I go to bed late by choice. Keeping a rhythm is the thing I find hardest, and I do not feel tired for it. For a long time I took that as proof that I had no sleep debt. The studies below say the opposite: not feeling it is the rule, not the exception.
Sleep debt is one of the rare health topics where the laboratory has run exactly the experiment you would want: take healthy people, remove one, two or four hours of sleep a night for days, measure what degrades, then let them sleep and measure what comes back. The answers are clear, and they contradict almost everything we feel.
the equivalent in attention deficits after 14 nights of 6 h in bed, in 48 adults who did not notice
Van Dongen 2003
of unrestricted sleep to clear one hour of debt, in 15 young men who said they were rested
Kitamura 2016
people followed for 13 years: short on weekdays and long at the weekend, mortality equal to the reference
Åkerstedt 2019
What the debt does, night after night
The reference experiment dates from 2003. Forty-eight adults aged 21 to 38 were randomised to spend fourteen nights at eight, six or four hours in bed, under continuous monitoring; another group went three nights without sleep. At six hours as at four, deficits accumulated day after day on every task, in proportion to the dose. After two weeks at six hours, the level of attention lapses matched that of two consecutive all-nighters.
The same year, a second team tested 66 volunteers at three, five, seven or nine hours in bed for seven nights. At seven hours, reaction speed first fell and then stabilised at a reduced level; at five hours, lapses were added; at three hours, everything degraded with no plateau. The authors conclude that the brain adapts to moderate restriction, but by running at reduced capacity.
What degrades first is sustained attention. The meta-analysis by Lim and Dinges, 70 articles and 147 tests, ranks the effects of less than 48 hours of deprivation: lapses in simple attention are the strongest effect, reasoning accuracy the weakest. You miss things before you reason badly, which explains why you do not notice.
Why you cannot feel it
In the 2003 experiment, felt sleepiness reacted on the first day, then barely moved, and did not even distinguish the six-hour group from the four-hour group. Meanwhile, lapses kept climbing. The authors write that participants were largely unaware of their growing deficits, and that this is probably why chronic restriction passes for benign.
A Japanese study turned the problem around. Fifteen young men aged 20 to 26, who slept 7.4 hours a night on average at home and complained of nothing, were given twelve hours in bed for nine nights. On the first night they slept 10.6 hours, 3.2 hours more than usual. Then the duration came down, night after night, towards an equilibrium value specific to each: 8.4 hours on average, from 7.3 to 9.3 depending on the person. Thirteen out of fifteen slept less at home than their optimum, with an average debt of one hour a night, and up to three. They had no awareness of it.
Add a difference between people that cannot be guessed either. Twenty-one adults went through 36 hours without sleep three times, two weeks apart, after a week of short or long nights. The differences in resistance between individuals were large, stable from one session to the next, and independent of sleep history: a trait vulnerability, not a state. What you can take is not what your neighbour can take, and neither of you knows before measuring it.
What gets repaid, and in how many nights
Repayment was measured with the same care as the debt. In 159 adults, five four-hour nights were followed by a single recovery night of zero to ten hours in bed, drawn by lot. The longer the night, the clearer the recovery, but at ten hours in bed, nine hours of actual sleep, attention lapses, sleepiness and fatigue had not returned to their starting level. One night, even a long one, is not enough.
With four nights of twelve hours in bed, in 83 adults after five four-hour nights, almost everything came back from the first recovery night, except lapses and speed on the attention test, which never fully recovered. In the 2003 seven-night experiment, the five- and seven-hour groups showed no sign of recovery during the three eight-hour nights that followed: they stayed at the reduced level at which they had stabilised.
Outside the laboratory, the picture is the same. A Polish team followed adults for 21 days of normal life: four baseline days, ten days of sleep cut by 30% relative to each person's need, then seven days of recovery, with daily electroencephalography. Every measure degraded during restriction. After a week of recovery, only one, mean reaction time, had returned to its starting value. And in the Japanese study, one hour of debt took four days of unrestricted sleep to clear.
So what about the weekend?
