In a sleep laboratory, 29 healthy adults aged 20 to 50 averaged 94 minutes of deep sleep in one recorded night, while the popular watch they wore that night counted 51. So “how much deep sleep do you need” has two answers, depending on which of those numbers is meant. Neither comes with an official target: sleep recommendations name total hours, not minutes of one stage.
Key takeaways
1. How much deep sleep you need has no official number: the recommendations cover total sleep only, 7 to 9 hours for adults and 7 to 8 for older adults.
2. Healthy sleepers in a lab spent a median of about a fifth of the night in deep sleep at 30 and younger and about a seventh above 60, with wide ranges; women had about twice the slow-wave sleep of men.
3. A watch figure is a different measurement: on one lab night, one watch counted 51 minutes where the lab counted 94.
4. Little deep sleep in the lab, or losing it over the years, went with new high blood pressure and a higher dementia risk, as associations; one low watch night is not a diagnosis.
What counts as deep sleep, in the lab and on a watch
Deep sleep is the sleep lab's stage N3, also called slow-wave sleep: the part of the night in which large, slow brain waves dominate the EEG. It is concentrated early, with slow-wave activity highest in the first sleep cycle and falling steeply after it. Apple Watch reports Awake, REM, Core and Deep; in one validation study, Core was scored as light sleep (lab stages N1 and N2) and Deep as N3.
The lab number comes from polysomnography, the overnight recording of brain waves, eye movements and muscle tone; since 2007 its rules count the former stages 3 and 4 as one stage, N3. Because N3 is defined by a wave pattern, only a recording shows when it happened. So Core is not deep sleep, and no recommendation sets a target for Core either. The stages are not ranked either: deep sleep fills the early cycles, as all four sleep stages at a glance shows.
How much deep sleep do you need? What sleep labs measured
No health agency sets a deep-sleep target: an expert panel recommends 7 to 9 hours of total sleep for adults and 7 to 8 for older adults, with no figure for any stage. Labs offer distributions instead. In 100 healthy sleepers, the median deep-sleep share was 20.7 percent at 30 and younger and 14.9 percent above 60, roughly 86 and 56 minutes; a second lab averaged about 94 minutes in adults aged 20 to 50.
Who stands behind those medians: 100 people aged 19 to 77, chosen from a population sample after a two-step health screening, one night each in a single Austrian laboratory. The groups are small (20 people at 30 and younger, 24 above 60), and the ranges run from lowest to highest value: 15.2 to 37.5 percent, and 2.4 to 35.6 above 60. Each minute figure multiplies a median share by a median total sleep, so it is an approximation. In 35 adults aged 20 to 50 after a week of at least 8 hours' sleep a night, the lab counted 94 to 95 minutes in one night (standard deviation 35 to 38); in 34 young adults over three nights, one deliberately disrupted, lab averages ran from 63.3 to 81.5 minutes.
| Deep sleep | Night | Share of sleep |
|---|---|---|
| 30 min | 7 h | 7.1% |
| 45 min | 7 h | 10.7% |
| 50 min | 8 h | 10.4% |
| 1 h | 7.5 h | 13.3% |
| 3 h | 8 h | 37.5% |
One hour of deep sleep in a 7.5-hour night is 13.3 percent: below the lab median above 60, yet well inside that group's range. Forty-three to 45 minutes in a 7-hour night comes to about 10 to 11 percent (45 minutes: 10.7), also inside the over-60 range, and a watch may be counting low. Thirty minutes in 7 hours, 7.1 percent, is low for every age group in that sample, though still inside the range above 60. Three hours in 8 hours, 37.5 percent, sits at the top of the youngest group's range. High values have known companions: slow waves return denser after a long stretch awake, and alcohol moves deep sleep into the first half of the night while REM shrinks from about two drinks.
