Sleep Stages: What NREM and REM Actually Are

This content is educational. It describes what sleep researchers measure when they study sleep architecture. It is not medical advice.

Why this matters

Almost every consumer wearable now reports sleep stages — "deep sleep," "REM," "light." The numbers look authoritative. The underlying science is genuinely interesting. What gets lost in the watch face is what each stage actually is, what research has measured about its function, and how reliably any of it can be detected without a sleep lab.

This page describes the structure of a night's sleep as researchers measure it. It is foundational background for understanding the rest of the sleep literature.

The basic model

Sleep is not one state. It is a cycle of physiologically distinct stages that recur through the night. The two broad categories that researchers describe:

• NREM sleep (Non-Rapid Eye Movement) — divided into three stages of increasing depth (N1, N2, N3). Brain activity slows, body temperature drops, and physical restoration processes intensify.
• REM sleep (Rapid Eye Movement) — characterised by fast brain activity (similar to wakefulness), rapid eye movements, and skeletal muscle paralysis. The primary stage for vivid dreaming.

A typical night cycles through these stages approximately every 90 minutes. The proportions shift across the night: early cycles contain more deep NREM (N3); later cycles contain more REM.

What researchers measure to identify stages

The gold standard is polysomnography (PSG) — the multi-channel recording used in sleep labs that we described in the sleep apnea diagnosis page. PSG identifies sleep stages using three primary signals:

• EEG (electroencephalogram) — brain electrical activity, the foundational signal for staging
• EOG (electrooculogram) — eye movements, used to identify REM and wake
• EMG (electromyogram) — muscle tone, used to confirm REM (muscle atonia) and distinguish from wake

Sleep technologists score the recording in 30-second epochs using rules published by the American Academy of Sleep Medicine [Berry et al. 2017]. Each epoch is assigned a stage based on the dominant patterns in that window.

This methodology is precise but expensive. Without these three signals — particularly EEG — stage classification becomes approximate.

The four stages described in the literature

N1 (light sleep, transitional)

The transition from wakefulness to sleep. EEG shows reduced alpha rhythm (the dominant pattern during relaxed wakefulness) and increased theta activity. Muscle tone reduces slightly. People woken from N1 often report they "weren't really asleep yet."

Typical proportion: 5-10% of total sleep Duration of episode: 1-7 minutes Function described in research: transitional state; minimal restorative function established Memorable feature: the brief muscle jerk many people experience just as they fall asleep (hypnic jerk) often occurs during N1

N2 (most of your sleep)

The largest single stage by total time. Characterised by two distinctive EEG features:

• Sleep spindles — brief bursts of 12-14 Hz activity lasting ~0.5 seconds
• K-complexes — high-amplitude negative-then-positive waves, often in response to external stimuli

Heart rate slows, body temperature drops, breathing becomes regular. Research has associated N2 with memory consolidation processes — particularly procedural memory and motor learning [Walker 2017; Diekelmann & Born 2010].

Typical proportion: 45-55% of total sleep Duration of episode: 10-25 minutes Function described in research: memory consolidation (procedural and declarative), sleep maintenance

N3 (deep sleep, slow-wave sleep)

What consumer apps call "deep sleep." EEG shows large, slow delta waves (0.5-2 Hz, high amplitude). The brain is at its furthest from wakefulness during N3 — the hardest stage to wake someone from, and the stage where waking up leaves people most disoriented.

Research has linked N3 to several restorative processes:

• Growth hormone secretion peaks during N3 [Van Cauter et al. 2000]
• Glymphatic clearance — the brain's waste-removal system — appears most active [Xie et al. 2013]
• Slow-wave activity correlates with subjective restoration after sleep [Bonnet & Arand 2003]

Typical proportion: 13-23% of total sleep, concentrated in the first half of the night Duration of episode: 20-40 minutes (longer in early cycles) Function described in research: physical restoration, growth hormone release, brain waste clearance, slow-wave-dependent memory consolidation

REM (rapid eye movement sleep)

A paradoxical state: EEG activity resembles wakefulness, but muscle tone is at its lowest point of the night (skeletal muscle atonia). Eye movements are rapid and frequent. Breathing and heart rate become irregular. Most vivid, narrative dreaming occurs during REM, though dreams also occur in NREM stages.

