Where the sleep-duration numbers come from

The answer looks settled. Seven to nine hours for adults, nine to twelve for a school-age child, eight to ten for a teenager. Agencies in the United States, Britain and China print numbers in that shape, close enough to look copied.

The reality is narrower. Almost every published figure traces back to four research papers and a handful of government notices from 2015, 2016 and 2019, each built by a different method, for different age bands, with different assumptions about what “hours of sleep” means. Some give a floor, others a range. Two Chinese documents give the same schoolchild different answers.

What follows is source-first.

The recommendations, as published

Two professional bodies published adult ranges in 2015, and they define “adult” differently.

Children and teenagers — four bodies, and the bands do not line up.

Every figure above is transcribed directly; what differs is the process behind it.

Where the adult number comes from

The “seven or more hours” line is the most repeated number here. It comes from a two-page document published in SLEEP in June 2015 as a joint statement of the American Academy of Sleep Medicine and the Sleep Research Society, covering ages 18 to 60 (consensus statement, 2015).

The method is in a companion paper: fifteen experts searched the literature with no date restriction, landing on 5,314 articles; exclusions for age, shift schedules and study type cut that to 1,266; 311 publications reached the panel, each Oxford-graded. Voting ran in three rounds, over one-hour increments (methodology paper, 2015).

Three details matter.

The floor was chosen over a range. A range implies that sleeping above a certain amount is harmful. The panel recorded uncertainty above nine hours and could not agree on how long sleep would cause poor health, so it published the lowest duration agreed appropriate: seven hours.

One grey zone was left open. Consensus held that 6 hours or less was inappropriate and 7 to 9 hours appropriate. On 6 to 7 hours there was no consensus, though the median vote fell in the inappropriate range. On 9 hours and above, appropriateness “could not be ascertained with certainty”.

It is a consensus statement, not a clinical practice guideline. The AASM’s own index says these come from an expert panel using a modified RAND Appropriateness Method “when little direct evidence is available”, and lists four sleep-amount papers, the newest dated November 2016 (AASM, read 2026-10-12).

The AASM board adopted a health advisory with the same seven-or-more position on 2014-02-17, updated 2015-08-14 (advisory) — before the consensus conference met in February 2015.

Children: overlapping bands, different edges

The pediatric statement is the same machinery on children: thirteen experts, a modified RAND process over 864 articles, Oxford grading, ten months, ending February 2016 (consensus statement). It was endorsed by the American Academy of Pediatrics, the Sleep Research Society and the American Association of Sleep Technologists.

It excludes infants under four months, citing “the paucity of published outcomes”: the intended 0-to-12-month band was cut to 4-to-12. And it publishes ranges rather than floors: for teenagers the literature showed worse outcomes both below 8 hours and above 10, while in other age groups the study bands were too wide to support a cut-off at the top (methodology paper).

The disagreements are structural, not numerical. For school-age children the AASM says 9 to 12 hours and the National Sleep Foundation 9 to 11, because the NSF band runs to 13 and the AASM’s stops at 12 (AASM, NSF). For the youngest group, the NSF and WHO both give 14 to 17 hours and China’s 2025 notice 13 to 18 (WHO, 全国爱卫办): same ages, three answers.

What the governments add

Most government pages republish, not re-derive. The CDC’s sleep-amount page prints nine rows whose footnotes point at three sources — the 2015 AASM/SRS adult statement for the 18-to-60 row, the 2016 AASM pediatric statement for the five rows under 18, and the 2015 National Sleep Foundation paper for the 0-to-3-month row and the two rows above 60 (CDC About Sleep, published 2024-05-15, reviewed 2023-11-20). One table, two professional bodies, three papers, age splits — 18–60, 61–64, 65+ — from different documents; read without the footnotes it looks like one CDC position.

Britain’s NHS publishes a shorter version: most healthy adults need around 7 to 9 hours, varying by person (NHS, undated page, read 2026-10-12). The old NHS how-much-sleep page returns a 404 (checked 2026-10-12). Retired numbers circulate too: one NHS trust leaflet issued 2022-06-09 lists 11 to 13 hours for a 3-to-5-year-old, crediting the National Sleep Foundation, whose 2015 update puts that band at 10 to 13 (leaflet).

WHO’s sleep-duration recommendation stops at age five: its 2019 guideline covers under-fives, bundling sleep with physical activity and screen time (WHO, 2019-04-02). The 2020 guidelines for children, adolescents, adults and older adults cover physical activity and sedentary behaviour, and the publication’s own overview does not mention sleep (WHO, 2020-11-25). A WHO adult sleep-duration recommendation is not something I could find.

China runs two tracks. The 2019 Healthy China Initiative set a population target — average adult sleep rising from a 6.5-hour baseline to 7–8 hours, marked 倡导性, an advocacy-type indicator rather than a binding one — and the same indicator table asks for at least 10, 9 and 8 hours a day for primary, junior and senior secondary students (健康中国行动 2019–2030, issued 2019-07-09). In March 2025 the National Patriotic Health Campaign Office published core sleep messages giving school-age children one band, 8 to 10 hours (全国爱卫办, 2025-03-19), while the education ministry’s 2021 notice sets three minimums of 10, 9 and 8 hours for the same pupils (教基厅函〔2021〕11号, dated 2021-03-30).

