Evolution of human sleep

Among the great apes, Homo sapiens is a conspicuous outlier in sleep behaviour. Humans sleep an average of approximately seven hours per night, far less than chimpanzees (roughly 9.5 hours), orangutans (approximately 12 hours), or some lemur species that may sleep 15–17 hours per day.¹ Yet human sleep is not merely shorter; it is structurally reorganized, with a substantially larger fraction devoted to rapid eye movement (REM) sleep than is typical among primates of comparable body size.² Understanding how and why this unusual sleep profile evolved requires examining a constellation of factors—the shift from trees to the ground, the control of fire, cooperative social organization, and the cognitive demands of an expanding brain—that together reshaped one of the most fundamental biological rhythms in mammalian life.

Sleep across the primate order

Comparative data on primate sleep reveal a broad phylogenetic pattern: species that sleep in more exposed, riskier environments tend to sleep less and with lighter, more fragmented bouts, while those that sleep in enclosed nests or tree holes sleep longer and more deeply.³ Great apes are unusual among primates in constructing elaborate sleeping platforms—nests of woven branches and leaves—in the forest canopy. Chimpanzees select firm, stable branches for their nests and often choose Ugandan ironwood or similar species whose structural properties maximize platform stability, suggesting that nest construction is not a casual behaviour but a deliberate engineering effort to secure safe, comfortable arboreal sleeping sites.¹³

Within this framework, humans represent an extreme. Phylogenetic analyses controlling for body mass, brain size, diet, and predation risk consistently show that human sleep duration is shorter than predicted by any standard allometric model of mammalian or primate sleep.^{1, 6} Samson and Nunn documented that humans sleep approximately two to four hours less per day than would be expected for a primate of our body size, making us, proportionally, the shortest-sleeping primate yet studied.¹ At the same time, the proportion of sleep time spent in REM is higher in humans than in any other primate surveyed—roughly 20–25% of total sleep time, compared with approximately 5–15% in other great apes.² This combination of short total sleep and high REM density is the core of what has been termed the human sleep paradox.

The transition to terrestrial sleeping

The most dramatic ecological change in the evolutionary history of hominin sleep was the abandonment of arboreal sleeping platforms in favour of ground-level sleeping sites. All extant great apes build and sleep in tree nests, and the phylogenetic distribution of this behaviour strongly suggests that the last common ancestor of humans and chimpanzees did so as well.¹³ At some point in the hominin lineage—almost certainly after the emergence of Homo erectus and possibly coinciding with the habitual use of fire—ancestral humans transitioned permanently to sleeping on the ground.

Ground sleeping imposed immediate costs. Terrestrial sleepers are far more vulnerable to predation by large carnivores, as well as to parasites, insects, and thermal stress, than are arboreal nesters elevated ten metres or more above the forest floor.^{3, 12} The fact that this transition occurred despite these risks implies that countervailing advantages were substantial. The most frequently invoked explanations involve fire and social cooperation, both of which would have fundamentally altered the cost–benefit calculus of sleeping on the ground.

The archaeological record of fire use indicates that controlled fire was present by at least one million years ago at Wonderwerk Cave, South Africa, with habitual use firmly established by roughly 400,000–350,000 years ago across multiple sites in Africa and Eurasia.¹⁵ Fire would have provided multiple sleep-related benefits simultaneously: warmth, light sufficient to deter nocturnal predators, smoke to repel biting insects, and a social focal point around which groups could cluster.⁷ Richard Wrangham has argued that fire-based predator deterrence was the critical enabling factor that allowed hominins to sleep safely on the ground, and that this shift was intimately linked to the broader suite of anatomical and dietary changes associated with cooking.⁷

The sleep compression hypothesis

Samson and Nunn proposed the sleep compression hypothesis to explain the human sleep paradox. Their central argument is that the transition to ground sleeping, protected by fire and cooperative vigilance, allowed ancestral humans to compress the most restorative elements of sleep—particularly REM and slow-wave sleep (SWS)—into a shorter total duration, effectively increasing sleep efficiency without sacrificing cognitive benefits.^{1, 2}

