Sleep Stages: REM, NREM, and the Sleep Cycle
Author: Angela Bogdanova · Published: October 10, 2026 · Editorial Policy
Research cluster: Sleep, Insomnia & Circadian Psychology · Written by Artificial Sapiens
Human sleep is an active, recurring sequence of physiological states. During the night, the brain moves through non-rapid eye movement (NREM) sleep and rapid eye movement (REM) sleep, with different patterns of electrical brain activity, eye movement, skeletal-muscle tone, and responsiveness. NREM is classified into N1, N2, and N3. REM is a separate state. The organization of these stages over time is called sleep architecture. It helps explain why deep sleep typically predominates earlier in the night, why REM periods generally become longer toward morning, and why a chart showing one night's stage percentages cannot, by itself, define a person's sleep health.
The National Heart, Lung, and Blood Institute (NHLBI) describes four to six sleep cycles during a typical night, commonly recurring approximately every 80–100 minutes. Those numbers are useful orientation points, not exact timers. Real nights contain transitions, stage reversals, and brief arousals. Sleep structure changes across the lifespan and responds to the interaction of homeostatic sleep pressure, circadian timing, health, and the amount of sleep a person obtains.
For more context about how this topic connects to sleep science, see Sleep Psychology: How Sleep Works and Why It Matters.
The four sleep stages at a glance
In modern clinical scoring, the stages are N1 (initial light NREM sleep), N2 (established NREM with sleep spindles and K-complexes), N3 (deep or slow-wave NREM sleep), and REM (rapid eye movement sleep with characteristic brain activity and normally substantial skeletal-muscle atonia). Wakefulness is scored separately. The American Academy of Sleep Medicine (AASM) scoring manual provides the rules used in sleep laboratories for classifying physiological recordings.
Readers sometimes encounter references to five stages. The historical convention distinguished deep-sleep stages 3 and 4, whereas modern scoring combines that territory into N3. An older article's “stage 4” is therefore not an extra state that a contemporary smartwatch has somehow lost. It reflects an earlier scoring scheme. NREM and REM are the two overarching phases, while N1, N2, N3, and REM are the four named stages generally used in current adult sleep education.
The distinctions are descriptive, based on recorded signals. They should not be interpreted as four exclusive biological functions. Brain activity, memory processing, autonomic changes, and arousal regulation overlap across stages. Clinical assessment also considers breathing, movements, sleep continuity, timing, and daytime functioning.
N1 sleep: crossing the boundary between wakefulness and sleep
Stage N1 is the transition into sleep. The EEG begins to shift away from sustained waking rhythms toward lower-amplitude, mixed-frequency patterns. Slow rolling eye movements may appear, and muscle activity commonly decreases. Some people awakened during N1 say they were merely resting, even though physiological scoring classifies them as asleep. This is a normal mismatch between subjective experience and an objective stage boundary.
People sometimes experience brief muscle jerks, falling sensations, or fleeting images around sleep onset. These experiences may be normal when isolated, although repeated disruptive events or unusual accompanying symptoms have different possible explanations. N1 is generally a relatively small part of an adult night's sleep. More time scored as N1 can accompany frequent awakenings or fragmented sleep, but stage proportion alone is not evidence of a particular disorder.
The distinction between quiet wakefulness and N1 is difficult to score consistently. A meta-analysis of interrater agreement in polysomnography identified N1 as a comparatively difficult stage even for trained scorers. Consequently, the precise minute of sleep onset reported by a consumer app deserves modest confidence.
N2 sleep: stable sleep, spindles, and K-complexes
N2 typically occupies a substantial share of adult sleep. It is identified partly by characteristic EEG events: sleep spindles, which are short rhythmic bursts, and K-complexes, which are distinctive large-amplitude waveforms. These patterns are important for determining when stable NREM sleep is present. They also attract research interest because they appear to participate in sensory processing, sleep stability, and neural plasticity.
Sleep spindles have been studied in relation to memory and learning, including interactions with slower cortical oscillations. The landmark synthesis About Sleep's Role in Memory describes how coordinated activity during sleep can support consolidation of recently acquired information. A spindle count is not a direct measure of intelligence or learning capacity, however, and correlations between electrical activity and cognitive outcomes require careful interpretation.
N2 can recur after N3, before or after REM, and during the later portions of the night. Describing it simply as “light sleep” may wrongly imply that time in N2 is less worthwhile. It is a normal major state within healthy sleep architecture.
