The puzzle of sleep

Sleep may seem paradoxical: Being inactive for a third of the day, in humans, means reduced chances to acquire resources or mate; plus it increases the odds of being preyed on, especially in our evolutionary past. Moreover, we also know that in our own case sleep is hard to avoid: Being awake for too long makes us tired and we progressively lose cognitive faculties, plus the more we stay awake the stronger there is of a drive to fall asleep, to the point it overrides the will not to. Even in those cases, lapses in wakefulnes known as microsleeps occur, so we have evolved mechanisms that do generate a strong and hard to elude need to sleep. Several theories have been proposed to explain why is it that we sleep, usually having to do with memory consolidation. Naturally, one wonders: Could we sleep less? If so why don't we? Elephants or horses for example sleep just 2-3 hours every day.

Back when I was reviewing prior to publication Alexey Guzey's takedown of Why We Sleep one of the claims that was discussed there was whether or not every animal1 sleeps. This is a background claim in the book that can be used as part of the evidence for sleep being very important. So, from Guzey's review

On page 6, Walker writes:

[E]very species studied to date sleeps

This is false, at least, according to Walker’s own source. When making this claim, he cites:

Kushida, C. Encyclopedia of Sleep, Volume 1 (Elsever, [sic] 2013)

….which turns out to be a 2,736 page book that costs $1,995. Fortunately, Walker tells us that we should search for this information somewhere in “Volume 1” or the first 638 pages of the book.

Anyway, page 38 reads:

It now appears that many species reduce sleep for long periods of time under normal conditions and that others do not sleep at all, in the way sleep is conventionally defined.

The Encyclopedia of Sleep (From 2013), says what Alexey says; and indeed Walker cited a source that contradicts him, but does not expand on examples of animals that do not sleep at all, one has to go to the end to follow some of the papers cited therein.

Does every animal sleep?

First, what is sleep? A paper from 2008 that is cited there, Do all animals sleep? (Siegel, 2008) defines various related terms (Like torpor or rest) as different from sleep, and sleep itself as "a rapidly reversible state of immobility and greatly reduced sensory responsiveness". Plus perhaps "To count as sleep it must also be homeostatically regulated; loss of sleep must be followed by an increased drive to sleep". One can then further define correlates of this behaviour in certain animals; those with a brain can also present various subtypes of sleep, as REM (Only observed in birds and mammals) or NREM.

That defined, what does the paper say about different organisms:

  • Unicellular organisms: No one has claimed that they sleep; but in some circadian rhythms have been found
  • Insects: Drosophila sleeps, cockroaches, bees, and scorpions meet the conditions of sleep except for homeostatic regulation. No evidence of REM sleep in insects.
  • Fish: Less than 10 species have been examined for rest or sleep behaviour. Zebrafish sleep, perches rest, but do not present the elevated response threshold to stimuli.
  • Amphibians: The bullfrog presents circadian rhythms rest but they are also more vigilant during rest, which would go against the definition of sleep above. The tree frog seems however to sleep.
  • Reptiles: Mixed evidence for REM sleep even for the same species. Turtles seem to sleep.
  • Mammals: Mammalian all seem to sleep, but this varies a lot across species and depending on the environment (eg. hunger decreases sleep). Sleep time shows great variation, with horses sleeping for 2h while bats sleep for 19 hours.
  • Marine mammals: Famously, dolphins sleep with half of their brain at a time, even closing the eye opposite to the hemisphere that is sleeping. Fur seals also sleep while on land, and show dolphin-style sleep. Importantly, dolphins have never been observed to sleep the way we do (i.e. bihemispheric sleep). Sleep deprived dolphins(after 5 days) do not show decline in performance in an accuracy(?) task, however dolphins required progressively more stimulation to stay awake.
  • Birds: Birds sleep, but they don't seem to suffer from cognitive impairment after being sleep deprived. A few species are also able to sleep one hemisphere of the brain at a time, even when in flight

So the author concludes that no, sleep is not universal; moreover given the great variation in sleeping time and behaviours, sleep may be serving different functions in each species.

