Sleeping isn't just about recharging your batteries. During sleep, memory consolidates. But how sleep changes memories remains unclear. To see more clearly, researchers from the University of Hong Kong conducted a study published in 2025 in Communications Biology. Using electroencephalograms (EEG) and in-depth neuronal analyses, they demonstrated that sleep, particularly paradoxical sleep (REM), does not just stabilize memories, it transforms them.
At the same time, a first study led by Jing Liu at the Hong Kong Polytechnic University, published the same year, had already observed this shift from detailed memories towards more general representations after sleep. By combining these results, the two studies reveal differentiated roles of slow-wave sleep and REM in memory: one preserves details, the other generalizes. An essential distinction for understanding the brain mechanisms of learning and forgetting.
A precise experiment to dissect the mechanisms of memory
The study, carried out at the University of Hong Kong and the Hong Kong Polytechnic University, involved 35 healthy young adults, with an average age of 22 years. All participated in an intensive learning test before sleeping.
Participants memorized 96 associations between verbs and images. The images belonged to four well-defined categories: animals, plants, electronic objects and means of transport. This structuring allowed researchers to distinguish two levels of memory. On the one hand we find item-specific memory (the exact memory of an image) and on the other hand categorical memory (the memory of the type of image).
During learning, then upon awakening, brain activity was recorded via EEG. In between, the participants slept a whole night under supervision. Some of the memories were reactivated during their deep slow-wave sleep, using the auditory targeting technique. Even if this aspect was not the heart of the analysis here.
What sets this study apart, according to researcher Jing Liu contacted by Live Science uses representational similarity analysis (RSA). In other words, a method of neuronal analysis that allows us to observe how information is stored and transformed in the brain. This method revealed subtle but measurable changes in the way memories are represented. The objective was not only to see what the participants remembered, but above all how their brain retained or modified information during the night. This precise point is central to distinguishing consolidation from memory transformation.
A shift from detail towards the concept observed in the brain
The results of the neuronal analyzes are clear. After a night's sleep, memories are transformed profoundly. Brain representations of specific elements are weakened. While more general representations remain stable, even reinforced.
Concretely, this means that a memory such as “a beagle” becomes, in the brain, a memory of “dog or animal”. The exact visual detail disappears, but the membership in a semantic category is preserved. This slip is measured using the RSA technique. It evaluates the similarity between the patterns of brain activation before and after sleep.
The first study showed that this transformation was correlated with the proportion of paradoxical sleep (REM) in the participant's sleep cycle. The higher the share of REM, the more “abstract” the memory becomes. And therefore the less faithful it remains to the initial details.
This phenomenon is not a random loss of information. This is an active process of memory restructuring. Even when participants still remembered the image, their brains no longer activated the same areas as during initial learning. This reconfiguration indicates an alternative, more synthetic representation of the information.
This suggests that the brain is not seeking to maintain the accuracy, but the cognitive utility of the memory. Generalizing allows you to relate new information to existing patterns. This promotes knowledge integration. This ability to extract the essence of an experience ultimately proves essential to the formation of knowledge. Thus, human memory appears not as a fixed archive. But it appears rather as a dynamic system, shaped by cognitive needs and nocturnal brain states.
Deep sleep and paradoxical sleep do not act in the same way
One of the major contributions of the recent study lies in the direct comparison of the effect of deep slow-wave sleep (SWS) to that of paradoxical sleep (REM) on memory. Contrary to the idea of a complementary action, the researchers here show opposite and differentiated effects.
Deep slow-wave sleep, characterized by slow brain waves (below 1.25 Hz), is associated with a stabilization of specific memories. On the other hand, REM promotes a semantic transformation. That is, a move towards a more generalized representation, as mentioned previously.
The researchers measured the ratio between REM and SWS in each participant. The higher this REM/SWS ratio was, the more the neuronal traces of item-specific memories disappeared, while the categorical representations intensified. The REM × SWS product, used in previous hypotheses to assess a synergy between the two, did not show a significant correlation here. This result calls into question the idea that the two phases of sleep always act together.
In addition, the intensity of theta (4–7 Hz) and beta (15–25 Hz) waves observed at the frontal level during REM proved to be a good predictor of this memory transformation. Strong activity in these frequency bands was associated with a loss of precision, but with better categorization.
Conversely, slow activity during deep sleep, particularly in the prefrontal cortex, was associated with a preservation of details.
A useful rather than exact memory for better daily functioning
These scientific results raise a fundamental question. What is the use of memory if it transforms the facts experienced? The answer, suggested by the authors, is that this transformation allows the brain to prioritize information, to link new elements to already existing structures and to facilitate decision-making.
By reducing specific details, the brain lightens the cognitive load. It thus creates more stable memories, more useful in daily life. Remembering that you saw a car in a scene is often enough to understand its meaning, even if you have forgotten its color. This process is therefore adaptive, not dysfunctional.
The implications are numerous. On an educational level, for example, understanding that REM promotes the integration of knowledge could guide the way in which we structure the revision phases. On a clinical level, this sheds light on the phenomena of false memories or memory distortions, particularly in people suffering from post-traumatic stress or sleep disorders.
Moreover, the memory thus transformed becomes more flexible. It allows generalization, creativity, even imagination. It is not a simple reproduction, but an active reconstruction. This plasticity remains essential to human intelligence.
Finally, this work reinforces the idea that sleep is not a passive phase of brain functioning. On the contrary, it is a key moment when the brain works, sorts, reorganizes. Good quality sleep, respecting the balance between REM and SWS, is therefore essential for optimal cognitive functioning.
This new understanding of nocturnal memory invites us to rethink the way in which we learn, we remember — and even what we believe to be true. And ultimately reminds us of the importance of quality sleep.
Source: Liu, J., Chen, D., Xia, T. et al. “Slow-wave sleep and REM sleep differentially contributes to memory representational transformation”. Common Biol 8, 1407 (2025).
With an unwavering passion for local news, Christopher leads our editorial team with integrity and dedication. With over 20 years’ experience, he is the backbone of Wouldsayso, ensuring that we stay true to our mission to inform.
