What Are Sleep Stages (Awake, Light, Deep, REM)?

Sleep stages are the distinct phases the body cycles through during a night's sleep — commonly grouped as awake, light sleep, deep sleep, and REM. In a clinical sleep lab these stages are defined by brain activity (EEG) alongside eye movement and muscle tone, but the sleep stages you see in a consumer wearable are estimated from movement and heart-rate patterns, not measured from brainwaves. Treat them as a directional wellness signal — useful for spotting trends across nights, not as clinical fact for any single night.

General information for developers, not medical advice. Nothing here is a diagnosis or a clinical measurement.

The four stages, defined

Clinically, sleep is scored into stages that repeat in roughly 90-minute cycles through the night:

• Awake. Periods of wakefulness, including brief arousals after you first fall asleep.
• Light sleep (clinically N1 and N2). The transitional and predominant stages; the largest share of a typical night. Heart rate and breathing slow, and you are relatively easy to wake.
• Deep sleep (N3, slow-wave sleep). Often described as the most physically restorative stage, marked by slow brainwaves. Hardest to wake from; tends to concentrate earlier in the night.
• REM (rapid eye movement). Associated with vivid dreaming, rapid eye movements, and relative muscle atonia. REM periods generally lengthen toward the morning.

Stage proportions vary by person, age, and night, so there is no single "correct" distribution to target.

How wearables estimate stages (the key point)

A clinical sleep study — polysomnography (PSG) — is the gold standard. It reads brainwaves (EEG), eye movement (EOG), and muscle tone (EMG) with electrodes and a trained technician scores the night.

Consumer wearables do not read brainwaves. They infer stages from proxy signals:

• Movement from an accelerometer (actigraphy).
• Heart rate and heart-rate variability (HRV) patterns.
• On some devices, respiration, skin temperature, or blood-oxygen (SpO2) as additional inputs.

A proprietary algorithm maps those signals to a probable stage minute by minute. That makes every stage label a model estimate/inference, not an EEG measurement — a distinction that matters when you present the number to a user.

Accuracy vs. a clinical sleep study — hedge this

This is the honest part that developers most need to internalize. Validation research comparing wearables to PSG finds a consistent pattern:

• Sleep vs. wake detection is fairly good — often around 90% or better at simply telling asleep from awake.
• Four-stage classification is much weaker — studies commonly report roughly 60–70% overall agreement with PSG when distinguishing light, deep, and REM.
• REM is often underestimated, and light-vs-deep is frequently misclassified (some devices overestimate light and underestimate deep).

Because of this, stage-level numbers are best read as rough directional trends over many nights, not per-night clinical truth. Accuracy also differs substantially between vendors and even between firmware versions of the same device, so the same night's sleep can be scored differently by two brands. Verify current per-device validation figures against a primary source before you publish any specific accuracy claim — they shift with study, population, and firmware.

Why this matters to a builder

If your app surfaces sleep stages, the framing you choose is a product and trust decision:

• Present stages as trends, not verdicts. "Your deep sleep has trended lower this week" is defensible; "You didn't get enough deep sleep last night" overstates what the data supports.
• Don't compare across brands. Different sensors, windows, and algorithms mean an Oura "deep sleep" minute is not interchangeable with a Fitbit one. Avoid implying portability of exact figures.
• Keep language non-diagnostic. Sleep-stage estimates are a general wellness signal, not a screen for any sleep disorder. Avoid clinical or prescriptive copy, and add a "general information, not medical advice" note wherever you show the numbers.

How sleep data appears in a wearable API

Most wearable and aggregator APIs return a nightly sleep object with two things: per-stage totals and a timeline of stage segments. The exact field names and units differ by vendor (seconds vs. minutes, snake_case vs. camelCase), so always confirm against the specific docs — but the shape is broadly consistent:

{
  "date": "2026-07-12",
  "time_in_bed": 28800,
  "total_sleep": 26400,
  "awake_duration": 2400,
  "light_sleep_duration": 14100,
  "deep_sleep_duration": 5400,
  "rem_sleep_duration": 6900,
  "sleep_score": 78,
  "stages": [
    { "stage": "awake", "start": "2026-07-12T23:10:00Z", "end": "2026-07-12T23:22:00Z" },
    { "stage": "light", "start": "2026-07-12T23:22:00Z", "end": "2026-07-12T23:58:00Z" },
    { "stage": "deep",  "start": "2026-07-12T23:58:00Z", "end": "2026-07-13T00:41:00Z" },
    { "stage": "rem",   "start": "2026-07-13T00:41:00Z", "end": "2026-07-13T01:19:00Z" }
  ]
}

Durations (here in seconds) let you show a summary; the stages array (sometimes called levels) gives you the hypnogram-style timeline of {stage, start, end} segments. Many APIs also include a summary sleep_score or efficiency. Field names, units, and aggregation windows vary — confirm them per API.

Measured vs. estimated, and individual variation

To be explicit: sleep stages are an estimate, produced by a model from movement and heart-rate signals — not a measurement of brain activity. They are directionally useful for tracking a person's own patterns over time (total sleep, consistency, rough time in deep and REM as trends), and far less reliable as a precise, per-night stage breakdown. Individual physiology, age, and night-to-night variation all move the numbers, so compare a person to their own history rather than to a population "ideal."

Where this fits

This page defines the concept. When you are ready to work with the data:

• To understand the APIs that deliver sleep and other wearable metrics, see wearable data APIs.
• For a hands-on integration that returns sleep stages, see the Oura API integration guide — Oura is a multi-sensor sleep tracker whose API exposes exactly the per-stage, timestamped structure described above.

General information for developers, not medical advice. Consumer sleep-stage estimates are a general wellness signal, not a clinical measurement or a diagnosis. Accuracy figures and field names above are illustrative and change with firmware and vendor — verify against current primary sources and API docs before you rely on them.