Dreaming in Non-REM Sleep: What Changed Since 1953
The first sleep laboratories found vivid dreams in REM sleep and almost nothing outside it, and that became the popular picture of dreaming. It was never the whole story. Change the question you ask a waking sleeper and non-REM sleep answers, and modern recordings find the same brain signature in both states.
Dreaming is not confined to REM sleep. Across the studies Nielsen reviewed in 2000, people woken from non-REM sleep reported mental activity about half the time, and in 2017 a high-density EEG study found the same posterior brain signature accompanying dreams in both REM and non-REM. The 1950s picture, in which REM was the state of dreaming, was partly an artefact of the question researchers asked.
At a glance
- Dement and Kleitman woke sleepers in 1957 and got dream reports from 152 of 191 REM awakenings and 11 of 158 outside REM. That result is real and still stands.
- David Foulkes changed one thing in 1962. He asked what had been going through their mind rather than whether they had dreamed, and non-REM recall rose to around 70 percent.
- In 2017 Siclari and colleagues found dreaming in both states goes with reduced low-frequency activity in a posterior region of the cortex, accurate enough to predict a report before asking.
Where the science stands
What is established: mental activity during non-REM sleep is real and reproducible. It has been found in every decade since Foulkes, at rates that vary with how the sleeper is asked and when in the night they are woken, and non-REM reports are on average shorter and less vivid than REM reports. Both halves of that sentence are solid.
What is open: almost everything about the mechanism. Whether one process generates all sleep mentation or two different ones do is unsettled, and Nielsen's proposal of covert REM activity inside non-REM sleep is a reconciliation, not a finding. The 2017 posterior hot zone result is influential and disputed, with a published commentary arguing the neural correlates of dreaming have not been identified yet. Where exactly a thought-like fragment stops being a thought and starts being a dream has no agreed answer either.
And underneath all of it sits the question this page does not touch. Why we dream, and what dreams mean, has no scientific consensus. Knowing that dreaming is not confined to one sleep stage tells you where it happens, not what it is for.
The research itself
If you are here for the studies rather than a summary, these are free to read in full:
- The neural correlates of dreaming
- The Neural Correlates of Dreaming Have Not Been Identified Yet
- What about dreams? State of the art and open questions
The books this page draws on are listed in Sources and References.
The experiment that set the picture
In 1953 Eugene Aserinsky and Nathaniel Kleitman found that sleep has a phase of rapid eye movement with an active cortex, and in 1957 William Dement and Kleitman turned that observation into a method. They woke sleepers inside and outside REM and compared what came back. Dream reports arrived from 152 of 191 awakenings during REM and from 11 of 158 outside it.
That is a strong result and it has never been overturned. The REM discovery gave dream research its first physical marker, and for the first time somebody could say when to wake a person to catch a dream.
What the experiment did not settle is why the non-REM awakenings came back nearly empty. Two explanations fit the same numbers. Either non-REM sleep has little mental activity in it, or it has mental activity that sleepers were not reporting. The field took the first reading, and for a while REM and dreaming were treated as the same thing.
One changed question, and the answer changed
In 1962 David Foulkes ran the awakenings again with eight sleepers over 57 nights, and altered one thing. Instead of asking whether they had been dreaming, he asked what had been going through their mind. Recall from non-REM sleep rose to around 70 percent.
The reason is ordinary and worth sitting with. Asked for a dream, people search their memory for something dreamlike, a scene with a story in it. Non-REM mentation is often none of those. It is closer to thinking, a fragment, a preoccupation, a sentence turning over. Asked for a dream, a sleeper discards it. Asked what was in their mind, they report it.
Foulkes also found that REM reports were more visual, more emotional and further from the details of waking life than the non-REM ones. So the two states really do differ. They differ in the kind of mental activity, not in whether there is any.
How often, and one generator or two
Tore Nielsen gathered the studies in 2000 and put the average recall from non-REM awakenings at about 50 percent across the literature. The figure moves with method. How the sleeper is asked matters, and so does how deeply asleep they were and how late in the night the awakening comes, since non-REM mentation grows more dreamlike towards morning.
That review also framed the argument that is still running. A one-generator model says a single process produces mental activity throughout sleep and the differences between states are differences of degree. A two-generator model says REM and non-REM have qualitatively different producers. Nielsen read the evidence as favouring two, though not cleanly. He then proposed a middle position, covert REM processes running inside non-REM sleep, which would account for the more dreamlike non-REM reports.
Covert REM is a proposal, not a result. It is worth knowing as the shape of the disagreement rather than as its resolution.
