How robust are the benefits of sleep on learning and memory?
Freitag, 04.06.2021:
8:30 - 10:00

Chair der Sitzung: David Philip Morgan, Central Institute of Mental Health
Chair der Sitzung: Gordon Feld, Zentralinstitut für Seelische Gesundheit - Universität Heidelberg
Ort: Learning, memory, and sleep

Zusammenfassung der Sitzung

Although the benefits of sleep for memory are well established, recent findings suggest that these effects may be more variable than expected. This symposium attempts to evaluate the robustness of the enhancing effect of sleep on memory. To this end, Chloe Newbury and Sabrina Berres will present two independent meta-analyses of the impact of sleep deprivation vs. wakefulness and sleep vs. wakefulness on declarative memory respectively. Both meta-analyses find that the meta-analytical effect of sleep on memory is not large but rather small to medium sized (approximately d = 0.4). Due to publication bias meta-analyses can inflate effect sizes making replication attempts of previous research necessary. In line with this, some well-known findings in sleep and memory research have proven difficult to replicate conceptually. For instance, David Morgan will present a large-scale (N = 4,000) registered report using online assessment showing no effect of sleep on recognition memory in an eyewitness identification paradigm. Finally, Gordon Feld will present data on an experiment using different word list lengths, which demonstrates the dependence of sleep-dependent memory consolidation on specific context factors. Taken together these findings demonstrate that the effect of sleep on memory may be less robust than expected from the literature. During the discussion, we will outline developments that could increase the robustness of sleep and memory findings in the future.


The impact of sleep deprivation on memory: A meta-analysis

Chloe Newbury, Rebecca Crowley, Kathleen Rastle, Jakke Tamminen

Royal Holloway, University of London, United Kingdom

A substantial number of studies suggest that acute sleep deprivation both before and after encoding has a detrimental effect on memory for newly learned material. However, there is as yet no quantitative analysis of the size of these effects. We conducted two meta-analyses of studies published between 1970 and 2020 that investigated effects of sleep deprivation on memory, one for deprivation occurring before and one for deprivation occurring after encoding. We found 55 effect sizes investigating sleep deprivation before encoding, which showed that sleep deprivation impairs encoding with a medium effect size (g = 0.62). However, there was evidence of publication bias, with a trim-and-fill procedure yielding an estimated g = 0.47. A post-hoc power analysis of each study found a mean power of 55% to detect the meta-analytic effect size. We found 117 studies showing a small effect (g = 0.30) of sleep deprivation after encoding impairing memory. The mean power to detect the meta-analytic effect was 16%. Our analyses suggest acute sleep deprivation may have an impact on memory, but both direct replications and better powered conceptual replications are needed to increase our ability to estimate the true effect size.

The benefits of sleep on episodic memory: A meta-analysis

Sabrina Berres, Edgar Erdfelder

University of Mannheim, Deutschland

People remember more previously learned information after sleep than after an equally long period of wakefulness. Although the positive effect of sleep on episodic memory is well established, its size is still unclear. In the present meta-analysis, we analyzed 824 effect sizes from 271 independent samples reported in 177 articles published between 1967 and 2019 to quantify the sleep benefit in episodic memory and to investigate potential moderator variables. We found a moderate overall sleep benefit in episodic memory (g = 0.45). Evidence for publication bias was ambiguous. However, when accounting for possible selective reporting, the effect is approximately half the size but still significant. For moderator analyses, we applied multilevel meta-regressions with robust variance estimation on the main dataset and on a subset containing only single words and word pairs as stimuli. According to these analyses, the sleep benefit increases when stimuli are studied multiple times compared to once. Another moderator is the memory task: The sleep benefit is largest in free recall, followed by cued recall and recognition tasks. Finally, the sleep benefit is more pronounced when controlling for memory performance right after learning. More specifically, using a direct measure of forgetting (memory performance right after learning - performance in a delayed memory task) yields a higher sleep benefit than when the memory performance in a delayed memory task is considered only. We discuss theoretical implications and provide guidance to increase the robustness of the sleep benefit in episodic memory in future research.

Drawing conclusions from null findings in sleep research: An investigation of the impact of sleep on eyewitness identifications, a registered report.

David Philip Morgan1, Jakke Tamminen2, Travis Seale-Carlisle3, Laura Mickes4

1Central Institute of Mental Health, Mannheim, Germany; 2Royal Holloway, University of London; 3Duke University School of Law; 4University of Bristol

Sleep is largely considered a brain state that is optimal for memory consolidation due to reactivation of memories encoded during preceding wakefulness. Given the importance of sleep in consolidating memories, we investigated the impact of sleep on eyewitness identification performance. Indeed, there is opportunity for sleep, and therefore consolidation, to occur between witnessing a crime and making a line-up identification. A line-up procedure is a recognition memory test that contains the police’s innocent or guilty suspect and fillers, who are known innocents. In our large-scale registered report (N = 4,000), we manipulated the presence and absence of sleep between witnessing a video of a mock crime and making a decision on a line-up. We predicted that sleep would improve discriminability (i.e., more identifications of the guilty suspect and fewer identifications of the innocent suspect). We did not have strong predictions about the impact of sleep on reliability (i.e., the probability that the identified suspect is guilty). Our results did not bore out our predictions, sleep did not benefit discriminability. Sleep also had no impact on reliability. These results are counter to the evidence pointing towards beneficial effects of sleep on memory. We discuss whether our findings indicate that the benefit of sleep on memory is smaller than is stated. Moreover, had we not taken the registered report route, these results may not have been published because of publication bias or they may not have been presented as originally intended because of pressure to conduct and present post-hoc analyses.

Does sleep-dependent memory consolidation depend on information load?

Gordon Feld

Zentralinstitut für Seelische Gesundheit - Universität Heidelberg, Deutschland

Sleep has been shown to benefit the consolidation of declarative memories. Many studies have used word-pair lists and cued recall procedures to elicit a robust benefit of sleep on long-term memory. Here, I will show a series of experiments, where manipulating the lengths of these word-pair lists affected this effect of sleep on memory. Specifically, a medium list-length of 160 word-pairs showed a robust effect of sleep on memory, whereas lists of length 40, 320 and 640 showed no benefit of sleeping during retention. I will demonstrate how theories of sleep dependent memory consolidation can be used to explain these counterintuitive results. Next, I will argue that many patterns of results from this research could have been aligned with theories of sleep and memory research. This theoretical flexibility needs to be tackled, e.g., by modelling approaches, to allow a better understanding of the processes underlying sleep-dependent memory consolidation.