Name: Sleep Duration Recommendations
Creator: Realised
Published: 2026-06-01

Summary

Most adults need somewhere in the 7-9 hour range of actual sleep, the floor (≈7h) is better established than the ceiling, both ends of the famous U-shaped mortality curve are real in the data but the long-sleep end is largely poor health showing up as a symptom rather than long sleep causing death — and while a real "natural short sleeper" gene exists, it is so rare (on the order of dozens of confirmed families worldwide) that if you are reading this and feel fine on five hours, the overwhelmingly likely explanation is that you are sleep-deprived and have stopped noticing.

Why Strong

"~7-9h is the right default; habitual <7h harms almost everyone" — Tier 1
• because: convergent expert consensus (AASM/SRS; NSF) built from structured review, plus replicated observational data and experimental sleep-restriction physiology that gives the short-sleep harm a believable causal engine. The direction is not seriously disputed.
• NOT a higher claim of precision because: the exact optimum and upper bound are measurement-fragile (below).

Short-sleep arm of the mortality U-curve — Tier 2
• because: replicated across large cohorts, present with objective measurement (Kripke 2011), mechanistically coherent.
• NOT Tier 1 because: most data are self-report; residual confounding (shift work, comorbidity, socioeconomics) is not fully excluded; MR evidence is mixed.
• NOT Tier 3 because: it survives better measurement and has experimental mechanistic support — more than "limited studies + plausibility."

**Long-sleep arm as a causal claim — Tier 3 / partly artefactual**
• because: it is a real and replicated association.
• NOT Tier 2 because: it attenuates or disappears under tight control for baseline health, is concentrated in self-report data, lacks a clean direct mechanism, and is best explained by reverse causation. We tier the association at Tier 2 but the causal interpretation at Tier 3 and flag it.

Natural short-sleeper genetics — Tier 2 (existence) / EMERGING (no-harm)
• existence is Tier 2: identified mutations (DEC2, ADRB1, NPSR1, GRM1), characterised wake-promoting mechanisms, replicated across labs.
• "no harm / healthy aging" is EMERGING: tiny samples, partial reliance on subjective report, animal-model lifespan data not yet matched by large human longitudinal safety data.

Regularity > duration for mortality — Tier 2
• because: a large objective UK Biobank analysis (Windred 2024) with a coherent circadian rationale.
• NOT Tier 1: single flagship cohort, awaiting wide independent objective replication.

Practical takeaway

The default to start from. Aim for 7-9 hours of actual sleep (most adults land 7-8.5h). Because time-in-bed overstates sleep, that usually means ~7.5-9.25 hours in bed to bank 7-9h asleep. The floor (~7h) is firmer than the ceiling; do not fear the upper end of the band.

Find YOUR number — the unmasking protocol (the single most useful action here).
1. Pick a low-demand stretch (holiday or 2-3 light weeks) where you can wake without an alarm.
2. Go to bed when genuinely sleepy; let yourself wake naturally. The first week you will likely oversleep — that is you repaying existing sleep debt (sleep_debt_payback), not your true need. Don't conclude anything from week one.
3. By weeks 2-3 the wake-up time stabilises. The duration you settle into, waking refreshed, alert through the day without caffeine-rescue or an afternoon crash, is your need. For most people it lands inside 7-9h.

What "the right amount" looks like (track these, not just the clock):
• Wake feeling reasonably restored most mornings; alert through mid-afternoon without needing caffeine to function (a coffee you enjoy is fine; a coffee you need to think is a deprivation flag).
• No reliance on weekend catch-up sleep or "sleeping in to recover" — large weekday/weekend gaps ("social jetlag") signal weekday under-sleep (sleep_consistency).
• Falling asleep in 15-20 minutes, not in under 5 (instant sleep onset = likely under-slept) and not in 45+ (possible orthosomnia, mistimed bedtime, or going to bed before sleepy).
• Stable mood and emotional regulation; intact focus and working memory.

