When we drift to sleep, there may be more to what our brain is doing than simply isolating us from the outside world, with evidence uncovered in a new scientific review suggesting that some mental processing, learning and even un-learning can occur as we lose wakefulness.

Dr Thomas Andrillon of Monash University in Melbourne has uncovered these findings in multiple past studies and a resultant review to be published in Current Opinion in Physiology1 which challenges conventional scientific wisdom that the brain simply turns off during sleep .

In the review, Andrillon and his team examined over 150 papers spanning more than seven decades and uncovered what researchers had learned about the brain’s activity in the transition period between when we shut our eyes and when we fall into truly deep sleep.

The research not only uncovers more about what our brains do as we fall asleep but could also pave the way for improved future treatment options for sleep disorders such as insomnia.

The brain’s gate, partially open

Andrillon told Lab Down Under that the reviewed studies tested the limits of what is called the ‘thalamic gating hypothesis’ — a prevailing neuroscience model that says that one part of the brain, the thalamus, acts to block the rest of the brain from processing information received from the outside world by sensory organs such as our eyes, ears and nose as we sleep.

The relationship between the thalamus and the cortex — the “cauliflower-like structure” that makes up the main part of the brain — changes during the day.

“When you’re awake, the thalamus is a relay of information. One function is that it passes information to the cortex and the cortex processes it. During sleep, that dynamic will change. This time, the thalamus will cease to relay information and will instead isolate the cortex from the outside world and prevent it from doing what it’s doing while we’re awake,” Andrillon said.

The hypothesised gating occurs because of what are known as thalamo-cortical oscillations. Through these oscillations, the thalamus is proposed to block the cortex’s ability to process external sensory information. The oscillations themselves occur on a continuous basis during deep sleep and can be triggered by external stimuli in light sleep.

Figure 1: Thalamus (red) and its position relative to other brain regions. Picture by Colder B (2015) used under the Creative Commons Attribution 4.0 International licence.

While the gating hypothesis was a good model on its own, how the brain actually worked was never simple, Andrillon said. One main reason was the fact that in larger mammals such as humans, the transition to sleep was not instantaneous like it was in smaller mammals such as rats and mice.

“Human brains take longer to go into deep sleep, so you have a transition period where these thalamo-cortical oscillations are building up. It’s not like the gate will shut instantaneously, but you could have a period during which the gate could be partially open. That’s where we were interested in.”

Traversing the brain’s archipelago

The research revealed that the brain processed certain pieces of information more easily than others during this transition period. For instance, whether a sound could be processed by the sleeping brain depended on its familiarity, relevance or contents.

More easily recognised sounds could span from those known by an individual person such as their name to those more generally recognised such as babies crying, Andrillon said.

“It could be the familiarity of a voice. It could be the emotionality of a voice. All these cues that make a sound potentially interesting seem to be processed more easily during sleep.”

Simple sounds seemed to be processed quickly by the sleeping brain, he added, while more complex signals such as groups of words or abstract ideas were more easily shut out.

“You could see it as an archipelago where you have the information going from one brain island to the other in a consistently more complex way. In the case of speech, you start with the very simple features of the sound and then you deal with a group of words and sentences and so on,” he said.

“It seems like when we sleep, this information can reach the first brain island but has a hard time jumping to another. Little by little you lose information, and the deeper you go into sleep, the less of this propagation of information you have. So in light sleep, you can do a bit but the deeper you go, the more you will be constrained to the very, very basic processing of sensory information.”

The stupidity of the sleeping brain

The fact that the sleeping brain could process external sounds suggested that it could also conduct basic learning at the same time as long as it was something simple and implicit such as the association between two sounds, Andrillon said.

Actually getting a sleeping person to learn an association between a sound and something more complex such as a concept or definition was more difficult, although some researchers have accomplished this.

“Even learning the meaning of a new word is still a fairly basic association because you just have to map a definition with a sound. If you have to learn a new grammatical rule or a historical fact, that requires an even more complex form of learning which has not been evidenced in sleep yet,” Andrillon told Lab Down Under.

Other research looked at how the brain could learn through odours, with one study building associations in the sleeper’s brain between the smell of cigarettes and other unpleasant odours, Andrillon said.

“The amazing part of this study is that doing the same thing while awake doesn’t work because the waking brain is not so stupid. When you have two odours at the same time, you don’t necessarily associate them. As a matter of fact, you have a lot of people who go outside to smoke. They may smoke near rubbish and in the street and can still enjoy their cigarettes.