From the same year however, Cirelli & Tononi assess the hypothesis that sleep may just be a default state when all the needs have been met. Assuming that, say, roaming around at night is more dangerous that staying with the rest of the pack in a safe place, then sleep would evolve for no reason other than to force itself upon the individual to induce a safer behaviour. The authors note that the quality of the evidence regarding sleep beyond mammals and birds is not very good; the bullfrog example noted before is based on a single study, so they don't agree that the evidence is strong here.

Regarding homeostatic regulation, the behaviours expected of sleeping beings were increased sleep pressure, and a compensatory rebound (after sleep deprivation, additional sleep time ensues). Siegel had argued that insects do not show this; but Cirelli & Tononi argue that sleep compensation may occur as deeper sleep rather than longer sleep. Again, the evidence for or against is not strong.

Regarding whether or not sleep loss leads to negative consequences, the authors say that given a sufficient amount of sleep deprivation, most animals studied die, with the possible exception of pigeons. This all said, the authors acknowledge that this may be due to constant stress, not lack of sleep itself, while in humans, a condition that causes chronic insomnia may be deadly for other reasons. However, sleep deprivation does lead to two universal conditions:

First, intrusion of sleep into wakefulness (Increased sleep pressure, urge to sleep) that goes to the point of being unavoidable, even with constant stimulation that aimed for total sleep deprivation (microsleeps); EEG wave patterns that are typically associated with sleep leak into wakefulness, so the wakeful state becomes more sleeplike. Here the authors conclude that there is no evidence that total sleep deprivation is possible for more than 24 hours without either microsleeps and mixed sleep-wakeful states occurring.

And second, cognitive impairment; here the authors note than in humans there is great variability in how susceptible individuals are to impairment, even at the task-level. Contra Spiegel, the authors claim there is evidence for cognitive impariment following deprivation in flies, birds, or rodents, in addition to humans.

The authors conclude that

The three corollaries of the null hypothesis do not seem to square well with the available evidence: there is no convincing case of a species that does not sleep, no clear instance of an animal that forgoes sleep without some compensatory mechanism, and no indication that one can truly go without sleep without paying a high price.

While they don't identify what function is sleep serving, they point to the fact that it may be an intrinsic requirement of neuronal activity, a requirement that also cannot be fulfilled while being awake, here they suggest memory consolidation without the disturbance of new memories coming in, or exercising/stimulating old memories to keep them recallable.

Okay, this was a while ago. What's the state of the art in 2020?

The Handbook of Behavioural Neuroscience, Chapter 24 (2019) mentions jellyfish as an example of animal that sleeps, even when no central nervous system is present. This behaviour is claimed to be full sleep (rest+higher responsivity threshold+homeostatic response); so jellyfish if sleep-deprived will then rest more when given a chance to do so.

While the chapter introduces sleep as "essential to nearly all studied animals to date" it does not mention examples where it does not occur (Perhaps as a form of hedging the claim). The chapter also talks about mexican cavefish, evolutionarily related fish in geographically close cave systems. These fish present substantial variation in sleep time from around 6 hours (In the case of their surface cousins) to almost nothing for some of the cavefish. Overall, the authors say that they would be tempted to conclude that all fish sleep, but stop short of doing so as we have only studied a small number of fish, so maybe some future study will find sleepless fish.

A 2018 study on the origins and evolution of sleep (Keene & Duboue) also notes great variation across and among species. To our list of animals that sleep they add another well known animal model, C. elegans which has just 302 neurons, and various mollusks, and octopuses. As with longevity, the genes that underpin sleep behaviours also seem conserved (They mention the genes Shaker, Sleepless, and Cyclin a which have human homologs). As with Tononi, the authors point to the fact that sleep being conserved across animals with neurons suggests that it may be a property that is also present at the single-cell level, i.e. in neurons.

For cockroaches and other insects the review does say that, contra Spiegel, they do present all the sleep markers, including homeostatic regulation.

Through the power of artificial selection, it has also been possible to study sleep variation within one species; in Drosophila it was possible to generate individuals that presented 90% sleep loss in 60 generations, though biomarkers present in sleep-deprived humans were present in higher concentrations in these flies, and these flies also showed reduced lifespan, same as wild type flies that are sleep deprived.

