Why Do We Dream?

Your child listens; you read. This page is here so you can answer the questions that come after the story. The script itself is available from the script button while the audio plays.

What this episode covers

The story starts from a question children really ask: "Why do we dream?" The honest answer is "We do not know yet." So the story spends its time on a different part. How did people find out when we dream? No one can see a dream. So researchers looked for a sign that a dream might be there: fast eye movements under closed lids. Then they woke the sleeper at that moment and asked. The story tells what this method found, what was later corrected, and what is still argued about. It ends with a small task: write down a dream each morning.

REM sleep is the part of sleep when the eyes move fast under closed lids. The name comes from "rapid eye movement."

How people found out

From babies to an eight-year-old son. Eugene Aserinsky worked at the University of Chicago. He reported REM sleep in 1953. In a 1996 account of his own, he wrote that he first watched sleeping babies with his own eyes. In each hour of sleep, their eyes stayed fully still for about twenty minutes. He wanted to see how this changed with age. In late 1951, he made the first all-night record of eye movement and brain waves together. The sleeper was his eight-year-old son. Instead of still periods, he found periods of about twenty minutes with strong eye movement. At first he wanted to call them "Jerky Eye Movement periods." His teacher, Nathaniel Kleitman, played the doubter. He asked for proof that no one could argue with. Aserinsky wrote that Kleitman must have had mixed feelings. This sleep, with an active brain, went against Kleitman's own idea that sleep was fully passive (PMID 11618742). This part comes from Aserinsky's own memory. So the script says "Aserinsky later wrote" and "As Aserinsky remembered it."

Waking people up to ask. A news article marked fifty years since the finding. It ran in Psychiatric News. It says Aserinsky was a graduate student in physiology, and Kleitman was the head of physiology. On 4 September 1953, they published a report of one and a half pages in Science. They recorded "rapid, jerky" eye movements in 20 sleeping adults. They saw the heart beat faster and breathing speed up during these movements, and then they woke the sleepers. They woke 10 people in 27 REM periods and heard 20 dream reports. When they woke the same people 23 times with still eyes, 19 times there was no dream recall (DOI: 10.1176/pn.39.1.0022). The article also says that the number of people was small by today's standards. The 1953 paper itself was not in our collection. We checked the numbers through this article.

Watching all night. In 1957, William Dement and Kleitman described the normal human sleep cycle. REM periods grow from 3 minutes early in the night to 50 minutes near the end. They come back about every 90 minutes. The article says this made it clear that everyone dreams several times every night (DOI: 10.1176/pn.39.1.0022). In 1959, Michel Jouvet in France saw REM sleep as a third state, apart from both non-REM sleep and waking. He called it "paradoxical sleep" (DOI: 10.1176/pn.39.1.0022). Aserinsky, Kleitman, Dement and Jouvet are named as the pioneers of REM sleep research (PMID 39284581).

Who was first? In 1926, Maria Denisova and Nicolai Figurin worked in the baby section of a brain institute in Leningrad, in the Soviet Union (now Russia). They studied breathing in sleeping babies. They wrote that body and eyelid movements gathered in the periods of fast breathing. One side says that study was about breathing. It treated eye movements as a side effect of breathing, and it did not link them to dreams. So credit should stay with the Chicago team (DOI: 10.1093/sleep/zsae027). The other side says the 1926 study described a new thing in a careful way. It also notes that Kleitman cited that study in his own 1939 book. So, this side says, the 1926 study found what we now call REM sleep (DOI: 10.1093/sleep/zsad292). A 2026 paper in a Russian journal says the Russian claim is beyond doubt (DOI: 10.17650/2222-8721-2026-16-2-81-85). The script gives this argument one short part: "People still do not agree."

Common answers

"We only dream in REM sleep" — not true. In the 1953 numbers, 19 of the 23 still-eye wakings gave no dream recall, so the other 4 did not. (The article does not state the 4 itself; we worked it out as 23 − 19.) The article says the idea of dreaming outside REM came up soon after REM was found. Visual, story-like reports can come from all sleep stages, but more often from REM (DOI: 10.1176/pn.39.1.0022). In one test, 11 healthy young men took a drug that cuts REM sleep. It cut REM sleep by 81% on average. The test was double-blind and cross-over. Still, the length, detail and strangeness of dream reports hardly changed. The authors suggested that mental activity in sleep does not depend on the sleep stage (PMID 22647346). A study with many brain-wave sensors found that dreaming, in both REM and non-REM sleep, came with lower slow activity in a back part of the brain, a "hot zone." Watching this zone in real time could predict whether a person in non-REM sleep would report a dream (PMID 28394322). Still, REM dreams tend to be longer, more vivid and more like a story (PMID 36572252). A study of 133 dream reports also found that REM reports were more connected in structure than reports from N2 sleep (PMID 32701992).

