Sleep Stages and Sleep Cycles — Dives into Deep Water, and a Sleep No Depth Gauge Can Measure

After Turning Off the Lights and Lying Down

Around midnight, you turn off the lights, lie down, and open your eyes to a seven o'clock morning alarm. To an observer beside you, those hours in between are a time when almost nothing happens. Your breathing is steady, you occasionally toss and turn, and if someone calls your name, you wake up.

During those few hours, your brain is not turned off. Sleep is an active process in which multiple systems of the brain and body work together (PMID 42754324). The brain changes how it works several times through the night, and those changes follow a fixed order.

This article follows that sequence by comparing it to going into the water. The brain repeats dives all night long, going down from shallow water into deep water and coming back up. The depth of each dive varies. And between dives, you enter a completely different kind of sleep that cannot be measured by depth. We will look in turn at what is used to measure that depth, what the dives and the sleep between them look like, what the brain is thought to do in each, and how the shape of the night changes as we age. When you feel sleepy and when you wake up are covered separately in another part of this series, "Sleep Pressure and the Biological Clock."

What Measures the Depth of Sleep

A person spending the night in a sleep laboratory lies down with several thin wires attached to their head and face. Brain waves are waves of electrical signals created collectively by countless nerve cells in the brain, captured by electrodes attached to the surface of the head. Dividing the stages of sleep begins with reading where on the head and in what shape these waves appear. Therefore, the spots where the electrodes are attached are also standardized (PMID 16303597).

Two more things are recorded along with the brain waves. These are the electrical signal generated when the eyes move and the tension of the muscles under the chin. These three must be read side by side to grade the stages of sleep (PMID 16303597). Polysomnography is a test that records brain waves, eye movements, muscles, breathing, and blood oxygen together while sleeping. The breathing and oxygen records are used to look at breathing problems during sleep, and the sleep stages are graded by the preceding three records.

When awake, brain waves have low height and tremble fast. As sleep deepens, they change into large and slow waves (PMID 22811426). Compared to diving, the shape of these waves is the depth gauge. The records of the eyes and muscles are used, above all, to recognize a different kind of sleep that cannot be distinguished by the depth gauge alone.

Polysomnography is the standard test used when diagnosing various sleep disorders (PMID 41074507). Still, this depth gauge is not without errors. Even for the same person, the record changes every night, and there are instances where two people grade the stages differently even when reading the same record. There is also the criticism that the very method of cutting sleep into several stages cannot capture the continuously changing appearance of sleep (PMID 41074507). There are a few early studies showing that agreement among readers has improved since the American Academy of Sleep Medicine's scoring manual, which unified the scoring rules, came out (PMID 19738470).

There is also a comparison where a computer instead of a person was given just a single line of brain waves to grade the stages. When comparing various methods using public data from 40 people, the best combination matched about 87% in the records of healthy people and about 69% in the records of patients (PMID 28254093). This figure belongs to automated brain wave scoring and is a different story from the accuracy of wrist-worn devices.

Devices worn on the wrist do not record brain waves. Sleep stages are designated to be graded by brain waves, eye movements, and chin muscles, so the stages shown by a wrist device are values guessed from other signals without those reference signals. Although the laboratory's records also have errors, the laboratory records graded with reference signals and the wrist device's graphs guessed without those signals are not on the same scale.

Two Kinds of Sleep, and From Shallow Water to Deep Water

If you trace a sleeping person's record through the night, two kinds of sleep appear in turns (PMID 23385692). One is sleep where the eyeballs move quickly beneath closed eyes, and the other is sleep without such movement. REM sleep is sleep where the eyeballs move quickly beneath closed eyelids, most of the body's muscles relax so they do not move, and people have many vivid dreams. REM is a phonetic rendering of the English acronym for "rapid eye movement." People also dream during non-REM sleep, and we will revisit this story a little later. Non-REM sleep is the remaining sleep that is not REM sleep, and it is divided into several stages from light sleep to deep sleep.

Sleep begins in non-REM sleep. The brain waves of a person lying down with their eyes closed gradually slow down from the small, fast ripples of being awake. The first stage is shallow water at the threshold between wakefulness and sleep, and it usually passes quickly.