Deep sleep by age and sex: why the numbers disagree
The age figure depends on who was measured. In healthy men scored to older rules, slow-wave sleep fell from 18.9 percent at 16 to 25 to 3.4 percent at 36 to 50. In 2,685 women and men recorded at home, women had 106 percent more slow-wave sleep than men on average. And a meta-analysis of 169 studies under current rules found no significant change in N3 share per decade, while total sleep fell 10.1 minutes per decade.
| Group | Median, percent of sleep | Range (minimum to maximum) | About, minutes a night |
|---|---|---|---|
| Age 30 and younger (20 people) | 20.7% | 15.2 to 37.5 | 86 |
| Above 60 (24 people) | 14.9% | 2.4 to 35.6 | 56 |
In the 100-sleeper sample the two age groups overlap widely, and the drop with age held after accounting for sex. Three other data sets explain why popular charts differ so much.
| Who was measured | Finding |
|---|---|
| 149 healthy men aged 16 to 83, older scoring rules | Mean share of slow-wave sleep 18.9 percent at 16 to 25, 3.4 percent at 36 to 50, no further significant drop to 71 to 83 |
| 2,685 women and men aged 37 to 92, home recording, older rules | Women had on average 106 percent more slow-wave sleep than men; poorer sleep with age in men, less consistent in women |
| 169 studies, 5,273 healthy adults, current rules | Change in N3 share per decade of age (minus 0.1 percentage points) not significant |
The men's data come from four laboratories, 1985 to 1999, under rules that count less deep sleep: when the same 72 recordings were rescored (ages 21 to 86, healthy people and patients), current rules found 9.1 minutes, or 2.4 percentage points, more. Those men are also the source some popular pages credit for “a fixed 2 percent less per decade” with bands such as 5 to 10 percent at 60. Their own figures show a steep fall by midlife, then no further significant drop up to ages 71 to 83.
The 106 percent means about twice as much on average, not in every woman; poorer sleep with age showed in the men of that cohort more clearly than in the women. The meta-analysis compares group averages between studies, so minus 0.1 percentage points per decade, not significant, does not prove the absence of decline, and N3 minutes still shrink when total sleep does. An older meta-analysis (65 studies, 3,577 people aged 5 to 102) found slow-wave share falling with age in adults, yet after 60 only sleep efficiency kept falling significantly, and the age effects depended on how carefully people had been screened for illness, medicines and sleep apnea. Followed over time, 346 Framingham participants aged 60 and over lost an average of 0.6 percentage points of slow-wave share a year.
Why deep sleep thins out from midlife
From midlife the slow waves shrink and thin out. The loss is greatest over the front of the head, early in the night: there, older adults' slow-wave activity averaged 75 to 80 percent below young adults'. Waves under the size threshold are scored as lighter sleep, so N3 minutes drop. After a night without sleep, the rebound was smaller in middle age.
Slow-wave activity measures EEG power, not minutes of N3, so 75 to 80 percent lower does not mean three quarters of deep sleep gone. The rebound data come from 29 young adults (mean age 27) and 34 middle-aged adults (mean age 51) who stayed awake for 25 hours and then slept by day. Sleep research has not settled whether older adults simply need less sleep or can no longer produce the sleep they still need.
Why a watch shows less (or more) deep sleep than a sleep lab
Watches and rings estimate sleep stages without recording brain waves. In one lab night, the Apple Watch Series 8 came to 51 minutes of deep sleep against 94 by polysomnography; a Fitbit Sense 2 reached 79, and an Oura Ring Gen3 showed 95 against 95, no significant difference. For single nights, minutes agreed poorly on all three. The useful reading is a week's average on one device.
| Device | Device, minutes | Sleep lab, minutes | Difference |
|---|---|---|---|
| Apple Watch Series 8 (29 people) | 51 | 94 | 43 min fewer |
| Fitbit Sense 2 (33 people) | 79 | 94 | 15 min fewer |
| Oura Ring Gen3 (35 people) | 95 | 95 | no significant difference |
Behind these figures stands one study (35 healthy adults aged 20 to 50, one inpatient night; usable data from 29 people for the watch, 33 and 35 for the other two devices). The watch's 51 minutes are 43 fewer than the lab's, or 54 percent of its figure, and 48.3 percent of the epochs the lab scored as deep it called light. The Fitbit result was 15 minutes short. Agreement per night scored 0.13 to 0.36 on a scale to 1, with that study's device software.
