REM Sleep

Abstract

REM sleep, which MeSH classifies under sleep, is the stage of sleep defined by rapid eye movements, a fast low-voltage brain activity resembling waking, and a near-complete loss of skeletal muscle tone. It recurs cyclically through the night in alternation with non-REM sleep, and it is the stage from which vivid, narrative dreaming is most often reported. This article defines REM sleep and separates it from the non-REM stages, traces its discovery in the electroencephalographic laboratory, describes the brainstem physiology that generates it and the atonia that accompanies it, sets out its place in the architecture of the night, reviews its long association with dreaming through the activation-synthesis account, and surveys the evidence that it contributes to emotional and procedural memory consolidation. Three interactive demonstrations make the generating switch, the cyclic structure, and the rebound after deprivation explicit.

Keywords: REM sleep, paradoxical sleep, sleep cycle, muscle atonia, memory consolidation

Of the states the sleeping brain passes through, REM sleep is the strangest. The body lies paralysed while the eyes dart beneath closed lids, the cortex runs at a pace close to waking, and the mind, if woken, reports a vivid dream. It was the discovery of this state in the 1950s that turned sleep from an apparent blank into a structured, cyclic process with its own internal architecture (Aserinsky & Kleitman, 1953). REM sleep recurs four or five times a night, lengthening toward morning, and it is physiologically so distinct from the quiet stages that surround it that Michel Jouvet called it paradoxical sleep: a brain that looks awake inside a body that cannot move (Jouvet, 1967).

What REM Sleep Is

REM sleep, named for the rapid eye movements that accompany it and also called paradoxical sleep, is one of the two broad kinds of sleep, set against non-REM sleep. It is identified by the convergence of three signs recorded together: on the electroencephalogram the brain shows fast, low-voltage, mixed-frequency activity close to that of an alert waking brain; the eyes make bursts of rapid movement under the closed lids; and the muscles of the body lose their tone almost entirely, a state of near-paralysis called atonia. No one of these alone marks REM sleep; it is their co-occurrence that defines the stage.

That definition matters because REM sleep is easily conflated with the states on either side of it. It is not simply light sleep: although its cortical activity resembles the drowsy transition at sleep onset, its atonia and eye-movement bursts are its own. It is not dreaming, although the two are tightly linked; dreams are also reported from non-REM sleep, and REM sleep is a physiological state that can be defined without any reference to mental content. And it is not a uniform block of the night but a recurring episode that returns several times, so that a sleeper moves in and out of REM repeatedly before waking. The sections that follow take these apart in turn, beginning with how the state was found.

The Discovery of REM Sleep

Before the 1950s sleep was widely treated as a single, passive state, a simple withdrawal from waking. That picture broke when Eugene Aserinsky, working in Nathaniel Kleitman's laboratory at the University of Chicago, noticed periodic bursts of eye movement recurring through the night and found that they coincided with a fast, activated electroencephalogram and with dream reports when sleepers were woken from them (Aserinsky & Kleitman, 1953). The observation revealed that sleep was not uniform at all but contained a distinct, recurring, physiologically active state.

William Dement and Kleitman then mapped the full structure of the night, showing that REM and non-REM sleep alternate in a regular cycle and that the eye-movement periods recur at roughly ninety-minute intervals, lengthening as the night goes on (Dement & Kleitman, 1957). The proportion of REM also turned out to change dramatically across the lifespan: Howard Roffwarg, Joseph Muzio, and Dement documented that newborns spend about half their long sleep in REM, a fraction that falls steeply through childhood to the adult level of roughly a fifth to a quarter, which suggested that REM plays some part in the development of the nervous system (Roffwarg, Muzio, & Dement, 1966). Figure 1 shows the cyclic architecture these studies revealed.

Figure 1

The alternation of REM and non-REM sleep across a night, with REM episodes lengthening toward morning.

The discovery reframed the whole subject. Sleep became an orderly succession of states with a measurable architecture, REM sleep became a target that could be selectively studied and deprived, and the tight link between REM and dreaming opened an experimental route into the mind during sleep that had not existed before.

The Physiology of REM Sleep

What generates REM sleep is not a single centre but a brainstem system that switches the state on and off. Jouvet localized the essential machinery to the pons, showing that structures there both trigger the activated cortical state and drive the active inhibition of the spinal motor neurons that produces atonia (Jouvet, 1967). REM sleep is therefore an actively constructed state, not merely the absence of waking or of deep sleep: the paralysis is imposed by descending inhibition, and the cortical activation is positively driven.