When you descend to the second stage, unique patterns interweave into the brain waves. Sleep spindles are bundles of fast-shivering waves that briefly interweave into brain waves during non-REM sleep and then disappear. They got this name because their shape is bulging in the middle like a spinning wheel's spindle, and they are a representative marker for identifying the second stage (PMID 21075230). Sleep spindles are thought to do the work of stabilizing memories (PMID 30300592).

The deepest water is the third stage. Deep sleep is the deepest stage among non-REM sleep, where brain waves change into large, slow ripples, so it is also called slow-wave sleep. These large, slow ripples are created when a wide area of the outer layer of the brain marches in step all at once (PMID 11119697). In animal studies, entire non-REM sleep is sometimes called slow-wave sleep, so the scope that this term points to varies by literature (PMID 31277860).

In hospital records, the three stages of non-REM sleep are written as N1, N2, and N3 starting from the shallow side (PMID 19738470). Older criteria divided deep sleep into two stages, the third and fourth, but the American Academy of Sleep Medicine's scoring guidelines grouped the two into one, and criticism also emerged over this grouping (PMID 22701392). How many stages to divide into is a convention decided to organize observations. Within non-REM sleep, the brain descends along a slope from shallow water to deep water. The depth scale in this article is used only here, within non-REM sleep.

Sleep That Cannot Be Measured by Depth — REM Sleep

When you stay in deep water for a while and come back up to the second stage, the record changes completely. Brain waves tremble small and fast again, eyes dart rapidly back and forth beneath closed eyelids, and signals in the jaw muscles almost disappear. In REM sleep, the outer layer of the brain becomes active, the eyes move rapidly, muscles relax, and we dream (PMID 22811426). REM sleep is also called paradoxical sleep (PMID 31277860). This means the brain is as busy as when awake, but the body does not move.

Because REM sleep is a different kind of sleep from non-REM sleep, it cannot be placed on the depth scale that runs from shallow water to deep water. Sleep physiology divides being awake, non-REM sleep, and REM sleep into three distinct states (PMID 31277860). During REM sleep, the brain is busy in a way similar to when it is awake, and the body loses its muscle strength more than at any other time of the night, barely moving at all except for occasional brief twitches. In terms of diving, it is a time when you stop diving and cross over into a different state that has nothing to do with depth underwater. When you look at brain waves, which act like a depth gauge, REM sleep reads like shallow water, but that scale was built to measure the depth of non-REM sleep, so it does not fit REM sleep. That is why the gauges for eyes and muscles tell us when REM sleep is happening.

Malfunctions can also happen in the way muscles relax. REM sleep behavior disorder is a condition where muscles that should be relaxed during REM sleep do not relax, causing a person to act out their dreams with their actual body. Seeing REM sleep where muscle tension remains during a polysomnography is the core marker for diagnosing this disease (PMID 34942893). The signal that relaxes muscles comes from a cluster of nerve cells deep in the lower part of the brain, and it is explained that REM sleep behavior disorder develops when this cluster or the pathways extending from it are damaged. This explanation comes mainly from experiments in mice (PMID 38060134). This disease and several other sleep disorders are covered in "Insomnia and Sleep Apnea".

It is true that there are many vivid dreams in REM sleep, but we also dream in non-REM sleep (PMID 23524011). The pattern in which brain activity spreads during REM sleep has long been thought to match the general characteristics of dreams. However, the brain regions that work together when we dream are still largely unknown (PMID 20870064).

Following a Single Night — Sleep Cycles

We follow the night of someone lying down at midnight from the beginning. They pass through shallow water to reach the second stage, where sleep spindles interweave, and arrive at the deep sleep of large, slow waves. After lingering there for a while, they rise back up to the second stage, stop diving there, and cross over into a different kind of sleep called REM sleep. Once REM sleep ends, the dive begins again. A sleep cycle is one full circuit going from non-REM sleep to REM sleep, and it repeats multiple times in a single night.

