Why do we dream? What sleep science has found so far

Every night, you close your eyes, the room goes dark, and your brain switches on a show nobody asked for. Scenes jump from one place to another, impossible things happen, and a friend can somehow have your mother’s face. Then you wake up, and most of it melts away.

In a 2010 review, Yuval Nir and Giulio Tononi called dreams a remarkable experiment that runs every night in every sleeping person. Cut off from the outside world, the brain builds a whole world of experience by itself. So why does it bother? Scientists don’t have one agreed answer yet, but they have learned a lot along the way.

What is a dream like?

When researchers collect dream reports, a clear picture comes out. In that review, Nir and Tononi describe dreams as highly visual, in full color, rich in shapes and full of movement. They are filled with the same things as waking life: people, faces, places, objects and animals. Dreams have sounds too, including speech and conversation. Touch, smell and taste show up more rarely.

One thing is usually missing. While we dream, we are not aware of where we are, lying in bed, or of what we are doing, which is sleeping. We take the events and characters as real, and we easily accept impossible things, like flying. The dream simply carries us along.

The night dreams were found

In 1953, Eugene Aserinsky and Nathaniel Kleitman described something odd in sleeping people. At regular times during the night, their eyes began moving quickly under their closed lids. These regular stretches came to be called rapid eye movement sleep, or REM sleep.

Early studies woke people up during REM sleep and asked what had been going through their minds. Nir and Tononi sum up those first results: 74 to 80% of awakenings from REM sleep brought vivid dream recall, compared with only 7 to 9% of awakenings from the rest of sleep, which is called non-REM sleep. For years, dreaming and REM sleep were treated as almost the same thing.

Dreams outside REM sleep, too

The picture turned out to be messier. Full dreams are sometimes reported after waking from the very first stretch of non-REM sleep, before any REM sleep has happened, and even after naps that contain only non-REM sleep.

In 2017, Francesca Siclari, Giulio Tononi and colleagues went looking for what dreaming looks like in the brain, whatever the stage of sleep. They recorded sleepers with 256 electrodes on the scalp and woke them many times through the night. In both REM and non-REM sleep, dreaming went together with a drop in slow electrical waves over a region near the back of the brain. They called it a posterior “hot zone.” Where faster activity showed up even matched what the dream held, such as faces, a sense of place, movement or speech.

Then they tried a little magic trick. In seven people in non-REM sleep, a computer watched the hot zone in real time and guessed, before each awakening, whether the person would say they had been dreaming. The guesses were right 87% of the time.

A peek inside the dreaming brain

In 1996, Pierre Maquet and colleagues used PET scans, which track blood flow, in seven people who stayed in steady REM sleep during the scan and remembered dreams on waking. Some regions were busier than usual, including the amygdala on both sides, a pair of small almond-shaped regions tied to emotion. Other regions were quieter, including a large area of the prefrontal cortex near the front of the head, which helps us plan and reflect.

Nir and Tononi suggest that the weak self-checking in dreams may be related to quieter regions like this one. Even awake, the prefrontal cortex quiets down when people are swept up in an intense sensory experience and become less self-aware. That may be part of why a dream rarely stops to ask, “Wait, is this real?”

So why do we dream?

Here is the honest answer: nobody knows for sure. In 2021, Erik Hoel wrote that scientists now understand much more about why we sleep, but no matching understanding of dreams has emerged. Many theories, he noted, treat dreams as a side effect. Still, a few ideas have shaped the debate.

Idea 1: the brain telling a story about its own noise

In 1977, J. Allan Hobson and R. W. McCarley proposed the activation-synthesis hypothesis. In dreaming sleep, a mechanism in the brainstem switches the forebrain on again and again, in a regular and automatic way. It also sends strong bursts of activity through circuits that handle eye movements and balance. The forebrain, they suggested, then builds the dream by matching those signals against what is stored in memory.

In this view, the plot of a dream is the brain trying to make sense of activity that started somewhere else. Ideas like this led some scientists to see dreaming as a by-product of REM sleep with no job of its own.

Idea 2: a safe place to practice danger

In 2000, Antti Revonsuo pushed back. Dreams, he argued, are not as random as that view suggests. They are organized and selective. Compared with waking life, some kinds of things show up in dreams far more often and others far less, and certain kinds of waking experience shape dreams strongly.

He proposed that dreaming evolved to simulate threats. Our ancestors lived short lives full of danger, he pointed out. A brain that rehearsed threatening events at night, over and over and in new combinations, might have been better at spotting and escaping danger by day. He pointed to studies of children’s dreams, recurring dreams and the dreams of hunter-gatherers as support. It remains a hypothesis that researchers still argue about.

Idea 3: helping new memories settle in

Sleep and memory have a long history together. We told part of that story in What happens to memories while you sleep. Do dreams play a part?

First, a surprise: dreams almost never replay a day exactly. In a 2003 study, Magdalena Fosse and colleagues asked 29 people to keep a diary of their days for two weeks and write down any dreams they remembered. Of 299 reports, 65% had something to do with recent waking life. But a full replay of an actual event showed up in no more than 1 to 2% of them. Dreams seem to borrow pieces of the day, not whole scenes.