Errors run both ways. In 34 young adults over three lab nights, six consumer devices missed by 6.0 minutes under to 35.9 over on average, mostly over; the “34 minutes too little” quoted online is not in that study. Six other wearables, in 53 adults (mean age 25), matched the lab's stage in only 50 to 65 percent of 30-second epochs. In 96 Japanese adults aged 20 to 70, an Oura Ring Gen3 showed no significant difference in deep-sleep time; a group mean can still hide large errors for one person. Eleven trackers tested on 75 people at a hospital and a sleep clinic reached stage agreement of 0.26 to 0.69 (macro F1, 0 to 1), and a 2026 review of 28 ring studies advised against rings replacing polysomnography for a detailed look at sleep stages.
The lab is noisy too: in 107 cognitively normal adults aged 54 to 84 over two consecutive nights, a change had to exceed 6.6 percent of total sleep to count as real. As an example, your watch shows 40 minutes in a 7-hour night: 9.5 percent. In the 100-sleeper data that sits below the median above 60 (14.9) but inside its range; in the blood-pressure cohort below it would fall in the lowest quarter (under 9.8 percent). Since one watch counted about half the lab's deep sleep, the lab figure could be higher. So read your week's average on the same device and settings, compare weeks rather than people, and treat a small shift as noise. The same devices report a sleeping heart rate too.
Does low deep sleep matter for health?
In cohorts, less N3 went with worse outcomes. In 1,850 adults free of hypertension, the lowest quarter of N3 (under 9.8 percent) had 1.69 times the odds of new hypertension over a mean of 5.3 years, against the third quarter. Above 60, each percentage point of slow-wave sleep lost per year went with a 27 percent higher dementia risk. Reasons for a low number include age, male sex, a watch that under-counts, sleeping pills, sleep apnea and little time awake.
| Share of N3 sleep at baseline | Odds ratio for new hypertension | 95% confidence interval |
|---|---|---|
| Under 9.8 percent N3 | 1.69 | 1.21 to 2.36 |
| 9.8 to 17.7 percent | 1.45 | 1.04 to 2.00 |
| 17.7 to 25.2 percent (reference) | 1.00 | the reference group |
An odds ratio compares odds, not counts of cases: 1.69 means 69 percent higher odds than in the reference quarter (17.7 to 25.2 percent N3), and the second quarter (9.8 to 17.7 percent) came to 1.45. The cohort was 59.4 years old on average, with one home recording at the start; the models allowed for sleep apnea, sleep duration and fragmentation, among others. The dementia figure is a hazard ratio of 1.27 (95 percent confidence interval 1.06 to 1.54) from 346 Framingham participants with 52 dementia cases in 17 years, adjusted for age, sex, APOE e4, smoking and several medicines. Neither study shows that losing deep sleep causes the disease. In the 149 men of the older lab data, growth hormone release tracked slow-wave sleep, independently of age.
Age and male sex lower it, one watch under-counted by 43 minutes, and most prescribed sleeping pills dampen slow-wave activity. In the 2,685-person cohort, sleep apnea went with lighter, poorer sleep, as did current smoking. Little time or activity awake before bed leaves less sleep pressure to discharge.
What raises deep sleep, and by how much
Deep sleep follows sleep pressure: it responds to how long and how intensely someone was awake, so more time and activity awake and less dozing raise it. Exercise adds little: one evening session raised slow-wave sleep by 1.3 percentage points against none, and a newer analysis found only a non-significant trend. Alcohol shifts deep sleep into the first half of the night and costs REM from about two drinks.
