Every health site publishes the same age-by-age sleep chart. Almost none of them read the document that explains how the numbers were reached — or the paragraph where the panel admits what it could not agree on.
You have seen the chart. Newborns, 14 to 17 hours. Teenagers, 8 to 10. Adults, 7 to 9. It appears on government websites, in pharmacy leaflets, and in roughly every article ever written about sleep, including several that used to appear on this site.
The chart is real. It comes from expert panels convened by the National Sleep Foundation and the American Academy of Sleep Medicine, and the people on those panels were not guessing. But somewhere between the source documents and the thousandth republication, three things got lost: how the numbers were actually produced, what the panels themselves said they were uncertain about, and the fact that the field’s own best evidence has since moved somewhere else entirely.
The adult recommendation is a floor, and that was deliberate
The 2015 joint consensus statement from the American Academy of Sleep Medicine and the Sleep Research Society does not say adults need seven to nine hours. It says: “Adults should sleep 7 or more hours per night on a regular basis to promote optimal health.”
Seven or more. The panel set no upper limit, and it explained why in a companion methodology paper that is one of the more candid documents in the field. Fifteen experts worked through a modified RAND appropriateness method — three rounds of voting on a nine-point scale — after screening 5,314 publications down to 311 they reviewed in detail. On the question of whether too much sleep is harmful, they could not reach agreement. In their words, the panel “was unable to come to a consensus regarding biologically plausible pathways by which long sleep could explicitly cause poor health,” and concluded that the association between long sleep and higher mortality “most likely represented the confounding effects of uncontrolled chronic illness.”
That is a significant finding stated plainly, and it is the opposite of what the chart implies. Every article warning you that oversleeping damages your heart is arguing against the reasoning of the panel whose numbers it is citing.
One panel refused to give a number at all
The AASM’s paediatric panel — thirteen experts, 864 articles, ten months of work — declined to issue any recommendation for infants under four months old. Their reason was the wide range of normal variation and insufficient evidence linking duration to outcomes at that age.
The chart that gets republished has no such gap. It confidently gives 14 to 17 hours for newborns. The number comes from a different panel; the refusal simply does not travel.
The National Sleep Foundation, whose figures those are, attaches its own caveat that rarely survives: it states on its own website that “an additional hour or two on either side of a given range may be appropriate, depending on the person,” and that some people function well at the bottom of a range while others need every minute of the top. Applied honestly, NSF’s acceptable adult band is not seven to nine hours. It is closer to six to eleven.
The two US federal agencies that publish the chart do not even match each other. The CDC gives adults aged 18 to 60 “7 or more hours” and attaches no caveat about individual variation. The National Heart, Lung, and Blood Institute gives adults 7 to 8 hours, and does note that needs vary. Same underlying consensus documents, different published numbers.
The number you would report is wrong by about an hour
Here is the problem underneath all of it. Nearly everything we think we know about sleep duration and health comes from asking people how long they sleep, and people are not good at this.
The CARDIA sleep study put wrist actigraphy on 669 middle-aged adults and compared the readings with what those adults said. Measured sleep averaged 6.0 hours. Reported sleep averaged 6.8. The correlation between the two was 0.47 — real, but loose.
The important part is that the error is not random. People whose measured sleep was five hours over-reported by 1.2 hours. People measuring seven hours over-reported by only 0.4. The error is worst precisely at the short end of the distribution — which is to say, in exactly the group that every “short sleep will kill you” study is about.
Measure sleep properly and the picture changes. A 2023 analysis of 5,027 people in the Sleep Heart Health Study, followed for eleven years, ran the comparison both ways in the same cohort. Using questionnaires, it found the familiar J-shaped curve: risk elevated at both short and long sleep. Using polysomnography on the same people, mortality declined steadily as sleep duration rose, with risk concentrated below five hours and no long-sleep penalty at all.
The right-hand arm of the J-curve — the part that generates the “oversleeping is dangerous” headlines — is substantially an artefact of how the question was asked.
What the objective data points to instead
In 2024, researchers analysed 60,977 UK Biobank participants who had worn accelerometers for a week — more than ten million hours of objectively recorded sleep — and followed them for a mean of 6.3 years. They compared two predictors of death: how long people slept, and how regularly.
Regularity won. Comparing the most to the least regular sleepers gave a hazard ratio of 0.70. Duration gave 0.76. And when the researchers tested whether adding duration to a model that already contained regularity improved it, the answer was no (p = 0.14 to 0.20). Duration contributed nothing once regularity was accounted for.
A Swedish cohort of 43,880 people found something in the same direction: short weekday sleep carried no excess mortality if it was followed by long weekend sleep. And among people aged 65 and over — the group the chart singles out for the narrowest range — there was no significant association with sleep duration at all.
So how much sleep do you need?
Nobody can tell you, and that is the honest answer rather than a rhetorical one. There is no clinical test for an individual’s sleep requirement. Twin studies put the heritability of sleep duration at around 46 per cent, but heritability is a statement about variation across a population, not about you.
The rare genetic short sleepers are real — mutations in DEC2, ADRB1 and NPSR1 have been traced through families by linkage analysis and exome sequencing, and carriers genuinely run on around six hours without accumulating debt. But they are rare familial variants found in pedigrees, worth roughly two hours. Common genetic variation, across 446,118 people, moves sleep duration by about one minute per allele. Feeling fine on six hours is not evidence that you are one of them.
What the evidence does support is more modest and more useful than a number:
- Consistency beats duration. Going to bed and waking at similar times outperformed sleep length as a mortality predictor in the largest objective study we have.
- Daytime function is the better signal. In a meta-analysis of 1.1 million people, one in four slept less than the age-based recommendation — but only 5.8 per cent fell outside the acceptable range, and poor sleep quality was more common than short duration. Among teenagers, half slept less than the recommended 8 to 10 hours while only 18 per cent reported daytime sleepiness.
- Persistent short sleep is worth attention. Below five hours consistently, the signal is real across both self-reported and objectively measured cohorts.
What we still do not know
Plenty. Whether short sleep causes poor health, or tracks something else, is not settled — the consensus panel itself noted that few experiments ran longer than a week, and that the six-to-seven-hour range is barely studied experimentally at all. Whether weekend catch-up genuinely compensates, or simply identifies people whose lives allow them to sleep in, has never been tested experimentally. And the study that flattens the J-curve used a single night of laboratory monitoring in a mostly older group, which is a thin basis for describing habitual sleep.
What can be said with confidence is narrower than the chart suggests: the panels never claimed to know your number, the figure you would report is probably wrong by about an hour, and the thing that best predicts outcomes in the largest objective dataset available is not how long you sleep but how consistently.
This article is general information, not medical advice. Persistent trouble sleeping, unrefreshing sleep, or daytime sleepiness that interferes with your life are worth raising with a doctor — they can indicate sleep apnoea, insomnia disorder or other treatable conditions that no amount of schedule adjustment will fix.