So the consensus seems to have moved towards the universality of sleep. If we look back at Kavanau (1998), the author points to numerous studies showing that, for example, a kind of cavefish does not sleep at all; including a one year long study where the fish were swimming at all times, and always disposed to accept food. But then again, a more recent review (Kelly et al. 2019), that cites the 1-year-long study notes that

Early research into sightless, cave-dwelling species of fishes [Pavan, 1946] suggested that these animals might be sleepless as they show evidence of continuous swimming and lack activity-based circadian rhythmicity. More recent studies, however, contradict this idea. Specifically, while cavefish (Astyanax mexicanus) do sleep, they in fact sleep very little, relative to their surface-dwelling conspecifics [Zafar and Morgan, 1992; Duboué et al., 2011, 2012; Yoshizawa et al., 2015; Jaggard et al., 2017, 2018]. [...] Furthermore, it remains unknown whether extensive periods of restfulness in buccal pumping sharks and rays can be considered sleep or simply quiet wakefulness. Unfortunately, none of the studies covered in this review answer these questions. To do so, future work must include a systematic investigation into the presence or absence of sleep, preferably on a broad range of elasmobranchs.

Chapter 20 of the 30th volume of the Handbook of Behavioural Neuroscience (2019) looks at various species. Some findings:

  • Larger animals sleep less (elephants sleep 2 hours, horses sleep 3 hours), and specifically animals with larger brains (in absolute terms) sleep less. Looking at brain size relative to body size shows the opposite pattern: Animals with relatively larger brains such as apes or dolphins, which tend to show advanced cognitive abilities, sleep more; humans in particular show a higher than expected amount of REM sleep. Animals with shorter gestation periods show more REM sleep, and animals with a higher antibody concentration in blood sleep more. Sleep may reallocate energy to the immune system?
  • Birds also sleep, including REM sleep, and they also show the compensatory response after sleep deprivation, including in pigeons, which Spiegel said above had not been observed to die from sleep deprivation; the Handbook notes they are particularly resistant to sleep deprivation compared to mammals.
  • Reptiles, evidence seems still mixed even within the same species studied.
  • For fish, and particularly the Pavan study, the Handbook also repeats the claim from Kelly et al., that newer evidence points to them actually sleeping. Some kinds of shark that do not require movement to breathe have been observed in an immobile state, but whether they are sleeping or not is unclear. For "ram ventilating" bony fish like sharks, which must remain swimming to "ram" water through their gills, they don't show periods of inactivity, forming the basis of the claim that they do not sleep at all; but the authors of the review do not take this as evidence that they don't sleep; rather sleep may be compatible with swimming. Unlike in the other sections, they don't mention here any EEG measurement. The Kelly paper above is a review of this specific case in particular, where he notes that the possibility that ram ventilating sharks and rays do not sleep, seems unlikely, and they posit that they sleep just like dolphins do.
  • For invertebrates they mention that sleep has been "convincingly demonstrated" in various species of arthropods, roundwordms, mollusk, flatworms, and jellyfish, suggesting that sleep probably evolved very early.

And they conclude,

Do all animals sleep? Sleep has been observed in all species studied by sleep scientists. There is a temptation to conclude that all animals sleep; however, no data exist for most animal groups. Around 30 animal phyla have yet to be tested for the presence of sleep (Lesku & Ly, 2017). Even within studied phyla, the phylogenetic coverage is often poor. This is true for invertebrates and also fishes and amphibians. That said, the existence of sleep in very simple animals, such as flatworms (Omond et al., 2017) and jellyfish (Nath et al., 2017), indicates that sleep evolved early in the lineage of animals. Whether it has persisted in all species over evolutionary time is unclear. Nonetheless, the apparent evolutionary longevity of sleep suggests that it fulfills a fundamental and inescapable need. This fundamental need is further revealed by (i) the persistence of sleep, despite the inherent vulnerability associated with this state (Lima et al., 2005); (ii) the evolution of unihemispheric SWS in marine mammals and birds (Lyamin et al., 2008; Rattenborg et al., 2000); and (iii) animals that can greatly reduce (but not eliminate) sleep when other demands favor sustained performance (Lesku et al., 2012; Rattenborg et al., 2016). It seems likely that sleep serves many functions, some of which might be evolutionarily “ancient,” present in jellyfish, flatworms, and vertebrates (Nath et al., 2017), while others might be evolutionarily new and present only in derived species (Lesku, Vyssotski, et al., 2011).