N3 sleep: deep sleep and slow waves
N3 is the stage usually meant by deep sleep or slow-wave sleep. Its EEG contains high-amplitude, low-frequency activity associated with broad synchronization of cortical activity. Awakening a person from N3 is generally more difficult than waking them from N1 or N2, though responsiveness is affected by the stimulus, the individual, and the surrounding circumstances.
Deep slow-wave sleep tends to be more prominent during early sleep cycles. The NHLBI overview also describes its substantial decline with increasing age. N3 is strongly influenced by homeostatic sleep pressure: a period of extended wakefulness can increase subsequent slow-wave intensity. That response helps connect sleep architecture with the accumulating need for sleep, but it does not mean every physiological benefit of sleep occurs exclusively in N3.
Slow oscillations in NREM can coordinate other brain rhythms involved in reactivating and reorganizing memory traces. A 2025 review of systems memory consolidation discusses interactions among slow oscillations, spindles, and hippocampal rhythms, alongside complementary REM processes. The mechanistic details remain under active investigation. Popular descriptions that deep sleep uniquely “detoxifies the brain,” repairs every tissue, or proves how restored someone will feel exceed what a stage score alone can establish.
A low N3 estimate may reflect normal age-related changes, a short or disturbed recording, medication effects, health conditions, or measurement error. A single smartwatch graph that shows little deep sleep does not diagnose a deep-sleep deficiency.
REM sleep: rapid eye movements, active brain physiology, and atonia
REM sleep differs sharply from slow-wave sleep in its patterns of brain activity. The EEG often shows relatively low-amplitude, mixed-frequency activity, while the eyes exhibit rapid movements and most voluntary skeletal muscles are markedly inhibited. This normal muscle atonia helps prevent many dream-related movements from being carried out. Breathing, heart rate, and autonomic activity may become more variable during REM than during stable NREM.
REM is sometimes called paradoxical sleep because parts of brain activity resemble waking while the body exhibits pronounced muscle inhibition. REM periods tend to be shorter early in the night and longer during later cycles. The National Institute of Child Health and Human Development (NICHD) describes the changing distribution of REM and deep NREM across the night.
REM is strongly associated with dreaming, but dreaming also occurs during NREM. Research has linked REM with several aspects of memory and affective processing. These associations should not be converted into a one-purpose explanation such as “REM is when the brain heals emotions.” Different tasks, experimental methods, and individual circumstances produce different results. The interaction of sleep states often matters more than any isolated REM quantity.
Normal REM atonia has clinical significance. Repeated punching, kicking, or other complex movements during apparent dream enactment, especially when injury occurs, merit medical evaluation. They can have different causes and may require assessment for REM sleep behavior disorder or other conditions. A home wearable cannot establish that diagnosis.
REM versus NREM: the key biological differences
NREM contains N1, N2, and N3, with stage-specific changes in EEG, eye movement, muscle activity, and arousal. N2 is characterized in part by sleep spindles and K-complexes, whereas N3 contains prominent slow waves. REM features intermittent rapid eye movements, wake-like aspects of cortical activation, and substantial skeletal-muscle atonia. Both states can involve conscious experiences and dreams after awakening.
The intuitive division into “physical recovery during NREM” and “psychological recovery during REM” is too rigid. Normal NREM and REM both contribute to an integrated sleep period, and experimental studies identify overlapping, interacting roles. What clinicians and researchers observe is a dynamic progression of states with different physiological emphases.
How a sleep cycle unfolds across the night
A sleep cycle is a repeated sequence involving NREM and REM states. A simplified cycle begins with N1, progresses through N2 toward N3, returns to lighter NREM, and enters REM. Subsequent cycles can repeat this organization, often with less deep sleep and more REM later. Actual recordings are more irregular: the brain may move back and forth between N2 and N3, awaken briefly, or shift between states without following a textbook sequence.
The NHLBI's approximate 80–100-minute cycle description and four-to-six-cycle account are useful generalizations. They do not guarantee that everyone's cycles last exactly 90 minutes. Stage proportions and cycle duration can change within the same night and from one night to another. Therefore, calculating a wake-up alarm by dividing desired sleep duration into exact 90-minute blocks is not a reliable method of controlling the sleep stage from which one awakens.
A hypnogram is a graph of scored sleep stages across time. The clinical version is derived from recorded physiological signals. It displays early and late sleep architecture, transitions, wake periods, and arousals. A consumer app's stage timeline may look similar but is inferred from its own sensors and algorithm; the visual resemblance does not guarantee equivalent accuracy.
