Sleep Stages

Abstract

The sleep stages, which MeSH classifies under sleep, are the distinct, recurring phases the brain passes through in a night, each defined by a characteristic pattern of electrical activity, eye movement, and muscle tone. They divide into non-REM sleep — the light transition of N1, the spindle-bearing N2, and the slow waves of N3 — and REM sleep, the wake-like stage of rapid eye movements and muscle atonia. The stages succeed one another in cycles of roughly ninety minutes, with deep non-REM concentrated early and REM lengthening toward morning. This article traces their discovery and scoring, the physiology behind their hallmark rhythms, the architecture they compose, and their roles in memory. Three interactive demonstrations make the defining waveforms, the night's architecture, and the lifespan shift in stage proportions explicit.

Keywords: sleep stages, non-REM sleep, slow-wave sleep, sleep spindles, sleep architecture

Sleep is not the uniform blank it long appeared to be. Across a night the brain moves through an orderly succession of states, each with its own electrical signature, and the mapping of those states is what turned sleep into a measurable, structured process (Loomis, Harvey, & Hobart, 1937). A sleeper descends from drowsy wakefulness into progressively deeper non-REM sleep, rises into a REM episode, and repeats the descent four or five times before morning. The stages are not arbitrary waypoints: each is defined by a convergence of signs recorded together on the polysomnogram, and the sequence in which they occur, and the proportions in which they fall, carry the whole functional story of why sleep is organized the way it is.

What the Sleep Stages Are

The sleep stages are the categories into which a night of sleep is divided when it is scored from physiological recording. The recording is the polysomnogram, which combines the electroencephalogram (EEG) measuring cortical activity, the electrooculogram measuring eye movement, and the electromyogram measuring muscle tone; it is the joint pattern across these three channels, read in thirty-second epochs, that assigns each moment of the night to a stage.

At the broadest level the stages divide in two. Non-REM sleep comprises the quiet stages, in which the EEG slows and deepens and the body retains some muscle tone, and it is itself graded by depth into three stages. REM sleep, set against the whole of non-REM, is the paradoxical stage whose fast, low-voltage EEG resembles waking even as the skeletal muscles fall into near-complete atonia. The distinction matters because the two kinds of sleep are generated by different mechanisms, serve partly different functions, and are distributed differently across the night: no single axis of depth runs smoothly from light sleep to REM, because REM is not the deepest non-REM but a categorically different state. The sections that follow take the stages apart in order, beginning with how the MeSH vocabulary files them.

Types of Sleep Stages

In the Medical Subject Headings, sleep stages sit directly beneath the broader heading of sleep, and the classification enumerates three narrower descriptors directly under it (Table 1). MeSH is an indexing vocabulary built for retrieving the biomedical literature, not a physiological taxonomy, so this formal tree is deliberately coarse: it lists the descriptors that have earned their own indexing terms and is silent on the finer stage divisions — N1, N2, N3 — that clinical scoring actually uses, which the sections below develop. The three MeSH children are named here as the classification files them; of the three, only REM sleep is yet a separate article on this site and so is the only one linked.

These descriptors are not mutually exclusive levels of one scale: REM sleep and slow-wave sleep are opposite poles of the night's electrical activity, while sleepiness is a drive state rather than a stage of the recording at all. The practical division of non-REM into the graded stages N1, N2, and N3 — which the scoring manuals use and the rest of this article follows — cuts finer than the MeSH tree, which carries only the single descriptor slow-wave sleep for the deepest of them.

Measuring the Stages

The stages exist as categories because the electroencephalogram made them visible. Alfred Loomis and colleagues, recording human brain potentials through a night in the 1930s, were the first to show that sleep passes through several distinct electrical states and to letter them A through E, establishing that sleep deepens in discrete steps rather than smoothly (Loomis, Harvey, & Hobart, 1937). The decisive addition came two decades later, when Eugene Aserinsky and Nathaniel Kleitman discovered that periodic bursts of rapid eye movement accompanied a fast, activated EEG and dream reports, identifying REM as a stage categorically unlike the quiet states around it (Aserinsky & Kleitman, 1953). William Dement and Kleitman then showed that REM and non-REM alternate in a regular cycle through the night, giving the stages their temporal structure (Dement & Kleitman, 1957).

