This study shows that falling asleep is an abrupt neural transition rather than a gradual drift: EEG recordings from thousands of people reveal a sudden change about 4.5 minutes before sleep onset. Researchers converted 47 EEG features into a mathematical space and traced trajectories that predict the moment of sleep like a ball rolling toward a fall. Using one night of data per person, the model forecasted later sleep timing with ~95% accuracy and located the tipping point within an average error of 49 seconds. The finding could help with insomnia diagnosis, drowsy-driving alerts, anaesthesia monitoring and brain-health assessment.
Brain Scans Find a Sharp 'Tipping Point' ~4.5 Minutes Before We Fall Asleep
This study shows that falling asleep is an abrupt neural transition rather than a gradual drift: EEG recordings from thousands of people reveal a sudden change about 4.5 minutes before sleep onset. Researchers converted 47 EEG features into a mathematical space and traced trajectories that predict the moment of sleep like a ball rolling toward a fall. Using one night of data per person, the model forecasted later sleep timing with ~95% accuracy and located the tipping point within an average error of 49 seconds. The finding could help with insomnia diagnosis, drowsy-driving alerts, anaesthesia monitoring and brain-health assessment.
03:44 AM, 11/13/2025Science
Brain scans reveal a rapid 'tipping point' minutes before sleep
New research from teams at Imperial College London and the University of Surrey shows that falling asleep is not a slow fade but an abrupt neural transition. Using electroencephalogram (EEG) recordings from thousands of volunteers, the researchers identified a pronounced change in brain electrical activity roughly 4.5 minutes before sleep onset.
"We found that falling asleep is a bifurcation, not a gradual process, with a clear tipping point that can be predicted in real time," said ICL neuroscientist Nir Grossman.
The team extracted 47 features from EEG signals and mapped them into an abstract mathematical space. Tracing an individual's trajectory through that space produced a pattern likened to a ball accelerating down a steepening slope until it 'tips' into sleep. The model, trained on a single night's recording for each person, predicted the timing of sleep on later nights with about 95% accuracy and located the tipping point with an average error of 49 seconds.
Why this matters
This finding deepens our understanding of how healthy sleep begins and has several practical implications. Potential applications include improved diagnosis and treatment of insomnia and excessive daytime sleepiness, development of in-vehicle alerts for drowsy driving, more precise monitoring of anaesthesia depth, and new markers for brain health.
The study was published in Nature Neuroscience. Grossman described the results to New Scientist reporter Grace Wade, emphasizing the ability to estimate, in real time and with high precision, how close an individual is to falling asleep.
Notes and caveats
While the results are promising, they are based on patterns found in EEG features and mathematical modeling. Wider clinical validation and testing across diverse populations and real-world conditions will be needed before the approach can be deployed in medical devices or safety systems.
Source: Study by researchers at Imperial College London and the University of Surrey, published in Nature Neuroscience.