The normal adult split: N1 ~5%, N2 ~50%, N3 deep ~13-23%, REM ~20-25%, cycling every ~90 min. With Ohayon normative ranges.
The ~90-minute cycle and the 4 stages (N1, N2, N3, REM)
Modern scoring (AASM) divides sleep into NREM stages N1, N2, N3 and REM, replacing the old Stages 1-4. A healthy adult cycles through them roughly every 90-110 minutes, completing 4-6 cycles per night (Carskadon & Dement). N1 is the brief drowsy transition (~5% of the night); N2 is the bulk of sleep (~45-55%), marked by sleep spindles and K-complexes; N3 is slow-wave/deep sleep; REM is the dreaming, low-muscle-tone stage. Cycle composition is not fixed: early cycles are N3-heavy, late cycles are REM-heavy.
N3 deep sleep (slow-wave): what it does and how much
N3 is defined by delta waves (0.5-2 Hz) occupying >20% of an epoch. It accounts for ~13-23% of total sleep time in healthy adults, weighted toward the first third of the night (Ohayon et al. 2004). Functionally it is the consolidation and clearance stage: slow oscillations and spindles coordinate hippocampal-to-cortical transfer of declarative memory (Rasch & Born 2013), and Xie et al. 2013 showed interstitial space expands ~60% during NREM sleep, accelerating glymphatic clearance of solutes including amyloid-beta. This is also when most growth-hormone pulse and the deepest arousal threshold occur.
REM sleep: what it does and how much
REM (rapid eye movement) runs ~20-25% of total sleep in healthy adults (Ohayon et al. 2004), with a desynchronized 'awake-like' EEG, vivid dreaming, and skeletal-muscle atonia. REM periods lengthen across the night: the first may last only ~10 minutes, the last (near morning) 30-60 minutes. It is implicated in emotional memory processing, procedural/motor learning, and overnight recalibration of emotional reactivity (Walker; Rasch & Born 2013). Selective REM deprivation produces a REM rebound on recovery nights, evidence the brain defends a REM quota.
How the stages shift across the night
Architecture is not uniform. Slow-wave (N3) pressure is highest at sleep onset and dissipates with each cycle, so deep sleep is concentrated in the first 2-3 hours. REM is reciprocally back-loaded: short and sparse early, long and dense in the final cycles, because REM propensity rides the circadian temperature/melatonin rhythm and peaks near the core-body-temperature minimum toward dawn (Carskadon & Dement). Practical consequence: cutting the night short (waking 2 hours early) costs you disproportionately more REM than deep sleep.
How much deep sleep and REM you actually need (Ohayon normative ranges)
The most-cited normative dataset is Ohayon et al. 2004, a meta-analysis of 65 polysomnography studies. For healthy adults it reports REM at ~20-25% and slow-wave (N3) at roughly 13-23% in young-to-middle adulthood, both declining with age, while N2 rises and wake-after-sleep-onset increases. For a 7-8 hour night that translates to roughly 70-100 min N3 and 90-120 min REM, though the band is wide and age-dependent. There is no single 'target number'; consistent total sleep and uninterrupted early-night sleep are what protect both fractions.
What wrecks your architecture (alcohol, caffeine, deprivation, age)
Alcohol shortens REM in the first half and fragments the second half with rebound REM and arousals (Ebrahim et al. 2013). Caffeine (half-life ~5-6 h) reduces total sleep, N3, and sleep efficiency when taken within ~6 hours of bed. Sleep deprivation and 'catch-up' fragmentation preferentially cost late-night REM. Aging is the dominant long-term driver: Ohayon et al. 2004 quantified a steady decline in slow-wave sleep across adulthood, with N3 minutes dropping markedly from the 20s through the 60s while WASO rises.
Questions logged on this protocol
How much deep sleep do I need?
