Sleep debt

Sleep debt is the running total between the sleep your body needed and the sleep it got. The metaphor comes from accounting, and like most borrowed metaphors it is right about one thing and misleading about two others. It is right that shortfalls accumulate rather than resetting each morning. It is misleading in suggesting that the balance is a single number, and that paying it off is a matter of depositing the missing hours.

The finding that matters most here is not about how much sleep you owe. It is about what you can feel. In the studies where people are kept short for weeks, the sense of being tired stops getting worse long before the impairment does — so the reliable subjective experience of chronic sleep debt is not exhaustion. It is thinking you have adapted.

The definitive experiment is Van Dongen and Dinges' 2003 study. Healthy adults were held to four, six or eight hours in bed for two weeks while their reaction times and working memory were tested daily, and a fourth group was kept awake for three days straight as a comparison.

Two results came out of it and they belong together.

The damage accumulates in a straight line. The six-hour group did not settle at a mildly worse level and stay there. Their lapses of attention kept getting worse every day for fourteen days, reaching by the end a level comparable to people who had been awake for a day and a half continuously.

The feeling does not. Those same people rated their own sleepiness as only slightly elevated, and their ratings flattened out after the first few days while their performance carried on down the slope. They had lost the signal that would have told them to go to bed.

Belenky's group, running a similar design the same year, found the other half of the picture: after three nights of unrestricted recovery sleep, the restricted groups had improved but had not returned to where they started. Some of what was lost came back quickly. Some of it did not come back on that timescale at all.

This is the question people are really asking, and the honest answer has two parts that point in different directions.

In the laboratory, the weekend is not enough. Depner and colleagues in 2019 gave one group five-hour nights, another five-hour nights with an unrestricted weekend, and a control group nine hours throughout. The weekend group slept more, as you would expect, and by the end of the following week their insulin sensitivity had dropped further than the group that never recovered at all — partly because the weekend pushed their body clock later and made the following week's mornings worse. The lie-in repaid one thing while charging another.

In the population, it looks better. Åkerstedt's Swedish cohort of nearly forty-four thousand people, followed for thirteen years, found that short weekday sleep paired with long weekend sleep was not associated with the raised mortality seen in people who slept short all seven days. That is an observational association, not a mechanism, and it cannot tell you whether the weekend helped or whether people whose weekends are free are simply different people. But it is real evidence and it does not support the harshest version of "you can never catch up".

The fair summary: alertness and mood repay fast; metabolic and circadian costs repay slowly or not on the same schedule; and the weekend lie-in buys back the first while deepening the third. Anyone telling you it is all recoverable, or none of it, is picking one study.

It is worth being precise, because the number does a lot of work in people's guilt.

The familiar range is the National Sleep Foundation's 2015 recommendation of seven to nine hours for adults. That figure came from a consensus panel reviewing the literature and voting — a defensible way to produce a public-health guideline, and not the same thing as an experiment that measured what an individual body needs. The panel itself flagged that some healthy adults sit outside the range.

The companion claim, that pre-industrial people slept much longer before electric light took it away, has been tested directly. Yetish and colleagues put activity monitors on three hunter-gatherer and horticulturalist groups in Tanzania, Namibia and Bolivia and found average nightly sleep of about six and a half hours, with no group averaging more than seven. They went to bed a few hours after sunset and woke before dawn. Whatever modern life has done to sleep, it did not take three hours off a natural nine.

So: seven to nine is a sensible default and a bad identity. What your own body needs sits somewhere in a range, is partly set by the same genetics that set your chronotype, and is best estimated from how the fourth or fifth day of a holiday goes rather than from a number on a chart.

Stop using tiredness as the gauge. It is the one instrument the Van Dongen study showed to be broken. Better indicators: how long it takes you to fall asleep (under five minutes is not a talent, it is a symptom), whether you need an alarm, and how different your free-day wake time is from your weekday one. The Sleep Quality test reads those rather than asking how tired you feel.

Repay in the mornings, not the evenings. Extending a lie-in pushes the clock later and makes Monday worse. Going to bed earlier does not. If there is only one hour available, put it at the front of the night.

Use the nap for the alertness half. A short afternoon nap buys back a good deal of the attention cost without moving the clock, which is the one lever that works mid-week — the entry on napping has the dose.

Know which problem you have. Sleep debt is not getting enough opportunity to sleep. Insomnia is having the opportunity and not being able to use it, and it does not respond to the same advice at all — see sleep hygiene for why the standard list underperforms there, and tired but wired for what that state actually is. If the nights have been bad for weeks, or if you sleep a full eight hours and wake unrefreshed every day, that belongs with a doctor: untreated sleep apnoea looks exactly like sleep debt from the inside and no amount of going to bed earlier touches it.