The modern account describes this as a switch between two opposing neuronal populations. REM-promoting neurons, many of them cholinergic, become active and generate the state; REM-suppressing neurons, using the aminergic transmitters noradrenaline and serotonin, fall silent during REM and resume between episodes, and the mutual inhibition between the two groups produces the roughly ninety-minute oscillation into and out of REM (Scammell, Arrigoni, & Lipton, 2017). The circuitry of the atonia is a central part of the picture, because its failure is clinically visible: when the descending inhibition breaks down, the sleeper acts out dreams, the defining feature of REM sleep behavior disorder (Peever & Fuller, 2017). The first demonstration models the reciprocal switch, showing how the two populations oscillate in antiphase to produce the REM rhythm.

REM and the Architecture of the Night

REM sleep is one part of a repeating cycle, and its meaning is clearest set against the non-REM stages it alternates with. A night of sleep is organized into cycles of about ninety minutes, each descending through the non-REM stages into deep slow-wave sleep and then rising into a REM episode before the next cycle begins. Early cycles are dominated by deep non-REM sleep and carry only brief REM episodes; as the night proceeds, deep sleep wanes and REM episodes lengthen, so that the last cycles before waking are richest in REM (Dement & Kleitman, 1957). This is why a person woken early is woken mostly from deep sleep, while one who sleeps a full night does most of their REM in the final hours. Table 1 sets the defining features of REM sleep beside those of non-REM sleep.

The cyclic structure also fixes how much REM a night contains. Because REM accumulates mainly in the later cycles, total REM time depends heavily on sleeping through to the morning: cutting a night short truncates exactly the cycles in which REM is densest, so a modest loss of total sleep can cause a disproportionate loss of REM. The second demonstration traces a hypnogram across an adjustable night and reports how the REM share builds up cycle by cycle.

Dreaming and REM Sleep

The link that made REM sleep famous is its association with dreaming. From the first studies, sleepers woken from REM reported vivid, story-like dreams far more often than those woken from non-REM sleep, and this correspondence became the standard experimental handle on dreaming (Aserinsky & Kleitman, 1953). The association is strong but not absolute: dreaming is reported from non-REM sleep as well, though such reports tend to be more thought-like and less vivid, so REM sleep is best described as the stage from which narrative dreaming is most reliably recovered rather than as the sole seat of dreams.

The most influential physiological account of REM dreaming is the activation-synthesis hypothesis of J. Allan Hobson and Robert McCarley, who proposed that dreams arise when the forebrain synthesizes a narrative from the largely random activation generated by the brainstem during REM (Hobson & McCarley, 1977). On this view the bizarre, shifting quality of dreams reflects their origin in internally generated signals rather than in coherent perception, and the emotional intensity reflects the activation of limbic circuits during REM. The hypothesis was a deliberate challenge to the idea that dreams are primarily disguised wishes, relocating their explanation from hidden meaning to brainstem physiology, and it set the terms for decades of subsequent debate about how much of dream content is physiological noise and how much is meaningful construction.

That debate turned sharply with evidence that dreaming and REM sleep are not the same phenomenon. Reviewing lesion cases, Mark Solms argued that the two are doubly dissociable: dreaming can be abolished by damage to specific forebrain regions while REM sleep continues intact, and dreaming can persist after damage to the brainstem structures that generate REM, which implies that dreaming depends on forebrain circuits—he emphasized a dopaminergic motivational pathway—rather than on the brainstem REM generator itself (Solms, 2000). On this account REM is a reliable trigger for dreaming but not its cause, and the activation-synthesis picture describes the state that usually accompanies dreams rather than the mechanism that produces them. The dissociation reframed the question the rest of this article returns to: whether REM exists for the dream, or the dream is one of several things a state with other purposes happens to carry.

The Functions of REM Sleep

Why the brain should construct such an elaborate state remains partly open, but the strongest evidence concerns memory. Sleep after learning improves later retention, and Robert Stickgold assembled the case that this sleep-dependent consolidation is real and that REM and non-REM sleep make distinct contributions, REM being especially implicated in procedural and perceptual skills (Stickgold, 2005). Susanne Diekelmann and Jan Born set this within a systems account in which newly encoded memories are reactivated and redistributed during sleep, with slow-wave sleep and REM sleep playing complementary roles in stabilizing and integrating them (Diekelmann & Born, 2010). A broad review of the field concluded that sleep actively benefits memory rather than merely protecting it from interference, and that REM sleep is particularly associated with the consolidation of emotional memory (Rasch & Born, 2013).

