MySleepMate / Sleep Numbers
The source index
Where sleep numbers come from
An index of the numbers people repeat about sleep, each traced back to the primary research it came from, mostly American, because most of the foundational sleep-duration and sleep-staging work was done in American labs and most of the guidance circulating in English descends from American consensus panels. Where the popular version and the source disagree, the gap is stated. Where no source exists at all, that is stated too.
- Every entry names a specific paper, panel or dataset. Not "studies show".
- The number in the source is quoted as the source gives it. Range and all.
- The gap between claim and source gets its own line. That is usually the interesting part.
- Three of the eight claims below do not survive the trace intact.
Sleep advice travels as round numbers. Ninety minutes. Eight hours. A twenty-minute nap. Stop caffeine ten hours before bed. Almost none of it arrives with a citation attached, so I started pulling the numbers apart to find out where each one actually came from. A surprising amount of it does not survive the trip. This page is what I found, and it is the citation the advice should have come with.
Third-year medical student at Manipal Tata Medical College, writing about sleep research. A student, not yet a licensed clinician, so everything here is sourced, and none of it is medical advice. Why you should or should not trust me →
Published 2026-08-31 · Last reviewed 2026-08-31 by Chaitanya Banerjee
Researched against primary sources, drafted with AI assistance, then reviewed and approved by the named author. How we write · Corrections log
How a claim gets traced
I follow the same four steps for every entry, in the same order, so you can check my work rather than take my word for it. If I have read something wrong, this structure should make it obvious where.
- The claim, as it actually circulates. Phrased the way you would hear it, not the way a paper would phrase it.
- Traced to. The specific primary source: authors, title, journal or issuing body, year. If the trail runs out at a blog post, the entry says it runs out at a blog post.
- What the source actually says. The number as published, with its range, its sample and its population.
- The gap. What changed between the paper and the version you heard. Sometimes nothing, and I say so. Usually something.
Where a claim has its own full article on this site, the entry links to it. Where it does not yet, the entry stands on its own and I am working on the article.
I have leaned American here, and that is a deliberate limit rather than a claim of superiority. The sleep-duration guidance circulating in English mostly descends from two American bodies, the American Academy of Sleep Medicine and the Sleep Research Society, and the foundational descriptive work on sleep architecture came out of Stanford, Harvard and a handful of other US labs. Tracing the American chain therefore explains where most of the advice you encounter came from. It does not mean non-American work is absent or lesser, and where I find a better source elsewhere I use it.
The index at a glance
Eight claims, the source each one traces to, and whether it survives the trip. Three of them do not, and those three are the reason I built this page.
| The claim | Traced to | Holds up? |
|---|---|---|
| A sleep cycle is 90 minutes | Carskadon & Dement, Principles and Practice of Sleep Medicine | Partly, 90 is a midpoint of 70–120, published with the range attached |
| You need 8 hours | Watson et al., AASM & SRS joint consensus, JCSM 2015 | No, the American consensus number is 7 or more |
| A power nap is 20 minutes | Rosekind et al., NASA Ames cockpit rest study, 1994–95 | Partly, the protocol was a 40-minute rest window; the famous 34%/54% figures are not printed in the report |
| No caffeine 10 hours before bed | Drake et al., JCSM 2013 (Henry Ford / Wayne State) | No, the trial supports 6 hours; 10 is a margin someone added |
| No screens 1 hour before bed | Chang, Aeschbach, Duffy & Czeisler, PNAS 2015 (Harvard / Brigham) | Direction yes, number no, the study used four hours of reading |
| No alcohol 3 hours before bed | Systematic review and meta-analysis, Sleep Medicine Reviews 2024 | Yes, a fair reading of the clearance and REM data |
| One in three Americans is sleep-deprived | CDC / NCHS Data Brief 559, 2024 NHIS | Yes, 30.5% under 7 hours, though two surveys get quoted interchangeably |
| The 10-3-2-1-0 rule | Craig Ballantyne, HuffPost, 13 May 2016 | No trial at all, it is a productivity formula, never tested as a package |
"A sleep cycle is 90 minutes"
Traced to
Mary A. Carskadon and William C. Dement, "Normal Human Sleep: An Overview," the opening chapter of Principles and Practice of Sleep Medicine (Kryger, Roth & Dement, eds.). Dement founded the Stanford sleep laboratory; this chapter is the descriptive baseline most later writing on sleep architecture rests on.
What the source actually says
NREM and REM sleep alternate with a period of approximately 90 minutes, but the chapter gives the spread in the same breath. The first cycle of the night runs roughly 70 to 100 minutes. Second and later cycles run longer, roughly 90 to 120 minutes. The first REM episode appears about 80 to 100 minutes after sleep onset, and REM periods lengthen as the night goes on.
