Why do we sleep?

Sleep is found across animals with nervous systems, from mammals and birds to flies and worms, and the pressure to sleep increases the longer it is postponed. It is regulated rather than merely permitted.

Loss of it degrades cognition, mood and memory, and animals deprived of sleep for several weeks die. Why they die is less settled than it looked: the older account was infection and tissue damage, and a 2020 study in flies and mice traced death to oxidative stress in the gut, and found that flies protected from it lived a normal lifespan with little or no sleep.

Learning followed by sleep is retained better than learning followed by the same length of time awake. Patterns of activity recorded while an animal learns a route reappear during subsequent sleep, compressed in time.

The effect looked specific rather than general: different stages of sleep have been linked to different kinds of material, and interfering with particular sleep rhythms has interfered with particular kinds of retention. Memory is the function with the most direct evidence, and that evidence has shrunk. A 2021 review reports key findings that could not be replicated and argues that the benefit is smaller, more dependent on the task, less tied to deep sleep and less lasting than was assumed.

In 2013 a study in mice found that the space between brain cells expands during sleep, and that fluid exchange and the clearance of metabolic waste products, including beta-amyloid, a protein linked to neurodegenerative disease, run faster than in the waking state.

That is a mechanism measured in a mouse, under anaesthesia in part of the work, and it has been both extended and contested since. It is a real finding and not yet a settled account of human sleep.

Contested is not a hedge here. A 2024 study using a different method to track a dye through the living mouse brain reported the opposite: clearance slower during sleep and under anaesthesia than in waking. A 2025 review answered that dye moving within the brain is not a measure of dye leaving it. Whether the disagreement is about the measurement or about what is being measured is unresolved, and this is the part of the sleep literature where confident summaries are least warranted.

Memory consolidation, waste clearance, energy conservation, immune regulation and synaptic downscaling all have supporting evidence. Any of them could be the original purpose with the others as passengers, and there is no accepted way to tell.

It is also possible that the question is wrong. A state that old and that widespread may have accumulated functions rather than had one, in which case asking what sleep is for is like asking what blood is for.

The review most of this rests on does take a side. Energy saving, tissue repair and immune work could all be done lying still and awake, so none of them explains why sleep switches off consciousness and the response to danger. That argument reads sleep as mainly for the brain.

Across mammals the amount and the kind of sleep track age, body size, diet and how safe the sleeping site is. Some birds sleep with one hemisphere at a time, and dolphins do the same while swimming. Great frigatebirds sleep in flight, about forty minutes a day, under a tenth of what they sleep ashore.

A single essential function should show less variation than this. Whatever sleep is doing, animals have found many different budgets for it.

The most vivid dreams are reported from REM sleep, a state with a distinctive signature, though dreaming is not confined to it. REM sleep is thought to play a part in some kinds of consolidation. Whether the experience of dreaming does anything, or accompanies something that does, has no accepted answer.

This is the same shape as the harder question about consciousness in general: a function can be described without the experience, and the experience is nonetheless there.