Whole-body procedural learning benefits from targeted memory reactivation in REM sleep and task-related dreaming

Highlights

• Targeted memory reactivation (TMR) applied in REM sleep improves whole-body procedural learning.

• Dreaming about kinesthetic aspects of procedural learning also predicts greater improvement.

• The coexistence of REM sleep TMR and task-dream reactivations predicts maximal improvement.

• Results offer insights into the mechanisms of memory reactivation during sleep.

Abstract

Sleep facilitates memory consolidation through offline reactivations of memory traces. Dreaming may play a role in memory improvement and may reflect these memory reactivations. To experimentally address this question, we used targeted memory reactivation (TMR), i.e., application, during sleep, of a stimulus that was previously associated with learning, to assess whether it influences task-related dream imagery (or task-dream reactivations). Specifically, we asked if TMR or task-dream reactivations in either slow-wave (SWS) or rapid eye movement (REM) sleep benefit whole-body procedural learning. Healthy participants completed a virtual reality (VR) flying task prior to and following a morning nap or rest period during which task-associated tones were readministered in either SWS, REM sleep, wake or not at all. Findings indicate that learning benefits most from TMR when applied in REM sleep compared to a Control-sleep group. REM dreams that reactivated kinesthetic elements of the VR task (e.g., flying, accelerating) were also associated with higher improvement on the task than were dreams that reactivated visual elements (e.g., landscapes) or that had no reactivations. TMR did not itself influence dream content but its effects on performance were greater when coexisting with task-dream reactivations in REM sleep. Findings may help explain the mechanistic relationships between dream and memory reactivations and may contribute to the development of sleep-based methods to optimize complex skill learning.

Introduction

Although dreams are thought to facilitate long-term memory mechanisms (Palombo, 1976, Scrima, 1984), supporting evidence is slim (see reviews in Cipolli, 1995, Smith, 2010, Stickgold et al., 2001, Wamsley and Stickgold, 2010, Wamsley and Stickgold, 2011). Recent paradigms for assessing the memory functions of sleep permit better controlled assessments of dreaming’s role in memory; several such paradigms are based on evidence that newly encoded memories are spontaneously reactivated (or ‘replayed’) at least in part during sleep. Targeted memory reactivation (TMR) is the readministration during sleep of sensory cues that were previously associated with learning. One expectation about TMR stimulation is that it will bias the content of replay events that occur during sleep (Alm et al., 2019, Bendor and Wilson, 2012, Wei et al., 2020) and will enhance consolidation above and beyond sleep itself (see review in Hu, Cheng, Chiu, & Paller, 2020). TMR thus offers a novel opportunity to test whether and how induced memory reactivations are related to dreaming. In the present study, we test whether TMR alters dream content and use the method to experimentally assess dreaming’s role in procedural memory learning.

Evidence indicates that neural replays, i.e., partial reactivations of neural circuits underlying newly encoded memories, occur spontaneously during sleep. In slow-wave sleep (SWS), these replays have been recorded in hippocampal place cells (e.g. Kudrimoti et al., 1999, Skaggs and McNaughton, 1996, Wilson and McNaughton, 1994) and in cortical circuits (Euston et al., 2007, Ji and Wilson, 2007, Peyrache et al., 2009, Qin et al., 1997, Xu et al., 2019). More indirect evidence of spontaneous reactivations have also been seen in human sleep using various brain imagery techniques (e.g. Bergmann et al., 2012, Deuker et al., 2013, Fogel et al., 2017, Jegou et al., 2019, Murphy et al., 2018, Peigneux et al., 2004, Schönauer et al., 2017, Zhang et al., 2018) and were shown to be related to memory improvement (e.g. Deuker et al., 2013, Fogel et al., 2017, Peigneux et al., 2004). Growing evidence also suggests that task-related patterns of brain activity spontaneously re-emerge during REM sleep both at the hippocampal (Kumar et al., 2020, Louie and Wilson, 2001, Pavlides and Winson, 1989) and cortical (Eckert et al., 2020, Maquet et al., 2000, Peigneux et al., 2003, Schönauer et al., 2017) levels, but these effects have not been consistently observed (Kudrimoti et al., 1999, Lansink et al., 2009).

While one study found that sequences lasting up to minutes can be replayed in REM sleep (Louie & Wilson, 2001), replays in SWS are usually compressed in time (Euston et al., 2007, Lee and Wilson, 2002, Nadasdy et al., 1999, Skaggs and McNaughton, 1996) and can even represent imaginary trajectories (Caze et al., 2018, Stella et al., 2019), suggesting a transformation or integration of memory traces during sleep (Fogel et al., 2017) rather than exact replays. Dreaming of a learning task may represent one type of imaginally transformed memory trace replay (Wamsley & Stickgold, 2011).

