The story you wake with: what dreams are, and who writes them

Your night has four or five dreaming windows

Sleep runs in cycles of about an hour and a half; in a Basel sleep laboratory the median cycle, across more than 6,000 recorded, lasted 96 minutes. Each cycle sinks toward deep sleep and climbs back to REM sleep, when the eyes move under closed lids and the body lies still. The REM stretches, in gold, grow longer through the night and are longest in its second half.

From 23:00 to 07:00 that is about five cycles. The longest dreaming window comes late in the night.

Those windows are where dreams were first caught. In 1953 Eugene Aserinsky and Nathaniel Kleitman, at the University of Chicago, woke sleepers while their eyes were moving: 20 times out of 27 the sleeper described a detailed dream, usually full of images. Woken while the eyes were still, 19 times out of 23 they recalled nothing.

The line is a schematic night drawn from the usual pattern of cycles, not a recording. Real nights vary from person to person and from night to night.

A dream, dealt at random

Here is a dream for that longest window: six fragments, drawn at random by this page. They come from the themes people most often say they have dreamed. In 2003, 1,181 Canadian students ticked which of 55 themes they had ever dreamed: being chased came first, at 81.5%; falling was ticked by 73.8% and school by 67.1%. Watch the fragments flare and fade. Nothing connects them yet.

A demonstration, not an experiment. The draw happens on your device and nothing is recorded.

You wake up. What was it about?

The alarm goes off. Someone asks what you dreamed. Pick the version that sounds right.

For most of history, a dream was a message

In Genesis, Pharaoh dreams of seven fat cows swallowed by seven thin ones. Joseph reads seven years of plenty and seven of famine, and is careful to say whose message it is: “It is not in me: God shall give Pharaoh an answer of peace.”

In the second century AD, Artemidorus of Daldis wrote the Oneirocritica, a manual for reading dreams. He says he spent many years with the fortune-tellers of the marketplace, in Greece, Asia, Italy and the big islands, listening to old dreams and what had followed them. Under your fragments are his readings, where he gave one, in the words of a 1644 English translation.

Sigmund Freud moved the sender inside the dreamer. In The Interpretation of Dreams, dated 1900, a dream is the fulfilment of a wish, and by the 1909 edition he called its interpretation the via regia, the royal road, to the unconscious. The sender changed. The idea that a dream is a coded message did not.

Then the sleep laboratory found an engine

In 1977 the psychiatrists J. Allan Hobson and Robert McCarley proposed that the dream starts in the brainstem, not in a hidden wish. During REM sleep, circuits there fire in intense, sporadic bursts, and the forebrain, switched on automatically, “synthesizes the dream by comparing information generated in specific brain stem circuits with information stored in memory.” They called it activation-synthesis.

It is the bluntest answer to the old question. The brain keeps itself running at night; the images are by-products, and the plot is whatever the waking parts of the mind can make of them.

Others proposed a purpose for it. In 2000 the Finnish psychologist Antti Revonsuo proposed that dreams rehearse escape from threats, which would explain all the chasing. In 2021 the neuroscientist Erik Hoel proposed that their strangeness keeps a learning brain from fitting itself too closely to the day, the way engineers add noise when they train neural networks. Both remain hypotheses.

But the fragments come from your life

If dreams were pure noise, their contents would be random. They are not. The psychologist G. William Domhoff, who studies long series of dreams kept over years (the DreamBank archive holds more than 20,000), finds that the people, conversations and activities that recur in them are “continuous with their waking interests and emotional concerns”. He adds that much dream content is still not understood.

In 2000 Robert Stickgold's team at Harvard Medical School had volunteers play the falling-block game Tetris. As they fell asleep, they reported images of the game. So did three patients with amnesia from damage deep in both temporal lobes, who could not remember having played at all: the game appeared as they fell asleep even though they could not consciously recall playing it.

Ten years later Erin Wamsley and Stickgold trained 99 people on a virtual maze. Of the 50 who then napped, four said their dreams had something to do with the maze. Those four improved about ten times more on the retest than the 46 who napped without dreaming of it. Thinking about the maze while awake gave no such boost. The authors are careful: the dream may be a sign that memory was being worked on, not the cause.

What the cycles may be for: priming and practice

Some researchers put the engine first and give it a purpose. In 2009 Hobson argued that REM sleep may be a “protoconscious” state: the brain running a virtual-reality model of the world that helps build and maintain waking consciousness. On that view the night is the brain priming itself, and the dream is what the priming looks like from inside.

The night also works on memory. In 1994 Matthew Wilson and Bruce McNaughton recorded cells in the hippocampus of three rats: cells that had fired together while an animal was in a particular place were more likely to fire together again in the deep sleep that followed. In a 2010 review, Susanne Diekelmann and Jan Born propose a division of labour: deep, slow-wave sleep may reactivate memories that depend on the hippocampus and redistribute them to the cortex, while REM sleep may help fix them there.

Not everyone agrees that sleep adds strength. Giulio Tononi and Chiara Cirelli argue that waking strengthens connections all over the brain and sleep turns them down overall, with the weaker connections losing most. Either way, the night works on what the day laid down. Whether the dream itself does any of that work, or is only how the work looks from inside, is still open.

And the dream happens while you sleep, not only when you tell it

Dreams are not confined to REM sleep. In 2017 Francesca Siclari, Giulio Tononi and colleagues recorded sleepers with 256 electrodes and woke them through the night, in both kinds of sleep. Dreaming went with quieter slow waves in a patch at the back of the brain, a “hot zone”. Watching that patch live during non-REM sleep, they predicted whether a sleeper would report a dream 87% of the time.