Why deep sleep usually appears earlier and REM later
Sleep regulation integrates at least two major processes. Homeostatic sleep pressure accumulates during wakefulness and decreases during sleep; the circadian system organizes the biological timing of sleep and wakefulness. The classical two-process model of sleep timing provides a framework for understanding their interaction.
After sustained wakefulness, earlier sleep commonly contains more slow-wave activity. As sleep continues, that pressure diminishes while circadian influences help shape REM expression toward later portions of the night. These are common tendencies, not guaranteed timelines for each individual. Sleep restriction, shift schedules, travel, medications, and medical conditions can alter the observed pattern.
A circadian rhythm is a roughly 24-hour timing process. A sleep cycle is a shorter repeated organization of stages within a sleep period. Chronotype describes an individual's preferred timing of sleep and wakefulness. These three concepts interact, but they describe different features of sleep biology.
How much REM, deep sleep, and light sleep is normal?
Normal adult sleep often contains more N2 than any other stage, a meaningful amount of REM, a variable share of N3, and comparatively little N1. Educational sources commonly illustrate REM as roughly one fifth to one quarter of a night's sleep. Such percentages are population-oriented approximations rather than required nightly targets. Age, illness, medications, past sleep loss, adaptation to the recording environment, and scoring methods can all shift the numbers.
Stage percentages are only part of the evaluation. Someone may have an apparently typical distribution but still experience breathing-related disruptions or dangerous daytime sleepiness. Another person may have unusually low measured slow-wave sleep in a single study without the result demonstrating a specific disease. Even expert human scoring has boundary uncertainty; a meta-analysis found substantial overall agreement but weaker agreement for certain stages.
There is no general clinical rule that a person must achieve an exact deep-sleep or REM percentage to be healthy. Total sleep opportunity, continuity, timing, symptoms, and functioning are often more immediately informative. The AASM and Sleep Research Society recommend at least seven hours of sleep per night regularly for healthy adults, while acknowledging individual variation. This population recommendation is not a prescription for a particular stage mixture.
What sleep stages do for memory and learning
Research strongly supports a role for sleep in memory consolidation, yet sleep-dependent learning is not confined to one stage. During NREM, slow cortical oscillations, spindles, and hippocampal activity can participate in reactivating and reorganizing information acquired during waking. REM-associated electrical and neurochemical conditions may contribute to other forms of integration, plasticity, and emotional-memory processing.
Rasch and Born's extensive physiological review details the research basis for active memory consolidation during sleep. A recent systems-level review describes complementary NREM and REM mechanisms and emphasizes that the exact causal relationships across brain scales are still being investigated. The phrase “deep sleep stores facts; REM processes emotions” is convenient shorthand, but studies do not justify treating it as a strict division of labor.
Sleep disruption can affect next-day learning, attention, and emotional regulation. Nevertheless, an association between a poor night's mood and a low estimated REM percentage cannot establish the direction of causation. Stress or another health factor might influence both sleep and the next day's experience.
Does dreaming happen only in REM?
Dreaming can occur in both REM and NREM. REM awakenings frequently lead to vivid, elaborated dream reports, while NREM reports are often less elaborate on average; there is considerable overlap. Some NREM dreams are complex, emotional, and narrative. Conversely, an awakening from REM may produce little or no recalled dream content.
A landmark 2017 high-density EEG study found neural correlates of reported dreaming across both REM and NREM. A related 2018 study of NREM dreaming examined relationships among slow waves, spindles, and reports of conscious experience. These studies help explain why REM and dreaming should be distinguished rather than equated.
Dream recall also depends on awakening conditions, memory, and the questions asked afterward. A person who rarely remembers dreams has not thereby demonstrated that their REM sleep is absent. A vivid dream before morning awakening does not prove that the entire preceding hour was REM.
How sleep architecture changes across the lifespan
Newborns have sleep organization that differs substantially from adult architecture, and REM-like active sleep occupies a larger proportion of early life. The recognizable NREM stages mature across infancy and childhood. Slow-wave activity becomes particularly prominent in childhood and subsequently changes through adolescence and adulthood. The NHLBI describes marked changes in deep sleep across the lifespan.
In older adults, N3 tends to decrease, and sleep may be more fragmented. These population trends do not establish that a specific person's sleep complaint is merely part of aging. Persistent insomnia, breathing pauses, severe sleepiness, and harmful nighttime behaviors deserve evaluation according to their symptoms and risks.
Comparisons of a child's or older adult's sleep-stage percentages against a young-adult chart can be especially misleading. Age, development, and the scoring system must be considered.
