How Much Sleep Do You Actually Need?

What the guidelines actually say

Two expert exercises converged on nearly the same answer by different routes.

In 2015 the American Academy of Sleep Medicine (AASM) and the Sleep Research Society (SRS) convened 15 sleep specialists who reviewed thousands of published studies linking sleep duration to health outcomes, graded the evidence, and voted in multiple rounds using a modified RAND Appropriateness Method. Their conclusion: adults aged 18 to 60 should sleep 7 or more hours per night on a regular basis. They went further than most guidelines dare and stated explicitly that 6 or fewer hours is inadequate to sustain health and safety.

Separately, the National Sleep Foundation ran a panel that produced age-banded ranges rather than a floor:

Age group	Recommended sleep
Newborns (0–3 months)	14–17 hours
Infants (4–11 months)	12–15 hours
Toddlers (1–2 years)	11–14 hours
Preschoolers (3–5)	10–13 hours
School age (6–13)	9–11 hours
Teenagers (14–17)	8–10 hours
Adults (18–64)	7–9 hours
Older adults (65+)	7–8 hours

Two things here are easy to miss. First, teenagers need more sleep than adults at exactly the life stage when school start times give them less — a scheduling problem, not a discipline problem. Second, older adults do not need dramatically less sleep. The belief that sleep need collapses with age confuses need with ability: older adults sleep more lightly and wake more often, so they often get less, but the requirement barely moves.

Neither panel set a hard upper limit. The AASM/SRS group noted that regularly sleeping more than nine hours may be appropriate for young adults, for anyone repaying accumulated sleep debt, and for people who are unwell.

Why the hours you cut are not average hours

This is the mechanism most summaries leave out, and it changes how you think about a short night.

Sleep runs in cycles of roughly 90 minutes, each moving through light sleep (N1, N2), deep slow-wave sleep (N3), and REM. Crucially, the mix is not constant across the night. Slow-wave sleep is front-loaded — it dominates the first two cycles. REM is back-loaded, with periods that lengthen as the night goes on, so the longest REM stretch of your night is typically the one just before you would naturally wake.

The consequence is arithmetic. Sleep six hours instead of eight and you have not shaved 25% off each stage evenly. You have kept nearly all your slow-wave sleep and deleted a large fraction of your REM, because REM was disproportionately living in the two hours you removed. Anyone who has cut sleep short for a week and noticed that memory and emotional regulation degraded faster than physical energy has felt this asymmetry directly.

It also explains why the popular "sleep in multiples of 90 minutes" trick is only half right. Waking at the end of a cycle does feel better than being hauled out of slow-wave sleep — but cycle length varies between people and between nights, and five well-timed cycles is still 7.5 hours. The tactic optimises the edges. It does not reduce the requirement.

The short-sleeper claim, and the base rate

Some people genuinely thrive on very little sleep. Research led by Ying-Hui Fu's group at UCSF has identified families carrying rare variants — in DEC2 (BHLHE41), and later ADRB1 and others — whose members sleep around four to six hours, show no measurable cognitive cost, and have done so lifelong without alarm clocks or caffeine dependence.

The honest framing is a base-rate problem. The trait is rare — well under 1% of the population — while the number of people who believe they have it is very much larger. The distinguishing features are unforgiving:

• A true short sleeper wakes spontaneously, refreshed, without an alarm.
• They do not sleep longer on weekends, holidays, or during a break.
• They have been this way since childhood, not since their job got demanding.
• They do not rely on caffeine to hold the pattern together.

If you sleep five hours on weekdays and nine on Saturday, that is not a genotype. That is sleep debt with a repayment schedule.

Duration is not the only variable

A newer strand of research suggests we may have over-indexed on total hours. Analyses of large wearable-device cohorts — most prominently work published in 2024 using UK Biobank accelerometer data — find that sleep regularity, meaning consistent bed and wake times, predicts all-cause mortality risk at least as strongly as sleep duration, and in some models more strongly.

This is an association in observational data, so causal claims should stay cautious. But it fits the underlying biology: your circadian system is a clock that entrains to regular signals, and an irregular schedule keeps it permanently mid-adjustment. It also yields the most actionable advice in this whole area, and it is not about hours at all — keep your wake time fixed, weekends included. Wake time anchors the clock far more firmly than bedtime does.

How to find your own number

Guidelines describe populations. Your requirement sits somewhere in a range, and there is a reasonable way to locate it.

Give yourself a two-week window with no early alarm; a holiday works well. For the first few days you will over-sleep, because you are repaying debt — ignore those. By the second week your spontaneous wake time typically stabilises, and the amount you sleep once the debt is cleared is a decent estimate of your actual need.

Then translate need into time in bed, which is the number you can actually control. Nobody is asleep for every minute they are lying down; healthy sleep efficiency runs around 85–90%. To get 7.5 hours of sleep you need roughly 8 to 8.5 hours of opportunity. Budgeting exactly seven hours in bed and expecting seven hours of sleep is the most common planning error in this area.

Signs of a deficit are mundane rather than dramatic: falling asleep within a couple of minutes of lights-out (fast sleep onset is a symptom, not a skill), needing an alarm every single day, and a weekend lie-in of more than an hour.

The honest limits of this evidence

Most of what links sleep duration to long-term health is observational. You cannot randomise thousands of people to six versus eight hours for a decade, so the long-horizon findings on cardiovascular disease and mortality carry the usual caveats about confounding and reverse causation — illness disrupts sleep as readily as poor sleep contributes to illness. Much of the data also relies on self-reported duration, which people estimate badly.

What is experimental, and consistent, is the short-term work. Randomised sleep-restriction studies reliably show degraded attention, slower reaction times, impaired glucose handling, worse mood regulation and reduced immune response within days. And one finding from that literature deserves to be far better known: under chronic restriction, participants' subjective sleepiness plateaus while their objective performance keeps declining. People adapt to feeling tired long before they stop being impaired.

Self-assessment is precisely the instrument you cannot trust here. That is the strongest argument for following the number rather than the feeling.

Further reading: The AASM/SRS position is set out in Seven or more hours of sleep per night: a health necessity for adults, with the full methodology in the Journal of Clinical Sleep Medicine. The National Sleep Foundation's age ranges come from Hirshkowitz et al. (2015) in Sleep Health.

This article is general information, not medical advice. Persistent insomnia, loud snoring with daytime sleepiness, or unrefreshing sleep despite adequate time in bed are worth discussing with a doctor — sleep apnoea in particular is common, under-diagnosed, and treatable.