New research into sleeping and wakefulness radically undermines a century of thought, as the two states are shown not to be as distinct as scientists previously believed.
While it’s a problem that over a third of Americans may not be getting at least the recommended seven hours of sleep per night, it turns out that some of us may be getting milliseconds of sleep without even noticing. Now, thanks to a new computer analysis, “flickers” of sleep have been detected in portions of otherwise wakeful brains, and vice versa.
An Unexpected Result
When Ph.D. candidates Alan Schneider of Washington University and David Parks of the University of Santa Cruz first presented the results of their work, it surprised Professor David Haussler, who ran the lab where Schneider was working. Professor Kenneth Hegen, director of the lab that Parks was operating out of, told them to go back again; they must have missed something. “It was an exciting process as a scientist to have my students tear down these towers brick by brick, and for me to have to be okay with that,” Hegen commented.
Overturning paradigms is never easy. Due to the extraordinary nature of what Schneider and Parks were uncovering, Hegen and Haussler demanded extraordinary evidence—and that’s just what they received. Hegen felt challenged to reexamine whether his long-held beliefs were based on actual evidence.
Data for the study was produced by monitoring the electrical signals in the brains of mice over many months, utilizing a small helmet that captured the activity of four to eight regions in their brains. Ten specific regions were studied across all of the mice. The scientists fed the data to a predictive learning algorithm to uncover what may be hidden from human eyes.
The data set was massive, at millions of gigabytes, requiring an advanced computing network housed at US San Diego to help process it. Over four years, the pair trained a neural network to discern whether a human brain was awake or asleep, comparing its answers to those provided by three human sleep experts.
MACHINE LEARNING HARNESSED TO UNDERSTAND SLEEP
Once the machine learning algorithm was trained, the researchers attempted to work backward and uncover exactly how it was learning. Parsing that out can be notoriously challenging for convolutional neural networks like the one used in this study. The researchers began feeding it smaller and smaller pieces of information in an attempt to discern how it was making decisions. To their surprise, they found it could determine wakefulness states in a brain-based on just milliseconds of information. This was much too short a time period for the multiple second long slow waves previously believed to govern sleep to provide any meaningful pattern.
The fast, high-frequency bursts even showed micro periods of sleep or wakefulness limited to just single portions of the brain, while the rest remained in a different state. Even while asleep, the mice’s brains would exhibit flickers of one form of sleep, such as REM, while the best of the brain reads as in NREM sleep.
Every possible combination of states was detected in the data, further undermining previous ideas that support a strict order of awakening from NREM sleep to REM sleep. Arguably, one of the most startling elements of the study was to see these detections reflected in visual observation of the subjects. While mice were observed to be awake and asleep, a flicker of the opposite state somewhere in the brain resulted in a noticeable change in behavior for the subject. If the mouse were asleep and quickly flashed as awake, it would coincide with it jerking in its sleep. Conversely, awake mice were seen to briefly stop and “zone out” when their brain flickered asleep for a moment.
The researchers’ interpretation of what may be occurring offers challenges to our existing understanding of sleep. Where it was previously believed that slow waves constituted sleep, they now hypothesize that sleep actually constitutes fast waves. The slow waves, in contrast, may only function to coordinate these fast waves when the brain is predominantly in one state or the other, which now appears to be more permeable than had been thought.
Our Future Understanding of Sleep
Where this research goes in the future is toward a much finer granularity of understanding how sleep works. Sleep issues are central to many neurological problems and diseases, and sleep is essential for neurological development in the first place. With Americans facing increasing cases of neurological diseases like Alzheimer’s, a disease that is known to include sleep issues as one of its components, grasping the intricacies of sleep becomes ever more crucial.
The new study, “A nonoscillatory, millisecond-scale embedding of brain state provides insight into behavior,” appears in the journal Nature Neuroscience.
Ryan Whalen is a writer based in New York. He has served in the Army National Guard and holds a BA in History and a Master of Library and Information Science with a certificate in Data Science. He is currently finishing an MA in Public History and working with the Harbor Defense Museum at Fort Hamilton, Brooklyn.