In 2013 Yukiyasu Kamitani's team at ATR laboratories in Kyoto trained a computer on brain scans of people looking at pictures, then applied it to scans recorded as sleepers drifted off. Its guesses about what kind of thing they were seeing matched the reports they gave when woken more often than chance.

In 2021 four laboratories went further and asked dreamers questions. A man in a lucid dream signalled that he knew he was dreaming with three left-right eye movements. He heard “8 minus 6” and within three seconds answered with two. Correct answers came in 18.4% of 158 attempts, and most questions went unanswered. A dreamer in France, asked questions in the same way, reported afterwards being at a party in the dream and hearing the experimenter's voice “as if you were a God.”

The night a dream seemed to come true

On 21 October 1966 a colliery spoil tip slid down the mountainside above Aberfan, in south Wales, and buried Pantglas Junior School. It killed 144 people, 116 of them children.

Afterwards, people came forward to say they had dreamed it. A psychiatrist, John Barker, and the Evening Standard journalist Peter Fairley asked the public for premonitions of the disaster. Barker knew the replies proved little, because they were collected after the fact, so the two men set up a Premonitions Bureau to log warnings before events. For 18 months it catalogued hundreds of them. Over 90% matched nothing, and none was used to prevent a disaster.

The warnings that might have saved the school were not dreams. They were letters. In 1963 the deputy borough engineer had written letters headed “Danger from coal slurry being tipped at the rear of the Pantglas Schools”. The National Coal Board went on tipping.

Tonight, about eight billion people will sleep

In 1989 the statisticians Persi Diaconis and Frederick Mosteller named the law of truly large numbers: “With a large enough sample, any outrageous thing is likely to happen.” Try it on dreams. Take any disaster in the news, and the week of nights before it. Set how often people wake remembering a dream, and how rare a dream close enough to the event to feel like a warning would be.

The arithmetic says only that such dreams are expected, whether or not anyone can see the future. The test of foresight is a warning recorded before the event. The Premonitions Bureau logged hundreds: a few seemed to match later disasters, most matched nothing, and none prevented one. In the laboratory, three preregistered attempts in 2012 to repeat a celebrated precognition experiment, run with people awake, found no effect.

Both settings are assumptions, not measurements; the population is the UN estimate for 2025. In a 2021 survey of 15,854 adults in 16 countries, 54% said they remember a dream at least once a week.

Now look again at your dream

Here are the six fragments the page dealt you, at random:

And here are the three versions you were offered. The page built each one from the same random draw, by shuffling the fragments and joining them with words such as so, because and until.

There was no story in the draw, only random fragments and a few joining words supplied by the page. Yet each version read like a dream someone might have had, and one of them probably felt right. Finding sense in fragments, even fragments joined at random, is something minds do quickly and well.

A link that deals you this same dream again

A storyteller that never stops

The neuroscientist Michael Gazzaniga watched the same skill in people whose two brain hemispheres had been surgically separated. One patient's speaking left hemisphere was shown a chicken claw; the silent right hemisphere was shown a snow scene. Asked to pick matching pictures, the right hand chose a chicken and the left hand a shovel.

Asked why, the left hemisphere, which had never seen the snow, answered at once: “Oh the chicken claw goes with the chicken and you need a shovel to clean out the chicken shed.” Gazzaniga called this part of the brain the interpreter.

Healthy brains do it too. In 2005 Petter Johansson, Lars Hall and colleagues asked 120 people to pick the more attractive of two faces, then secretly handed them the other one. Only 13% of the swaps were noticed at the time, and people explained choices they had not made. Whether the same interpreter writes dreams is not proven; it is a likely family resemblance.

Retold, the story became more sensible

The psychologist Frederic Bartlett, in his 1932 book Remembering, reported giving people a Native American tale, “The War of the Ghosts”, from the Kathlamet texts recorded by Franz Boas, and had each one retell it to the next, like a game of telephone. In the original, the warriors turn out to be ghosts, and at sunrise “something black came out of his mouth.”

Link by link, the strange parts went. Names changed, the ghosts disappeared, the black thing became a soul and then a spirit, and by the tenth version a man is simply wounded in a fight and dies. Bartlett wrote that the story had become “more coherent, as well as much shorter.” Bartlett studied a tale passed from person to person, not dreams. But remembering works by rebuilding, so a dream retold over breakfast may drift the same way, toward a tidier story.

The same gift makes novelists and conspiracies

Robert Louis Stevenson spent two days hunting for a plot. On the second night he dreamed the scene at a window and the scene in which Hyde takes the powder and changes in front of his pursuers. “All the rest was made awake, and consciously,” he wrote in “A Chapter on Dreams”, published in 1888. Mary Shelley's vision of the “pale student of unhallowed arts” came as she lay in bed, and she was careful to say: “I did not sleep.”

Pointed at the world, the same habit of joining dots can mislead. In five studies published in 2018, Jan-Willem van Prooijen, Karen Douglas and Clara De Inocencio found that people who saw patterns in random coin tosses and chaotic paintings were more likely to believe in conspiracy theories and the supernatural, though the link was not equally clear in every test.

Storytellers and believers use the same equipment. The difference is what they do next: the novelist knows the story was made, and the believer takes it for something found.

So what is a dream?

Tested, the prophetic dream has not held up: the bureau that collected premonitions logged mostly misses, and a laboratory test of foresight failed to replicate. Pure static has not held up either: dreams draw on your worries, your people, the game you played before bed. The likeliest picture is a brain working through its own material at night, and a mind that joins what comes up into a story, while you sleep and again, more neatly, when you tell it.

That makes a dream a window, but one that looks in. It can reflect what is on your mind, and it shows how readily a brain turns fragments into meaning. The story you wake with is yours. The message, if there is one, is about you.