The next step in understanding the function of dreams

The function and meaning of dreams continue to fascinate and allude humankind. Despite the near universal intrigue, the pendulum of public and scientific opinion on dream research seems to have regularly oscillated over the past 150 years between fringe parapsychology and mainstream scholarly pursuit. One contributing factor is that dream research has always and continues to primarily rely on retrospective reports of mentation from a different state of consciousness. As such, each exciting breakthrough in our understanding of dreams is typically preceded by either an advancement in technology, such as the discovery of dream-rich rapid eye movement (REM) sleep [1], or a new creative study design—often at the expense of investigators as the funding support for dream research has frequently oscillated with that of public opinion.

Fortunately, for contemporary sleep and dream researchers, a number of prominent scientists over the years have aided in chipping away at the veil of the dreaming brain [2–11], and the work of Picard-Deland et al.[12] recently published in SLEEP marks another step forward in our understanding.

The scholarly pursuit of the mechanisms behind and functions of dreaming has already produced substantive results. For instance, previous work has revealed that dreaming occurs both during REM and non-REM (NREM) sleep [13–16] and that a vast majority of dream content (over 80%) reflects memories for past events or anticipated future events [17, 18]. Furthermore, rather than replaying past events in their entirety, dreams typically incorporate elements of past memories into unique simulations [18–20]. While some still posit that dreams merely reflect the byproduct of neural activity related to other processes [21], the consistent presence of memory content in dreams and the relationship that has been identified between dreams and learning [13, 22–24] have led many contemporary theorists to suggest that at least one potential function of dreams is to aid in the consolidation and integration of memories into long-term storage and to facilitate novel connections and associations (e.g. the NEXTUP model proposed by Stickgold and Zadra [25]). In this way, dreams may aid in developing the framework of our autobiographical memories [26], helping to shape who we are and how we view ourselves and the world around us. Moreover, the incorporation of past events in various novel scenarios may also act as preparation for the next time we are faced with similar challenges [18].

While this view of the function of dreams is tantalizing, more evidence is needed to support it. In their recent report, Picard-Deland et al. utilize a serial-awakening protocol to investigate how the memory content of dreams evolve over the course of a night. Early versions of this protocol have already proven fruitful in helping to peer behind the curtain of dreaming. For instance, previous studies suggest that memory type (episodic vs. semantic) and source (recent vs. remote) tend to vary across the night [18, 27, 28], and that early NREM dreams tend to include more episodic memory traces compared to later REM sleep [29, 30]. These findings, however, have been sewn together piecemeal from a variety of studies with different objectives and a range of sample sizes. As such, one of the primary goals of Picard-Deland et al. was to conduct a well-powered, systematic study in which the memory type and temporal period (recent past, distant past, or anticipated future) of dreams were assessed from each stage of sleep (N1, N2, N3, and REM) across three periods of the night (early, mid, and late) all within a single, within-subject study design.

The first notable finding was that the structured serial-awakening protocol employed by the authors was quite successful in their group of 20 participants. An average of 11.4 awakenings generated 8.2 dream reports across the night. Furthermore, despite the change in the prevalence of sleep stages across the night, the authors successfully collected multiple dream reports from each sleep stage across the three periods of the night (min = 6, max = 18). With regard to dream content, over 87% of the reports contained memory sources, replicating previous research [17, 18]. As such, the authors demonstrated that this type of systematic dream assessment can successfully generate a high proportion of dream reports with frequent memory content from each stage of sleep across the night.

From this successful dream collection, the authors go on to describe a host of replicated and novel findings regarding the evolution of dream content across a night of sleep. For instance, the likelihood of recalling a dream varied by sleep stage (e.g. dreams were more frequently recalled in REM compared to N3 sleep), but not by the period of the night (i.e. dream reports were equally likely in early, mid, and late sleep). With regard to the memory content of dreams, the authors boiled down their rich dataset to three key findings: (1) memory sources were most frequently reported during N1 and REM sleep, (2) they replicated the finding that independent of sleep stage, memory sources identified in dreams become increasingly remote across the night, and (3) memory sources of dreams tend to adhere to similar themes through the different sleep stages and across the night. The authors also failed to replicate the finding that REM dreams tend to incorporate more semantic memories compared to NREM dreams.

Picard-Deland et al. conclude by describing how these results may align with contemporary theories of sleep’s role in memory processing in general, and while speculative, the parallels highlighted generate intriguing questions for future research. For instance, the increase in memory sources during N1 and REM dreams may reflect a unique milieu of neural conditions that facilitate the development of novel associations between past and present events, supporting past research suggesting that these stages are important for insight and creativity [31–33]. Additionally, the increased remoteness of memory content across the night—regardless of sleep stage—may indicate that recent events are organized during early sleep and then integrated with past memories and schemas during the latter half of the night. The authors suggest potential roles for both circadian effects and synaptic downscaling in this process (e.g. after weaker recent memories are downscaled in the early portion of the night, the remaining stronger memory traces may be integrated into long-term memory networks). Finally, the consistent themes in dream content across the night suggest a complementary or sequential relationship between sleep stages in the processing of memories that evolves across the night.

Dream research is always difficult given its inherent challenges, such as the disruption of sleep and reliance on self-reported narratives. It is entirely possible that probing dreams early in the night may modulate the content of subsequent dream reports later in the night [2]. As demonstrated by Picard-Deland et al., however, taking a standardized approach to investigating dreams can result in a deeply rich dataset and it is only through replication and follow-up studies that we will be able to distinguish these potentially confounding effects from the true function of dreaming. The methods used here should be built upon and combined with measures of behavior as well as current and future advances in neural and physiological measurement. Dream research will not only tell us about the function of dreams. Ultimately, it will help us to understand the purpose of sleep in general.

Funding

TJC is currently funded by the National Institute of Mental Health (NIMH K23MH127464) and by the Sleep Research Society Foundation Career Development Award.

Disclosure Statement

The author has no financial or non-financial disclosures related to this work.