What Influences REM Sleep? 37 Factors Checked Against the Research
Updated: 1 day ago
Last Updated: October 16, 2026
You slept a full eight hours. Your deep sleep looked fine on the watch. So why do you feel foggy, irritable, and emotionally raw, like you barely slept at all?
That feeling often traces back to REM. The single biggest thing that cuts REM is a short night with a fixed alarm, because REM is back-loaded into the last hours of sleep. After that, the clearest REM suppressors are alcohol (REM delayed 18 minutes and cut 11.3 minutes on average across 27 studies), nicotine, a hot bedroom, altitude and hard late workouts. Several things people blame turn out to be well-measured nulls: caffeine, a late dinner, daytime naps and your mattress did not change REM, and neither did most sleep supplements. Below are 37 factors, each checked against its primary source.
I rebuilt this list on October 15, 2026 after re-checking every source, and a few claims from the first version did not survive. The biggest corrections: caffeine and naps do not reduce REM, and REM barely changes with age. Sample sizes now sit in the table, so the small studies are easy to spot.
REM (rapid eye movement) sleep is the stage tied to dreaming, emotional processing and memory. It makes up roughly a fifth to a quarter of a healthy adult's night and lands mostly in the final hours, which is why it is the easiest stage to lose.
Key takeaways
A short night with a fixed alarm cuts REM first, because REM is back-loaded into the last hours of sleep.
Alcohol is the most reliable REM suppressor: REM delay starts around 2 drinks and REM loss around 3 (Gardiner 2024).
Nicotine, a hot bedroom, altitude, sleeping upright and a hard workout ending close to bed all reduce REM.
Caffeine, a late dinner, daytime naps, your mattress and phone signals did not change REM in controlled studies.
Of five supplements with sleep-lab data, only melatonin (in people with abnormally low REM) and a ketone ester (after very hard training) moved REM.
Age and sex change REM far less than most people assume. Deep sleep is where those differences show up.
Why REM is the easiest stage to lose
REM propensity peaks just after your body temperature low point, on the rising part of the curve, which puts most of it in the last hours of the night (Czeisler 1980). Anything that shortens the back end of your night, or shifts your clock, hits REM first. Jet lag and shift work misalign it for the same reason.
The clearest demonstration: 5 nights of 4 hours in bed instead of 9 cut REM and stage 2 minutes, while deep sleep's share rose (Shechter 2012). Your brain protects deep sleep and lets REM go, then tries to repay the debt. After 3 nights of selective REM deprivation, REM rose to 140% of baseline on the first recovery night (Endo 1998).
Deep sleep sits at the other end of the night and responds to a partly different set of inputs, ranked in our guide to the factors that increase deep sleep.
What influences REM sleep: 37 factors checked
Each row was checked against its primary source on October 15, 2026. "No change" rows are well-measured nulls, not missing data. The sample size sits next to each study, so you can see which findings rest on 6 people and which rest on thousands.
Demographics and body
Factor | REM effect | What it means | What the research found |
Age | Small or no decline | REM changes little with age. Total sleep and sleep efficiency carry most of the age decline | REM% fell about 0.6 points per decade through most of adulthood, levelling off in the mid-70s; a 2019 meta with modern scoring found no significant change (0.0 points per decade) (Floyd 2007 (382 studies, n=4,171); Boulos 2019 (169 studies, n=5,273)) |
Sex / hormones | Little difference | REM is about the same in men and women. Deep sleep is where the sexes differ | REM% did not vary with the share of men across 169 PSG studies; the clear sex gap is deep sleep, which is higher in women (Boulos 2019 (n=5,273); Redline 2004 (SHHS, n=2,685)) |
Menstrual cycle (luteal) | Slightly lower in luteal | REM dips a little in the two weeks before a period, likely because body temperature runs higher | REM was reduced in the mid-luteal phase at specific circadian phases; reviews call it a minor luteal decrease, linked to higher body temperature (Shechter 2010 (n=8); Baker 1999) |
Late pregnancy | Negative | REM tends to be lower late in pregnancy, along with lighter sleep | 27 women in the third trimester had fewer REM minutes than 24 non-pregnant women on lab PSG; small longitudinal studies are mixed (Wilson 2011) |
Obesity (BMI 30+) | Slightly negative | Extra weight is linked to at most a few minutes less REM once sleep apnea is ruled out | With moderate or severe sleep apnea excluded, BMI 30+ had 7.8 min less REM (p=0.003); in 2,745 older men, REM was not linked to body composition (Pessoa 2025 (Baependi, n=1,014); Rao 2009 (MrOS)) |
Acute illness | Negative (lab model) | Fighting an infection can trim REM for a night, but real colds haven't shown a clear REM drop | Endotoxin given in the evening cut REM share in healthy men (P<0.05); the same dose mid-sleep did not; a real cold cut sleep about 23 min with no REM change reported (Pollmächer 1993 (n=15); Haack 2001 (n=10); Drake 2000 (n=21)) |
Genetics | Little to none | How much REM you get looks driven more by circumstances than genes | REM time showed no significant heritability (h2 0.07, p=0.19) while deep sleep did (0.35); some twin studies disagree (Leocadio-Miguel 2025 (n=684 family members)) |
The surprise in this group is how little REM depends on who you are. Age, sex and genes move it far less than they move deep sleep.