Prioritise regularity over hitting the exact number. Per Windred 2024, a consistent 7 hours likely beats a ragged 6-to-9 average. Anchor your wake time first (it's the strongest circadian lever; circadian_rhythm_optimization, morning_sunlight_exposure), let bedtime follow your genuine sleepiness, and keep both within roughly an hour day to day, weekends included.

If you genuinely need a lot (regularly 9h+ to feel okay) — investigate, don't restrict. Sudden or rising long-sleep need, or 9-10h that still leaves you unrefreshed, is a flag to get checked: screen for depression, OSA (snoring, witnessed apnoeas, morning headache, unrefreshing sleep — osa_diagnostic_lifestyle), anaemia, and thyroid issues. Quality fragmentation, not quantity, is the usual culprit. Do not force yourself to sleep less to chase a "healthier" number — that inverts the evidence.

If you suspect you're a natural short sleeper. The bar is high and lifelong: 4-6h since childhood/early adulthood, with no caffeine dependence, no catch-up sleeping, no daytime sleepiness, no mood/cognitive cost — sustained for years, ideally across people in your family line. If that is not unmistakably you, assume you are not one and treat your short sleep as a deficit to close. Self-diagnosis here is the most expensive mistake in this entry.

Response windows. Subjective improvement from closing a chronic deficit shows within 3-7 nights of adequate sleep; full recovery of accumulated debt can take weeks (sleep_debt_payback). Finding your stable natural duration takes the 2-3 weeks of the unmasking protocol.

Evidence detail

Why This Entry Exists

A Realised user asks one of these, and the honest answer is more interesting than the slogan:
• "Do I really need 8 hours? I feel fine on 6."
• "I read that 'you need 8 hours' is a myth invented by the lightbulb / the mattress industry."
• "I have the short-sleeper gene" (almost always self-diagnosed, almost never true).
• "Is sleeping 9 hours bad for me? A study said long sleepers die younger."

This entry exists to hold the honest middle against two opposite errors. The first error is the rigid eight-hour mandate — the idea that everyone must hit a specific clock number or they have failed, which breeds clock-anxiety and orthosomnia (anxiety about not sleeping perfectly that itself wrecks sleep). The number is a population range, not a personal target you fail. The second, more dangerous error is the fashionable contrarian dismissal — "8 hours is industrial nonsense, sleep less, grind more" — which weaponises a few legitimate nuances (individual variation, historical segmented sleep, the rare short-sleeper gene) to license chronic under-sleeping that the evidence says is genuinely harmful for almost everyone.

It also exists to do something the slogan can't: separate what the duration number actually predicts from what causes what. The U-shaped mortality curve is one of the most-cited findings in sleep science and one of the most misread. Realised's job is to give you the real shape, name which arm is solid and which is mostly an artefact, and then point you at the thing that probably matters more than your average duration anyway — regularity (see sleep_consistency).

What bad advice does this protect against? The slogan that everyone needs exactly 8h on the dot (breeds orthosomnia). The opposite slogan that 8h is a myth and you can train yourself down (breeds chronic deprivation dressed up as discipline). The self-flattering belief that you personally are a natural short sleeper. The fear that 9 hours is killing you. And the conflation of "time in bed" with "sleep" — the recommendations are for actual sleep, and most people's time-in-bed overstates it by 30-60 minutes (see sleep_architecture_and_stages).

(Boundary note: why_sleep_matters owns the consequences of poor sleep across organ systems and is the home of the broad mortality/CVD/cancer evidence; this entry owns the how-much question, the shape and honest causal reading of the duration curve, individual variation, and the short-sleeper genetics. sleep_debt_payback owns recovery from a deficit; sleep_consistency owns regularity, which this entry defers to as probably the bigger lever.)