The gap
The 90 is real, and I want to be fair to it. What went missing in transmission is that it was published as the centre of a distribution, not as an interval you can count in blocks. The range was there from the beginning; the popular version dropped it and kept the midpoint. A five-cycle night at 70 minutes and a five-cycle night at 120 minutes differ by more than four hours.
"You need 8 hours of sleep"
Traced to
Watson NF, Badr MS, Belenky G, et al. "Recommended Amount of Sleep for a Healthy Adult: A Joint Consensus Statement of the American Academy of Sleep Medicine and Sleep Research Society." Journal of Clinical Sleep Medicine 2015;11(6):591–592. The panel used a modified RAND Appropriateness Method over roughly twelve months, concluding in February 2015.
What the source actually says
Adults should sleep 7 or more hours per night on a regular basis to promote optimal health. It is written as a floor, not a target, and the accompanying methodology paper is explicit that sleeping less than 7 hours regularly is associated with adverse outcomes.
The gap
Eight is not in the recommendation. I went looking for it and it is not there. The two American bodies that actually convened a panel on this landed on seven, expressed as a minimum. "Eight hours" predates the consensus, and it converts a floor into a target, which is why so many people who sleep 7 hours 15 minutes believe they are failing at something.
Full article: how much sleep you need, by age →
Or check a specific pair of times against the 7-hour floor →
"A power nap should be 20 minutes"
Traced to
Mark R. Rosekind, R. Curtis Graeber, David F. Dinges, Linda J. Connell and colleagues at the NASA Ames Fatigue Countermeasures Program. Published as NASA Technical Memorandum Crew Factors in Flight Operations IX: Effects of Planned Cockpit Rest on Crew Performance and Alertness in Long-Haul Operations, and held by the US DOT National Transportation Library. Rosekind and colleagues also discussed this work in a broader 1995 review of strategic napping in the Journal of Sleep Research, which is a related paper rather than a restatement of the trial.
What the source actually says
Twenty-one volunteer pilots flying scheduled overnight trans-Pacific Boeing 747 routes. Twelve were given a planned 40-minute rest opportunity during low-workload cruise; nine continued normal duties. Inside that 40-minute window the rest group took about 5.6 minutes to fall asleep and slept a mean of roughly 25.8 minutes.
The outcomes the report actually prints are counts, not percentages, and they are more concrete than the version you usually see. Across the final 90 minutes of flight, the rest group recorded 34 microsleep events against 120 in the no-rest group. On psychomotor vigilance testing, the rest condition produced 81 lapses against 124.
About the 34% and 54% figures
These are the numbers everybody quotes, including me until I checked. They are attached to this study everywhere, and Rosekind himself has described the finding in those terms since.
But they do not appear as a stated result in the technical memorandum. I went looking for the sentence containing them and could not find it in the accessible text of the report. What the report contains are the microsleep and vigilance-lapse counts above. The percentages are a later formulation of that work, repeated onward from secondary coverage, and every page that presents them as a direct quotation from a NASA study, this one included, until 31 August 2026, is passing along something it has not verified.
So the defensible version is: a NASA cockpit-rest study found substantial improvements in pilot performance and alertness after a planned rest period, and that work is widely cited as showing a 34% improvement in performance and 54% in alertness. Not: NASA measured 34% and 54%.
The gap
Three things got smoothed, and the last one bothers me most. First, the protocol was a 40-minute scheduled rest window, not a 20-minute nap; roughly 26 minutes is what the pilots actually slept inside it, and 20 is a tidied-down version of a measurement rather than a designed dose. Second, the population was commercial pilots partway through an overnight long-haul flight, already accumulating fatigue on a disrupted schedule, a genuinely different situation from an office worker napping at 2pm, and the effect sizes should not be assumed to transfer. Third, the two percentages that make this study famous are not in the study.
And the deeper problem: no NASA study compared nap lengths at all. The one experiment that did, Brooks and Lack, 2006, 24 adults each taking every condition, found a 10-minute nap more effective than a 20-minute one, with the 20-minute nap producing nothing measurable until 35 minutes after waking.
Full article: how long a nap should actually be, and the study that compared durations →
"Stop caffeine 10 hours before bed"
Traced to
Drake C, Roehrs T, Shambroom J, Roth T. "Caffeine Effects on Sleep Taken 0, 3, or 6 Hours before Going to Bed." Journal of Clinical Sleep Medicine 2013;9(11):1195–1200. Conducted at the Sleep Disorders & Research Center at Henry Ford Hospital with Wayne State University School of Medicine, Detroit.
What the source actually says
A fixed 400 mg dose of caffeine, roughly two to three cups of coffee, was given at 0, 3 and 6 hours before habitual bedtime, against placebo, with sleep measured at home. Caffeine taken even 6 hours before bed significantly disrupted sleep. The authors' stated conclusion is that this provides empirical support for refraining from substantial caffeine for a minimum of 6 hours before bedtime.
