Investigating the long-term trajectory of targeted memory reactivation during sleep at home

It is thought that sleep helps stabilise newly formed memories (Klinzing et al., 2019; Rasch & Born, 2013). During slow-wave sleep, new memories are reactivated and gradually reorganised into durable neural representations. One method to investigate memory reactivation during sleep is called targeted memory reactivation (TMR): during wake, participants learn associations between items and sounds, and in ensuing sleep, a subset of these sounds is played, prompting memory reactivation without disrupting sleep. While research has repeatedly observed that TMR benefits retention of those specific memories in ensuing wake (Hu et al., 2020), little is known about whether these benefits last beyond a day or a week, and how the neural representations of the memories change over time. In a previous lab-based study (unpublished, N=15), we investigated the long-term effects of TMR and found no significant benefits of TMR on memory performance at any delay (1 day, 1 week, 3 weeks). Considering the previous evidence suggesting TMR consistently improves memory performance (Hu et al., 2020), we can think of a few factors possibly contributing to this null result: 1) The large number of control sounds: Most studies that incorporate control sounds (i.e., sounds that have not been presented during the learning task) have either only one or two control sounds which are repeated, or they are presented on a different night (Hu et al., 2020). In our previous study, we wanted to include a balanced number of control sounds to TMR sounds for better time-locked EEG comparison. However, in doing so, this might have inadvertently led to the lack of the behavioural TMR bias effect, i.e., an overload of various sounds being played resulting in less room for specific biasing towards the reactivation of the target memories. 2) Small sample size: Although TMR effects are repeatedly observed (Hu et al., 2020), the effect sizes are often not very large. Considering that the sample size in the previous study was restricted to 15 (due to limited resources), it is unclear whether the null result was due to a lack of statistical power to detect an effect or if it was a true null effect. 3) Disrupted sleep: It may be the case that the unfamiliar sleep lab environment or sounds being played resulted in disrupted sleep which in turn could have prevented the TMR benefit (Göldi & Rasch, 2019; Whitmore & Paller, 2023). 4) The strength of the association between the sound and the image location was unknown in the previous study. Although participants were exposed to the stimuli several times (through passive and active learning) and sounds were semantically linked to the images (e.g., a picture of a cat and the sound of a cat meowing), we cannot rule out the possibility that the association between sound and image-location was weak. If this was the case, the sounds presented during sleep may not have sufficiently triggered reactivation, resulting in a null effect of TMR. This study will endeavour to address some of these factors in a remote setting to enable a larger sample of participants in a more representative sleeping environment than in the lab: 1) No control sounds will be played during the post-learning night. Control sounds will instead be played during the adaptation night prior to learning, to avoid interfering with any possible benefits of playing target sounds during post-learning sleep. 2) The remote setup enables data collection of larger cohorts, thus improving statistical power to detect any possible TMR benefit. 3) To reduce the likelihood of sleep disruption, the overnight sleep session will be conducted in the participant’s own home. To increase familiarity with the sleep EEG kit as well as hearing sounds during sleep, the participants will complete one or more adaptation nights. 4) To test the link between the sound and location of the image, we have added a test at baseline assessing memory performance in the same way as the main test, but without any image stimuli to rely on. In this test, the participants hear the sound and are asked to move an empty box (replacing the image) to the remembered location. Based on the accuracy of this sound-location test, we will be able to determine whether the sound on its own prompts retrieval of the stimulus location. An alternative explanation is that the delayed memory test may have abolished the benefit of TMR. Most previous studies assess memory performance immediately after the sleeping phase. In our previous study, all memory tests were administered in the evening to keep timings across sessions consistent. This meant that there was a 24-hour delay between the immediate and delayed test which included a delay of sleep and wakefulness. It may be the case that TMR benefits are only observable for delays of sleep. However, we would argue that if TMR theoretically enhances reactivation and thereby strengthens and preferentially consolidates target memories during sleep making them less prone to interference, the TMR benefit should still be observable after a longer delay including wakefulness. An alternative theory is that sleep reduces contextual interference (Yonelinas et al., 2019). Thus, there is the possibility that the null effect is true and wakefulness after TMR abolishes the effect due to contextual interference. Considering that the TMR effect should theoretically be present after a 24-hour delay, this is not considered a confound and the timings for the memory tests will be kept the same time of day. This study will address whether cueing visuospatial memories in an at-home setting during sleep improves behavioural memory performance (as compared to uncued memories) in the longer term. Considering that TMR is starting to be used in patient populations, we urgently need to understand whether the technique renders any lasting benefits.