Across 23 studies in healthy adults, one evening session added 1.3 percentage points, about 5 minutes in a 7-hour night, not the “up to 13 percent” that circulates. Pooling 28 studies, a 2022 network analysis found that evening exercise at any intensity did not measurably disrupt the night; moderate sessions showed plus 0.84 points, not significant (credible interval minus 1.17 to 2.78). Earlier, 66 studies put the effect of a single bout on slow-wave sleep at small.
In healthy volunteers, alcohol raised slow-wave sleep in the first half of the night and disrupted the second. A 2025 meta-analysis of 27 studies found REM delayed and shortened from about two drinks, with faster entry into N3 only after about five. Benzodiazepines shorten deep sleep and REM. Alcohol and short nights come up again in what moves HRV overnight.
Vigorous exercise ending an hour or less before bed may impair falling asleep, so hard sessions are better ended earlier, and your huuman Coach plans every session into your week. Turning these levers into a routine, with sizes for each, is the job of a separate guide on how to get more deep sleep.
More health topics to explore
References
- Hirshkowitz M et al. — National Sleep Foundation's sleep time duration recommendations: methodology... (2015)
- Mander BA et al. — Sleep and Human Aging (2017)
- Robbins R et al. — Accuracy of Three Commercial Wearable Devices for Sleep Tracking in Healthy... (2024)
- Mitterling T et al. — Sleep and Respiration in 100 Healthy Caucasian Sleepers--A Polysomnographic... (2015)
- Chinoy ED et al. — Performance of seven consumer sleep-tracking devices compared with... (2021)
- Van Cauter E et al. — Age-related changes in slow wave sleep and REM sleep and relationship with... (2000)
- Redline S et al. — The effects of age, sex, ethnicity, and sleep-disordered breathing on sleep... (2004)
- Boulos MI et al. — Normal polysomnography parameters in healthy adults: a systematic review and... (2019)
- Moser D et al. — Sleep classification according to AASM and Rechtschaffen & Kales: effects on... (2009)
- Ohayon MM et al. — Meta-analysis of quantitative sleep parameters from childhood to old age in... (2004)
- Lafortune M et al. — Reduced slow-wave rebound during daytime recovery sleep in middle-aged subjects (2012)
- Miller DJ et al. — A Validation of Six Wearable Devices for Estimating Sleep, Heart Rate and... (2022)
- Svensson T et al. — Validity and reliability of the Oura Ring Generation 3 (Gen3) with Oura... (2024)
- Lee T et al. — Accuracy of 11 Wearable, Nearable, and Airable Consumer Sleep Trackers:... (2023)
- Jin Y et al. — Performance evaluation of finger-worn devices for sleep stage classification... (2026)
- Mullins AE et al. — The stability of slow-wave sleep and EEG oscillations across two consecutive... (2025)
- Javaheri S et al. — Slow-Wave Sleep Is Associated With Incident Hypertension: The Sleep Heart... (2018)
- Himali JJ et al. — Association Between Slow-Wave Sleep Loss and Incident Dementia (2023)
- Dijk DJ et al. — Slow-wave sleep deficiency and enhancement: implications for insomnia and... (2010)
- Stutz J et al. — Effects of Evening Exercise on Sleep in Healthy Participants: A Systematic... (2019)
- Yue T et al. — Different Intensities of Evening Exercise on Sleep in Healthy Adults: A... (2022)
- Kredlow MA et al. — The effects of physical activity on sleep: a meta-analytic review (2015)
- Ebrahim IO et al. — Alcohol and sleep I: effects on normal sleep (2013)
- Gardiner C et al. — The effect of alcohol on subsequent sleep in healthy adults: A systematic... (2025)
- de Mendonça FMR et al. — Benzodiazepines and Sleep Architecture: A Systematic Review (2023)
About this article · Written by the huuman Team. Our content is based on peer-reviewed research and clinical guidelines. We follow editorial standards grounded in scientific evidence.
This article is for educational purposes only and does not constitute medical advice. Health and training decisions should be discussed with qualified professionals.