Evidence

Read the duration↔outcome literature with one caveat held throughout: most of it measures self-reported sleep, which correlates only modestly with physiologically measured sleep, and is contaminated by reverse causation (sick people sleep differently because they are sick). Those two facts shape every number below.
A. The recommendation itself — "7 or more hours" (Tier 1 consensus, Tier 2 underlying outcome data)

The anchor expert consensus.
Watson NF, Badr MS, Belenky G, et al. (2015). "Recommended Amount of Sleep for a Healthy Adult: A Joint Consensus Statement of the American Academy of Sleep Medicine and Sleep Research Society." SLEEP / J Clin Sleep Med. A 12-month structured process by a 15-member expert panel. Professional-society / academic (non-industry).
• Headline recommendation: adults aged 18-60 should regularly sleep 7 or more hours per night to support optimal health. Note the exact wording — it is a floor ("7 or more"), not the popular "7-9 box." The panel was confident sleeping <7h habitually is associated with adverse outcomes; it was deliberately more cautious about an upper bound, flagging that >9h "may be appropriate" for young adults, those recovering from sleep debt, and the ill, and that the harms of long sleep are less clearly causal.
• The widely-quoted "7-9 hours" range comes from the National Sleep Foundation's parallel consensus (Hirshkowitz et al., 2015, Sleep Health; same era, expert-panel methodology), which gives an age-banded "recommended" band of 7-9h for adults 18-64 and 7-8h for 65+, with "may be appropriate" wings on either side. So "7-9" is a reasonable practical band; "7 or more, and don't read too much into the ceiling" is the more defensible scientific statement.

Funding/bias note: these are professional-society consensus statements, not industry products. The recurrent internet claim that the 8-hour figure was invented to sell mattresses or by Edison's lightbulb is addressed in CONTROVERSY — it is false as stated, though it gestures at a real point about historical sleep patterns.
B. The U-shaped mortality curve — real in the data, but the two arms are NOT equally trustworthy (MIXED Tier 2 / Tier 3)

The large meta-analyses that built the U-curve:
• Cappuccio FP, et al. (2010). "Sleep duration and all-cause mortality." SLEEP. 27 cohorts, 1.38 million participants, 112,566 deaths. Independent/academic. Short sleep RR ≈ 1.12 (95% CI 1.06-1.18); long sleep RR ≈ 1.30 (95% CI 1.22-1.38).
• Yin J, et al. (2017). Dose-response meta-analysis. JAHA. ~474,684 participants for CV outcomes. Independent. U-shaped curve, lowest all-cause mortality at ~7h; below 7h, RR ≈ 1.06 per hour lost; above 7h, RR ≈ 1.13 per hour gained. CVD outcomes echo the U (long sleep tracks especially with stroke).
• A 2025 GeroScience meta-analysis (search through Oct 2024) reports short sleep (<7h) HR ≈ 1.14 and long sleep (≥9h) HR ≈ 1.34, the long-sleep signal larger and more pronounced in women.

On their face these say "both too little and too much kill you, optimum ~7h." That is the slogan. Here is the correction.

Correction 1 — the long-sleep arm is largely reverse causation / confounding (this is the single most important nuance in the entry).
Kurina LM, et al. (2013). "Sleep duration and all-cause mortality: a critical review of measurement and associations." Ann Epidemiol. 42 prospective studies, 35 populations. Independent. Verbatim conclusion: it is "premature to conclude, as previous reviews have, that a robust, U-shaped association between sleep duration and mortality risk exists across populations," because most studies use self-report (only modestly correlated with physiologic sleep), and because "careful attention must be paid to measurement, response bias, confounding, and reverse causation." Multiple cohorts that controlled tightly for baseline health found the long-sleep mortality signal attenuates or disappears — i.e., long sleep is frequently a marker of existing illness (cancer, heart failure, depression, infection, frailty, undiagnosed sleep disorder) rather than a cause of death. Long sleep is what sick bodies do.

Correction 2 — objective measurement flips the trust between the arms, and shortens the optimum.
Kripke DF, et al. (2011). "Mortality related to actigraphic long and short sleep." Sleep Med. Actigraphy, not self-report. Academic. When sleep is measured objectively rather than self-reported, objectively-measured durations predicted mortality where self-reported durations did not, the U-shape persists but the optimum sits shorter than the self-report optimum, and the short end carries real risk (≈90% survival at the objective optimum vs markedly lower at the short and long extremes). The lesson: the short-sleep harm signal is the one that survives better measurement; the long-sleep harm signal is the one that erodes when you control for illness. They are not symmetric.

