By Deep sleep and REM sleep are opposites in almost every respect. Deep sleep (stage N3, slow-wave sleep) comes first, concentrated in the two or three hours after you fall asleep; the EEG slows into large synchronised waves, heart rate and blood pressure fall to their nightly minimum, and you are very hard to wake. REM comes later and lengthens toward morning; the EEG looks close to waking, the eyes move, almost every muscle below the neck is paralysed, and this is where long narrative dreams happen. In 100 healthy adults scored to AASM standards, median deep sleep was 20.7% of the night under age 30 and 14.9% over 60, while REM ran 15.5% and 10.3% in the same groups. Because they sit at opposite ends of the night, cutting sleep short at the alarm removes mostly REM, while going to bed very late removes mostly deep sleep. No, deep sleep is not REM sleep: they are different stages, and the confusion is made worse by Apple Watch, which reports a third label called Core sleep for the lighter NREM stages between them. Neither stage is the important one: both are estimated rather than measured by any consumer device, so the useful reading is the trend over weeks, not one night's split.
Most sleep apps draw deep sleep and REM as two bars of different colours and leave it at that, which quietly suggests they are two flavours of the same thing. They are closer to opposites. One shuts the brain down into its slowest, most synchronised state; the other runs it at close to waking speed while paralysing the body so it cannot act on what it is doing.
Getting the difference straight is worth ten minutes, because it tells you what a particular bad night actually cost you.
Deep sleep vs REM: the short version
Sleep is scored into four states: N1 and N2 (light non-REM), N3 (deep, slow-wave), and REM. A night runs through them in cycles of roughly 90 minutes, but the mix inside those cycles changes dramatically from the beginning of the night to the end.
The measured proportions come from a study of 100 healthy sleepers aged 19 to 77, recorded for a full night and scored to current American Academy of Sleep Medicine standards. Median deep sleep was 20.7% of sleep in the under-30 group and 14.9% in the over-60 group. REM was 15.5% and 10.3% in the same two groups. Both stages shrink with age; the light N1 and N2 stages and time awake in bed expand to fill the gap.
Two things follow immediately. First, most of a normal night is neither deep sleep nor REM: light non-REM sleep is the majority of it, and that is not a defect. Second, the textbook figure of 20-25% REM that circulates online sits above what these carefully screened adults actually recorded on a laboratory night. Round numbers travel further than measured ones.
When each stage happens, and why that is the whole story
Deep sleep is front-loaded. Slow-wave activity is at its maximum in the first cycle and decays across the night, so the great majority of your N3 arrives in the two to three hours after you fall asleep. REM does the reverse: the first REM period of the night may last only a few minutes, and the last one before your alarm can run twenty or thirty.
That asymmetry has been measured directly rather than just inferred from charts. In a sleep-restriction study, healthy men were limited to three hours of sleep under two different schedules: one group window ran from midnight to 03:00, the other from 03:00 to 06:00. After four days the same men recorded 17.0% REM in the early window against 25.7% in the late one, and after one day their latency to the first REM period was 78.8 minutes early versus 45.5 minutes late. The late-night hours are simply REM-richer.
Practically, this is the most useful thing on this page. A night that starts three hours late costs you deep sleep. A night that ends three hours early costs you REM. Both are called "not enough sleep", and they are not the same injury.
Is deep sleep REM sleep? No, and here is why the question keeps coming up
Two habits of language cause this. The first is that everyday English uses "deep sleep" to mean "properly asleep", which includes REM. The second is that both stages get sold as the valuable ones, so people assume the valuable stage must be a single thing with two names.
On a sleep study they are almost unmistakable. Deep sleep shows large slow waves at under 4 Hz covering most of the recording, no eye movement, and steady muscle tone. REM shows low-amplitude mixed-frequency activity that resembles wakefulness, bursts of rapid eye movement, and a near-flat chin muscle trace because the brainstem has switched off almost every voluntary muscle below the neck. If you want the anatomical detail, the Wikipedia overview of rapid eye movement sleep is accurate and well sourced.
The paralysis is the detail people find most surprising, and it is not a curiosity. When it fails, the sleeper acts out the dream: that is REM sleep behavior disorder, and it is the clearest possible demonstration that REM and deep sleep are different machinery.
Core sleep vs deep sleep: Apple's third label
If you use an Apple Watch you have a fourth word to deal with. Apple's own support page states that you can view the time and percentage you spent "Awake or in REM, Core, or Deep sleep". Core is not a term from sleep medicine. It maps onto stages N1 and N2: the lighter non-REM sleep that makes up the bulk of a normal night.