Thus, the claim that "every animal observed so far sleep" is probably correct; although REM sleep is not universal. Walker, in his reply to Alexey Guzey's comments cites some of the same evidence I mention here.

Hence we can answer that "every animal sleeps" is compatible with the current evidence, although not confirmed yet; "every animal we have studied in depth sleeps"is probably right, and I add that future studies (EEG on sharks, bullfrogs, etc) will point in this direction too.

Why we (really) sleep

This post was prompted by this thread where Michael Nielsen complains we don't have a good explanation for sleep and I wondered what a good explanation would be. Here I am not going to explain why we sleep, I'll leave that for some other post, but will offer instead what an answer must meet to count.

A way to explain why we sleep would have an evolutionary level (Why the behaviour or trait evolved) and that level would talk about the benefits of sleep that make up for the downtime it requires. But also it would require underpinnings at the levels of the neuron and the synapses and the proteins and regulatory pathways therein to explain what is it that it is actually doing for us.

Evolutionary explanations can be underwhelming; it's all too easy to say that "legs evolved to move". That feels underwhelming because it ignores the possible counterfactuals; as a matter of fact we know that some animals don't have legs to move, so if "movement" is the evolutionary requirement, the phenotype of having "legs" is not the only means, thus leading to the question "Why legs and not something different?". So we would then need to complement the straight evolutionary answer with a counterfactual one, thinking about different ways the phenotype could have been acquired. Biology is a very rich domain, so there are usually multiple ways of meeting the same "design goal" so we may end up having to repeatedly think of counterfactuals.

In this article I've suggested that sleep is universal. This in turn is some evidence that sleep is a necessity for all animals that possess neurons. Now, is this true?

For example, assume that sleep is a need of systems of neurons and we need to explain why we observe sleep. Can there be systems of neurons that do not require sleep at all? There is a debate in biology about whether or not neurons evolved more than once, but the end-result is in any case similar, we all have the same basic kind of neurons (Though we have different kinds of ion channels; every animal have K but only some have Ca and Na; the more variation the better as it shows that even different designs do not banish sleep completely). Animals have gained and lost limbs, scales, wings and other traits, yet they all share the same basic neuron design, and they all sleep. If so, the argument chain would be:

  1. Animals in a sufficiently complex environment need systems of neurons to successfully occupy their ecological niche (This is more of a premise that one has to buy. Here we stop the explanation at the fact that some animals evolved to have neurons and others do not. We could further try to explain this but not now)
  2. There is only one way to make neurons in the animal kingdom. Evidence is that it is the case so far, they all share the same basic architecture, and while as noted above there is debate about how many times they evolved, it seems such a fixed feature that they are along in the possibility-space for a system that meets what is required of a control and information processing system.
  3. Either neurons themselves, or any system of neurons requires sleep. Evidence is that every animal with neurons we have looked at so far presents some form of sleep + Some evidence I have not presented here about the molecular basis of sleep.
  4. What are the exact needs of neuronal systems that are being met with sleep? We do know that lack of sleep causes various forms of cognitive impairment, so we would have to get into the molecular biology of those to see exactly what is going on.
  5. But why do we sleep 8 hours and not 2 like elephants? I haven't looked into it yet; but it will be a combination of brain-specific architecture and environment. For a fixed brain architecture (e.g. number of neurons and synapses, overall connectomic patterns, etc) there will be an optimal amount of sleep for it to do its functions. However, other evolutionary requirements will shift this around. You could see this with the flies described above that slept 90% less; that gains was possible by paying the cost of shorter lives and effects akin to those of sleep deprivation. In any case there are healthy individuals that sleep just 4 hours without any obvious drawback. The causes of individual sleep variation merit more research.

So what's next? Identifying the exact systems that are affected by sleep and how those relations vary across species. Then see what happens to them during sleep deprivation. What exactly happens to animals that die from it? If we saw them dying from, say, pathogens then that would be a clue to the immune system playing a core role. But if we observe a general dysregulation, and if malfunctioning brains are able to induce it then it is perhaps the brain alone that has that need.


By animal throughout the article I refer to animals that have neurons. An example of animal without neurons are the sponges.