Net honest reading of the curve:
• Short sleep (<7h) raising risk: Tier 2, holds up. Replicated, present with objective measurement, mechanistically coherent (see MECHANISM and why_sleep_matters), and consistent with experimental sleep-restriction physiology.
• **Long sleep (>9h) as a cause of harm: Tier 3 / partly artefactual.** Real as an association, weak as causation; best read as "if you suddenly need 9-10h, get checked — it's often a symptom (depression, OSA, anaemia, hypothyroidism, infection, cardiac disease), not a habit to fear in itself." The correct response to long sleep is investigation, not forced restriction.
C. Individual variation is real and substantial (Tier 1-2)

Sleep need is genuinely distributed, not a single point. The consensus statements explicitly acknowledge that need is shaped by genetics, age, health, recent sleep debt, physical activity, and life stage. Need is higher in adolescence, during illness, after intense training (see sleep_and_testosterone_the_nonnegotiable_foundation, physical_activity_sleep), and after accumulated debt (see sleep_debt_payback). The practical implication: the band is a starting hypothesis, and your own rested-without-alarm duration is better personal data than the population mean.
D. The rare genetic short sleeper — real, fascinating, and almost certainly not you (Tier 2 for existence, EMERGING for "no harm")

Familial Natural Short Sleep (FNSS).
Fu/Ptáček lab, UCSF — multiple papers. DEC2/BHLHE41 (He et al., 2009, Science); ADRB1 (Shi et al., 2019, Neuron); plus NPSR1 and GRM1/mGluR1. Academic/government-funded. A handful of genetic mutations produce people who sleep ~4-6h apparently without the cognitive or health penalties that under-sleeping normally imposes. Mechanistically they converge on enhanced wake-promoting signalling (DEC2 de-represses orexin; ADRB1 boosts noradrenergic wake drive; NPSR1 enhances arousal) — i.e., the trait is increased efficient wakefulness, not "needing less recovery." Animal work (DEC2 in Drosophila) even suggests the mutation can extend healthspan, which is genuinely interesting biology.

The two honest caveats that make this NOT a licence:
1. It is vanishingly rare. Only on the order of dozens of confirmed FNSS families have ever been identified worldwide; one population estimate for a confirmed short-sleep variant is ~4 per 100,000. The loose claim "about 1% are natural short sleepers" is the optimistic ceiling and includes much softer phenotyping. Your prior probability of being a true FNSS carrier is extremely low.
2. The "no harm" finding is preliminary, not proven. The systematic review of these genes is explicit: sample sizes are tiny ("difficulty finding such subjects"), some evidence leans on subjective report, and the no-detriment conclusion is emerging — based on small families observed "without any related health concerns," not large-population longitudinal safety data. We do not have the evidence to certify long-term safety even for genuine carriers, let alone for someone assuming they are one.

The fatal asymmetry for self-diagnosis: true FNSS carriers feel genuinely, durably fine on little sleep with no caffeine crutch, no weekend catch-up, and no daytime sleepiness — across years. The chronically sleep-deprived also report feeling fine, because sleep restriction blunts the very self-perception that would detect the impairment (people restricted to 6h for two weeks perform like the totally sleep-deprived but rate their sleepiness as only mildly elevated — Van Dongen et al., 2003). "I feel fine on 5 hours" is exactly what both groups say. The feeling cannot distinguish them. The base rates can: one group is one-in-tens-of-thousands; the other is everywhere.
E. Regularity may beat duration anyway (Tier 2, and growing)

Windred DP, et al. (2024). "Sleep regularity is a stronger predictor of mortality risk than sleep duration." SLEEP. UK Biobank, 60,977 participants, ~10 million hours of accelerometry. Independent/academic. More regular sleep was associated with 20-48% lower all-cause mortality across the top SRI quintiles vs the least regular, and sleep regularity out-predicted sleep duration for all-cause mortality in nested models. Translation: a consistent 7 hours probably beats a ragged average that swings between 6 and 9. This is why Realised treats the duration number as necessary-but-not-sufficient and routes hard to sleep_consistency.