Reading that same single circuit as three lines on a laboratory monitor looks like this. Right after lying down, the brainwave line trembles small and fast. Upon entering the second stage, sleep spindles briefly interweave into that line, and upon reaching deep sleep, the entire line undulates with large, slow waves. After rising back to the second stage and moving on to REM sleep, the brainwave line trembles small and fast again, but rapid shaking appears on the eye movement line, and the jaw muscle line becomes almost flat. Looking only at the brainwave line can make REM sleep and wakefulness confusing, so the eye and muscle lines must be read together.

Because the group of nerve cells that summons sleep and the group that wakes us up suppress each other, wakefulness and sleep are described as switching over and staying on one side like a switch. The first stage at the threshold of falling asleep is a short doorway, and once that doorway is passed, sleep settles firmly on one side. The transition between non-REM sleep and REM sleep is explained by the same principle (PMID 23385692). Within non-REM sleep, the space between the shallow stage and the deep stage connects gradually, like the slope we saw earlier. The pattern of a healthy adult staying awake during the day and alternating between non-REM sleep and REM sleep at night comes from this switch (PMID 23385692).

As the night progresses, the shape of a single circuit changes. The dives in the early part of the night go down deeply into deep sleep, and toward dawn, deep sleep becomes shorter while the REM sleep between dives grows longer (PMID 26888926). Therefore, even for the same circuit, a circuit around midnight and a circuit at dawn look different. Around dawn, many dives return without going all the way down to deep sleep.

The figure that a single circuit lasts 90 minutes is often passed around. On average it is around that mark, but the length of a circuit varies by person and according to the earlier and later parts of the same night, so it is not fixed at exactly 90 minutes. What determines that length and the number of circuits completed in a single night remains an unsettled question (PMID 32179430). Rather than counting how many circuits of a certain number of minutes were completed and how many times, it is better to remember the pattern: dives of varying depths and the REM sleep between them are connected multiple times.

Different forces determine when sleep starts and when it ends. The force that builds up in proportion to the time spent awake and the body's internal clock that rises and falls on a daily cycle work together to create sleepiness (PMID 9248514). These two forces are covered in "Sleep Pressure and the Biological Clock."

What the Brain Does During Deep Sleep and REM Sleep

After memorizing a pair of an unfamiliar face and name and then sleeping, the benefit of sleep appears most clearly in the task of recalling that pair. Deep sleep was linked to scores on recalling names when looking at faces, while the effect of sleep was weak and erratic in the task of simply recognizing a face one had seen (PMID 42380545, synthesis of 19 studies). At least for facial memory, sleep does not help all memories equally, and its effect stands out for memories that need to be bound together.

There is a widely accepted model that explains this benefit. The explanation is that memories learned during the day are replayed while asleep, and this replay happens alongside sleep-specific brain waves such as the slow waves and sleep spindles of deep sleep (PMID 24765066). When the slow waves, sleep spindles, and brief signals from the inner area of the brain where new memories are initially stored match their rhythms, it is thought that memories are transferred from the inner part of the brain to the outer layer (PMID 40842704).

This model has reached the stage of hands-on experiments. There are results showing that boosting the slow waves of deep sleep using sound or mild electricity improves memory processing during sleep (PMID 24765066). However, such stimulation techniques may have limitations for widespread use in hospitals due to technical constraints (PMID 32005347). Non-REM sleep is also explained as a time when the brain conserves energy (PMID 22811426).

The realm of REM sleep has a broader range of hypotheses. During REM sleep, brain regions that handle emotions are strongly activated, leading to the presumption that REM sleep is involved in emotional regulation (PMID 22811426). There is also an explanation that REM sleep inserts new memories into existing frameworks of knowledge and refines the texture of emotions (PMID 42090911). However, in a meta-analysis gathering whether sleep specially preserves memories of emotionally charged scenes, such an effect was not seen overall. Looking separately at only eight studies that directly compared REM sleep and deep sleep, a tendency emerged where memories charged with emotion were better retained after REM sleep, and memories without emotion were better retained after deep sleep (PMID 32179376).