For the stages to be studied across laboratories they had to be scored the same way everywhere. Allan Rechtschaffen and Anthony Kales edited the 1968 manual that fixed the rules — the R&K standard — defining wakefulness, four non-REM stages numbered 1 through 4, and REM, each by explicit EEG, eye-movement, and muscle-tone criteria read in fixed epochs (Rechtschaffen & Kales, 1968). The standard held for nearly forty years until the American Academy of Sleep Medicine revised it, merging the two deepest non-REM stages into a single stage N3 and renaming the series N1, N2, and N3 to reflect that stages 3 and 4 differed only in the quantity of slow-wave activity, not in kind (Iber, Ancoli-Israel, Chesson, & Quan, 2007). The modern scheme is therefore wakefulness, N1, N2, N3, and REM. The first demonstration shows the hallmark EEG waveform that defines each of these stages.

Each stage carries its own signature. Stage N1, the light transition from wakefulness, shows a low-voltage mixed-frequency trace as the waking alpha rhythm breaks up. Stage N2 is marked by two transient events on a background of theta activity: the sleep spindle, a brief burst of waxing-and-waning oscillation around 11 to 16 hertz, and the K-complex, a large biphasic wave. Stage N3, slow-wave sleep, is defined by high-voltage delta waves below about 4 hertz filling at least a fifth of the epoch. REM returns the EEG to a fast, low-voltage, wake-like pattern, punctuated by sawtooth waves and the rapid eye movements that name it.

The Architecture of the Night

The stages are not scattered at random but arranged in a repeating structure. A night of sleep is organized into cycles of about ninety minutes, each descending from N1 through N2 into the slow waves of N3 and then ascending back through lighter sleep into a REM episode before the next cycle begins (Dement & Kleitman, 1957). The cycles are not identical, and the way they differ is the central fact of sleep architecture: deep N3 sleep dominates the early cycles and wanes as the night proceeds, while REM episodes are brief at first and lengthen toward morning. A sleeper therefore does most of their slow-wave sleep in the first few hours and most of their REM in the last few. Table 2 sets out the defining features of the modern stages, and Figure 1 shows the hypnogram that charts their succession.

Figure 1

A hypnogram of one night, showing deep N3 sleep concentrated in the early cycles and REM episodes lengthening toward morning.

What governs this architecture is the interaction of two drives. Alexander Borbély's two-process model holds that a homeostatic pressure for sleep (Process S) builds during waking and discharges during sleep, especially through slow-wave activity, while a circadian process (Process C) sets the timing; the depth and distribution of the stages across the night reflect the two processes working together (Borbély, Daan, Wirz-Justice, & Deboer, 2016). The early dominance of N3 is the signature of a sleep pressure that is highest at sleep onset and dissipates as the night proceeds. The second demonstration lets the night vary in length and charts the resulting hypnogram, reporting how the stages redistribute as sleep is cut short or extended.

The Physiology of the Stages

The hallmark rhythms that distinguish the stages are not cortical accidents but the output of a thalamocortical system that switches its firing mode with the depth of sleep. Mircea Steriade and colleagues showed that as the brain descends from waking into deep sleep, thalamic and cortical neurons shift from the tonic firing of the alert state into a synchronized bursting mode, and it is this synchronization that generates the spindles and slow oscillations seen on the EEG (Steriade, McCormick, & Sejnowski, 1993). The sleep spindle of stage N2 arises from the reticular nucleus of the thalamus pacing thalamocortical cells into rhythmic bursts around 11 to 16 hertz, while the slow oscillation of stage N3 reflects near-synchronous alternations between depolarized up states and silent down states across the cortex.

The spindle has become a focus in its own right because it indexes the integrity and function of N2 sleep. Luigi De Gennaro and Michele Ferrara's review set out the spindle's defining parameters — its frequency, topography, and density — and its sensitivity to age and pathology (De Gennaro & Ferrara, 2003). More recent work by Laura Fernandez and Anita Lüthi has detailed the thalamic circuit mechanisms that generate spindles and the evidence that they both protect sleep from arousal and support the consolidation of memory, tying the microstructure of a single stage to sleep's larger functions (Fernandez & Lüthi, 2020). The circuitry that assembles and switches between whole stages, in turn, is a brainstem and hypothalamic system of mutually inhibiting wake- and sleep-promoting populations whose transitions produce the orderly succession of states (Scammell, Arrigoni, & Lipton, 2017).