The logic of the hypothesis rests on the observation that not all sleep stages are equally valuable. REM sleep is disproportionately associated with memory consolidation, emotional processing, and synaptic homeostasis, while SWS is critical for declarative memory formation and metabolic recovery.^{8, 14} If ecological conditions (specifically, the safety provided by fire and group vigilance) reduced the total amount of time that had to be spent sleeping, natural selection could have favoured individuals who packed more REM and SWS into fewer hours, discarding lighter, less restorative NREM stage 1 sleep. The result would be a sleep profile that is both shorter and more cognitively productive per unit time—precisely the pattern observed in modern humans.²

This hypothesis generates a testable prediction: human sleep should be more “efficient” than that of other great apes, with higher ratios of REM and SWS to total sleep time. Comparative polysomnographic data, though limited for wild apes, are broadly consistent with this expectation.⁶ The hypothesis also predicts that the selective pressure driving sleep compression was ultimately cognitive—that the freed waking hours were devoted to social learning, tool production, and other activities that enhanced fitness, creating a positive feedback loop between shorter sleep, greater REM density, and enhanced cognition.

Social sleeping and the sentinel hypothesis

A complementary explanation for the evolution of human sleep centres on the role of social groups as a collective defence system. The sentinel hypothesis, drawn from behavioural ecology, proposes that in communally sleeping groups, not all individuals need to be asleep simultaneously; the natural variation in sleep timing across group members ensures that at least some individuals are awake and vigilant at any given time, providing passive predator surveillance without requiring any formal watch system.⁵

Samson and colleagues tested this hypothesis using actigraphy data from Hadza hunter-gatherers in Tanzania. They found that across a group of roughly 33 adults, there were only 18 minutes in any given night when all individuals were simultaneously asleep.⁵ The variation in chronotype—some individuals being naturally early sleepers and others late sleepers—was sufficient to provide near-continuous wakefulness across the group throughout the night. The researchers termed this pattern “sentinel-like behaviour” and argued that the wide distribution of chronotypes observed in human populations may itself be an evolved trait, maintained by selection because groups with greater chronotype diversity experienced lower predation risk.⁵

This finding has implications for understanding why chronotype variation persists so robustly in human populations and why it appears to shift predictably with age. Older individuals in the Hadza sample tended to be earlier chronotypes, while younger adults were later chronotypes, creating a natural relay of vigilance across the night. Rather than being a dysfunction of aging, the age-related shift toward earlier sleep timing may be an adaptive feature that enhances group-level sentinel coverage.⁵

Sleep, memory, and the evolution of cognition

The disproportionate allocation of human sleep to REM stages has significant implications for cognitive evolution. REM sleep is the phase during which the brain is most active, exhibiting electroencephalographic patterns that closely resemble waking activity, and it is during REM that some of the most critical processes of neural maintenance and memory processing occur.⁸

Extensive experimental evidence demonstrates that both REM and SWS contribute to memory consolidation, though through different mechanisms. SWS is particularly important for consolidating declarative (fact-based) memories and transferring information from the hippocampus to neocortical long-term storage, while REM sleep appears to play a preferential role in consolidating procedural and emotional memories, as well as in extracting generalizable patterns and rules from experience.⁸ The synaptic homeostasis hypothesis further proposes that sleep serves to downscale synaptic strength that builds up during waking learning, restoring the brain’s capacity for new plasticity the following day.¹⁴

If the sleep compression hypothesis is correct, the evolutionary increase in REM proportion would have amplified precisely those cognitive functions—pattern recognition, procedural learning, creative problem-solving, and emotional regulation—that are most distinctive of human cognition.² Samson and Nunn have speculated that this created a co-evolutionary feedback loop: enhanced REM sleep supported greater cognitive complexity, which in turn enabled more sophisticated social structures and technologies (including better fire management), which further improved sleeping conditions and allowed still greater sleep efficiency.^{1, 2} This model positions sleep architecture not merely as a passive consequence of ecological change but as an active driver of the cognitive evolution that distinguishes Homo sapiens from other primates.

