What Sleep Is For

Sleep is dangerous. Hours of immobility and drastically reduced awareness, in a world containing things that eat you, repeated every day for a lifetime. Allan Rechtschaffen put the point as sharply as it can be put: if sleep does not serve an absolutely vital function, it is the biggest mistake the evolutionary process ever made.

Every animal with a nervous system does it, across lineages that share almost nothing else, for six hundred million years. Traits that costly do not survive without a function. What that argument establishes is that there is a reason; it does not tell us what the reason is, and the honest state of the field is that we know a great deal about what happens during sleep and much less about which of it is why sleep exists.

The best-evidenced mechanism is memory consolidation. During slow-wave sleep the hippocampus replays sequences of the day's activity in compressed form (this has been recorded directly in animals, watching place cells fire in the same order they fired while the animal ran a maze) and the replay appears to transfer memories towards cortical storage. Disrupt the slow waves experimentally without waking the sleeper and retention the next day drops. It is an unusually clean mechanistic story.

The second candidate is synaptic homeostasis. Tononi's argument is that learning during waking consists largely of strengthening synapses, which cannot continue indefinitely: the system saturates and the energy cost becomes unsustainable. Sleep scales everything back proportionally, restoring capacity while preserving the pattern of relative strengths, so what was learned survives and the accumulated noise does not. It is elegant, it competes with the consolidation account rather than contradicting it, and both could be true.

The third is clearance. The finding that the space between brain cells widens substantially during sleep, increasing the flow of cerebrospinal fluid through the tissue and washing metabolic waste out, was dramatic enough to be reported everywhere when it appeared. It offered a satisfyingly concrete answer. The details have since been argued over considerably, including the magnitude of the effect and how it was measured, which is the normal fate of a striking single result and not a reason to dismiss it.

Notice that these are mechanisms rather than purposes. A process can do several things without any of them being the reason it evolved, and distinguishing a function from a consequence is one of the genuinely hard problems in biology. Anybody who tells you the function of sleep has been settled is describing one laboratory's research programme.

One practical consequence follows from the evolutionary argument alone, without needing to resolve any of this. Claims to have trained the need away should be treated with the scepticism appropriate to a claim that somebody has out-engineered six hundred million years. A very small number of people carry variants permitting short sleep with no measurable deficit. The number who believe they belong to that group is enormously larger, and the reason is in the next chapter but one: impaired people report feeling fine, because the thing that recovers is the sense of being alert rather than the performance.