The gap
The best-known American trial on caffeine timing tested a 6-hour cutoff and found it was already too late. It did not test 10 hours at all. Ten is someone's safety margin stacked on top of a 6-hour finding, and it is presented as though a study produced it. The direction of the advice is sound; the specific number is not a result.
Worth noting what the study does not establish: it used one large fixed dose in one group, so it says little about how a single morning cup behaves in a fast metaboliser.
"No screens 1 hour before bed"
Traced to
Chang AM, Aeschbach D, Duffy JF, Czeisler CA. "Evening use of light-emitting eReaders negatively affects sleep, circadian timing, and next-morning alertness." Proceedings of the National Academy of Sciences 2015;112(4):1232–1237. Division of Sleep and Circadian Disorders, Brigham and Women's Hospital, and Harvard Medical School.
What the source actually says
Participants read on a light-emitting eReader in the hours before bed, against reading a printed book. The light-emitting condition suppressed evening melatonin by about 55%, pushed dim-light melatonin onset more than 1.5 hours later, lengthened the time taken to fall asleep, reduced and delayed REM, and left participants less alert the next morning.
The gap
The effect is real and well measured. The number attached to the advice is not from this study. Participants read for four hours before bed, in a controlled laboratory, on a device held at reading distance. "One hour" is a practical compromise someone chose; it is not the exposure the experiment tested, and nothing in the paper identifies one hour as the threshold at which the effect disappears.
"Stop drinking 3 hours before bed"
Traced to
The systematic review and meta-analysis literature on alcohol and subsequent sleep in healthy adults, most recently synthesised in Sleep Medicine Reviews (2024), which pools the controlled dosing studies rather than resting on any single trial.
What the source actually says
Alcohol suppresses REM in the first half of the night and produces rebound and fragmentation in the second, in a clear dose-dependent pattern. Disruption appears at doses as low as roughly two standard drinks, and low doses taken within about 3 hours of bedtime delay REM onset and cut total REM. A moderate dose clears over roughly 4 to 5 hours, which lines up with the point at which the second half of an eight-hour night starts breaking up.
The gap
Not much of one, and I think it is worth saying so plainly rather than manufacturing a problem. This is the entry that survives the trace most intact. Three hours is a defensible reading of the clearance and REM data rather than a number invented for symmetry. If anything the honest version is "three hours is the floor, and more helps."
"One in three Americans is sleep-deprived"
Traced to
The Centers for Disease Control and Prevention, through the National Center for Health Statistics. The current figure comes from NCHS Data Brief No. 559 (April 2026), drawing on the 2024 National Health Interview Survey, a nationally representative household survey of the US civilian noninstitutionalized population. An older and separately quoted figure comes from the Behavioral Risk Factor Surveillance System.
What the source actually says
In the CDC's own words: "In 2024, 30.5% of adults had less than 7 hours of sleep on average in a 24-hour period." The brief defines short sleep duration as 0 to 6 hours. The BRFSS, a separate survey using a 6-hours-or-fewer threshold, put the figure at 33.2% in 2022.
The gap
"One in three" is close enough to be fair, and I checked this one against the brief itself rather than a summary of it. Two small things to keep honest. The CDC measures sleep per 24-hour period, not per night, which is not the same thing for shift workers and nappers, so "30.5% sleep under seven hours a night" is a paraphrase, not a quotation. And the NHIS and BRFSS figures come from different surveys with different thresholds and different years, yet get quoted interchangeably as if they were one measurement. Both are self-reported, which is its own limitation: people are poor estimators of their own sleep.
"The 10-3-2-1-0 rule"
Traced to
Craig Ballantyne, a fitness and productivity author, writing in HuffPost on 13 May 2016, presenting it as part of his "Perfect Day Formula." It spread through Inc. and Medium before going viral on TikTok years later, and is now routinely described as something sleep experts recommend.
What the source actually says
A five-step countdown, quoted as originally published: no caffeine 10 hours before bed, no food or alcohol 3 hours before, no work 2 hours before, no screens 1 hour before, and zero hits of the snooze button. It is framed as a routine for getting more done: the stated goal is waking "well rested and ready for battle," not treating a sleep disorder.
The gap
This is the largest gap on the page. There is no clinical trial of the 10-3-2-1-0 protocol, because it was never a protocol. It is five sleep-hygiene ideas of varying evidential strength, arranged into a countdown by someone who writes about productivity, and it has since acquired a medical authority it never had. Some components hold up well. At least one, the 10-hour caffeine cutoff, overshoots the study it is usually credited to by four hours.