The idea that the brain's waste products are washed away during sleep, particularly during non-REM sleep, is also receiving attention. A close look at this cleaning flow comes mainly from mouse experiments. In humans, similar signals resembling a large pulsation of the fluid surrounding the brain during sleep have been seen at the imaging stage, and better methods to properly measure human cleaning functions are needed (PMID 41325105). There is also the hypothesis that sleep deletes and organizes unnecessary connections among those accumulated during the day (PMID 28100731).

A high amount of slow brain waves in deep sleep and feeling refreshed do not overlap immediately. The slow brain wave activity of non-REM sleep does not always align with the depth of sleep a person feels or how refreshed they are the next day (PMID 41325105). And why sleep must be divided into two types remains largely unsolved well over half a century since REM sleep was discovered (PMID 32179430).

Returning to memory, the evidence for deep sleep is thicker. One review writes that the central role of deep sleep—from maintaining bodily balance to memory—has been established. The same review also points out the problem that terms like deep sleep and slow wave activity are defined differently across studies (PMID 41424515).

Among what is considered to be the work of sleep, the most accumulated evidence is for deep sleep and memory consolidation (in facial memory, it stood out especially for memories that need to be bound together), while the role of REM sleep and the cleaning function of sleep still have a wide realm of hypotheses. What happens to memory and the body when sleep is insufficient, and to what extent that evidence holds in humans, is covered in "Sleep Deprivation".

As We Age, the Shape of Our Dive Changes

If you draw the nights of a thirty-year-old and a seventy-year-old side by side, the shape of their dive is different. As adults age, their total sleep time, the actual percentage of time spent sleeping out of the time lying down, and the proportions of deep sleep and REM sleep all decrease. Conversely, the time it takes to fall asleep, the proportion of light first and second stages, and the time spent awake in the middle after falling asleep all increase (PMID 15586779, a meta-analysis of 65 studies involving 3,577 healthy people aged 5 to 102). Even among children and adolescents, the older they were, the lower the proportion of deep sleep was (PMID 15586779).

In the same analysis, the only thing that continued to clearly decrease even after age 60 was the actual percentage of time spent asleep. The magnitude of the change also varied greatly depending on how carefully the participants were selected. Depending on whether sleep disorders such as psychiatric illnesses, other diseases, medications and alcohol, and conditions where breathing repeatedly stops during sleep were filtered out, the association with age was reduced or even masked (PMID 15586779).

The patterns of brainwaves also change. As people age, sleep spindles become smaller, rarer, and shorter, and older individuals had shorter durations of both deep sleep and REM sleep. Women had greater sleep spindle amplitude and longer deep sleep (PMID 42058455, a meta-analysis of 42 studies involving 1,878 healthy individuals). In another meta-analysis gathering data from people without sleep problems, women's sleep spindles were also denser (PMID 42331757).

Superimposing a seventy-year-old's dive onto a thirty-year-old's dive, they spend more time lingering in shallow water, stay less in deep water, and frequently poke their heads out of the water in the middle of the night. The reduction in deep sleep and frequent disruptions in sleep as we age are changes that appear even in healthy older adults, and the body often adapts to this without feeling major discomfort (PMID 41963174). Still, about half of older adults complain of sleep difficulties, which are commonly intertwined with other co-occurring illnesses, various medications, and lifestyle changes such as a decrease in outdoor activities (PMID 41963174). Cases where sleep difficulties stem from sleep disorders are covered in "Insomnia and Sleep Apnea".

Where to Look Further

This article followed the sleeping brain by dividing it into stages and cycles. Why sleep is divided into two kinds and what determines the length of one cycle remain open questions.

• When we feel sleepy and when we wake up, the two forces that call for sleep are covered in "Sleep Pressure and Biological Clock" of this series.
• What happens to daytime attention and the body when sleep is short, and how far the evidence for memory consolidation and brain cleaning goes in humans, continues in "Sleep Deprivation".
• Sleep disorders such as insomnia, the disease where breathing repeatedly stops during sleep, and REM sleep behavior disorder are covered in "Insomnia and Sleep Apnea".
• The work of sleep moving memories to long-lasting places can be revisited from the side of memory in "Three Stages of Memory — Encoding, Consolidation, and Retrieval".
• The pathway where the brain washes away waste products while sleeping is covered in detail in "Glymphatic System — Sleep Is the Time to Wash the Brain".