The Functions of the Stages

That sleep is divided into stages at all suggests they do different work, and the strongest evidence for a division of labour concerns memory. Björn Rasch and Jan Born assembled the case that sleep actively consolidates memory rather than merely shielding it from interference, and that the non-REM and REM stages contribute differently, with slow-wave sleep particularly implicated in the consolidation of declarative memory (Rasch & Born, 2013). The mechanism they describe is an active-systems consolidation in which the slow oscillations of N3, the thalamic spindles of N2, and hippocampal sharp-wave ripples are nested in time so that newly encoded memories are repeatedly reactivated and redistributed to the cortex during deep sleep (Klinzing, Niethard, & Born, 2019). On this account the stages are not interchangeable: the spindle-rich and slow-wave-rich stages carry a specific computational role that the architecture of the night is arranged to deliver.

The proportions of the stages are not fixed but change profoundly across the lifespan. Howard Roffwarg, Joseph Muzio, and William Dement documented that newborns spend about half their long sleep in REM and a large share in deep non-REM, fractions that fall steeply through childhood toward the adult distribution and continue to decline into old age, when slow-wave sleep in particular becomes scarce (Roffwarg, Muzio, & Dement, 1966). This developmental trajectory is itself evidence about function: the abundance of REM and slow-wave sleep in early life points to a role in building and shaping the nervous system that differs from whatever the stages do in the mature brain. Jerome Siegel's comparative work cautions, however, against assuming any single universal function, documenting how widely the amount and structure of the stages vary across species as a function of ecology rather than of a common computational need (Siegel, 2005). The third demonstration shows how the proportions of the stages shift across the human lifespan.

Worked Example

Because the stages fall in fixed cycles but in shifting proportions, the makeup of a night is worth computing explicitly, and the second demonstration reproduces the arithmetic. Consider a typical eight-hour night of 480 minutes organized into five cycles. In a healthy adult the stages fall roughly as N1 5 percent, N2 50 percent, N3 20 percent, and REM 25 percent. Converting to minutes: N1 is 0.05 times 480, or 24 minutes; N2 is 0.50 times 480, or 240 minutes; N3 is 0.20 times 480, or 96 minutes; and REM is 0.25 times 480, or 120 minutes. The four sum to 24 plus 240 plus 96 plus 120, which is 480 minutes — the whole night, as they must.

Now look at how those totals are distributed across the five cycles rather than spread evenly. Slow-wave N3 is front-loaded: suppose its 96 minutes fall as 45, 30, 15, 6, and 0 across the five cycles, so that the first two cycles alone carry 75 of the 96 minutes, about 78 percent of the night's N3. REM is the mirror image: suppose its 120 minutes fall as 10, 20, 25, 30, and 35, so the last two cycles carry 65 of the 120 minutes, about 54 percent of the night's REM. Waking after the fourth cycle, losing the fifth entirely, removes 35 minutes of REM but zero minutes of N3 — because the final cycle held no slow-wave sleep at all.

That asymmetry is the quantitative heart of sleep architecture. The early cycles are where slow-wave sleep is paid out and the late cycles are where REM accumulates, so truncating the start of sleep and truncating the end remove entirely different stages. An early night cut short loses REM; a delayed sleep onset against a fixed waking time loses the same late REM; only total sleep loss touches the front-loaded N3. It is the same logic the demonstration makes visible by letting the number of completed cycles vary and recomputing each stage's total.

Discussion

The sleep stages are among the best-characterized structures in physiology at the level of description and among the still-debated at the level of purpose. Their electrical signatures are precisely defined and reliably scored (Rechtschaffen & Kales, 1968; Iber et al., 2007), their generating circuitry is mapped from the thalamocortical rhythms that produce spindles and slow waves to the brainstem switches that assemble whole states (Steriade et al., 1993; Scammell et al., 2017), and the architecture that arranges them across the night is well understood as the output of homeostatic and circadian drives (Borbély et al., 2016). What remains contested is why sleep should be divided into these particular stages in this particular order.

The memory-consolidation account gives the clearest answer for the non-REM stages, tying the slow oscillations and spindles of deep sleep to the active redistribution of memory (Rasch & Born, 2013; Klinzing et al., 2019), yet the comparative diversity of the stages across species warns against treating any one function as the reason sleep is staged at all (Siegel, 2005). The developmental evidence deepens the puzzle rather than resolving it: the enormous share of REM and slow-wave sleep in early life implies a role in neural development that may be distinct from their role in the adult brain (Roffwarg et al., 1966). The stages are thus a case in which the form is almost fully solved and the function is only partly understood — which is precisely why the microstructure of a single stage, the spindle, has become so active a research front.

