benefited WM performance, supporting H1a. However, we found mixed evidence for offloading (H2a; H2b). Although WM performance was worse on trials containing a distractor task, this effect did not interact with the presence of previously learned items, which provides some support for the boosting hypothesis (H2b). At the same time, certain list types (e.g., those containing 25% Old and 75% Old items) showed no significant performance difference between blank and distractor conditions, providing some support for the offloading hypothesis (H2b). Thus, we are left with an imperfect account of whether offloading or boosting account better fit our results. Lastly, WMC was not significantly associated with offloading, failing to support H3a and instead aligning with H3b. This finding may suggest that individual differences in WMC stem from other cognitive abilities, such as attentional control or domain-specific skills, rather than from differences in WM offloading. These findings highlight the complex interplay between WM and LTM and suggest that future research should explore offloading under different experimental conditions to further disentangle these accounts.