The pieces can come from surprising places. In 2000, Robert Stickgold and colleagues had people play the falling-blocks computer game Tetris. As they drifted off to sleep, players reported images of the game popping into their minds. So did three people with amnesia caused by damage on both sides of the medial temporal lobe, the deep area that holds the hippocampus. They could not remember playing the game at all, yet the images still came. Some other players even saw images from older versions of the game they had played before.

In 2010, Erin Wamsley, Robert Stickgold and colleagues asked whether those dreamy pieces go along with learning. They trained 99 college students to find their way through a virtual maze and tested them again 5 hours later. In between, 50 of them took a nap and 49 stayed awake. The nappers improved more. And four nappers who mentioned the maze when asked about their dreams improved about ten times more than the other nappers. They had started out among the weaker players, but they also improved more than other weak starters. Thinking about the maze while awake brought no such boost.

None of the maze reports, from nappers or from people who stayed awake, was an exact replay. Two people heard the music from the task, and three described other maze-like places. The researchers do not claim the dreams did the learning. Their reading is that dreams may reflect memory work that the sleeping brain is doing anyway.

What happens to dreaming when the hippocampus itself is damaged? In 2020, Goffredina Spanò, Eleanor Maguire and colleagues woke four people with damage to the hippocampus on both sides, plus ten matched volunteers, at different points through the night. The patients were woken just as often but reported fewer dreams, and their dreams had less in them. One patient never reported a dream, and usually said, “I can’t picture it.” The team concluded that a healthy hippocampus may be needed for dreaming as we usually know it. It is the same region that was largely removed in patient H.M., the man who could no longer form new memories.

Idea 4: a little noise to keep the brain flexible

The newest idea borrows from artificial intelligence. When computer neural networks learn, they can fit their training examples too tightly and then stumble on anything new. This is called overfitting. Programmers often fix it by adding noise, feeding the network strange or scrambled inputs.

Erik Hoel’s 2021 “overfitted brain hypothesis” suggests that dreams do the same job for us. After a day of learning, the brain makes up odd, jumbled experiences at night, and that may help what we learned stretch to new situations. Hoel set out predictions that could test the idea in the lab. For now, it is an interesting proposal rather than a result.

Can scientists see what you dream?

Just a little. In 2013, Horikawa, Kamitani and colleagues scanned people with fMRI as they drifted off to sleep, woke them, and asked what they had just seen. They also recorded each person’s brain while looking at real pictures. Computer models trained on those waking pictures could then pick out the contents of the images people saw while falling asleep. The team suggested this could become a way to study dream contents with objective brain measurements.

In short

We dream in REM sleep and also in non-REM sleep, and a hot zone at the back of the brain seems to track when a dream is happening. While we dream, emotional regions are busy and a self-checking region at the front is turned down. Why we dream is still an open question. Dreams might be the brain’s story about its own noise, a rehearsal for danger, a glimpse of memories settling in, or a bit of helpful nonsense that keeps learning flexible. More than one of these may be true at once. Tonight, your brain will run the experiment again.

#dreams #sleep #REM sleep #memory #hippocampus #everyday brain

Sources

1. Dreaming and the brain: from phenomenology to neurophysiology. Nir Y, Tononi G. Trends in Cognitive Sciences 14, 88–100 (2010)
2. Regularly occurring periods of eye motility, and concomitant phenomena, during sleep. Aserinsky E, Kleitman N. Science 118, 273–274 (1953)
3. The neural correlates of dreaming. Siclari F, Baird B, Perogamvros L, Bernardi G, LaRocque JJ, Riedner B, Boly M, Postle BR, Tononi G. Nature Neuroscience 20, 872–878 (2017)
4. Functional neuroanatomy of human rapid-eye-movement sleep and dreaming. Maquet P, Péters JM, Aerts J, Delfiore G, Degueldre C, Luxen A, Franck G. Nature 383, 163–166 (1996)
5. The brain as a dream state generator: an activation-synthesis hypothesis of the dream process. Hobson JA, McCarley RW. American Journal of Psychiatry 134, 1335–1348 (1977)
6. The reinterpretation of dreams: an evolutionary hypothesis of the function of dreaming. Revonsuo A. Behavioral and Brain Sciences 23, 877–901 (2000)
7. Dreaming and episodic memory: a functional dissociation?. Fosse MJ, Fosse R, Hobson JA, Stickgold RJ. Journal of Cognitive Neuroscience 15, 1–9 (2003)
8. Replaying the game: hypnagogic images in normals and amnesics. Stickgold R, Malia A, Maguire D, Roddenberry D, O'Connor M. Science 290, 350–353 (2000)
9. Dreaming of a learning task is associated with enhanced sleep-dependent memory consolidation. Wamsley EJ, Tucker M, Payne JD, Benavides JA, Stickgold R. Current Biology 20, 850–855 (2010)
10. Dreaming with hippocampal damage. Spanò G, Pizzamiglio G, McCormick C, Clark IA, De Felice S, Miller TD, Edgin JO, Rosenthal CR, Maguire EA. eLife 9, e56211 (2020)
11. The overfitted brain: dreams evolved to assist generalization. Hoel E. Patterns 2, 100244 (2021)
12. Neural decoding of visual imagery during sleep. Horikawa T, Tamaki M, Miyawaki Y, Kamitani Y. Science 340, 639–642 (2013)