"Some people never dream" — still argued. In one sleep clinic, 6.5% of patients said on a form that they had never dreamed. After phone interviews, this fell to 0.38% of all patients. (The abstract writes "N=534" next to 6.5%. For 0.38% to still hold 16 people, 534 must be the number of form answers, not all patients.) Sixteen people still said they had never dreamed. They were woken 36 times during lab sleep. Not once did they report a dream. On sleep tests, health and background, these 16 did not differ from 12 "rare dreamers," who sometimes reported a dream when woken (3 of 32 wakings). The authors said these people might not experience dreaming (PMID 14592328). Another study looked at 289 patients with a sleep illness in which people act out dreams with their body during REM sleep. Eight of them did not recall dreams, and four had never recalled a dream. Yet on video, these eight showed dream-like actions, such as arguing, fighting and speaking. The authors called this strong evidence that they do dream, but do not recall it (PMID 26307463). The two studies looked at different groups, and both are small. The script does not name the illness. But it keeps the order: first, rare people who move along with their dreams; then, a few of them who recall no dreams.

"Dreams happen in a blink" — not shown; small studies point both ways. We searched our files and ran three new PubMed searches: dream length and REM period length, time sense in dreams, and time in lucid dreams. We found no study that measured how long an ordinary dream really lasts. One 1975 study leans toward the common belief. Dreams were started with a signal of known length. Simple, shapeless dreams were judged about right. But complex dreams, with many things at once, were judged much longer than they were, as if time sped up (PMID 50182). On the other side, five skilled lucid dreamers (people who know they are dreaming) counted inside a dream. It took about as long as counting while awake, and movements like squats took longer (PMID 15739850). In an online survey of 571 lucid dreamers, a lucid dream lasted about 14 minutes on average, by self-report (PMID 24934010). All three are narrow and small. So the script says "This is not clear yet" and gives the first two studies side by side.

"Things from the day show up in dreams" — yes, but not as they were. In a Montreal sleep lab, 343 people reported 528 dreams. Of these, 35.8% had the lab in them. From REM sleep, the figure was 44.2% (PMID 34614021). Dement said that dreams mirror worries from the day, but they are never just like the day (DOI: 10.1176/pn.39.1.0022).

Still unknown: why do we dream?

At a meeting for the fiftieth year of REM research, Jouvet said "the function of dreams remains an unsolved mystery" (DOI: 10.1176/pn.39.1.0022). A 2023 review also says the purpose of dreams is unclear. It lists several ideas linked to memory and feelings (PMID 36572252). The script puts four ideas side by side. It does not call any one of them the answer.

Sorting memories. This idea says that in sleep, new memories are played again and made stable, and that this shows up in dreams (PMID 24477388). In one study, people practiced a computer maze before a night of sleep. The maze showed up in dreams from all stages of sleep. Dreams that clearly showed the task went with better maze scores the next morning (PMID 30091247). A meta-analysis joined 16 studies with 45 results. It found a strong link between task dreams and better memory (SMD 0.51). The link was clear for non-REM dreams, but not for REM dreams. The authors called the possible stronger link in non-REM early evidence (PMID 37058584). These are all links, so the script says, "this only shows the two things went together." It also gives the non-REM result in one sentence. There is also a case against the idea that REM sleep sorts memories. A 2000 review says that animal studies that took away REM sleep split about half and half. Three major kinds of antidepressant drugs cut REM sleep a lot. On the whole, they do not harm learning or memory. A few patients with brain-stem damage lost REM sleep, but they lived normal lives. So the review said REM sleep serves no role in processing or keeping memories (PMID 11515146). This review is an argument built from many sources, not one experiment. It is aimed at REM sleep, not at dreams. In the meta-analysis above, the clear link between dreams and memory came from non-REM dreams. So this case does not answer the whole "dreams and memory" idea. The script also limits the case to "the idea that REM sleep sorts memories."

Practice for danger. In 2000, the dream researcher A. Revonsuo proposed an idea. He said dreams act out threats over and over, so we can practice noticing danger and getting away (PMID 11515147). One study supports this. Kurdish children who had lived through severe trauma reported more dreams. Their dreams had more threats, and the threats were more severe, than in children with less trauma or Finnish children. The authors said the results support most of the predictions of the idea (PMID 15766897). But another study found the opposite. In South Africa, an area with much crime, 208 people had faced more danger to life than 116 people in Wales. Yet they reported fewer threat dreams. Overall, less than 20% of dreams had a real threat to survival. Getting away from a threat happened in less than 2% of dreams. The authors said this goes against key parts of the idea (PMID 17702604). The same group later found that "approach" dreams were more common than "avoid" dreams (PMID 22196966). People who support the idea replied that much of the disagreement comes from reading the theory, and its tests, in different ways (PMID 18325788).