The clearest causal evidence came from targeting REM directly. Richard Boyce and colleagues, in Antoine Adamantidis's laboratory, used optogenetics to silence the theta rhythm specifically during REM sleep in mice and found that this selectively impaired the consolidation of contextual and emotional memories, without disturbing the rest of sleep (Boyce, Glasgow, Williams, & Adamantidis, 2016). That result moved REM's role in memory from correlation toward cause. Function is not confined to memory, however: comparative work by Jerome Siegel has emphasized how widely sleep and REM vary across species and cautioned against assuming a single universal function, pointing instead to a diversity of roles shaped by each animal's ecology (Siegel, 2005). A distinct line of evidence is the rebound that follows selective REM loss, in which REM presses to recover the lost amount, implying that the brain regulates REM as a quantity worth defending. The third demonstration shows this rebound building across nights of selective REM deprivation and discharging on recovery.

Worked Example

Because REM accumulates unevenly across the night, its share of total sleep is worth computing explicitly, and the second demonstration reproduces the arithmetic. Consider a typical eight-hour night organized into five cycles. The REM episode in each cycle lengthens as the night goes on, running roughly 10, 20, 25, 30, and 35 minutes. The total REM time is their sum: 10 plus 20 plus 25 plus 30 plus 35 is 120 minutes. As a fraction of the 480 minutes of an eight-hour night, that is 120 divided by 480, which equals 0.25, or 25 percent—squarely within the fifth-to-a-quarter range typical of healthy adults.

Now shorten the night. Suppose the sleeper wakes after the fourth cycle, losing the final cycle entirely. The REM obtained is only 10 plus 20 plus 25 plus 30, which is 85 minutes, and the lost cycle alone would have carried 35 minutes of REM—nearly a third of the night's total. The total sleep fell by just one cycle of ninety minutes, from 480 to 390 minutes, a drop of about 19 percent; but the REM fell from 120 to 85 minutes, a drop of about 29 percent. The loss of REM is disproportionate to the loss of sleep, because the truncated cycle was the one richest in REM.

That asymmetry is the quantitative heart of REM's vulnerability. Any force that trims the end of the night—an early alarm, a late bedtime against a fixed waking time, fragmented final hours—removes exactly the cycles in which REM is densest, so REM is lost faster than sleep itself. It is the same logic the demonstration makes visible by letting the number of completed cycles vary and recomputing the REM share each time.

Discussion

REM sleep occupies a peculiar place in the science of sleep: it is the most distinctive of the sleep states and in some ways the least understood. Its physiology is now mapped in detail, from the brainstem switch that turns it on to the descending inhibition that paralyses the body (Scammell et al., 2017; Peever & Fuller, 2017), and its architecture across the night is thoroughly characterized (Dement & Kleitman, 1957). What remains contested is its purpose. The memory-consolidation account has the strongest experimental support, including direct causal evidence from selective interference with REM (Boyce et al., 2016; Rasch & Born, 2013), yet the comparative diversity of REM across species warns against treating any one function as universal (Siegel, 2005).

The relationship with dreaming captures the same tension between well-described form and uncertain function. That vivid dreaming issues mostly from REM is not in doubt, and the activation-synthesis hypothesis gave that fact a physiological grounding (Hobson & McCarley, 1977); but whether the dream is the point of REM or a by-product of a state that exists for other reasons is unsettled. The developmental evidence adds a further clue without resolving it: the enormous share of REM in early life suggests a role in building the nervous system that may differ from whatever REM does in the mature brain (Roffwarg et al., 1966). REM sleep is thus a state whose mechanism is largely solved and whose reason is still argued over—an unusual position in physiology, and part of why it continues to attract research.

Current Directions

The most consequential recent work has made REM's contribution to memory causal rather than correlational. By silencing the REM theta rhythm precisely while leaving the rest of sleep intact, optogenetic studies have shown that REM-specific activity is necessary for consolidating particular classes of memory, turning a decades-old correlation into a manipulable mechanism (Boyce et al., 2016). The current question is which memories depend on REM and through what cellular processes, with emotional memory the leading candidate for a specifically REM-dependent role (Rasch & Born, 2013).