Source Papers

• "Sleep stage scoring in the adult population." (2005) — PMID 16303597 — Sleep stage scoring reads standard-placement brain waves along with eye movement and submental muscle recordings together (polysomnography).
• "Control of sleep and wakefulness." (2012) — PMID 22811426 — Wakefulness features low, fast brain waves, while sleep features large, slow brain waves. Non-REM sleep suggests roles in brain energy conservation and memory consolidation, while REM sleep suggests roles in muscle atonia, rapid eye movements, dreaming, cortical activation, and limbic system activation for emotional regulation.
• "REM Sleep Is Causal to Successful Consolidation of Dangerous and Safety Stimuli and Reduces Return of Fear after Extinction." (2016) — PMID 26888926 — Cited as background fact only: human nights feature more slow-wave sleep in the first half, with REM sleep taking up a larger share toward the second half (the REM sleep and fear memory results of this paper are not used).
• "Relationships between REM and NREM in the NREM-REM sleep cycle: a review on competing concepts." (2020) — PMID 32179430 — Why two types of sleep are necessary remains largely unclear, and what determines the number and length of nighttime cycles is an open question.
• "The neurobiology of sleep." (2013) — PMID 23385692 — Switch circuits where sleep and wake neuron groups mutually inhibit each other create fast and complete state transitions, with the same principle applying to non-REM-REM transitions.
• "Sleep-wake physiology." (2019) — PMID 31277860 — Three arousal states: wakefulness, non-REM sleep, and REM sleep (paradoxical sleep). This review follows the animal research convention of calling all non-REM sleep slow-wave sleep (this article uses slow-wave sleep to refer only to the deepest stage, per human scoring usage).
• "Why do we sleep?" (2000) — PMID 11119697 — Hypothesis that slow waves represent large-scale synchronization across the neocortex, and spindles open the gate for plasticity.
• "Cognitive neuroscience of sleep." (2010) — PMID 21075230 — Connection between learning and memory and the neurophysiology of each sleep state, including non-REM stage 2 spindles.
• "Sleep spindles." (2018) — PMID 30300592 — Sleep spindles are non-REM brain oscillations considered to have a memory stabilization function.
• "The AASM scoring manual: a critical appraisal." (2009) — PMID 19738470 — AASM scoring guidelines increase N1 and N3 while decreasing N2, and improve inter-rater reliability.
• "The AASM Scoring Manual four years later." (2012) — PMID 22701392 — Combining stages 3 and 4 into N3 was one of the main criticisms.
• "Polysomnography in Transition: Reassessing Its Role in the Future of Sleep Medicine." (2025) — PMID 41074507 — Limitations of polysomnography: night-to-night variability, inter-rater variability, and macroscopic stages failing to capture sleep dynamics well.
• "A comparative review on sleep stage classification methods in patients and healthy individuals." (2017) — PMID 28254093 — Single-channel EEG automated scoring, showing 87.06% accuracy in healthy individuals and 69.05% in patients across 40 public datasets.
• "Neurophysiological Aspects of REM Sleep Behavior Disorder (RBD): A Narrative Review." (2021) — PMID 34942893 — REM sleep without muscle atonia is a polysomnography marker and mandatory diagnostic finding for REM sleep behavior disorder.
• "Ultradian and circadian modulation of dream recall: EEG correlates and age effects." (2013) — PMID 23524011 — Dreams occur during both non-REM and REM sleep.
• "Neuroimaging of dreaming: state of the art and limitations." (2010) — PMID 20870064 — REM sleep brain activity topography was once thought to match general dream features, but the neural correlates of dreaming remain largely unexplored.
• "Human sleep/wake regulation." (1997) — PMID 9248514 — A homeostatic component that builds up depending on prior wakefulness combines with a 24-hour circadian component to regulate sleep and wakefulness.
• "The role of sleep in strengthening face learning and memory consolidation: A systematic review." (2026) — PMID 42380545 — 19 studies: benefits of sleep are most pronounced in associative tasks like face-name learning, slow-wave sleep/N3 is associated with improved cued recall, and simple recognition shows weak and inconsistent effects.