That task-related reactivations occur spontaneously at the cortical level, and in both REM sleep and SWS, raises the possibility that they may be accompanied by the subjective experience of task-related dreaming (Stickgold et al., 2001, Wamsley and Stickgold, 2010). This possibility is consistent with evidence that dreams frequently integrate elements of recent, as well as temporally distant, experiences (e.g. Malinowski and Horton, 2014, Schredl and Hofmann, 2003, Vallat et al., 2017). The incorporation of recent waking-life experiences in dreams also correlates with frontal theta activity in REM sleep (Eichenlaub et al., 2018, Malinowski and Horton, 2014, Vallat et al., 2017), an oscillatory activity related to synaptic plasticity (Bramham, Maho, & Laroche, 1994) and offline memory consolidation (Boyce et al., 2016, Diekelmann and Born, 2010, Hutchison and Rathore, 2015). Nonetheless, research assessing whether task-related dreaming predicts improved learning remains sparse and inconsistent. Compelling evidence for a direct role of dreaming in learning is provided in studies by Wamsley, 2019, Wamsley et al., 2010a), in which subjects who dreamed about a virtual maze task in any stage of sleep had improved performance the next day. Several other studies report an association between task-related dreams and performance (De Koninck et al., 1990, De Koninck et al., 1988, De Koninck et al., 1996, Fiss et al., 1977; Schoch, Cordi, Schredl, & Rasch, 2019; Plailly et al., 2019). In contrast, some studies find that task-related dreaming from different sleep stages is unrelated to learning (Cipolli et al., 2004, Fogel et al., 2018, Wamsley et al., 2016), while others uncovered too few task-related dreams to even assess an associated consolidation effect (Nguyen et al., 2013, Schredl and Erlacher, 2010). Such seemingly negative results point to a need to look beyond direct task incorporations as a marker of memory replay to more general dream content changes that might facilitate learning; such changes might include dreams of non-obvious task-related, kinesthetic, associative or metaphorical features.

Evidence regarding the role of task-related dreaming in motor learning is also scarce and results are mixed. Fogel et al. (2018) showed that semantic similarities between participants’ descriptions of a tennis-playing task and their subsequent sleep-onset dream reports predicted better performance at that task. In contrast, dreaming about balance tasks did not (Nefjodov et al., 2016, Solomonova et al., 2018). The bizarreness, length and emotional intensity of REM dreams are all related to poorer speed on a mirror-tracing task, but not to decreased errors (Schredl & Erlacher, 2010), suggesting that dreaming supports memory consolidation in indirect, unexpected ways.

Further evidence supporting a role for dreaming in motor learning is that, among expert lucid dreamers, intentional practice of motor tasks in the dream state led to more improvement than did no such practice; tasks included dart-throwing (Schadlich, Erlacher, & Schredl, 2017), coin-tossing (Erlacher & Schredl, 2010) and finger-tapping (Stumbrys, Erlacher, & Schredl, 2016). Whether spontaneous motor and kinesthetic dream imagery are also related to motor learning is assessed in the present study.

The beneficial effects of TMR hold for different types of learning (see review in Hu et al., 2020), including procedural learning (e.g. Antony et al., 2012, Koopman et al., 2020, Laventure et al., 2016, Schonauer et al., 2014). TMR’s effects on procedural skills is based primarily on studies employing similar finger-focused sequential tasks, i.e., a serial reaction task (Cousins et al., 2014, Cousins et al., 2016, Diekelmann et al., 2016, Koopman et al., 2020) or a finger-tapping task (Antony et al., 2012, Laventure et al., 2016, Laventure et al., 2018, Pereira et al., 2017, Rasch et al., 2007, Schonauer et al., 2014). There is more limited evidence that TMR in SWS improves performance on gross procedural skills such as a non-dominant-arm throwing task (Johnson et al., 2019, Johnson et al., 2020, Johnson et al., 2018).

The sleep stage in which TMR cues are administered may affect performance. Sequential motor tasks depend on NREM, rather than REM sleep, physiology, including sleep spindle activity (e.g. Antony et al., 2012, Cousins et al., 2014, Laventure et al., 2016, Laventure et al., 2018) and slow-wave sleep duration (e.g. Antony et al., 2012, Cousins et al., 2016). Yet, continuous cueing in all stages of sleep also improves procedural learning (Schonauer et al., 2014). Further, changes in motor-related brain activity following cueing in SWS relate to both SWS and REM sleep duration (Cousins et al., 2016), suggesting a cross-stage processing of skill learning. Procedural learning has in fact been associated with both NREM (e.g. Barakat et al., 2013, Barakat et al., 2011, Korman et al., 2007, Morin et al., 2008, Nishida and Walker, 2007, Smith and MacNeill, 1994, Walker et al., 2002) and REM (e.g. Fischer et al., 2002, Karni et al., 1994, Nitsche et al., 2010, Plihal and Born, 1997, Schredl and Erlacher, 2007, Smith, 2001) sleep; REM sleep may be especially important for complex and cognitively demanding tasks (Greenberg and Pearlman, 1974, Pearlman, 1979, Smith, 2001). And although REM sleep dreaming commonly exhibits richer simulations of whole-body movements and kinesthetic sensations than does NREM sleep dreaming (Hobson, 1988, Occhionero and Cicogna, 2011, Porte and Hobson, 1996), literature on TMR applied in REM sleep to improve motor skills is sparse.