A second active front is the circuitry of the state itself and its clinical breakdown. Detailed maps of the wake-sleep switch and of the neurons that impose REM atonia have made it possible to ask how the state fails (Scammell et al., 2017), and the study of REM sleep behavior disorder—in which atonia is lost and sleepers enact their dreams—has become especially urgent because it frequently precedes neurodegenerative disease, making the integrity of REM circuitry an early window onto conditions that appear years later (Peever & Fuller, 2017).

Common Misconceptions

Dreaming happens only in REM sleep.

Vivid narrative dreams are reported most often from REM sleep, but dreaming also occurs in non-REM sleep, where reports tend to be more thought-like and less vivid, so REM is the stage dreaming is most reliably recovered from rather than its sole source (Hobson & McCarley, 1977).

REM sleep is the same as deep sleep.

REM sleep is the opposite of deep slow-wave sleep on the electroencephalogram: its activity is fast and wake-like, whereas deep non-REM sleep shows slow, high-voltage waves. They are distinct stages that alternate within each cycle (Dement & Kleitman, 1957).

The body is at rest during REM sleep.

The brain is highly active during REM sleep while the skeletal muscles are actively paralysed by descending inhibition; the stillness is an imposed atonia, not simple rest, and its failure produces dream-enactment disorder (Jouvet, 1967; Peever & Fuller, 2017).

REM sleep is spread evenly through the night.

REM episodes are brief early in the night and lengthen toward morning, so most REM falls in the final cycles and a shortened night loses REM out of proportion to the sleep lost (Dement & Kleitman, 1957).

Glossary

Activation-synthesis hypothesis.

The account proposing that dreams arise when the forebrain synthesizes a narrative from largely random brainstem activation generated during REM sleep, rather than from disguised wishes.

Atonia.

The near-complete loss of skeletal muscle tone during REM sleep, produced by active descending inhibition of the spinal motor neurons.

Electroencephalogram (EEG).

The recording of the brain's electrical activity from the scalp whose frequency and amplitude distinguish the sleep stages; REM sleep shows a fast, low-voltage, mixed-frequency trace resembling waking.

Hypnogram.

A graph plotting sleep stage against time across a night, which makes the alternation of REM and non-REM sleep and the lengthening of REM episodes visible.

Memory consolidation.

The process by which newly encoded memories are stabilized and integrated over time, to which sleep—REM sleep in particular for emotional and procedural material—contributes.

Non-REM sleep.

The sleep stages other than REM, ranging from light sleep to deep slow-wave sleep, characterized by slow high-voltage brain activity and retained muscle tone.

Paradoxical sleep.

Jouvet's term for REM sleep, capturing the paradox of a brain whose activity resembles waking housed in a body rendered immobile by atonia.

PGO waves.

Ponto-geniculo-occipital waves, bursts of brainstem-driven activity that herald and accompany REM sleep and travel to the visual thalamus and cortex.

Rapid eye movements.

The bursts of quick conjugate eye movement under closed lids that give REM sleep its name and are one of its three defining signs.

Reciprocal-interaction model.

The model in which REM sleep arises from the mutual inhibition between REM-promoting cholinergic neurons and REM-suppressing aminergic neurons, producing the cyclic oscillation into and out of REM.

REM rebound.

The increase in REM sleep that follows a period of selective REM deprivation, indicating that the brain regulates REM as a quantity it acts to recover.

REM sleep behavior disorder.

A parasomnia in which the atonia of REM sleep fails and the sleeper physically enacts dreams, often preceding neurodegenerative disease.

REM sleep.

The sleep stage defined by rapid eye movements, fast wake-like cortical activity, and muscle atonia, recurring cyclically through the night and most associated with vivid dreaming.

Sleep cycle.

The roughly ninety-minute unit in which sleep is organized, descending through non-REM stages into deep sleep and rising into a REM episode before repeating.