“So when we’re awake, we can make a disassociation between two things occurring at the same time. We are smart enough to understand that they are not necessarily related to each other. It seems the sleeping brain is a bit stupider and if you present two pieces of information at the same time, they will tend to be associated with each other.”

While these studies were still in a proof-of-concept phase, they showed potential therapeutic benefits in the future where the brain could be tweaked during sleep and produce results on patients that could not be achieved during wakefulness, Andrillon said.

Forgetting something by listening to it

One of the more unexpected results from the review was that the sleeping brain could unlearn, or actively forget information that had been learned previously.

In one of the studies reviewed, sounds were played over and over to test subjects as they transitioned into deep sleep. Over time, these sounds became familiar and became what are called ‘auditory objects’ in the brain — objects that the brain is familiar with and that have an identity of their own.

While the brain could conduct this kind of learning in light sleep, the reverse occurred when sounds were played in deep sleep with the brain forgetting or erasing sounds that had been previously learned, Andrillon said.

“That was quite surprising to us because it was the first time we had been confronted with this sort of negative learning. Even conceptually, we had a hard time imagining how you could learn something negatively or how you could forget something by listening to it.”

However while these results were strange, they fit into the standard model of why we sleep, with certain theories suggesting that we consolidate memories in light sleep and then erase useless information during deep sleep, Andrillon said.

“The two things put together make sure that important information is reinforced after a night of sleep while making your brain ready for new learning by cleaning what’s useless. We weren’t really expecting those results in our study but they were really going in that direction. We evidenced parts of sleep where learning was possible and parts where forgetting was actually happening.”

Paradoxical sleep

The studies incorporated in the review also revealed the differences between light sleep and what is called rapid eye movement (REM) sleep that occurs while people are typically dreaming.

Andrillon said that while information could be processed in REM sleep — which the French call ‘paradoxical sleep’ — thalamic gating did not occur as it did in light sleep.

In light sleep, or non-REM sleep, a piece of information was initially processed well with this ability decreasing over time as the brain fell into deeper sleep. In REM sleep, the opposite happened with the brain missing the initial bit of information and then processing later portions, the research showed.

“The reason why we think it takes much more time is because there is always something present in the brain, which is usually what people are dreaming about. People are already conscious, already processing information, and then another piece of information comes along. To switch to this new piece of information, they need a bit of time,” Andrillon said.

This was a bit like the sensation of dreaming when the sound of an alarm clock pierces the dream, Andrillon noted, as there is a delay between when the dreamer hears the alarm and when they realise that it is actually time to wake up.

“What we found was really close to that picture. In REM sleep, things take time, not because you don’t have access to the information, but more because you’re already busy processing something else and you need time to switch your attention back to the external world.”

Unlike thalamic gating in light sleep, the exact mechanism by which information is processed or blocked is still not known for REM sleep. Current theories include the dream hypothesis above, changes in chemistry that modulate brain activity, or the work of other brain oscillations linked to specific layers in the cortex which relay or block sensory information.

Tackling sleep disorders

Andrillon told Lab Down Under that the review opens up many potential questions and possible research areas in the realm of sleep. One obvious area to explore is why humans need to switch off in deep sleep at all and why they could not just remain in light sleep while still processing signals from the outside world.

Another question involves looking at what the consequences are of our brains remaining in light sleep and connected to external information, and how this differs from any benefits of deep sleep.

Andrillon also expressed an interest in studying the differences between individuals and their access to light and deep sleep, saying that this could lead to further therapeutic treatment for sleep disorders.

“It will be very interesting to understand how light, deep and REM sleep can be explained in terms of brain activity and look at whether we can use any knowledge gained on, for example, how the brain switches off from external perturbators to help people who are too sensitive to these perturbators sleep better,” he said.

“Sleep disorders are usually taken lightly either by the health system or people themselves because insomnia, for example, hasn’t killed anyone. However, it still has a dramatic impact on life quality. There are sometimes no satisfying responses to insomnia or people reporting bad sleep quality, so it will be very interesting to see if we can do something about it.”


1 Andrillon T, Kouider S. The vigilant sleeper: neural mechanisms of sensory (de)coupling during sleep. Current Opinion in Physiology, Volume 15, June 2020, Pages 47-59.

Featured image: Arrows Brain Alternatives by Gerd Altmann used under the Simplified Pixabay Licence.