Mechanism

Why a floor exists (why short sleep genuinely harms). Sleep is not idle downtime; specific restorative processes are time-gated and don't all complete if the window is cut short. Slow-wave sleep concentrates in the first half of the night and drives glymphatic clearance, growth-hormone release, and metabolic/immune restoration; REM concentrates in the back half and drives emotional and memory consolidation (see sleep_architecture_and_stages). Curtailing total sleep disproportionately amputates the back-loaded stages (you wake before the last REM-rich cycles), so a "lost hour" is not a uniform 1/8 trim — it can remove a disproportionate share of REM. Experimentally, sleep restriction reliably raises sympathetic tone, blood pressure, cortisol, and inflammatory markers (IL-6, CRP), impairs glucose tolerance and insulin sensitivity, increases appetite (ghrelin up, leptin down), and degrades attention, working memory, and emotional regulation — a coherent multi-system picture that matches the short-sleep mortality signal. This is the part of the U-curve with a believable causal engine.

Why the ceiling is mushier (why long sleep is mostly a symptom). There is no clean, dose-dependent mechanism by which an extra hour of physiological sleep in a healthy person damages the body — which is exactly why the long-sleep arm is so suspect as causation. Instead, the conditions that cause people to spend longer in bed and asleep are themselves lethal or pathological: depression (hypersomnia is a core symptom), heart failure and chronic fatigue, anaemia, hypothyroidism, occult infection or inflammation, and fragmented, non-restorative sleep from undiagnosed OSA (long time in bed compensating for poor-quality sleep; see osa_diagnostic_lifestyle). The body sleeps long because it is unwell; the unwellness, not the sleep, raises mortality. Where any direct long-sleep mechanism is proposed (more sedentary time, circadian misalignment, low light exposure, depressive physiology), it is weak and largely shared with the underlying illness.

Why "natural short sleepers" are possible at all. Their mutations bias the brain's arousal systems toward more efficient wakefulness — the same restorative work appears to be accomplished in less time, or the wake-drive is high enough that less sleep is tolerated without the usual penalty. This is a genuinely different mechanism from "training yourself to need less sleep," which does not exist: deliberate restriction does not lower your need, it accrues unpaid debt (sleep_debt_payback) while blunting your ability to perceive the resulting impairment.

Why the feeling lies (the orthosomnia/perception trap). Two perception failures bracket this topic. At the short end, chronic restriction degrades the metacognition that would tell you you're impaired — you feel adapted while measurably declining. At the "I must hit 8h" end, anxious clock-watching and effortful trying activate the arousal system that prevents sleep onset (orthosomnia), so chasing the number can destroy the thing being chased. Both are reasons to treat the number as a soft target and to weight how you feel across a rested fortnight and regularity over hitting a nightly figure.

Risks And Contraindications

• Orthosomnia (the perfectionism risk). Treating the number as a pass/fail target, and especially using a wearable's sleep "score" as a verdict, can create anxiety that itself impairs sleep onset and quality. The number is a soft population guide, not a nightly grade. If sleep tracking is raising your anxiety rather than informing you, stop scoring and judge by daytime function. Realised's register is recovery, not optimisation — this is not a metric to maximise.
• Do not chase the curve into restriction. The single dangerous misreading of the U-curve is "long sleep is bad, so sleep less." For an unwell or genuinely long-sleeping person, forced restriction removes needed recovery and can worsen the underlying condition. Long sleep is a prompt to investigate, not to cut.
• Do not use short-sleeper genetics as permission. Assuming you are a natural short sleeper when you are not means normalising a chronic deficit with real cardiometabolic, cognitive, mood, and safety costs (drowsy driving among them). The trait is rare; the assumption is common and harmful.
• Don't conflate time in bed with sleep. Recommendations are for actual sleep. Spending 9h in bed sleeping 6 is not "long sleep" — it is fragmented or insufficient sleep, often pointing at OSA, pain, anxiety, or alcohol (alcohol_and_sleep).
• Population scope. The 7-9h band is for healthy adults 18-64. Adolescents need more (8-10h); over-65s often consolidate slightly less (7-8h) with more fragmentation; pregnancy, illness, and heavy training raise need. This entry does not set paediatric guidance.
• Older adults — don't over-medicalise normal change. Lighter, more fragmented, somewhat shorter sleep is partly normal ageing, not automatically pathology; but new excessive daytime sleepiness still warrants a look.