Calming feelings. A 2023 study compared 896 dreams. Some came from two forager groups, the BaYaka in Congo and the Hadza in Tanzania. Others came from groups in the Global North, meaning richer, Western countries. Both forager groups also had a lot of threat in their dreams. The study concluded that, in people without a sleep or mental illness, dreams can help control feelings. They link possible threats to calm settings, and this lowers fear and bad feelings (PMID 37783728).

A side effect. Some theories say dreams are a random side effect of REM sleep, with no job of their own (PMID 11515147). The 1977 "activation-synthesis" idea says the brain stem turns on the front of the brain again and again. The front of the brain then builds a dream by matching those signals with stored memories (PMID 21570).

People who recall dreams, and people who do not

One study compared 46 high recallers and 46 low recallers. High recallers reported dreams on 5.98 days a week on average, and low recallers on 0.34 days. These numbers compare two chosen groups. They do not show how all people spread out (PMID 30319519). Another study scanned the brain 5 minutes after people woke from a nap. In 20 high recallers, links in and around the brain's "default mode" network were stronger than in 18 low recallers. The authors suggested this may help keep the dream in short-term memory while waking up (PMID 32597973). This is one small study's idea, so the script says "The team guesses." Dream research also has a basic limit. We can only reach a dream through the report of someone who woke up (PMID 35417930).

Questions your child may ask

"Do animals dream?" This is hard to check. We know about human dreams because people tell us. We cannot ask animals. One review suggests three ways to study it in animals: brain activity linked to dreams, the replay of new memories, and acting out dreams (PMID 34653784). Another view says that since REM sleep is found across mammals, other mammals likely dream too (PMID 15843983). You can link this to the story: "We cannot use the story's method, waking up and asking, with animals."

"Can I choose my dreams?" Children aged 3 to 5 tended to think dreams can be controlled, like make-believe. With age, more people believed that dreams are not under our control (PMID 11806699). Instead of correcting your child, you can ask back: "Did you choose last night's dream?" That fits the story's method.

What is known and what is not yet

Known. REM sleep, with fast eye movements, comes back again and again through the night, and waking people from it gives many dream reports (DOI: 10.1176/pn.39.1.0022). People also dream outside REM sleep, and REM dreams tend to be longer and more vivid (PMID 22647346, PMID 36572252). People differ a lot in how often they recall dreams (PMID 30319519).

Not yet. Why we dream (PMID 36572252). Whether people who recall no dreams really do not dream, or just forget (PMID 14592328, PMID 26307463). Who should get credit for finding REM sleep (DOI: 10.1093/sleep/zsae027, DOI: 10.1093/sleep/zsad292).

Things to try together

• A one-week dream diary. Put paper and a pencil by the bed. As soon as your child opens their eyes, before getting up, they write one line. If nothing comes, they write "none." Researchers also use morning diaries to measure dream recall. One study used a two-week sleep diary (PMID 30206818). In a study of patients with brain damage, the morning diary matched well with asking people after waking them during sleep (PMID 3880176). We did not find, in this collection, evidence that keeping a diary makes you recall more dreams. So the script only says "count."
• Compare days you write at once and days you write later. This links to the study that suggests the first few minutes after waking may matter for dream memory (PMID 32597973). It is one study's guess, so tell your child it is fine if the result is different.
• Share dreams. In 88 pairs of mothers and children (average age 9.9), there was a small but clear link (0.19) between how often mothers and their children recalled dreams. The authors said this fits the idea that families shape dream recall (PMID 17326485). The link is small, and the cause is not known. If you write your own dreams too, your child has something to compare with.