Full article: where the 10-3-2-1-0 rule came from, number by number →
Numbers with no source at all
Some claims I could not trace, because there is nothing at the end of the trail. I am listing them separately rather than quietly dropping them, because "I looked and found nothing" is itself a finding, and leaving them out would make this index look tidier than the evidence actually is.
- "You cannot catch up on lost sleep." Circulates without attribution. The research on recovery sleep is real but considerably more nuanced than the slogan, and the slogan does not trace to a paper making that claim.
- "Everyone dreams four to six times a night." Derived loosely from cycle counts rather than from a study measuring dream frequency, which is much harder to do than it sounds.
- "The hour before midnight is worth two after." An aphorism, widely repeated as physiology. Deep sleep genuinely is front-loaded, which gives the saying a plausible shape, but no source establishes a two-to-one exchange rate.
This page exists to be checked. If I have misread a source, cited the wrong paper, or missed a better one, the corrections log is public and every fix is dated. Tell me at the contact page and it gets logged whether or not it is flattering.
References
- Carskadon MA, Dement WC. Normal Human Sleep: An Overview. In: Kryger MH, Roth T, Dement WC, eds. Principles and Practice of Sleep Medicine. Source for cycle period and the 70–120 minute spread.
- Watson NF, Badr MS, Belenky G, et al. Recommended Amount of Sleep for a Healthy Adult: A Joint Consensus Statement of the American Academy of Sleep Medicine and Sleep Research Society. J Clin Sleep Med. 2015;11(6):591–592. Source for "7 or more hours."
- Rosekind MR, Graeber RC, Dinges DF, Connell LJ, et al. Crew Factors in Flight Operations IX: Effects of Planned Cockpit Rest on Crew Performance and Alertness in Long-Haul Operations. NASA Technical Memorandum (available via the US DOT National Transportation Library). Source for the 40-minute rest protocol, the ~25.8 minutes of actual sleep, the 34-versus-120 microsleep counts and the 81-versus-124 vigilance lapses. Not the printed source of the 34% / 54% figures, which are a later formulation.
- Drake C, Roehrs T, Shambroom J, Roth T. Caffeine Effects on Sleep Taken 0, 3, or 6 Hours before Going to Bed. J Clin Sleep Med. 2013;9(11):1195–1200. Source for the 6-hour caffeine cutoff.
- Chang AM, Aeschbach D, Duffy JF, Czeisler CA. Evening use of light-emitting eReaders negatively affects sleep, circadian timing, and next-morning alertness. PNAS. 2015;112(4):1232–1237. Source for the 55% melatonin suppression figure.
- The effect of alcohol on subsequent sleep in healthy adults: a systematic review and meta-analysis. Sleep Medicine Reviews. 2024. Source for the dose-dependent REM findings.
- National Center for Health Statistics. Short Sleep Duration and Sleep Difficulties Among Adults: United States, 2024. NCHS Data Brief No. 559, April 2026. Source for the 30.5% figure. Verified against the brief itself.
- Ballantyne C. The 10-3-2-1-0 Formula. HuffPost, 13 May 2016. Origin of the 10-3-2-1-0 rule.
Questions people actually ask
Is a sleep cycle 90 minutes or not?
Ninety minutes is the average period, and it is a real number from real research. But Carskadon and Dement published it alongside a range of roughly 70 to 120 minutes, with the first cycle of the night usually the shortest at 70 to 100 minutes. Treating 90 as a fixed block you can count is where the error creeps in.
Do adults need 7 hours of sleep or 8?
Seven, as a minimum. The 2015 joint consensus statement from the American Academy of Sleep Medicine and the Sleep Research Society recommends that adults sleep 7 or more hours per night on a regular basis. "Eight hours" is not the consensus figure and does not appear in the recommendation.
How long before bed should I stop drinking coffee?
The strongest American evidence is Drake et al. 2013, which found that 400 mg of caffeine still measurably disrupted sleep when taken 6 hours before bed, and concluded that people should refrain for a minimum of 6 hours. The commonly repeated 10-hour figure was not tested in that study.
Is the 10-3-2-1-0 sleep rule backed by science?
Not as a package. It originated with Craig Ballantyne, a fitness and productivity author, in a 2016 HuffPost article, and has never been tested as a protocol in a clinical trial. Individual components have research support of varying strength, and the 10-hour caffeine cutoff goes considerably further than the study usually cited for it.
How many Americans do not get enough sleep?
According to CDC / NCHS Data Brief No. 559, drawing on the 2024 National Health Interview Survey, 30.5% of US adults averaged less than 7 hours of sleep per 24-hour period. A separate CDC survey, the BRFSS, put the figure at 33.2% in 2022 using a 6-hour threshold.
Now put a number on your own night
The calculator on this site is built on the 90-minute average documented above, and it states that assumption on the page rather than hiding it. It will show you bedtimes for a given wake-up time, along with what it is guessing at.
Open the sleep calculator