So the comparison people actually want is this: Core sleep is not a failure state, and it is not deep sleep. It is where sleep spindles occur, it typically occupies around half of the night in a healthy adult, and seeing a large Core block is not a sign that you slept badly. Reading Apple's labels as "REM and Deep are the good bars, Core is the filler" is the single most common misreading of that screen.
Other brands relabel the same thing differently ("light sleep" is the usual alternative), which is why two devices worn on the same night can produce percentage splits that look like they came from different people. They are not necessarily disagreeing about the physiology; sometimes they are drawing the same night with different boundaries.
Core sleep vs REM sleep vs deep sleep: what is what
The difference in REM and deep sleep is easiest to see in one place, with Apple's Core label added, because the three names describe the same night in three vocabularies.
| Core sleep (Apple) | Deep sleep | REM sleep | |
|---|---|---|---|
| Sleep-medicine stage | N1 and N2, light non-REM | N3, slow-wave sleep | REM |
| When it dominates | All night, between the other two | The first two or three hours | The last hours before waking |
| Share of sleep in healthy adults | The largest share of the night | 20.7% under 30, 14.9% over 60 (median) | 15.5% under 30, 10.3% over 60 (median) |
| Waking someone | Fairly easy | Hard; grogginess afterwards | Easy; alert within seconds |
| Dreams | Fragments | Sparse, thought-like | Long, vivid stories |
The shares in the third row come from the 100 healthy sleepers aged 19 to 77 already cited above; a meta-analysis pooling 65 studies and 3,577 people aged 5 to 102 found the same direction, with slow-wave sleep percentage, REM percentage and REM latency all falling with age in adults, so both stages shrink as you get older.
People ask what's more important deep sleep or REM as if one could stand in for the other. It cannot, for the reasons in the next section: the body prioritises deep sleep early and REM late and rebounds on whichever it lacks.
When does REM sleep end? Each REM episode ends with a shift to lighter sleep or a brief arousal, and the night's last REM period ends with the morning wake-up. That is why the final hour decides how much REM you get, and why an alarm set early removes REM first. Do we dream in deep sleep? Sometimes, but the dreams are sparser and less story-like than REM dreams; the FAQ below gives the short version of the evidence.
What each stage is for, and which one matters more
Deep sleep is when the brain does its housekeeping. The most cited demonstration is a 2013 study in mice, in which natural sleep or anaesthesia was accompanied by a 60% increase in the space between brain cells, a striking rise in the exchange of cerebrospinal fluid with interstitial fluid, and a faster clearance rate of beta-amyloid. That work was done in mice and the human picture is less settled than headlines suggest, but it gave the restorative function of sleep a plausible physical mechanism for the first time.
REM has a harder-to-summarise job: memory consolidation (particularly emotional and procedural memory) and something like overnight emotional processing. What it also has is an unusually direct link to long-term outcomes. A study of two independent cohorts, 2,675 older men followed for a median of 12 years and 1,386 middle-aged adults followed for nearly 21, found a 13% higher mortality rate for every 5% reduction in REM sleep after adjustment, with similar results for cardiovascular and other causes. People with less than 15% REM had higher mortality than those with 15% or more, and a random forest model picked REM out as the sleep stage most strongly associated with survival.
That is an association in observational cohorts, not proof that raising your REM percentage extends your life. Low REM may well be a marker of something else, such as undiagnosed illness or medication. It is still the strongest single argument against treating REM as the optional stage.
Which leaves the answer to "which is more important": you do not get to choose, so the question is academic. Both are regulated, both rebound when lost, and both are estimated rather than measured by everything you own.
Take it to a doctor rather than to an app if the stage split is not the real problem. Specifically: if you snore loudly, stop breathing, or wake gasping or choking, get assessed for sleep apnea. It shreds both stages and no amount of sleep hygiene fixes it. If you physically act out dreams, shout, punch, kick or leave the bed and can describe a matching dream afterwards, that is dream enactment and belongs in front of a neurologist or sleep physician, particularly over the age of 50. And if you sleep seven hours or more and are still exhausted every day, the cause is a medical question, not a tracking question. Stage percentages from a watch, ring or phone app cannot diagnose or exclude any of this.