For healthy adults, N3 (slow-wave/deep) sleep runs roughly 13-23% of total sleep time per the Ohayon et al. 2004 normative meta-analysis. On a 7-8 hour night that is about 70-100 minutes, concentrated in the first third of the night. The fraction declines with age, so a healthy 55-year-old may run lower than a 25-year-old at the same total sleep time. There is no validated 'minimum target'; protecting total sleep duration and the uninterrupted early-night window is what preserves N3, since slow-wave pressure is highest at sleep onset.
How much REM sleep do I need?
REM is about 20-25% of total sleep time in healthy adults (Ohayon et al. 2004), so roughly 90-120 minutes across a 7-8 hour night. Because REM is back-loaded toward morning and the last REM period can run 30-60 minutes, the final 1-2 hours of sleep contribute disproportionately to your REM total. Cutting sleep short by waking early removes mostly REM, not deep sleep. As with N3, there is no single target number to hit on a tracker; total sleep time and an unbroken late-night window are the variables that matter.
What is the difference between REM and deep sleep?
They are different stages with opposite EEG signatures and different jobs. Deep sleep (N3) shows high-amplitude slow delta waves (0.5-2 Hz), the highest arousal threshold, and drives declarative-memory consolidation and glymphatic clearance (Rasch & Born 2013; Xie et al. 2013). REM shows a fast, desynchronized 'awake-like' EEG, vivid dreaming, and skeletal-muscle atonia, and supports emotional and procedural processing. Timing differs too: N3 dominates the first third of the night; REM dominates the last third. They are not interchangeable, and a night can be deficient in one while normal in the other.
Why do I get more REM toward morning?
Two systems stack. First, slow-wave (N3) pressure is highest at sleep onset and dissipates with each cycle, freeing later cycles for REM. Second, REM propensity is gated by the circadian clock and peaks near the core-body-temperature minimum, which falls in the last hours before habitual wake (Carskadon & Dement). So the final cycles are both released from N3 competition and riding the circadian REM peak, which is why your longest, most vivid dreams and your easiest awakenings cluster toward morning.
Can you increase deep sleep?
Partly, and mostly by not suppressing it rather than by 'boosting' it. The reliable levers are: keep total sleep adequate (deep sleep is taken first when sleep is short), avoid alcohol and large late meals, and keep caffeine outside a ~6-hour pre-bed window. Acute sleep deprivation produces a slow-wave rebound on the recovery night, evidence N3 is homeostatically defended. [VERIFY: claims that consumer interventions such as cooling mattresses, exercise timing, or auditory closed-loop slow-oscillation stimulation reliably raise N3 in healthy adults are mixed and mostly small-sample; treat 'deep sleep boosters' skeptically.]
Do sleep trackers (Oura, Whoop) measure sleep stages accurately?
They estimate stages from heart rate, heart-rate variability, movement, and temperature, not from EEG, so stage-level output is an inference. Against polysomnography, consumer wearables are generally good at total sleep time and reasonable at distinguishing sleep from wake, but markedly less accurate at classifying specific stages, with deep-sleep and REM estimates the weakest. [VERIFY: published validation accuracy varies by device and firmware version; epoch-by-epoch stage agreement for N3/REM is commonly reported well below wake/sleep agreement, but exact figures differ across studies and update frequently.] Use trends across weeks, not a single night's stage breakdown.
- [01]
- [02]
- [03]
- [04]
- [05]
Other stacks in this hub
Why You Wake at 3am
The cortisol awakening response, the two-process model, and the actual causes of mid-sleep awakening, with the protocols.
Sleep Onset Latency: What It Is, Why It Matters, How to Measure
Sleep onset latency (SOL) is the time from lights-out to the first epoch of N1. Why under-15-minute SOL is normal, what longer SOL signals, and how the actigraphy estimate misses by 10–20 minutes.
Caffeine and Sleep: Half
Caffeine's ~5-6h half-life, the ~10-12h quarter-life, and why the last dose belongs 8-10 hours before bed.