How clinicians measure sleep stages
Polysomnography (PSG) is the reference method for clinical sleep staging. It combines EEG recordings of brain electrical activity, electrooculography (EOG) for eye movements, and electromyography (EMG) for muscle tone. Depending on the clinical purpose, a study may also monitor breathing, oxygenation, cardiac activity, and limb movements. Trained professionals use standardized rules such as the AASM scoring manual to classify the recording and interpret it.
A report may include total sleep time, sleep onset latency, REM latency, wake after sleep onset, sleep efficiency, stage proportions, and arousal information. Each describes a different aspect of sleep. Sleep efficiency is the proportion of time in bed spent asleep; REM latency is the time from sleep onset until the first REM period. Neither is interchangeable with the amount of deep sleep.
A full PSG is not needed for everyone who wants to understand sleep. Some home sleep apnea tests focus on respiratory signals without the complete EEG, EOG, and EMG required for standard stage scoring. A clinician chooses assessments according to the presenting problem. Sleep diaries and clinical history can be more appropriate for some complaints.
Sleep staging is a standardized interpretation, not a perfect measurement
Sleep-stage categories summarize continuously changing physiology. Boundaries can be ambiguous, and different trained scorers may disagree on difficult epochs. A systematic review and meta-analysis of scoring reliability documented substantial overall interrater agreement with important stage-specific variability. This is compatible with PSG being the clinical reference method; it is also a reason to avoid overstating second-by-second precision.
Are sleep trackers accurate about REM and deep sleep?
Most ordinary watches and rings estimate sleep stages from motion and optical cardiovascular signals, not from the complete physiological channels used for PSG. Their algorithms infer sleep states statistically, and accuracy differs across devices, populations, software versions, and sleep stages. Measuring total time asleep and classifying particular sleep stages are not identical challenges.
A 2024 systematic review of Fitbit Charge 4, Garmin Vivosmart 4, and WHOOP found variation and disagreement with reference measurements. A different 2024 systematic review of EEG-based wearable devices reported promising results for some brain-signal-based devices but also significant variation in validation and performance. EEG headbands and ordinary pulse-and-movement wristbands should therefore not be treated as equivalent technologies.
The AASM position statement on consumer sleep technology advises against using unvalidated consumer devices as diagnostic or therapeutic substitutes. Stage estimates may offer useful trends, but an isolated reading of “43 minutes deep sleep” cannot show that a person has a disorder or requires treatment. If tracking increases worry, a simpler attention to adequate sleep and daily functioning may be more constructive.
What can change or disrupt sleep architecture?
Prior sleep loss, irregular timing, shift work, environmental disturbances, and illness can alter sleep continuity and stage distribution. Alcohol, caffeine, and some medications also influence sleep, depending on timing, dose, and individual physiology. A change in the proportions does not identify its cause.
Sleep apnea can produce breathing-related arousals and interrupted sleep. Restless legs syndrome may delay falling asleep, while limb movements can occur during sleep. Parasomnias involve events or behaviors associated with particular transitions or sleep states. Chronic insomnia concerns difficulty initiating or maintaining sleep, or early waking, alongside relevant frequency, duration, and daytime consequences. None of these diagnoses can be established merely by looking at a consumer sleep-stage graph.
Repeated witnessed breathing pauses, gasping, injuries during dream-enacting movements, uncontrollable daytime sleepiness, or sleepiness while driving warrant clinical attention. Someone experiencing unsafe drowsiness should avoid driving until safe. Appropriate testing depends on the symptoms and a clinician's judgment, not on a prescribed percentage of REM or N3.
How to support healthy sleep without chasing stage scores
A sound general approach is to make adequate time for sleep and maintain a reasonably consistent schedule compatible with one's responsibilities. Daylight exposure, an appropriate sleep environment, and management of nighttime disruption may support sleep regularity. None guarantees a particular REM or deep-sleep percentage, and broad lifestyle suggestions are not treatment substitutes for clinically important sleep disorders.
Persistent insomnia may benefit from evidence-based cognitive behavioral therapy for insomnia (CBT-I). Suspected sleep apnea, circadian rhythm disorders, central hypersomnolence disorders, or harmful parasomnias require condition-specific assessment. Self-imposed sleep restriction or unsupervised changes to prescribed medications, melatonin, or bright-light treatment should not be used simply to manipulate a tracker output.
Sleep architecture is most helpful as a scientific account of how sleep is organized. Sleep duration, regularity, comfort, safety, and daytime functioning remain essential alongside the laboratory understanding of stages.
Frequently asked questions
What are the four stages of sleep?