Controversy

Three live disputes sit on this topic; Realised takes an explicit position on each.

Controversy 1 — Is the U-shaped duration↔mortality curve causal, or an artefact?
• Position A (the strong-claim reading): both short and long sleep causally raise mortality; aim for ~7h. Best evidence: the large self-report meta-analyses (Cappuccio 2010; Yin 2017), broadly U-shaped and replicated.
• Position B (the methodological critique): the curve is real as an association but fragile as causation — self-report is a poor proxy for physiologic sleep, and the long-sleep arm is largely reverse causation (sick people sleep more). Best evidence: Kurina 2013 (premature to call it robustly U-shaped); cohorts where the long-sleep signal dies after controlling for baseline health; Kripke 2011 (objective measurement keeps the short-sleep risk, shifts the optimum shorter).
• Funding/bias dimension: low commercial stakes either way; this is a genuine methodological dispute, not a money-driven one. If anything, the clean shape is the more clickable headline, so the simple "both ends kill you" story is over-circulated.
• Realised Position: The short arm is real and causal-enough to act on (close deficits). The long arm is mostly your body telling you something is wrong — treat 9-10h need as a symptom to investigate, not a cause to fear or a habit to forcibly cut. Optimum ~7-8h for most, with regularity weighted above exact duration.

Controversy 2 — "8 hours is an Industrial-Era / commercial myth."
• Position A (contrarian): the 8-hour norm is modern invention; pre-industrial humans slept in two segments ("first" and "second sleep," per historian Roger Ekirch) and historical/forager sleep was shorter than assumed (e.g., Yetish et al. 2015 found three pre-industrial societies averaging ~6-7h with no fixed 8h block). Therefore the 8-hour mandate is overstated.
• Position B (mainstream): segmented sleep and ~6.5-7h forager totals are real findings, but they do not show humans thrive on chronic 5-hour nights; the foragers slept consistently, with strong circadian anchoring and daylight exposure, and their totals still land near the lower end of the recommended band, not below it.
• Funding/bias dimension: the contrarian claim is amplified in hustle/productivity culture (sleep-as-weakness framing) and is convenient for anyone wanting to rationalise overwork. The mattress-industry origin story for "8 hours" is essentially folklore and not supported.
• Realised Position: The contrarians are right about a small thing and wrong about the conclusion. Sleep need is a range, not a sacred 8.0, and a single consolidated block is cultural — fine. But "8 is a myth" is routinely (mis)used to justify chronic under-sleeping, which the evidence does not support. The historical data point toward regular, daylight-anchored sleep totalling roughly 7h, which is exactly the modern floor.

Controversy 3 — Short-sleeper genetics as licence to sleep less.
• Position A: "Some people are genetically fine on 4-6h (DEC2/ADRB1), so maybe I am / maybe anyone can train toward it."
• Position B: the trait is real but extraordinarily rare, the no-harm evidence is preliminary/small-N, the mechanism is efficient wakefulness (not reduced recovery need), and training yourself to need less sleep is not a thing — restriction accrues debt while hiding its own cost.
• Realised Position: Celebrate the biology, refuse the licence. Base rates make self-diagnosis almost always wrong, and the perception-blunting of sleep restriction makes "I feel fine" worthless as evidence. Default to the band unless your lifelong, no-caffeine, no-catch-up, no-daytime-sleepiness history is unmistakable.

