Sources & Methodology

The science in this article draws on the American Academy of Sleep Medicine (AASM) official scoring manual (Berry et al., 2012), standard sleep physiology references (Carskadon & Dement, 2011), published normative sleep parameter data (Ohayon et al., 2004), and independent consumer tracker validation research (Chinoy et al., 2021). GreatHealthGear does not conduct clinical research. Where the science is contested or preliminary, this is noted explicitly.

Sleep stage terminology follows the AASM R&K system (2007 revision), which replaced the original Rechtschaffen & Kales (1968) classification. Some older literature uses different stage numbering; this article uses the current N1/N2/N3/REM system.

The direct answer: Sleep is structured into four stages that cycle through the night in approximately 90-minute units. N3 (deep sleep) dominates early cycles and is critical for physical recovery; REM dominates later cycles and is critical for memory and emotional processing; N1 and N2 form the transitions between them. A healthy adult night typically includes 15–25% N3 and 20–25% REM, with these proportions declining with age.

The Four Stages of Sleep

N1 — The Transition Stage

N1 is the lightest stage of sleep, occupying the transition between wakefulness and deeper sleep. In EEG recordings, it is characterised by a shift from waking alpha waves to slower theta waves. N1 typically lasts only a few minutes and is easily disrupted — a noise or light stimulus during N1 usually wakes the sleeper without the person having felt genuinely asleep.

N1 accounts for approximately 2–5% of total sleep time in healthy adults. Consumer trackers often classify extended N1 as “light sleep” or “awake,” which contributes to variation between devices and clinical measurements.

N2 — Stable Light Sleep

N2 is the most abundant single stage, accounting for approximately 45–55% of total sleep in healthy adults. It is characterised by two distinctive EEG features: sleep spindles (bursts of 12–15 Hz oscillations lasting 0.5–3 seconds) and K-complexes (sharp, high-amplitude waveforms).

Sleep spindles during N2 are associated with procedural memory consolidation — the encoding of skills and sequences. Research suggests that the density of sleep spindles correlates with learning performance on motor tasks. N2 is not a “wasted” state between deep sleep and REM; it is physiologically productive.

Consumer trackers typically group N1 and N2 together as “light sleep.” The distinction between them is only visible via EEG — optical and movement sensors cannot reliably differentiate N1 from N2.

N3 — Deep Sleep (Slow-Wave Sleep)

N3 is the most physically restorative stage. It is defined in clinical EEG by the presence of delta waves — high-amplitude, low-frequency (0.5–4 Hz) oscillations occupying at least 20% of the epoch being scored. The AASM scoring manual (Berry et al., 2012) sets this as the threshold for N3 classification.

During N3, the body does its most intensive physical repair work:

  • Growth hormone secretion peaks during the first slow-wave sleep period
  • Immune function consolidation — cytokines are produced and memory T-cells are produced during deep sleep
  • Glymphatic clearance — cerebrospinal fluid flows through the brain, flushing metabolic waste (including tau protein and amyloid-beta, associated with Alzheimer’s disease in accumulation studies)
  • Memory consolidation of declarative memories (facts and events)

N3 is concentrated heavily in the first half of the night. According to Ohayon et al. (2004), in adults aged 20–30, N3 accounts for 15–25% of total sleep time. This proportion declines with age — adults over 60 typically show 5–10% N3, and some elderly individuals show very little measurable slow-wave activity in clinical EEG.

Ohayon et al. (2004), in a meta-analysis of 65 studies covering 3,577 healthy individuals from childhood to old age, found that slow-wave sleep (N3) declines progressively from adolescence through adulthood, with the most dramatic decline occurring between ages 20 and 60. By age 60, the average adult's N3 proportion is approximately half what it was at age 25. This is a normal age-related change, not a disorder.

REM Sleep

REM sleep is physiologically distinct from all other stages. Brain activity in REM closely resembles wakefulness — the EEG shows a low-amplitude, mixed-frequency pattern similar to the waking brain. What distinguishes it is muscle atonia (near-complete paralysis of voluntary muscles) and the rapid eye movements visible beneath closed eyelids.

The dominant theory of REM function relates to emotional memory processing. Research from the Walker group and others suggests that REM sleep provides a kind of “emotional recalibration” — the brain replays emotional memories in a neurochemical environment low in noradrenaline, which is thought to reduce the emotional charge of difficult experiences over time. This is why REM-deprived individuals show increased emotional reactivity.

REM sleep also supports the integration of complex information — pattern recognition, creative connections between concepts, and the consolidation of learning that occurred during the previous day.