How firm the numbers are

Neither AASM statement is a graded clinical practice guideline for sleep amount. The NSF panel is more explicit about its basis: panellists noted whether each score rested on convincing scientific evidence, weaker evidence, expert opinion, or their own experience (NSF methodology and results, 2015). The ranges still stand: in June 2026 the foundation reaffirmed all nine after reviewing 133 meta-analyses (NSF, 2026-06-10).

The measurement problem underneath

The recommendations rest mainly on self-reported sleep. The AASM methodology paper lists what that costs: most studies rely on retrospective self-report, which can over- or under-estimate measured sleep; survey questions differ (“typical”, “average”, “weeknight”, “24-hour”); most items have not been psychometrically validated; “short sleep” is defined inconsistently as under 5, 5, 6 or 8 hours; and the measures do not capture regularity, timing or quality. The same paper states that objective laboratory measures “correlate weakly with self-reported sleep duration, which is the method most relevant to clinical, public health, and policy recommendations”.

The National Sleep Foundation flags the gap from the other side: population studies often do not separate time in bed from time asleep, and actual sleep is usually shorter than time in bed, so “this biases data toward higher sleep duration estimates” (NSF, 2015).

The gap is measurable. One validation study compared the NHANES sleep question with wrist-worn accelerometer data in 8,438 adults: self-report averaged 6.88 hours a day, accelerometer 6.01 — over-reporting of about 52 minutes, correlation 0.14 (Lee, Sleep and Breathing 2022). An earlier study of 669 middle-aged adults found the same direction, 6.8 reported against 6.0 measured, correlation 0.47 (Lauderdale, Epidemiology 2008).

Objective measurement then disagrees with itself. A separate analysis of the same NHANES accelerometer files, 11,279 participants aged 6 and over, reported median sleep durations of 7 hours 50 minutes to 8 hours 58 minutes by age group (Scientific Reports 2022) — well above the 6.01-hour mean the validation study calculated from the same waves.

The prevalence figures inherit the problem: they answer a question about reported sleep. The 2014 US survey found 65.2 percent of adults reporting seven hours or more, and estimated 83.6 million adults below that line (MMWR, 2016-02-19). Healthy People 2030 tracks the share getting at least seven hours a day: 72.3 percent baseline in 2020, 69.5 percent in 2024, a 73.3 percent target, status “getting worse” (ODPHP SH-03, read 2026-10-12); a 2024 NHIS brief puts short sleep at 30.5 percent of adults (NCHS Data Brief 559, April 2026). Those two agree, and both describe reported sleep, on a question that varies by instrument — a review of 53 surveys found 44 sleep-duration questions split across nocturnal sleep, the past 24 hours, the major sleep episode, and unaided recall (Barger et al., 2021). China’s 2025 white paper comes from wearables rather than recall: night-time sleep of 6.85 hours on average, from more than 150,000 volunteers (中国睡眠研究会, published 2025-03-25; that figure also appears in the society’s own conference report, an industry-common figure, unverified for that paper’s sample).

The recommendation and the measurement are separate things; only one gets quoted.

How I checked this

Every number above comes from a primary document read on 2026-10-12: the consensus and methodology papers, the CDC and NCHS pages, the NHS page, two WHO guidelines, the Chinese notices, three measurement studies. Undated documents are marked as such.

Two cross-checks ran. The American prevalence figures reconcile exactly: 69.5 percent getting enough sleep in 2024 implies 30.5 percent below seven hours, matching the separate NCHS brief. The two NHANES accelerometer analyses did not; both appear above as they stand. Chinese documents were read in the original, figures transcribed not translated.

What I could not check

• Whether the AASM position changed between 2014 and 2015. Only the updated advisory text is online, no version history.
• Why the CDC’s table splits at 61–64. The row points at the NSF’s 26-to-64 band; I found no CDC note on the split.
• Whether an NHS page on child sleep duration still exists. The old page 404s; the trust leaflets disagree with each other.
• The NSF’s 2015 final report, in full. Only its abstract is public; the discussion sections are behind a paywall, so nothing here rests on its body text.
• The 2025 Chinese white paper’s own text. The society’s page announces it without figures and will not serve the page to a script; the 6.85-hour average and the 150,000-volunteer sample come from the society’s separate conference report as indexed, corroborated by news reports. The sample is volunteers with wearables, not a population sample.
• A WHO recommendation on adult sleep duration. The 2020 overview covers physical activity and sedentary behaviour and does not mention sleep; whether guidance exists elsewhere in WHO’s documents, I could not establish.
• Whether any body has graded these recommendations formally. I found no GRADE-graded sleep-duration guideline; absence of a positive result is not proof of absence.