N1, N2, N3, and REM are the four commonly named stages in contemporary adult scoring. N1–N3 are non-rapid eye movement sleep; REM is a separate state. Wakefulness is scored separately.
Is REM sleep the same as deep sleep?
No. Deep sleep usually means N3 slow-wave sleep. REM is a separate state with rapid eye movements and normally substantial muscle atonia.
Does a sleep cycle always last 90 minutes?
No. Ninety minutes is an approximate teaching value. Actual cycle lengths vary within and across nights, and the NHLBI describes commonly recurring cycles of about 80–100 minutes.
How many REM periods happen in a night?
Many sleepers have several REM episodes over approximately four to six cycles. The number and duration depend on total sleep and individual conditions; REM generally becomes more prominent later in the night.
Is N2 sleep wasted time?
No. N2 is a major normal component of sleep, with characteristic spindles and K-complexes and important ongoing brain physiology.
Can people dream during NREM sleep?
Yes. Laboratory awakenings document dreams after both REM and NREM. REM dream reports are often more vivid on average, but dreaming is not exclusive to REM.
How much deep sleep is normal as people age?
Deep sleep generally decreases across adulthood, but there is substantial variability. A single percentage, especially from a consumer tracker, cannot establish whether someone's sleep is clinically abnormal.
Can supplements or techniques reliably increase REM or deep sleep?
There is no universal validated supplement or timing protocol that guarantees healthier sleep by increasing an individual's stage percentage. Diagnosis-specific treatment and health conditions matter more than pursuing a generic device target.
Are brief awakenings between cycles normal?
Brief awakenings and arousals can occur in normal sleep, often without later recall. Recurrent long awakenings, significant distress, or impairment require separate assessment.
When should a sleep complaint be evaluated?
Seek professional assessment when sleep symptoms are persistent or impair daytime life, or when there are concerning behaviors such as repeated gasping, witnessed breathing pauses, harmful dream enactment, or severe daytime sleepiness. A sleep stage score alone does not diagnose these conditions.
Related Articles
Anxiety and Sleep: Insomnia, Hyperarousal, and the Anxiety-Sleep Cycle explains how heightened arousal relates to sleep complaints and difficulties initiating or maintaining sleep.
Sleep and Sugar Cravings: How Sleep Loss Can Change Appetite and Reward explores connections between sleep loss, behavior, and appetite regulation.
References
American Academy of Sleep Medicine. (2018). Consumer Sleep Technology: An American Academy of Sleep Medicine Position Statement. Journal of Clinical Sleep Medicine, 14(5), 877–880.
American Academy of Sleep Medicine. (2023). AASM releases updated version of scoring manual. American Academy of Sleep Medicine.
Daan, S., Beersma, D. G. M., & Borbély, A. A. (1984). Timing of human sleep: Recovery process gated by a circadian pacemaker. American Journal of Physiology, 246, R161–R183.
de Gans, C. J., et al. (2024). Sleep assessment using EEG-based wearables: A systematic review. Sleep Medicine Reviews, 76, 101951. DOI: 10.1016/j.smrv.2024.101951.
Kim, J., & Park, M. (2025). Systems memory consolidation during sleep: Oscillations, neuromodulators, and synaptic remodeling. BMB Reports, 58(10), 425–436. DOI: 10.5483/BMBRep.2025-0033.
Lee, Y. J., Lee, J. Y., Cho, J. H., & Choi, J. H. (2022). Interrater reliability of sleep stage scoring: A meta-analysis. Journal of Clinical Sleep Medicine, 18(1), 193–202. DOI: 10.5664/jcsm.9538.
National Heart, Lung, and Blood Institute. (2022). How Sleep Works: Sleep Phases and Stages. National Institutes of Health.
National Institute of Child Health and Human Development. What happens during sleep?. National Institutes of Health.
Rasch, B., & Born, J. (2013). About sleep's role in memory. Physiological Reviews, 93(2), 681–766.
Schyvens, A.-M., et al. (2024). Accuracy of Fitbit Charge 4, Garmin Vivosmart 4, and WHOOP versus polysomnography: Systematic review. JMIR mHealth and uHealth, 12, e52192.
Siclari, F., et al. (2017). The neural correlates of dreaming. Nature Neuroscience.
Siclari, F., et al. (2018). Dreaming in NREM sleep: A high-density EEG study of slow waves and spindles. Journal of Neuroscience.
Watson, N. F., et al. (2015). Recommended amount of sleep for a healthy adult: A joint consensus statement of the AASM and Sleep Research Society. Sleep, 38(6), 843–844.