REM sleep is distributed differently across the night from N3. According to Carskadon & Dement (2011), REM periods are short (10–15 minutes) in the first sleep cycles and lengthen progressively, with the final REM period of the night often lasting 45–60 minutes. This is why the final hour of an 8-hour sleep period contains disproportionately large amounts of REM — and why an alarm clock set an hour early disproportionately cuts REM rather than total sleep.

How Sleep Cycles Work

Sleep does not progress from N1 to N2 to N3 to REM once and stay in REM. It cycles.

A typical adult night consists of 4–6 complete sleep cycles of approximately 90 minutes each. Each cycle contains elements of N1, N2, N3 (or a reduced version of it), and REM, but the proportions shift across the night:

  • Cycles 1–2 (hours 1–3): Long, deep N3 periods; short, shallow REM periods
  • Cycles 3–4 (hours 3–6): Shorter N3; longer, more intense REM periods
  • Cycle 5–6 (hours 6–8): Minimal N3; long, extended REM periods

This is why “making up” lost early-night sleep with late morning sleeping preferentially restores REM rather than deep sleep — the morning hours are architecturally weighted toward REM.

If you go to bed 1 hour later than usual, you compress your first deep sleep cycle. If you wake 1 hour earlier than usual, you cut your final REM cycle. Both are disruptive, but in different ways. Consistent bedtimes protect deep sleep; consistent wake times protect REM.

How Age Changes Sleep Architecture

Sleep architecture is not static across the lifespan. The most significant changes occur in:

Slow-wave sleep (N3): Substantial decline from early adulthood through age 60, then relative stability. According to Ohayon et al. (2004), N3 decreases by approximately 2% per decade from age 20 to 60.

REM sleep: More stable across adulthood than N3, but declines modestly in later life (after 65). The proportion of REM relative to total sleep time remains relatively consistent from age 20 to 50, then declines slowly.

Total sleep time: Gradual reduction from approximately 7.5–8 hours in young adults to 6.5–7 hours by age 65 in published population studies.

Sleep efficiency: (Time asleep vs. time in bed) declines with age due to increased wake time after sleep onset. More frequent awakenings and longer time to return to sleep become common.

If you are over 60 and your tracker consistently shows minimal N3, this may be age-expected rather than a sleep disorder — but it is worth discussing with a doctor if accompanied by daytime fatigue, cognitive complaints, or mood changes.

What Consumer Trackers Actually Measure

Consumer sleep trackers — whether rings (Oura Ring 4, Samsung Galaxy Ring) or smartwatches (Garmin Venu 3, Fitbit Sense 2) — do not measure brain waves. They measure movement (accelerometry), heart rate (photoplethysmography), and sometimes skin temperature. Sleep stage classification is inferred from these proxy signals.

Chinoy et al. (2021), in a validation study published in npj Digital Medicine, compared seven consumer devices to clinical PSG. Key findings:

  • Sleep duration: Most consumer devices performed well (within 15–25 minutes of PSG in most participants)
  • Deep sleep detection: Moderate sensitivity — devices correctly identified approximately 50–70% of clinically measured N3 epochs
  • REM detection: Most problematic — devices systematically overestimated REM sleep duration by approximately 15–25 minutes on average
  • Accuracy varied by device: Ring-based devices (with finger-based PPG) generally outperformed wrist-based trackers
Chinoy et al. (2021) — the most comprehensive independent consumer sleep tracker validation study at time of publication — found that no consumer device achieved clinical-grade sleep staging accuracy. However, the paper noted that devices performed well enough for "monitoring sleep duration and detecting broad changes in sleep patterns" — which is precisely what most consumer users need.

The practical implication: trust your tracker’s sleep duration data. Treat stage percentages as directional trend indicators, not clinical measurements. Compare your own night-to-night and week-to-week data, not your percentages against clinical reference values.

What This Means for You

Understanding sleep architecture changes how you interpret and act on tracker data.

Deep sleep percentage: If yours is consistently below 10–12%, check for the main suppressors first: alcohol within 3 hours of sleep, inconsistent bed timing, and bedroom temperature above 20°C. If these are already well-managed and the low readings persist with daytime fatigue, speak to a doctor.

REM percentage: If yours appears low on your tracker, remember the overestimation bias — the actual percentage may be close to normal. REM is most reliably protected by not cutting your final sleep cycle short (alarm setting and consistent bedtime).

Total sleep duration: This is your most reliable metric. If the tracker consistently shows under 6.5 hours across the week, this is a signal worth taking seriously — the sleep science on chronic short sleep is substantial and consistent.

Age context: If you are over 50 and concerned about declining deep sleep percentages, check the age-normative data. Some decline is expected and does not indicate pathology.

Further Reading