Although some research does support a role for dreaming in sleep-dependent learning enhancement, whether TMR-induced reactivations alter dreaming or interact with specific dream contents remains completely unknown. At best, some studies suggest that such effects or interactions are possible.

First, stimuli presented during sleep, though not administered in a TMR design, are often incorporated into dream content. Dreams are influenced by olfactory stimuli (Okabe et al., 2018, Schredl et al., 2009, Trotter et al., 1988), visual stimuli (Dement and Wolpert, 1958, Rechtschaffen and Foulkes, 1965), auditory stimuli (Berger, 1963, Burton et al., 1988, Dement and Wolpert, 1958, Hoelscher et al., 1981), and—most effectively—somatic/kinesthetic stimuli (Dement and Wolpert, 1958, Koulack, 1969, Leslie and Ogilvie, 1996, Nielsen, 1993). It is noteworthy that dream incorporations of such stimuli are rarely direct, episodic memories, but are rather partial representations of the stimuli and their context, involving substantial associative material. Second, some TMR-like experiments do report influences on dream content. Early pioneer Hervey Saint-Denys (1867) demonstrated that a music box playing, during his sleep, waltzes that had been associated with previous dance partners triggered dreams about those dance partners. Similarly, De Koninck and Koulack (1975) produced the dreamed reactivation of a presleep stressful film when the film’s soundtrack was readministered during REM sleep. More recently (Smith & Hanke, 2004), when a ticking clock sound was paired with a mirror tracing task and then readministered during phasic REM sleep, task-related dream changes were evoked. Specifically, participants reported longer dreams and dreams with references to competition sports and driving—both activities analogous to the mirror tracing task. With olfactory stimuli, when the smell of roses was associated with rural images and readministered during sleep, more dreams of countryside imagery were evoked (Schredl, Hoffmann, Sommer, & Stuck, 2014). Recently, Carr et al.’s (2020) novel ‘targeted lucid dreaming reactivation’ protocol—in which visual and auditory stimuli are paired with pre-sleep lucid dreaming training and readministered in REM sleep—successfully increased the occurrence of lucid dreams. A similar ‘targeted dream incubation’ protocol successfully shaped the content of sleep-onset imagery (Haar Horowitz, Cunningham, Maes, & Stickgold, 2020). Notwithstanding this evidence of dream reactivation, however, it remains to be shown how these reactivations are related to sleep-dependent memory processing.

Virtual maze tasks have been used in both TMR studies (Shimizu et al., 2018) and studies assessing dreaming’s role in spatial memory consolidation (Fogel et al., 2018, Nielsen et al., 2007, Solomonova et al., 2015, Stamm et al., 2014, Wamsley and Stickgold, 2019, Wamsley et al., 2010b). Our recent work suggests that a whole-body-immersion VR task substantially increases the likelihood of affecting both visual and kinesthetic dream content, i.e., brief exposure to the VR flying task used in the present study significantly increased dreams of flying (Picard-Deland, Pastor, Solomonova, Paquette, & Nielsen, 2020). In addition to enhancing the probability of successful dream reactivation, a room-scale VR environment may allow for a more ecological learning context that engages visuo-spatial, motor and balance systems. That VR-based interventions facilitate balance and motor rehabilitation in clinical populations (Cano Porras et al., 2019, de Amorim et al., 2018, de Araujo et al., 2019, Karamians et al., 2020) suggests they may be well-suited for assessing TMR efficacy with complex procedural learning.

In sum, while procedural learning may benefit from NREM sleep, REM sleep, or both, use of TMR in REM sleep has been largely overlooked, especially concerning its efficacy for improving whole-body and complex motor skills. In addition, dreaming’s role in sleep-dependent learning—and in TMR-mediated learning—remains unclear. Although considerable research shows that task-relevant sensory stimuli delivered during sleep can, on the one hand, improve related memories and, on the other, wend their way into dream content; whether these phenomena interact and share common memory reactivation processes remains to be investigated. An immersive VR task within a TMR protocol may be especially suitable for assessing both whole-body motor learning and a possible role for dreaming.

The objectives of the current study are threefold (see Fig. 1); (I) to test whether TMR administered during the SWS or REM sleep of a morning nap benefits complex procedural skill learning; (II) to investigate if dreaming about kinesthetic aspects of a VR task is related to improved task performance; (III) to assess whether TMR promotes direct and indirect dreaming about the VR task (task-dream reactivations). If the above tested relationships are supported, we will further test whether dreaming about the VR task mediates the relationship between TMR and improved performance.