Key Researchers

Antoine Adamantidis

. University of Bern (Department of Biomedical Research) and Inselspital; used optogenetics to show that the theta rhythm of REM sleep is causally required for the consolidation of contextual and emotional memory, moving REM's role from correlation toward cause. ORCID

Eugene Aserinsky

(1921-1998). University of Chicago; discovered rapid eye movement sleep in 1953 as a graduate student, observing periodic eye-movement bursts coinciding with an activated electroencephalogram and dream reports. Wikipedia - Wikidata

Jan Born

(b. 1958). University of Tübingen (Institute of Medical Psychology and Behavioural Neurobiology); established the active-systems account of sleep-dependent memory consolidation, dissecting the complementary roles of slow-wave and REM sleep in stabilizing and integrating memory. ORCID - Faculty Page

William Dement

(1928-2020). Stanford University; co-mapped the cyclic alternation of REM and non-REM sleep across the night, named REM sleep, and founded sleep medicine as a clinical discipline. Wikipedia - Wikidata

J. Allan Hobson

(1933-2021). Harvard Medical School; proposed with Robert McCarley the activation-synthesis hypothesis of dreaming, grounding dream content in brainstem REM physiology rather than in disguised wish. Wikipedia

Michel Jouvet

(1925-2017). University of Lyon; identified the pontine mechanisms that generate REM sleep and the muscle atonia that defines it, and named the state paradoxical sleep. Wikipedia

Nathaniel Kleitman

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

John Peever

. University of Toronto (Department of Cell and Systems Biology); dissected the brainstem circuitry generating REM-sleep atonia and its breakdown in REM sleep behavior disorder, linking the state's motor control to neurodegeneration. Faculty Page

Jerome Siegel

. University of California, Los Angeles (Center for Sleep Research); advanced comparative and evolutionary analyses of sleep, documenting the diversity of mammalian REM across species and questioning single-function accounts. Faculty Page

Robert Stickgold

. Harvard Medical School (Center for Sleep and Cognition); provided much of the experimental evidence for sleep-dependent memory consolidation and the distinct contributions of REM sleep to procedural and emotional memory. Wikipedia

Frequently Asked Questions

What is REM sleep?

REM sleep is the stage of sleep defined by rapid eye movements, fast low-voltage brain activity close to that of waking, and a near-complete loss of skeletal muscle tone called atonia. It recurs cyclically through the night and is the stage from which vivid dreaming is most often reported (Aserinsky & Kleitman, 1953).

Why is REM sleep called paradoxical sleep?

Michel Jouvet named it paradoxical sleep because it presents a paradox: the brain shows fast, activated electrical activity resembling waking, yet the body is almost completely paralysed by atonia, so an active brain sits inside an immobile body (Jouvet, 1967).

How much of the night is spent in REM sleep?

In healthy adults REM makes up roughly a fifth to a quarter of total sleep. It is distributed unevenly, with brief episodes early in the night lengthening toward morning, so most REM occurs in the final cycles before waking (Dement & Kleitman, 1957). Newborns spend a far larger share, about half, of their sleep in REM (Roffwarg, Muzio, & Dement, 1966).

Does dreaming happen only in REM sleep?

No. Vivid, narrative dreams are reported most often from REM sleep, but dreaming also occurs in non-REM sleep, where reports tend to be more thought-like and less vivid. REM is the stage dreaming is most reliably recovered from, not its only source (Hobson & McCarley, 1977).

What generates REM sleep in the brain?

REM sleep is produced by a brainstem switch between REM-promoting neurons, many of them cholinergic, and REM-suppressing aminergic neurons; the mutual inhibition between these populations produces the roughly ninety-minute cycle into and out of REM (Scammell, Arrigoni, & Lipton, 2017).

Why is the body paralysed during REM sleep?

During REM sleep the spinal motor neurons are actively inhibited, producing the atonia that prevents the sleeper from acting out dreams. When this inhibition fails, the result is REM sleep behavior disorder, in which dreams are physically enacted (Peever & Fuller, 2017).

What is REM sleep for?

Its best-supported role is in memory: sleep after learning aids consolidation, and REM sleep is especially implicated in emotional and procedural memory, with direct causal evidence from experiments that selectively disrupt REM activity (Stickgold, 2005; Boyce, Glasgow, Williams, & Adamantidis, 2016). Comparative evidence cautions against assuming a single universal function (Siegel, 2005).

What happens after a person is deprived of REM sleep?

Selective loss of REM sleep is followed by a rebound: on subsequent nights REM increases above its usual level, as though the brain is recovering a defended quantity. This regulation is part of the evidence that REM serves a function the brain acts to protect (Rasch & Born, 2013).

References