Research has associated REM with:

• Emotional processing and consolidation of emotional memories [Goldstein & Walker 2014]
• Creative problem-solving — REM-rich sleep is associated with better performance on certain creative tasks [Cai et al. 2009]
• Brain development in infants (newborns spend ~50% of sleep in REM; adults ~25%)

The muscle atonia of REM is functional: it prevents physical enactment of dream content. When this mechanism fails, the result is REM Behaviour Disorder (RBD) — a clinical condition where people physically act out their dreams, sometimes with self-injury [Schenck & Mahowald 2002]. RBD also has clinical significance because it is strongly associated with later development of Parkinson's disease and related synucleinopathies.

Typical proportion: 20-25% of total sleep, increasing across the night Duration of episode: 10-60 minutes (longest in late-night cycles) Function described in research: emotional memory consolidation, brain development, creative consolidation

How stages cycle through the night

A typical adult night follows a recognisable pattern:

The implication: deep sleep is concentrated in the first half of the night; REM dominates the second half. Cutting sleep short by 2-3 hours disproportionately reduces REM (and therefore emotional processing); cutting the first half short reduces deep sleep (and therefore physical restoration).

What changes the architecture

Several variables affect the proportions of each stage:

• Age — N3 declines steadily with age. By age 60, N3 may account for less than 5% of total sleep [Ohayon et al. 2004]
• Sleep deprivation — recovery sleep prioritises N3 first, then REM, often at the expense of N2
• Alcohol — increases initial N3 in the first half of the night, suppresses REM in the second half (covered in sleep hub research)
• Caffeine — suppresses N3 and reduces total sleep time
• Antidepressants — many SSRIs and SNRIs significantly suppress REM
• Sleep medications — benzodiazepines and z-drugs increase N2 at the expense of N3
• Sleep disorders — sleep apnea fragments sleep, particularly affecting REM and N3
• Pregnancy — sleep architecture shifts significantly, with reduced N3 and N2

What wearables can and can't measure

This is where consumer products overpromise. We covered this in detail in the wearables piece, but specific to sleep stages:

• Wearables don't measure EEG. They infer stages from heart rate patterns and movement
• Total sleep time estimation is reasonable (within ~30 minutes of PSG)
• Sleep onset detection is reasonable (within ~20 minutes)
• Specific stage classification has wide error margins — Apple Watch agrees with PSG on stages 50-60% of the time; Oura and Whoop are similar

Practical translation: trust the broad night-by-night totals from your wearable. Don't read too much into single-night stage breakdowns ("I only got 12 minutes of deep sleep") — that number is more approximate than the watch face suggests.

What is reasonable: tracking trends in your own data over weeks. Comparing your average "deep sleep" estimates across months can show real changes, even if the individual numbers are noisy.

What clinical sleep disorders affect stage architecture

Several conditions show characteristic abnormalities in sleep architecture, identifiable on PSG:

• Obstructive sleep apnea — fragmented sleep with frequent micro-arousals, reduced N3 and REM
• Insomnia — increased N1, reduced sleep efficiency
• Narcolepsy — abnormal REM regulation; REM may occur within minutes of sleep onset
• Depression — characteristically shortened REM latency, longer initial REM episodes
• REM Behaviour Disorder — preserved muscle tone during REM
• Periodic Limb Movement Disorder — repetitive limb movements fragmenting sleep

If a clinician suspects any of these based on symptoms, PSG is the diagnostic standard. Wearable data is not adequate for diagnosis.

Related Proco pages

• How sleep apnea is diagnosed
• Wearables: what they can and can't measure
• How to read a clinical trial

Sources

Proco provides educational, research-based information. This page describes sleep physiology as researchers measure it. If you experience persistent sleep difficulty, consult a sleep medicine specialist.

Schema (for implementation)

{
  "@context": "https://schema.org",
  "@type": "MedicalWebPage",
  "headline": "Sleep Stages: What NREM and REM Actually Are",
  "description": "Sleep is not a single state but a cycle of physiologically distinct stages — N1, N2, N3 (deep sleep), and REM. This page describes what researchers measure to identify each stage and what each does.",
  "datePublished": "2026-06-01",
  "dateModified": "2026-05-31",
  "author": {"@type": "Organization", "name": "Proco"},
  "publisher": {"@type": "Organization", "name": "Proco", "url": "https://procohq.com"},
  "medicalAudience": "Patient",
  "about": {"@type": "Thing", "name": "Sleep architecture and sleep stages"}
}