Light sleep vs deep sleep: the same night under three names
Light sleep vs deep sleep looks like a choice between a weak stage and a strong one. In the lab it is a split inside non-REM sleep: N1 and N2 are the light stages and N3 is the deep one. Apple Watch calls them Core and Deep, and most other apps say light and deep. The table lines up the names with minutes from a home recording study of 400 Swedish women. Its scoring rules are older and snorers were over-sampled, so its deep-sleep share is lower than in the healthy sample used earlier on this page.
| Light sleep | Deep sleep | REM | |
|---|---|---|---|
| Lab name | N1 and N2 | N3, slow-wave sleep | REM |
| Apple Watch label | Core | Deep | REM |
| Mean minutes, 400 women | 275 (31 + 244) | 41 | 76 |
| Share of the 392 minutes, calculated from those figures | about 70% | about 10% | about 19% |
Sahlin C, Franklin KA, Stenlund H, Lindberg E. Sleep in women: Normal values for sleep stages and position and the effect of age, obesity, sleep apnea, smoking, alcohol and hypertension. Sleep Med 2009. PMID 19345643
Two conclusions hold across both samples. Most of the night is light sleep, whichever way it is counted, and light sleep is not wasted time. And the split between deep sleep and REM depends on who is measured and how it is scored: 10% against 19% here, 20.7% against 15.5% in the healthy adults under 30 above. A single night's bar for deep sleep from a device is an estimate against numbers that vary this much.
Can you dream in light sleep?
Yes. REM is the stage of the long, story-shaped dreams, but dreaming is not limited to it. A review of mentation in REM and NREM sleep concluded that experience occurs in NREM sleep too, with an average recall rate of about 50% after NREM awakenings.
Nielsen TA. A review of mentation in REM and NREM sleep: "covert" REM sleep as a possible reconciliation of two opposing models. Behav Brain Sci 2000. PMID 11515145
A high-density EEG study went further. In both NREM and REM, reports of dreaming went with a local drop in slow activity at the back of the brain, and monitoring that region in real time predicted whether a person woken from NREM sleep would report a dream. Most NREM time is light sleep, so dreaming in light sleep is common, though those dreams tend to be shorter and less story-like than REM dreams.
Siclari F, Baird B, Perogamvros L, Bernardi G, LaRocque JJ, Riedner B, et al. The neural correlates of dreaming. Nat Neurosci 2017. PMID 28394322
A 2026 study woke people from N2, the largest light-sleep stage, and found that perceived sleep depth was highest during immersive dreaming or deep unconsciousness and lowest during minimal awareness, a mere sense of presence. How deep a night feels is therefore not the same as how deep it is, which is one more reason to read stage bars as rough.
Michalak A, Marzoli D, Pietrogiacomi F, Bergamo D, Elce V, Pedreschi B, et al. Immersive NREM2 dreaming preserves subjective sleep depth against declining sleep pressure. PLoS Biol 2026. PMID 41875080
What are the sleep cycles?
A sleep cycle is one stretch of non-REM sleep followed by an episode of REM. So what are normal sleep cycles? The best recent numbers come from the Centre for Chronobiology in Basel, which pooled 369 healthy people who each had an 8-hour sleep opportunity in controlled laboratory studies between 1994 and 2020. Across 6,064 recorded cycles the median cycle lasted 96 minutes, with large differences between people, and the first cycle was consistently shorter than later ones. Eight hours therefore holds about five cycles, not a fixed number, and the 90-minute textbook figure is an average that no one's night follows exactly. For bedtimes counted back from an alarm, the sleep cycle calculator does the arithmetic.
Cajochen C, Reichert CF, Münch M, Gabel V, Stefani O, Chellappa SL, et al. Ultradian sleep cycles: Frequency, duration, and associations with individual and environmental factors-A retrospective study. Sleep Health 2024. PMID 37914631
Deep sleep cycles: why they thin out through the night
Deep sleep cycles are not equal. In the same Basel data set the spread of non-REM episodes grew narrower and the spread of REM episodes wider later in the night, and after sleep deprivation the first cycle held longer non-REM sleep. That fits the front-loading of N3 described above: the first one or two cycles carry most of the deep sleep, and the later cycles are mostly light sleep and REM.
Where SleepTrace fits
Everything above is scored from electrodes. SleepTrace is not: it estimates stages from sound with a phone (iPhone or Android) that records the night, which puts it in exactly the same category as the watches and rings: an estimate, useful as a trend and unreliable as a single night's verdict. We would rather say that clearly than sell you a percentage.
Where a microphone genuinely adds something is the part a stage chart cannot show: what happened. A REM block that keeps breaking up at 04:30 means something different if you can hear snoring, a partner, a dog or a truck underneath it. If you want the numbers for each stage on their own, start with how much deep sleep do you need and how much REM sleep do you need, or work backwards from your alarm with the sleep calculator.
Frequently asked questions
No. They are two different stages with almost nothing in common apart from both being sleep. Deep sleep is stage N3, part of non-REM sleep: slow brain waves, low heart rate, no eye movement, very hard to wake from. REM is a separate stage with fast brain activity close to waking, darting eye movements and near-total muscle paralysis. The confusion is understandable (both get called the deep or important part of sleep in everyday speech), but on a sleep study they look nothing alike.
