Neuroscientists discover brain rhythms slow down during sleep in two distinct ways

A new study published in The Journal of Neuroscience has uncovered how the human brain’s internal timing mechanisms shift during sleep. Researchers analyzed electrical recordings from 106 patients and found that two different kinds of brain timescales—based on broad and gamma-frequency brain signals—both became longer during sleep. However, these timescales followed opposite organizational patterns across the cortex. While broadband timescales increased along a known sensory-to-association hierarchy, gamma timescales decreased. The study suggests that the brain’s natural rhythms slow during sleep, but in distinct ways depending on the type of signal and brain region.

The researchers were interested in a fundamental question: how does the timing of brain activity—its so-called “timescales”—change during sleep? In neuroscience, timescales refer to how long neural signals remain autocorrelated, or how long a signal “remembers” its past. These timescales vary across the brain and are thought to be linked to how information is processed and integrated.

Sensory areas, for example, usually show shorter timescales, suitable for quickly responding to external input, while association areas, involved in more complex processing, typically display longer ones. Previous studies had shown that timescales lengthen during sleep, but the nature of these changes and how they are organized across the brain remained unclear.

To investigate this, the research team analyzed intracranial electroencephalography (iEEG) recordings from 106 individuals with epilepsy, who had electrodes implanted for clinical purposes. The data came from a publicly available database and included brain activity recorded during wakefulness, deep non-REM sleep (also called NREM3), and REM sleep. The participants ranged in age from 13 to 62 years, with 48 women in the sample. Researchers focused on 1-minute recordings during each state and examined activity from over 1,000 electrode sites distributed across the cortex.

The team extracted two distinct kinds of timescales from the data. One came from the broadband iEEG signal, which captures a wide range of brain activity from slow to fast frequencies (0.5–80 Hz). The other was derived from gamma-band power (40–80 Hz), a faster rhythm associated with sensory processing and cognitive functions. They measured the timescales using the autocorrelation function, which shows how long a signal is similar to itself over time, and fitted these with an exponential decay model. They also assessed spatial correlations between brain regions by calculating how synchronized signals were across different electrode pairs.

The researchers found that both broadband and gamma timescales became longer during sleep compared to wakefulness. In deep non-REM sleep, broadband timescales increased on average by about 105 milliseconds across brain areas, while gamma timescales increased by around 31 milliseconds. During REM sleep, both also showed slight increases, but the effects were much smaller. Importantly, these changes followed different patterns depending on the signal type.

Broadband timescales followed a known hierarchy in the brain. In both wakefulness and deep sleep, they increased from sensory areas like the visual cortex to higher-order regions such as the medial temporal lobe and orbitofrontal cortex. This pattern matched previous findings and supports the idea that association areas naturally operate on slower timescales, integrating information over longer periods.

Gamma timescales, however, followed the opposite trend. During sleep, they became longer in sensory regions like the visual and auditory cortices, and shorter in associative areas. This reversal of the typical hierarchy suggests that gamma-based timescales may serve different functions during sleep, possibly related to the reprocessing of sensory input or memory consolidation.

To understand what might be causing the increase in timescales during sleep, the researchers focused on slow waves—large, slow oscillations that are a hallmark of deep sleep. Using a separate 10-minute recording of NREM3 sleep, they identified over 85,000 individual slow wave events. They found that both types of timescales increased during these events, especially broadband ones.

Broadband timescales rose by 171% around the trough of slow waves, while gamma timescales increased by only 51%. Moreover, brain regions with higher overall densities of slow waves showed longer broadband timescales, suggesting a direct link between slow-wave activity and temporal dynamics.

The study also looked at how brain regions interact across space during different states. By measuring spatial correlations—how much two brain regions fluctuate together—they found that long-range synchronization was strongest during deep sleep, particularly for broadband signals. Interestingly, while broadband spatial correlations were highest over long distances during wakefulness, gamma spatial correlations peaked during NREM sleep but only at short distances. These findings indicate that the brain’s ability to integrate information across space and time varies depending on the sleep state and type of neural activity.

One of the key takeaways from this work is that the brain’s intrinsic timing mechanisms are more complex than previously thought. Rather than a single timescale that lengthens during sleep, the study reveals the presence of multiple timescales that vary by frequency range and cortical location. This layered organization suggests that different rhythms may serve distinct functions, even within the same brain region.

The researchers also highlight that while sleep is often thought of as a disconnected state, their findings complicate this view. Deep sleep showed increased spatial correlations and longer timescales in several areas, indicating enhanced local coordination rather than global shutdown. However, REM sleep showed reduced spatial correlations, possibly reflecting a more fragmented or internally focused mode of activity.

Like all research, this study has limitations. The sample consisted of individuals with epilepsy, which may affect brain dynamics, though the researchers used resting-state recordings outside of seizure activity. In addition, while intracranial recordings offer high spatial and temporal precision, electrode placement is limited by clinical needs, meaning some brain areas are underrepresented. Future studies could explore how timescales change over longer periods or during specific cognitive tasks, as well as whether similar patterns are seen in non-clinical populations.

The study, “Sleep Modulates Neural Timescales and Spatiotemporal Integration in the Human Cortex,” was authored by Riccardo Cusinato, Andrea Seiler, Kaspar Schindler, and Athina Tzovara.