Hadza sleep and pre-industrial sleep patterns

Much of what is assumed about “natural” human sleep derives from studies of industrialized populations, but research on small-scale societies has substantially revised this picture. Samson and colleagues conducted actigraphic sleep studies among the Hadza of Tanzania, one of the last remaining populations of full-time hunter-gatherers, and found that Hadza adults slept an average of 6.25 hours per night, with considerable individual variation.⁹ Sleep was generally consolidated into a single nocturnal bout, with very little daytime napping, contradicting the common assumption that pre-industrial peoples routinely napped or practised biphasic sleep.⁹

A landmark cross-cultural study by Yetish and colleagues extended these findings by comparing sleep patterns across three pre-industrial societies on three continents: the Hadza (Tanzania), the San (Namibia), and the Tsimane (Bolivia). Despite vast differences in ecology, latitude, and culture, average sleep duration across all three groups converged on a remarkably similar range of 5.7–7.1 hours per night.⁴ None of the three groups regularly napped, and sleep onset was typically several hours after sunset rather than coinciding with darkness, challenging the assumption that artificial light is the primary driver of modern short sleep.

Perhaps most strikingly, Yetish and colleagues found that the strongest predictor of sleep timing was ambient temperature, not light exposure. In all three groups, sleep onset occurred during the period of declining nighttime temperature, and waking coincided with the temperature nadir just before dawn.⁴ This suggests that thermoregulation—not photoperiod—is the primary environmental cue governing human sleep timing in natural conditions, a finding with implications for understanding how sleep regulation evolved in concert with bipedalism, hairlessness, and thermoregulatory adaptations.

Biphasic and segmented sleep in historical perspective

The historian A. Roger Ekirch documented extensive textual evidence from pre-industrial Europe indicating that segmented or “biphasic” sleep—two distinct sleep bouts separated by a period of quiet wakefulness in the middle of the night—was common practice from at least the medieval period through the early modern era.¹¹ References to a “first sleep” and “second sleep” appear in legal documents, literature, prayer manuals, and medical texts across multiple European languages, suggesting that the consolidated eight-hour sleep block now considered normal is itself a relatively recent cultural invention, emerging only with the spread of affordable artificial lighting in the 18th and 19th centuries.^{10, 11}

How this historical European pattern relates to the evolutionary baseline is debated. The Hadza, San, and Tsimane data show predominantly consolidated monophasic sleep, which may reflect the distinct thermal and social environments of equatorial and subtropical regions where human sleep patterns originally evolved.⁴ Segmented sleep in higher latitudes, where winter nights are extremely long, may represent a flexible behavioural response to environmental conditions rather than a conserved ancestral pattern. The capacity for flexible sleep architecture—shifting between monophasic, biphasic, and polyphasic patterns depending on season, social context, and ecological conditions—may itself be the truly characteristic human adaptation, enabling Homo sapiens to colonize an extraordinary range of environments from the equator to the Arctic.^{3, 4}

Synthesis and open questions

The evolution of human sleep is best understood as the product of multiple interacting selective pressures rather than any single cause. The transition from arboreal to terrestrial sleeping, enabled by fire and social cooperation, removed the ecological constraints that had kept primate sleep long and light. The resulting sleep compression concentrated restorative REM and slow-wave sleep into fewer hours, potentially enhancing cognitive functions that were under strong positive selection in the genus Homo.^{1, 2} Chronotype diversity within social groups provided passive sentinel coverage, reducing the individual cost of shorter, deeper sleep.⁵ And the flexibility of human sleep architecture allowed expanding populations to adapt their sleep behaviour to diverse environmental conditions as they dispersed across the globe.

Several important questions remain open. The precise timing of the transition to ground sleeping is poorly constrained, as sleep leaves almost no direct trace in the archaeological record. Whether Homo erectus, with its evidence of fire use and increasingly terrestrial anatomy, had already adopted a human-like sleep pattern remains speculative.¹² The relative contributions of fire, group size, and technology to the evolution of sleep compression are difficult to disentangle. And the degree to which modern sleep disorders reflect a mismatch between evolved sleep biology and contemporary environments—an extension of the broader evolutionary mismatch paradigm—is an active area of research with significant public health implications.⁶ What is clear is that human sleep, far from being a passive biological given, is a deeply evolved adaptation shaped by the same forces of ecology, sociality, and cognition that define the broader trajectory of human evolution.