MUMultiple speciesBmoderateT1hypothesis only
Ten of sixteen clinically unremarkable horses stayed under 30 minutes of REM sleep: measured lying times and what sleep loss actually shows.
Executive summary
In REM sleep the muscles lose their tone, so a horse reaches this phase reliably only when lying down. Across the published work, roughly 231 minutes of total sleep and roughly 40 minutes of REM sleep per 24 hours have been measured, with very wide scatter between individual animals and between methods. The widespread minimum of 30 minutes of lying time is poorly supported: clinically unremarkable horses often fall below it.
Key points
In REM sleep, the phase in which dreaming occurs, the skeletal muscles lose their baseline tone. That relaxation is not a side effect of the state, it belongs to it. In the rodent it has been shown experimentally that several inhibitory messenger systems must act on the motor neurones at the same time for tone to fall away; if they fail, the result is either REM sleep with muscle tone preserved or, conversely, a loss of tone during waking. In the horse this mechanism has never been tested experimentally. It is a well-argued transfer from the rodent experiment, not a measurement made on the horse.
In the stable this has one simple consequence. A review of equine sleep records that, because of this loss of tone, REM sleep can occur only in lateral or sternal recumbency. Individual horses reach it standing as well, but only momentarily: the vanishing muscle tone makes the animal buckle, and the buckling itself ends the sleep again.
Possibly, but it is documented only in individual animals. In an overnight recording with brain-wave measurement in five neurologically unremarkable horses, one animal showed REM sleep while standing, accompanied by numerous partial collapses. The same work found that drowsiness and REM sleep can barely be told apart electrically, whereas deep sleep stands out clearly from both.
Across the published work a horse sleeps markedly less than a human being, but above all it distributes its sleep differently. The share of REM sleep within total sleep time comes out almost the same in both species. What differs is the packaging: the horse's sleep cycle is roughly three times shorter, and within a cycle it wakes briefly many times over, which the adult human practically never does.
| Parameter | Horse | Human |
|---|---|---|
| Total sleep per 24 hours | roughly 231 minutes | roughly 430 minutes |
| REM sleep per 24 hours | roughly 40 minutes | roughly 73 minutes |
| Share of REM sleep in total sleep | 17.5 per cent | just under 17 per cent |
| Length of one sleep cycle | roughly 41 minutes | just under 120 minutes |
| Brief waking episodes per cycle | roughly seven | fewer than one |
These averages, however, feign a precision that does not exist. The same review gives the ranges separately by method of measurement: in studies with brain-wave recording, total sleep time lay between 172 and 262 minutes, in studies relying on behavioural observation alone, between 65 and 382 minutes. The difference does not lie with the horse. It arises from whether sleep while standing is counted at all, and from how sleep is defined in the first place. The review therefore calls explicitly for a unified methodology for the behavioural measurement of sleep.
This problem is not confined to the horse. The analysis of 65 studies that established human normative values across the lifespan likewise records that the relationships found depended heavily on how strictly healthy participants had been selected: a normal value always describes the procedure that produced it as well.
Countless advice pages state that a horse must lie down for at least 30 minutes a day. In the literature reviewed here, that figure can be traced to exactly one place: a Swiss housing trial from 2017 cites it and refers for it to earlier work, not to a finding of its own. The dedicated review of equine sleep carries no such threshold. A Finnish study from 2026 uses it as the dividing line between short-sleeping and normally sleeping horses, without giving a source or a justification for it.
How little the line holds is shown by that same Finnish work. Of sixteen clinically unremarkable horses, ten fell below it, averaging roughly eleven minutes of apparent REM sleep, with individual values from zero to 28 minutes. The remaining six lay between just under 37 and roughly 66 minutes. None of these horses showed signs of a sleep disorder. A line below which the majority of a healthy group sits makes a poor requirement figure.
That does not follow. Short lying times occur in clinically unremarkable horses too, and the scatter between individual animals is considerable. The picture becomes notable only when a horse stops lying down altogether for days on end, or when falls, night-time injuries or daytime sleepiness appear alongside it. That judgement belongs in veterinary practice.
Most figures on equine sleep come not from the brain but from the camera. What is measured is lying, by video, by accelerometers or by position sensors. The Finnish work too assigned REM sleep from filmed posture, that is, from lateral or sternal recumbency with a relaxed neck no longer carrying the head. That is a surrogate measure: it describes a precondition of REM sleep, not REM sleep itself. A horse lying down may be awake, and a very brief episode of REM sleep while standing stays invisible.
Genuine polysomnography is rare. A German study recorded electrical brain activity, eye movements and muscle tone in ten foals for 48 hours each, together with video, in their own stable and with wireless equipment. It distinguished four states of wakefulness and four body postures. Its most important finding for the present text: the waveform of foals differs from the one published for adult horses. Figures from foalhood therefore cannot be carried over to adult horses.
How unreliable observation alone can be is shown by a study of eleven horses on the first-night effect. The first night in an unfamiliar place differed measurably from the following ones in electrical brain activity, most clearly over the back of the head, while behavioural observation did not pick that difference up. Anyone counting lying minutes after transport or at a show is therefore probably underestimating the actual disturbance of sleep.