Evidence

• "'53 REM Discovery Launched Study Of Sleep Disorders, Treatment" (2004) — DOI: 10.1176/pn.39.1.0022 — A Psychiatric News article for the 50th year of REM sleep. It gives the 1953 waking numbers, the 1957 sleep cycle, and Jouvet's "unsolved mystery."
• "The discovery of REM sleep." (1996) — PMID 11618742 — Aserinsky's own account: watching babies, the 1951 all-night record of his eight-year-old son, and a doubting teacher.
• "Who discovered rapid eye movement sleep? The claims for precedence" (2024) — DOI: 10.1093/sleep/zsae027 — The 1926 study is not proof of being first; credit stays with the Chicago team.
• "Reply to Silber: on discoveries in science" (2023) — DOI: 10.1093/sleep/zsad292 — Argues that the 1926 study found what we now call REM sleep.
• "On the 100th anniversary of the discovery by Russian scientists of periodic phenomena and the phase activity of eye movements during sleep" (2026) — DOI: 10.17650/2222-8721-2026-16-2-81-85 — Claims the Russian researchers were first.
• "The 70 years since the discovery of rapid eye movement sleep: history, electroencephalographic features and unsolved issues." (2025) — PMID 39284581 — Aserinsky, Kleitman, Dement and Jouvet laid the base for REM sleep research.
• "Dreaming without REM sleep." (2012) — PMID 22647346 — Cutting REM sleep by 81% hardly changed dream recall (11 men).
• "The neural correlates of dreaming." (2017) — PMID 28394322 — In both REM and non-REM sleep, dreaming came with activity in a back "hot zone."
• "What are the neural mechanisms and physiological functions of dreams?" (2023) — PMID 36572252 — REM dreams are longer and more vivid; the purpose of dreams is unclear.
• "Structural differences between REM and non-REM dream reports assessed by graph analysis." (2020) — PMID 32701992 — REM reports were more connected in structure than N2 reports.
• "Non-dreamers." (2003) — PMID 14592328 — 16 people who said they never dream gave no dream reports when woken (no test differences from 12 rare dreamers); they might not dream.
• "Evidence that non-dreamers do dream: a REM sleep behaviour disorder model." (2015) — PMID 26307463 — 8 people who recall no dreams showed dream-like actions; they dream but do not recall.
• "The characteristics of EEG activity and the subjective estimation of time during dreams of different structure." (1975) — PMID 50182 — In started dreams, simple dreams were judged about right; complex dreams were judged much too long.
• "Time required for motor activity in lucid dreams." (2004) — PMID 15739850 — In 5 skilled lucid dreamers, counting in a dream took about as long as awake; movements took longer.
• "The phenomenology of lucid dreaming: an online survey." (2014) — PMID 24934010 — 571 lucid dreamers reported that lucid dreams last about 14 minutes on average.
• "Dreaming of the sleep lab." (2021) — PMID 34614021 — 35.8% of 528 dreams had the lab in them.
• "Dreaming and offline memory consolidation." (2014) — PMID 24477388 — The idea that memory replay in sleep shapes dreams.
• "Dreaming of a learning task is associated with enhanced memory consolidation: Replication in an overnight sleep study." (2019) — PMID 30091247 — Dreams that clearly showed a maze task went with better scores the next day.
• "A meta-analysis of the relation between dream content and memory consolidation." (2023) — PMID 37058584 — The link between task dreams and memory (SMD 0.51) was clear in non-REM, not in REM.
• "The case against memory consolidation in REM sleep." (2000) — PMID 11515146 — Argues that REM sleep plays no role in memory.
• "The reinterpretation of dreams: an evolutionary hypothesis of the function of dreaming." (2000) — PMID 11515147 — The idea that dreams practice for threats.
• "The threat simulation theory of the evolutionary function of dreaming: Evidence from dreams of traumatized children." (2005) — PMID 15766897 — Children with severe trauma had more and worse threats in dreams, supporting most predictions.
• "Threat in dreams: an adaptation?" (2008) — PMID 17702604 — The group facing more danger reported fewer threat dreams; under 20% of dreams had real survival threats.
• "Approach/avoidance in dreams." (2012) — PMID 22196966 — "Approach" dreams were more common than "avoid" dreams.
• "How to test the threat-simulation theory." (2008) — PMID 18325788 — Much disagreement comes from reading the theory and its tests differently.
• "Evidence for an emotional adaptive function of dreams: a cross-cultural study." (2023) — PMID 37783728 — 896 dreams, including forager groups, support a feeling-control role for dreams in non-clinical groups.
• "The brain as a dream state generator: an activation-synthesis hypothesis of the dream process." (1977) — PMID 21570 — The brain stem turns on the front brain, which builds dreams from memory.
• "Dream Recall Frequency Is Associated With Medial Prefrontal Cortex White-Matter Density." (2018) — PMID 30319519 — 46 high recallers: 5.98 days a week; 46 low recallers: 0.34 days.
• "Brain functional connectivity upon awakening from sleep predicts interindividual differences in dream recall frequency." (2020) — PMID 32597973 — Stronger brain links just after waking may help high recallers keep the dream.
• "What about dreams? State of the art and open questions." (2022) — PMID 35417930 — We can only reach dreams indirectly; this is a basic limit.
• "Do animals dream?" (2021) — PMID 34653784 — Three ways to study dreams in animals we cannot ask.
• "Mammalian sleep." (2005) — PMID 15843983 — Since REM sleep is common in mammals, other mammals likely dream.
• "Development of beliefs about the origins and controllability of dreams." (2002) — PMID 11806699 — Young children think dreams can be controlled; this belief fades with age.
• "Brain structural basis of individual variability in dream recall frequency." (2019) — PMID 30206818 — Dream recall was measured with a two-week sleep diary.
• "Dream recall after sleep interruption in brain-injured patients." (1985) — PMID 3880176 — Morning diaries matched well with waking people to ask.
• "Frequency of dream recall by children and their mothers." (2006) — PMID 17326485 — A small but clear link (0.19) between mothers' and children's dream recall.