Core sleep is Apple's label, not a scientific stage. Apple Watch reports four categories (Awake, REM, Core, Deep); Core covers the lighter non-REM stages that sleep medicine calls N1 and N2. Deep is stage N3. So Core sleep is not shallow wasted time and it is not deep sleep either: it is the majority of a normal night, typically around half of it, and it is where sleep spindles happen.
Neither, and the question does not really have an answer, because you cannot trade one for the other. The body regulates both, prioritises deep sleep early and REM late, and rebounds on whichever one it was short of. The only practical version of the question is which one a given behaviour costs you: a late bedtime mostly costs deep sleep, an early alarm mostly costs REM, and alcohol reliably costs REM.
Both, but differently. The long, vivid, story-shaped dreams people remember come overwhelmingly from REM, which is also where dream reports are longest and most emotional. Non-REM sleep, including deep sleep, produces mentation too, but it tends to be sparse, static and thought-like rather than narrative. Waking someone out of deep sleep usually produces confusion rather than a dream report.
Each REM episode ends with a shift to lighter sleep or a brief arousal, and the last REM period of the night ends at the morning wake-up. REM episodes get longer toward morning, which is why waking early removes far more REM than deep sleep.
They differ in timing, brain activity and function. Deep sleep comes first, with slow brain waves and a heart rate at its lowest, and in healthy adults it makes up a median 20.7% of sleep under age 30 and 14.9% over 60. REM comes later, with fast brain activity and long dreams, at a median 15.5% under 30 and 10.3% over 60.
Yes, but less than in REM. A review of mentation in REM and non-REM sleep found that experience occurs in non-REM sleep too, with an average recall rate of about 50% after awakenings from it, and a high-density EEG study found that reports of dreaming in both stages went with a local drop in slow activity at the back of the brain. Deep sleep dreams tend to be sparser and more thought-like than REM stories.
A cycle is a stretch of non-REM sleep followed by REM. In 6,064 laboratory-recorded cycles from 369 healthy people, the median was 96 minutes, with large differences between individuals and a shorter first cycle. That is about five cycles in an 8-hour night, but no cycle count is a target.
Yes. A review found an average recall rate of about 50% after awakenings from NREM sleep, and a high-density EEG study linked dream reports in NREM to the same posterior brain activity change seen in REM. Dreams in light sleep tend to be shorter and less story-like than REM dreams.
Light sleep is N1 and N2, and deep sleep is N3, the slow-wave stage. In a home study of 400 women, light sleep took about 275 of 392 minutes and deep sleep about 41. Deep sleep is concentrated in the first half of the night, light sleep fills most of the rest, and in Apple Watch terms light sleep is Core.
References
- Mitterling T, Högl B, Schönwald SV, Hackner H, Gabelia D, Biermayr M, Frauscher B. Sleep and Respiration in 100 Healthy Caucasian Sleepers--A Polysomnographic Study According to American Academy of Sleep Medicine Standards.. Sleep (2015). Europe PMC
- Xie L, Kang H, Xu Q, Chen MJ, Liao Y, Thiyagarajan M, O'Donnell J, Christensen DJ, Nicholson C, Iliff JJ, Takano T, Deane R, Nedergaard M. Sleep drives metabolite clearance from the adult brain.. Science (2013). Europe PMC
- Leary EB, Watson KT, Ancoli-Israel S, Redline S, Yaffe K, Ravelo LA, Peppard PE, Zou J, Goodman SN, Mignot E, Stone KL. Association of Rapid Eye Movement Sleep With Mortality in Middle-aged and Older Adults.. JAMA Neurol (2020). Europe PMC
- Wu H, Stone WS, Hsi X, Zhuang J, Huang L, Yin Y, Zhang L, Zhao Z. Effects of different sleep restriction protocols on sleep architecture and daytime vigilance in healthy men.. Physiol Res (2010). Europe PMC
- Ohayon MM, Carskadon MA, Guilleminault C, et al. Meta-analysis of quantitative sleep parameters from childhood to old age in healthy individuals: developing normative sleep values across the human lifespan. Sleep (2004). Europe PMC
SleepTrace is a wellness app, not a medical device. This article is general information, not medical advice. If your symptoms are frequent, severe or worrying, please talk to a doctor.
Deep sleep and REM do different jobs, and the balance matters more than one number. SleepTrace estimates light sleep, deep sleep and REM from the sound of your night and shows them as a timeline. These are estimates from audio, not an EEG, so read trends across a week rather than one night.
Hear your own night. SleepTrace turns a night of audio into your sleep phases, the sounds you made, and how it all trends. Works without a wearable, better with one. Your phone on the nightstand is enough to start. Get it for iPhone or Android →