How common the problem is can so far be estimated only through surveys. In a Nordic survey with 1,749 responses from owners and carers, a sleep disorder was suspected in 5.0 per cent of horses. Most strongly linked to that suspicion were observations that trigger the suspicion themselves: an unexpected loss of balance or a fall, a horse never seen lying down, difficulty getting down and getting up, night-time injuries and daytime sleepiness. That is partly circular reasoning and not an independent confirmation.
One feature of the environment nevertheless remained: the lying surface. A hard, wet surface went with a roughly ninefold higher chance of a suspected case, a hard, dry one with a roughly three-and-a-half-fold higher chance. Age and height at the withers acted weakly in the same direction. Since participation was voluntary and a case was defined solely by the owner's suspicion, these are associations, not causes.
What did not hold up is instructive. In 83 horses at a sanctuary fitted with automatic position sensors, neither age nor lameness from chronic orthopaedic disease had a meaningful influence on lying time, contrary to the researchers' expectation. Lying times ranged from zero to 319 minutes a day, averaging roughly 67. Only one group of eight horses with clinical signs of REM sleep deficiency stood clearly apart: they lay for roughly eight minutes on average, the others for roughly 74.
A very small trial points towards physical causes. Eight horses, four of them with angular limb deformity, were observed twice over several days. The affected animals lay for about half as long as the comparison animals, and under a painkiller the difference disappeared. The researchers accordingly put it cautiously: a shorter lying time could go together with musculoskeletal discomfort.
Swiss animal welfare legislation prescribes minimum dimensions for the bedded area in group housing. A trial with 38 horses in eight groups tested four variants: without bedding, and with half, one and one and a half times the minimum area, the rest of the floor being covered with hard rubber mats. Without bedding, lying became rare, and the majority of horses stayed under 30 minutes in 24 hours. As the bedded area grew, lying time rose, yet even at one and a half times the minimum area some individual animals still fell below.
What mattered was less the space than the availability of a soft, yielding surface, and rank within the group played its part. Low-ranking horses were forced to get up more often than high-ranking ones at half and at the full minimum area; only at one and a half times the area did the two become alike. The legal minimum dimensions therefore do not secure undisturbed lying for every member of a group.
Whether turnout and company additionally encourage lying is open. A trial with ten young riding horses compared a box, an individual paddock and a paddock shared in pairs. The expected increase in lying could not be confirmed statistically, and the researchers speak expressly of indications. All that is secure is that lying is heavily concentrated in the night, roughly eleven times longer than by day. Variation from day to day was very large within one and the same horse.
In the search engine the two terms often appear in the same headline, yet they describe different things. As early as 2006, two horses were described whose striking sleepiness was attributed to a lack of sleep while lying down and expressly not to narcolepsy. That case report yields nothing further: no abstract is indexed for it, and the figures cannot be checked. It establishes the distinction, not how common it is.
Familial narcolepsy has been described in the horse, but it is rare and built differently from the human condition. In three Lipizzaner half-sisters by the same sire it was assumed on clinical grounds and through a drug test. Hypocretin, the messenger whose deficiency defines human type 1 narcolepsy, lay within the normal range in the cerebrospinal fluid of all the animals examined. The signs had been present from an early age and gradually decreased in two of the three fillies from their second year of life.
The only randomised trial on this question found nothing. Ten Lusitano horses were kept from lying down over three consecutive days and then tested in two tasks, one on spatial memory and one on attention. In both, only a trend towards longer times appeared, with no statistically meaningful decline. With ten animals of a single breed, however, a medium-sized effect is barely detectable that way. An absence of evidence is not evidence of harmlessness.
The Finnish observational study supplies a counterweight. Horses with short apparent REM sleep performed worse in a learning test in which the rewarded side was switched repeatedly, and they persisted less. The relationship concerned the number of switches mastered, not the error rate. This is an observation in sixteen animals and not proof of a cause.
In humans the direction is established, but the magnitude is modest. An analysis of 61 studies of short-term sleep restriction found an effect for three domains only: executive control, sustained attention and long-term memory. For multitasking, impulsive decisions and intelligence the data were insufficient. And the widespread notion that REM sleep serves above all to process emotion does not survive an analysis of more than a thousand comparisons: the advantage for emotional material was no greater after sleep than after staying awake.
The comparison across species calls for further caution. A phylogenetic analysis of published sleep data showed that published sleep durations depend markedly on the laboratory conditions under which they were collected. If the analysis is restricted to data gathered in standardised fashion, the classical explanations of sleep no longer hold, namely energy saving, cognitive performance and development. The differences between species reflect ecological constraints instead, above all the competition between feeding and sleeping. The short sleeper among the equids named there is, incidentally, the donkey with roughly three hours, not the horse.