Watching the brain instead of asking
In 2017 Francesca Siclari, Benjamin Baird, Lampros Perogamvros and colleagues approached it from the other end. Rather than asking sleepers more cleverly, they recorded with high-density EEG and looked for what distinguished the awakenings that produced a dream report from those that did not.
They found it in a posterior region of the cortex, taking in the occipital areas, the precuneus and the posterior cingulate. Dream reports went with reduced low-frequency activity there. The finding held in both REM and non-REM sleep. The signature tracked dreaming, not sleep stage, and that is the part that matters here. Monitoring that region in real time let them predict whether a sleeper would report a dream before they asked.
The result is influential and it is disputed. A commentary published afterwards argues the neural correlates of dreaming have not been identified yet, and it is linked above alongside the paper. We put both in front of you because a page that shows only the striking half of a live argument is not being honest about it.
What this does and does not change for a dreamer
The practical part is small and real. If you wake in the small hours with something in your mind that does not feel like a proper dream, more of a fragment or a train of thought, you have not failed to dream. You have most likely caught non-REM mentation, and it counts. Writing it down is worth as much as writing down a scene with a story in it.
It also explains an ordinary frustration. People who wake early or sleep lightly often say they never dream. What they often mean is that they rarely wake out of REM, so what they catch is the quieter register and they do not recognise it.
What it does not change is meaning. None of this research says what a dream is for or how to read one. It says where dreaming happens, and the answer is more of the night than the textbooks used to allow.
How we hold this at Somniary
This page exists because we had the older picture on the site and left it standing. Our page on the REM discovery reports the 1957 result accurately and stops there. That was the state of knowledge in 1957 and has not been the state of knowledge for a long time. Correcting our own pages is part of the work, not an afterthought.
Our interpreter never asks which sleep stage a dream came from, and after reading this you can see why that would be the wrong question. What a dreamer brings is what they remember, in whatever register they remember it, and that is what gets read.
Frequently Asked Questions
Do you dream in non-REM sleep?
Yes. Mental activity during non-REM sleep is well documented, and across the studies Tore Nielsen reviewed in 2000 the average recall rate from non-REM awakenings was around 50 percent. Non-REM reports are on average shorter, less vivid and more thought-like than REM reports, which is why the earliest experiments found so little of them, but they are dreams by any ordinary use of the word.
Why did early research say dreams only happen in REM sleep?
Because that is what the first experiments found. Dement and Kleitman woke sleepers in 1957 and got dream reports from 152 of 191 awakenings during REM and from only 11 of 158 outside it. The finding was real and still stands. What it did not settle was whether non-REM sleep is empty or simply harder to report from, and the answer turned out to be the second one.
What did Foulkes find in 1962?
David Foulkes changed the question. Instead of asking whether sleepers had been dreaming, he asked what had been going through their mind, and recall from non-REM sleep rose to around 70 percent. People had been discarding quiet, thought-like mental activity because it did not match what they understood a dream to be. The study appeared in the Journal of Abnormal and Social Psychology in 1962.
Can scientists tell when someone is dreaming?
Partly, and not by sleep stage. In 2017 Siclari and colleagues used high-density EEG and found that dream reports in both REM and non-REM sleep went with reduced low-frequency activity in a posterior region of the cortex. Monitoring that region live let them predict whether a sleeper would report a dream. The result is influential and contested: a published commentary argues the neural correlates of dreaming have not been identified yet.
Where can I read the non-REM dreaming research for free?
Siclari and colleagues 2017 in Nature Neuroscience is available as a free PDF from the Universidad de la Republica sleep laboratory, and the commentary disputing it is open access on PubMed Central. The 2022 review by Scarpelli, Alfonsi, Gorgoni and De Gennaro in the Journal of Sleep Research, also open access on PubMed Central, covers where the whole question stands.
Go Deeper
The study that changed the question and raised non-REM recall to around 70 percent.
View in Sources ↗Where the roughly 50 percent figure comes from, and the one generator against two argument in full.
View in Sources ↗The posterior hot zone, found in both REM and non-REM sleep and able to predict a dream report.
View in Sources ↗A recent map of the whole field, useful for seeing how much of this is still argued over.
View in Sources ↗One dream is a night.
Kept together, they begin to show a thread. The Dream Journal letter arrives on the 1st and the 15th: one symbol read closely, one honest idea from the science of sleep.
Free · Twice a month max · One-click unsubscribe
And if you want to keep the dreams themselves: the Companion plan keeps what you cannot see alone.