Cross-Pillar Connections

• Sleep — why_sleep_matters: the home of the downstream-consequence evidence (CVD, metabolic, cancer, cognition, mood) that explains why the short-sleep arm has a causal engine; this entry owns the how-much and the curve's honest shape.
• Sleep — sleep_consistency: the regularity lever this entry repeatedly defers to as probably mattering more than exact duration (Windred 2024). Anchor wake time first.
• Sleep — sleep_debt_payback: why week-one of the unmasking protocol overshoots (debt repayment), and how to recover a deficit rather than train down a need.
• Sleep — sleep_architecture_and_stages: why a "lost hour" disproportionately costs back-loaded REM, and why time-in-bed ≠ sleep.
• Sleep — osa_diagnostic_lifestyle: the prime suspect when someone "needs" 9-10h but wakes unrefreshed — fragmented quality masquerading as long-sleep need.
• Mental (mood/cognition): short-sleep degrades emotional regulation, working memory, and metacognitive self-assessment — the perception-blunting that makes "I feel fine on 5h" untrustworthy (see why_sleep_matters for the cognitive evidence).
• Physical — sleep_and_testosterone_the_nonnegotiable_foundation, physical_activity_sleep: training raises sleep need and short-sleep blunts recovery and hormonal output; the duration question is upstream of physical recovery.
• Diet (metabolic): short sleep impairs glucose tolerance and shifts appetite hormones (ghrelin↑/leptin↓), linking the floor to metabolic and weight outcomes.

What would change our mind

Falsifiability: explicit upgrade/downgrade criteria from source

We would UPGRADE the long-sleep arm toward causal (Tier 3 → Tier 2) if:
• Large objective-measurement (PSG/actigraphy) cohorts with rigorous control for baseline disease and an active illness washout still showed a dose-dependent long-sleep mortality signal — i.e., the signal survived the reverse-causation correction.
• A credible direct mechanism for harm from extra physiological sleep in healthy people were established.

We would DOWNGRADE the short-sleep arm if:
• Mendelian-randomisation or rigorous objective cohorts showed the short-sleep mortality association is itself largely confounded (e.g., by shift work, poverty, comorbidity) rather than reflecting sleep loss per se. (Some MR work already complicates the simple story; this is watched.)

We would revise the short-sleeper-gene framing if:
• Large longitudinal cohorts of confirmed FNSS carriers demonstrated robust long-term health-and-cognition equivalence (would firm the "no harm" claim from EMERGING toward solid), or
• Population genomics showed the trait is far more common than current estimates (would change the base-rate argument against self-diagnosis).

We would soften the "regularity beats duration" emphasis if:
• The Windred-type finding failed to replicate in independent objective cohorts, or duration re-emerged as the dominant predictor once regularity was better measured.

Industry bias note

Structural incentives the evidence base may reflect

Bias risk on this topic is moderate and unusually two-directional, so naming the vectors is the value-add:
• Low classic-suppression risk. Sleeping the right amount is free and unpatentable; there is no pharma incentive to bury it. (The adjacent markets — sleep aids, mattresses, wearables — actually benefit from people caring about sleep, so the structural bias runs toward amplifying sleep concern, not suppressing it.)
• The wearable/score vector inflates the "perfect number" framing. Devices selling nightly sleep "scores" have an incentive to make duration/quality feel like a gradeable performance metric — which feeds orthosomnia. Realised explicitly resists treating the number as a score to maximise.
• **The hustle-culture vector inflates the opposite error. "Sleep is for the weak / 8 hours is a myth / successful people sleep 4 hours" is a productivity-culture meme with no commercial product but a strong ideological pull, and it cherry-picks the genuine nuances (variation, segmented sleep, rare genes) to license harmful under-sleeping. This is the more dangerous bias for a Realised user because it flatters and is socially rewarded.
• The cleanest evidence is the non-conflicted evidence**, and it lands in the honest middle: independent/academic meta-analyses and critiques (Cappuccio, Yin, Kurina, Kripke, Windred) and government-funded genetics (UCSF) — none with a product to sell — converge on: a real ~7h-centred floor, a methodologically soft ceiling, large individual variation, and regularity as a probably-bigger lever. Realised holds all of those rather than the slogan at either pole.

Sleep Duration Recommendations