Current Directions

The most active recent work has moved from scoring the stages to reading their microstructure as a functional signal. The sleep spindle in particular has become a biomarker: its density, frequency, and coupling to the slow oscillation track memory consolidation and vary with age and disease, so the fine structure within stage N2 now carries information that the coarse stage label discards (Fernandez & Lüthi, 2020). The current questions concern how precisely the slow oscillation, the spindle, and the hippocampal ripple must be timed relative to one another for consolidation to occur, and whether that timing can be enhanced (Klinzing et al., 2019).

A second front is the automation and refinement of staging itself. Machine classification of the polysomnogram now matches expert scorers, which both removes a laborious manual step and raises the question of whether the thirty-second epoch and the five-stage scheme are the right units at all, or whether a finer, continuous description of sleep would capture function better than the discrete stages inherited from the scoring manuals (Iber et al., 2007).

Common Misconceptions

The sleep stages run on a single scale from light to deep to deepest.

Non-REM sleep does deepen from N1 through N3, but REM is not the deepest stage on that scale; it is a categorically different state whose EEG resembles waking, so the stages do not form one continuous axis of depth (Aserinsky & Kleitman, 1953).

There are still four non-REM stages.

The older Rechtschaffen and Kales standard numbered non-REM sleep 1 through 4, but the 2007 AASM revision merged stages 3 and 4 into a single N3, so modern scoring recognizes three non-REM stages, N1, N2, and N3 (Iber et al., 2007).

The stages are spread evenly through the night.

Deep slow-wave sleep is concentrated in the early cycles and REM lengthens toward morning, so the composition of sleep changes systematically across the night rather than repeating identically each cycle (Dement & Kleitman, 1957).

Everyone needs the same amount of each stage.

Stage proportions change dramatically with age — newborns spend about half their sleep in REM and the elderly obtain little slow-wave sleep — so there is no single fixed distribution that holds across the lifespan (Roffwarg, Muzio, & Dement, 1966).

Glossary

Delta waves.

The high-voltage, low-frequency waves below about 4 hertz that define stage N3, slow-wave sleep, and reflect synchronized cortical slow oscillations.

Electroencephalogram (EEG).

The recording of the brain's electrical activity from the scalp whose frequency and amplitude distinguish the sleep stages epoch by epoch.

Epoch.

The fixed interval, conventionally thirty seconds, into which a night's recording is divided and to each of which a single sleep stage is assigned during scoring.

Hypnogram.

A graph plotting sleep stage against time across a night, which makes the cyclic succession of the stages and their shifting proportions visible.

K-complex.

A large biphasic wave appearing in stage N2, often following a sensory stimulus, and one of the two transient events that mark the stage.

Memory consolidation.

The stabilizing and integration of newly encoded memories over time, to which the non-REM stages contribute through nested slow oscillations, spindles, and ripples.

Non-REM sleep.

The sleep stages other than REM — N1, N2, and N3 — characterized by progressively slower, higher-voltage EEG activity and retained muscle tone.

Polysomnogram.

The combined overnight recording of EEG, eye movement, and muscle tone from which each epoch of sleep is scored into a stage.

REM sleep.

The stage defined by rapid eye movements, fast wake-like cortical activity, and muscle atonia, set against the whole of non-REM sleep.

Sleep architecture.

The organized pattern in which the stages succeed one another across a night, including the cyclic alternation and the front-loading of N3 and late-loading of REM.

Sleep cycle.

The roughly ninety-minute unit in which the stages are organized, descending through non-REM into N3 and rising into a REM episode before repeating.

Sleep spindle.

A brief burst of waxing-and-waning oscillation around 11 to 16 hertz that marks stage N2, generated by the thalamic reticular nucleus and linked to memory consolidation.

Slow-wave sleep.

The deepest non-REM sleep, stage N3, dominated by delta waves and concentrated in the early cycles of the night; the stage most tied to homeostatic sleep pressure.

Two-process model.

Borbély's account in which a homeostatic sleep pressure (Process S) and a circadian process (Process C) together determine the timing and depth of the sleep stages.