• "Sleep for cognitive enhancement." (2014) — PMID 24765066 — Widely accepted model: memory reactivation during sleep is linked to slow oscillations, spindles, and ripples, and enhancing slow oscillations via electrical or acoustic stimulation promotes memory processing.
• "Slow-wave sleep as a key player in offline memory processing: insights from human EEG studies." (2025) — PMID 40842704 — Temporal coupling of slow oscillations, spindles, and hippocampal ripples assists memory transfer from the hippocampus to the neocortex.
• "Brain Stimulation for Improving Sleep and Memory." (2020) — PMID 32005347 — Stimulation techniques have the potential to enhance slow oscillations and spindles to improve memory, though technical considerations may limit broad clinical application.
• "Multi-scale sleep dynamics in memory consolidation: From microstructural oscillations to macro-architecture." (2026) — PMID 42090911 — Non-REM sleep stabilizes declarative memory, while REM sleep modulates emotional tone and integrates memories into existing schemas.
• "Sleep's impact on emotional recognition memory: A meta-analysis of whole-night, nap, and REM sleep effects." (2020) — PMID 32179376 — The hypothesis that sleep generally enhances emotional memory is not supported; an analysis of 8 studies points to selective enhancement of emotional memory following REM sleep.
• "Is glymphatic clearance the secret to restorative sleep?" (2026) — PMID 41325105 — Glymphatic clearance has been established mainly in rodents, with parallel signals in humans, and measurement methods need improvement. Slow-wave activity does not consistently align with subjective sleep depth and refreshment.
• "Neural Control of REM Sleep and Motor Atonia: Current Perspectives." (2023) — PMID 38060134 — Primarily rodent genetics research: a pontine neuron group drives REM sleep and muscle atonia, and damage to those cells or muscle relaxation pathways results in REM sleep behavior disorder.
• "From macro to micro: slow-wave sleep and its pivotal health implications." (2024) — PMID 41424515 — States that the central role of slow-wave sleep from homeostasis maintenance to memory is established, while raising the issue of inconsistent definitions for slow-wave sleep, slow-wave activity, and slow oscillations.
• "Sleep Is for Forgetting." (2017) — PMID 28100731 — Hypothesis that sleep clears information accumulated during the day to tidy up synaptic networks.
• "Future horizons: Toward an integrated neuroscience of sleep." (2026) — PMID 42754324 — Sleep is an active process involving multiple brain and body systems.
• "Meta-analysis of quantitative sleep parameters from childhood to old age in healthy individuals: developing normative sleep values across the human lifespan." (2004) — PMID 15586779 — 65 studies, 3,577 individuals: in adults, advancing age decreases total sleep, sleep efficiency, slow-wave sleep, and REM sleep percentages, while sleep latency, stage 1, stage 2, and wake after sleep onset increase. After age 60, continuous decline affects only sleep efficiency, and screening criteria drastically alter effect sizes.
• "Differences in sleep spindles and polysomnography in humans: a meta-analysis on the influence of age, sex, and cognitive ability." (2026) — PMID 42058455 — 42 studies, 1,878 individuals: aging decreases spindle amplitude, density, and duration; older adults have shorter slow-wave sleep and REM sleep; and females show greater spindle intensity and slow-wave sleep.
• "A Systematic Review and Meta-Analysis of Biological Sex Differences in Sleep Spindles and Slow Wave Activity in Adults With and Without Insomnia." (2026) — PMID 42331757 — 13 normal sleepers studies: females exhibit higher spindle density, sigma, and delta power.
• "Sleep disorders in the elderly and their management." (2026) — PMID 41963174 — Healthy older adults experience disrupted sleep and reduced slow-wave sleep, but often adapt well enough that discomfort is minimal; nevertheless, about 50% complain of sleep difficulties, frequently linked to comorbidities, polypharmacy, and lifestyle changes.