It is everyone's strategy, and it has been tested. Thirty-six healthy young adults were split into three groups: nine hours of sleep, five hours for nine nights, or five hours on weekdays then two unrestricted weekend nights then two short nights. The weekend brought only about 1.1 hours of extra sleep in total. It reduced evening snacking at the time, but as soon as the short nights resumed, the body clock was delayed, snacking and weight rose again, and insulin sensitivity had dropped by 9 to 27% depending on the tissue. The group without a weekend had lost 13% on the same index. The catch-up prevented nothing.
And yet the Swedish cohort of 43,880 people followed for thirteen years says something else, at another scale. Before 65, sleeping five hours or less every day raised mortality by 65% compared with six or seven hours, and short weekend sleep alone raised it by 52%. But short weekday sleep with long weekend sleep did not differ from the reference. After 65, no association at all. The authors conclude cautiously that long weekend sleep may compensate for the week's shortfall.
The two results fit together once you read what each measures. The cohort speaks of survival over thirteen years, from a single self-reported duration. The laboratory speaks of what the body does during the week after the weekend: insulin, weight, clock. The lie-in may protect the first. It does not repair the second.
How to measure yours
There is a simple test, taken from the Japanese study. The first-night rebound, that is how much longer you sleep than usual when nothing wakes you, was closely linked to the accumulated debt: the larger the debt, the longer the first night. In people who said they were rested, that rebound exceeded three hours. The protocol, at home: one night without an alarm, without alcohol, without late screens, and a sleep-duration figure to compare with your weekday average.
Your watch can supply that figure, with one caveat. Across seven consumer devices compared with polysomnography in 34 young adults, sleep detection was good, wake detection average, and stage estimation inconsistent from one device to the next. Total duration is therefore usable for a rebound or an average. The displayed stages and recovery scores are not.
For the order of magnitude, the guidelines give seven to nine hours for an adult, adding that the need varies from person to person. In France, in a representative sample of 1,004 adults aged 25 to 45, 18% slept less than six hours on weekdays, and 20% reported sleeping at least an hour and a half less than they would need to be in good shape.
What I do
| Rule | Why | Source |
|---|---|---|
| Do not trust the feeling | Felt sleepiness plateaus in two days, lapses keep climbing | Van Dongen 2003 |
| Measure the rebound of the first free night | The rebound closely tracks the accumulated debt, even in people who say they are rested | Kitamura 2016 |
| Count in nights, not in one lie-in | One 10 h night is not enough after five short nights; one hour of debt takes four days | Banks 2010, Kitamura 2016 |
| Use the watch's duration, not its stages | Sleep detection good, stages inconsistent across seven devices | Chinoy 2021 |
I will not give a debt figure here. What I have are watch durations and an impression, and the impression, as we have just seen, is the worst of instruments. What I take away: bedtime regularity weighs more than I thought, and creatine after a short night is a patch on one day, not a repayment.
What this article does not say
- That everyone needs eight hours. The optimum ranged from 7.3 to 9.3 hours in fifteen people, and vulnerability to sleep loss is a stable trait that varies between individuals.
- That the debt is never repaid. Most measures come back within one to four nights. It is the attention lapses that linger, and that is precisely what you do not notice.
- That the weekend is useless. It was not associated with excess mortality in a cohort of 43,880 people. It does not prevent the week's metabolic effects in the laboratory.
- That your watch measures your debt. It measures a duration, and that is already useful. Stages and recovery scores are another matter, and another article.
- That I have measured mine properly. I describe a test taken from a study, not a personal result.
« Sleep debt cannot be felt. It is read in what you sleep when nothing stops you. »
What Helix does
Sleep is the variable where feeling misleads the most, and where a simple measurement is enough to correct it. Helix reads the watch's sleep duration, not its stages, computes the weekly average and the rebound of free nights, and translates the gap into debt rather than into a recovery score.
When the debt builds up, the engine says so in nights, not in percentages, and it does not count a long weekend as a full repayment. When a free night is missing to measure a rebound, it says that too.
A sleep debt counted in nights, from what you actually sleep.
Join the beta