Key Researchers

Alexander A. Borbély

. University of Zürich (Institute of Pharmacology and Toxicology); author of the two-process model of sleep regulation, which explains how homeostatic and circadian drives together shape the depth and timing of the sleep stages across the night. ORCID - Faculty Page - Wikipedia

Jan Born

(b. 1958). University of Tübingen (Institute of Medical Psychology and Behavioural Neurobiology); established the active-systems account of sleep-dependent memory consolidation, assigning the slow oscillations and spindles of non-REM sleep a specific role in stabilizing and redistributing memory. ORCID - Faculty Page

Mary A. Carskadon

. Brown University (Alpert Medical School, Department of Psychiatry and Human Behavior); developed the Multiple Sleep Latency Test and documented the developmental changes in sleep architecture and sleep need across adolescence. Faculty Page - Wikipedia

William C. Dement

(1928-2020). Stanford University; with Kleitman mapped the cyclic alternation of REM and non-REM sleep across the night, giving the stages their temporal architecture, and founded sleep medicine as a clinical discipline. Wikipedia - Wikidata

Nathaniel Kleitman

(1895-1999). University of Chicago; founder of modern sleep research, supervised the 1953 discovery of REM sleep and co-described the cyclic succession of the stages across the night. Wikipedia - Wikidata

Anita Lüthi

. University of Lausanne (Department of Fundamental Neurosciences); defined the thalamic circuit mechanisms of the sleep spindle and its dual role in protecting sleep from arousal and supporting memory, linking the microstructure of stage N2 to sleep's functions. ORCID - Faculty Page

Allan Rechtschaffen

(1927-2021). University of Chicago; lead editor of the 1968 standardized scoring manual that defined the sleep stages and their criteria, making inter-laboratory sleep research possible. Wikipedia - Wikidata

Mircea Steriade

(1924-2006). Université Laval (Faculty of Medicine); established the thalamocortical circuit mechanisms that generate the sleep spindle and the slow oscillation, giving the EEG hallmarks of the non-REM stages a cellular basis. Wikipedia

Frequently Asked Questions

What are the stages of sleep?

Sleep is divided into non-REM sleep, graded by depth into stages N1, N2, and N3, and REM sleep. N1 is the light transition from waking, N2 is marked by sleep spindles and K-complexes, N3 is deep slow-wave sleep, and REM is the wake-like stage of rapid eye movements and muscle atonia (Iber, Ancoli-Israel, Chesson, & Quan, 2007).

How are the sleep stages measured?

They are scored from the polysomnogram, which records the EEG, eye movement, and muscle tone together; the joint pattern across these channels in each thirty-second epoch assigns that moment of the night to a stage, following standardized scoring rules (Rechtschaffen & Kales, 1968).

How long is one sleep cycle?

A full cycle through the stages lasts about ninety minutes, descending from light sleep into deep N3 and rising into a REM episode before the next cycle begins. A typical night contains four or five such cycles (Dement & Kleitman, 1957).

Why is deep sleep concentrated early in the night?

The homeostatic pressure for sleep is highest at sleep onset and discharges through slow-wave activity as the night proceeds, so deep N3 sleep dominates the early cycles and wanes toward morning while REM lengthens (Borbély, Daan, Wirz-Justice, & Deboer, 2016).

What is a sleep spindle?

A sleep spindle is a brief burst of waxing-and-waning oscillation around 11 to 16 hertz that is one of the two transient events defining stage N2. It is generated by the thalamic reticular nucleus and is linked to the consolidation of memory during sleep (Fernandez & Lüthi, 2020).

How did the modern five-stage scheme replace the older one?

The 1968 Rechtschaffen and Kales standard defined four non-REM stages plus REM, but because stages 3 and 4 differed only in the amount of slow-wave activity, the 2007 AASM manual merged them into a single stage N3, giving the modern scheme of N1, N2, N3, and REM (Iber et al., 2007).

Do the sleep stages change with age?

Yes. Newborns spend about half their sleep in REM and a large share in deep non-REM, fractions that fall steeply through childhood; slow-wave sleep in particular continues to decline into old age, so the distribution of the stages is strongly age-dependent (Roffwarg, Muzio, & Dement, 1966).

What are the sleep stages for?

Their best-supported role is in memory: the slow oscillations and spindles of non-REM sleep drive the consolidation of declarative memory, while the stages differ in their contributions (Rasch & Born, 2013). Comparative evidence cautions against assuming a single universal function across species (Siegel, 2005).

References