Even as we drift into what feels like unconscious rest, our brains might not be entirely asleep. A groundbreaking study titled DREAM: The Dream EEG and Mentation Database, published in Nature Communications, reveals that certain parts of the brain can remain alert even during the deepest stages of sleep. This research challenges the long-held belief that dreaming is confined strictly to REM sleep. The DREAM project, a large-scale international collaboration, has consolidated decades of dream research into a unified, open-access platform.
Coordinated by Monash University in Australia, the DREAM project analyzed data from 2,643 awakenings involving 505 volunteers, drawing insights from 20 separate studies conducted across 13 countries. This extensive analysis provides a comprehensive view of how our brains operate during various sleep stages, highlighting the complexities of sleep science and dream experiences.
For many years, REM sleep has been considered the exclusive realm of dreaming, characterized by vivid imagery, muscle paralysis, and rapid eye movements. However, the findings from the DREAM study challenge this long-standing assumption. Among the 2,643 awakenings analyzed, participants reported dream experiences in approximately 85% of REM cases, but intriguingly, also in about 40% to 60% of NREM (non-REM) awakenings. Even during the slow-wave sleep phase, which is known as the deepest and most restorative phase of sleep, some individuals described experiencing emotional thoughts or fragmented sensations.
The researchers utilized EEG and MEG technologies to examine the last 30 seconds of brain activity prior to each awakening. Their findings were unexpected: in instances of NREM dreaming, brain waves displayed faster and smaller oscillations, closely resembling the patterns seen during quiet wakefulness. This suggests that the brain — or at least certain regions of it — may temporarily transition into an “awake-like” state, despite the body remaining in a state of deep sleep. These episodes may represent brief bursts of conscious processing, supporting the notion that consciousness during sleep exists on a spectrum rather than as a simple switch.
The strength of the DREAM project is not only in its expansive data set but also in its innovative methodology. Each awakening recorded in the study was classified into one of three categories: a clear dream (experience), an experience without recall (often referred to as a white dream), or no experience at all. This classification allowed researchers to train artificial intelligence models to predict whether a person had been dreaming based solely on brain activity.
Initial findings showed that simpler brain-wave features provided moderate accuracy in detecting dreams during deep NREM sleep, while more complex models significantly improved accuracy during REM sleep. As highlighted by Discover Magazine, this represents one of the first large-scale efforts to develop non-invasive tools for dream detection. With further advancements, such technology could potentially monitor consciousness during sleep or even serve as an early marker for cognitive disorders.
In fact, related research indicates that difficulties in reaching REM sleep could serve as an early warning sign for Alzheimer’s disease, emphasizing the broader medical relevance of sleep-stage mapping and dream monitoring. As co-author Giulio Bernardi noted in a press release, the collaborative work presented in this article, spearheaded by Monash University, brought together the expertise of 53 authors from 37 institutions across 13 countries, marking a pivotal advancement in the scientific exploration of human consciousness.