How far this can go is shown by two free-roaming elephant cows. They rested for roughly two hours a day on average, the shortest duration yet measured in a mammal, and lay down only every third or fourth day. After stretches of up to 46 hours without sleep they showed no recovery sleep at all. A large herbivore can therefore go without lying down for a long time with no compensation following. For the horse that is no model, but it is a reason not to assert a general daily lying requirement for large animals.
How young this research field is comes out in a marginal note. For the sleep of farmed ungulates, the protocol for a first systematic mapping of the literature was published only in 2026; at that point, then, no such overview existed at all. That protocol does not, however, name the horse explicitly.
Original analysis
| Method of measurement | What it captures directly | What it cannot settle | Largest number of animals in the literature used here |
|---|---|---|---|
| Polysomnography in the home stable: brain waves, eye movement, muscle tone and video at once | Sleep stages from the brain signal, synchronised with body posture | How common a finding is in a population | 10 foals for 48 hours each |
| Overnight brain-wave recording with video away from familiar surroundings | Whether drowsiness, deep sleep and REM sleep are electrically separable | How a horse sleeps at home | 5 adult horses |
| Brain-wave comparison across several nights in the same place | Whether the first night runs differently from the following ones | How much daytime performance suffers as a result | 11 adult horses |
| Video observation of body posture | Lying time and position, from which apparent REM sleep is derived | Whether the animal was actually in REM sleep at the time | 16 horses with five recordings each |
| Accelerometer or position sensor worn by the animal | Lying time and number of lying bouts over days to weeks | Sleep stages, and whether an interruption was forced | 83 horses at a sanctuary |
| Controlled housing trial with varying lying area | How lying time responds to surface and area | Whether more lying also means more REM sleep | 38 horses in eight groups |
| Survey of owners and carers | How often a suspicion is voiced and what it goes with | Whether the suspicion is correct in an individual case | 1,749 responses |
Roughly four hours, though the figure is less certain than it sounds. Averaged across the published work, a horse reaches roughly 231 minutes of total sleep in 24 hours, of which roughly 40 minutes are REM sleep. The range is wide and depends on the method of measurement: 172 to 262 minutes in studies with brain-wave recording, 65 to 382 minutes in purely behavioural observation. The difference arises mainly from whether sleep while standing is counted and from how sleep is defined. A single value for the horse therefore describes the method rather than the animal.
Because muscle tone falls away in REM sleep. If a horse reaches this phase while standing, the supporting musculature no longer holds, the animal sags, and the fall ends the sleep. In an overnight recording with brain-wave measurement in five horses, one animal showed exactly this pattern, accompanied by numerous partial collapses. Taken on its own, however, buckling proves no cause: pain on getting down, a restless or unsafe place to lie, and neurological disease are equally possible. What lies behind it in an individual case is settled only by a veterinary examination.
The figure is widespread and weakly supported. In the literature reviewed here it appears traceably only in a Swiss housing trial, which itself takes it over from earlier work. The dedicated review of equine sleep knows no such threshold, and a Finnish study from 2026 uses it as a dividing line without naming a source. In that same study, ten of sixteen clinically unremarkable horses fell below it. As a prompt to look at the lying place at all, the figure serves; as a boundary between healthy and ill, it has not been tested.
Usually not. Narcolepsy has been described in the horse, but it is rare: documented cases include a familial form in three Lipizzaner half-sisters, in which hypocretin, decisive in humans, was normal in the cerebrospinal fluid. The more frequent explanation for striking daytime sleepiness is a lack of sleep while lying down, a distinction already described in 2006 in two horses. How often each of the two causes applies is not captured in reliable figures. Working that out belongs in veterinary practice and not in a search engine.
Harm is not established, but neither can the all-clear be given. The only randomised trial kept ten horses from lying down for three days and found no statistically meaningful decline in two cognitive tasks, merely a trend towards longer times. With ten animals of one breed, a medium-sized effect is barely detectable. Conversely, a study of eleven horses showed that the first night in an unfamiliar place is measurably disturbed in electrical brain activity, with nothing visible in the behaviour. The disturbance is therefore real, its consequences unquantified.
The data offer two pointers, no more. In a Nordic survey, a hard, wet lying surface was the only feature of the environment clearly associated with suspected sleep disorders, a hard, dry one considerably less so. In a Swiss housing trial, lying became rare without bedding, and even at one and a half times the legal minimum area some individual animals stayed under 30 minutes in 24 hours. Low-ranking horses were forced to get up more often. Both are observations from group housing; no recipe for an individual stable follows from them.
Terms defined in this text
Equine Health
In 392 Quarter Horses, 57 per cent of the healthy horses carried a P variant: what the two assessments of P2, P3 and P4 actually found.
HOHorseBmoderate Open access
Movement & Recovery
Of 750 horses at competitions, 44 per cent left no finger under the noseband: what the taper gauge, pressure sensors and a cadaver head show.
HOHorseBmoderate Open access
Comparative Health
At gallop a foreleg carries 14.0 newtons per kilogram: why in the horse it is the supporting limb, not the fracture, that decides the outcome.
MUMultiple speciesBmoderateT0not transferable Open access
Once a month: new syntheses, checked studies, and what has changed in the state of the evidence.