Exercise
Factor | REM effect | What it means | What the research found |
Evening intense exercise | Small negative (that night only) | Hard late workouts trim REM a little and push it later. Morning or moderate sessions don't | One hard session (77%+ HRmax, ending 0.5 to 4 h before bed) cut REM by 1.95 points of total sleep vs no exercise; 2.5 points when it ran over 30 to 60 min. Moderate and light sessions: no significant change. Evening exercise also delayed REM onset by 7.7 min. Morning exercise showed no REM change in timing studies (Yue 2022 (network meta, 28 trials, n=325); Frimpong 2021 (-2.34 points); Stutz 2019 (23 studies)) |
Cold water immersion after exercise | No change | An ice bath after an evening workout didn't change REM | No difference in stage proportions or REM onset vs exercise alone (Robey 2013 (n=11)) |
Environment and habits
Factor | REM effect | What it means | What the research found |
Short sleep / early alarm | Negative (REM first) | REM is back-loaded, so a short night with a fixed alarm cuts it first | 4 h vs 9 h for 5 nights cut REM and stage 2 minutes, while deep sleep's share rose (Shechter 2012 (n=27, crossover, PSG)) |
Bedroom heat | Negative | A hot, especially humid, bedroom costs you REM. Staying cool protects it, but going colder won't add more | A 32°C bedroom cut total sleep by 24.8 min and lowered REM vs 22°C; a cooling mattress topper restored 21.4 min of sleep and the lost REM. At normal room temperature, cooling added deep sleep, not REM (Pasquier 2025 (n=15, crossover, PSG); Kräuchi 2024 (n=72)) |
Unfamiliar room (first night) | Negative (first night only) | Your first night in a hotel or new bedroom costs some REM. It recovers by night two | Across 53 studies, the first night in a new sleep setting had less REM and a later REM onset than the second night (p ≤ 0.001); smallest at ages 20 to 29 (Ding 2021 (meta, 53 studies, n=1,422)) |
Continuous background noise | Negative | A constant noise machine or hum can trim REM. Earplugs mostly protect deep sleep, not REM | 50 dB pink noise all night cut REM by 18.6 min (118.1 to 99.5 min); 32 dB wind turbine noise cut REM 11.1 min and delayed it 16.8 min. On-and-off traffic noise mainly cut deep sleep instead (Basner 2026 (n=25, randomized crossover); Smith 2020 (n=48); same research group) |
Evening screens | Negative at 4 h; none at 90 min | Hours of bright screen time before bed can trim REM. A normal evening on your phone did not | 4 h of iPad reading at full brightness for 5 evenings: REM 109 vs 121 min with a printed book (p=0.03). 90 min of phone reading ending 50 min before bed: REM unchanged (about 19 to 20%) (Chang 2015 (n=12, crossover); Höhn 2024 (n=62)) |
Presleep stress | Negative | Stress before bed chops REM up. People whose sleep reacts to stress lose the most | A lab stressor before bed lowered REM% and raised REM density; REM minutes fell only in people with stress-reactive sleep (Feng 2023 (n=62, PSG)) |
Altitude | Negative | Sleeping high cuts REM, and higher cuts more | At 4,559 m, REM and deep sleep share fell across nights 1 to 4 vs 490 m; simulated 2,000 and 4,000 m cut REM share step by step (Nussbaumer-Ochsner 2011 (n=14); Li 2021 (n=25)) |
Sleeping sitting up | Negative | Sleeping upright on a plane or in a car cuts REM most. Recline as far as you can | In a 4 h daytime nap, an upright seat gave 79% less REM than a flat seat; a 40° recline gave 37% less (Roach 2018 (n=6, within-subject)) |
Hot bath right before bed | Negative (first REM period) | Heating your body hard right before bed pushes REM back. An earlier bath didn't | A bath at 21:00 raising core temperature about 2°C cut REM; the same bath at 17:00 changed nothing. 1 h at 41°C ending just before sleep shortened the first REM period (Horne & Shackell 1987 (n=6); Bunnell 1988) |
Sleeping next to a partner | Positive (small) | Sharing a room with your partner was linked to slightly more, less broken REM | REM share 23.0% beside a partner vs 21.0% alone, with fewer REM interruptions (8.5 to 5.4); no other stage changed (Drews 2020 (n=24, 12 couples, within-subject)) |
Presleep learning | Positive (small study) | REM tracks how much you learn. Small study | Students who learned French fastest in a 6-week immersion course showed the biggest rises in REM% (De Koninck 1989 (n=10)) |
REM deprivation | Rebound | Lose REM and your brain pays it back the next night | 3 nights of selective REM deprivation cut REM to 9.2% of baseline; it rose to 140.1% on the first recovery night (Endo 1998 (n=8 men)) |
Circadian timing | Sets when REM happens | Most REM lands in the last hours of the night. Jet lag and shift work misalign it | REM propensity peaks just after the body temperature low point, on the rising part of the curve (Czeisler 1980 (n=10, 306 subject-days)) |
Daytime napping | No change in amount | Naps don't eat into your nighttime REM. A late nap may shift when it starts | Naps from 09:00 to 18:00 did not reduce the next night's REM in young or older adults; an early-evening nap made REM start sooner (Campbell 2005 (n=38); Werth 1996) |
Mattress type | No change | Mattress choice didn't move REM | REM unchanged across 4 mattresses (all P>0.05); deep sleep did change (Chen 2014 (n=16; run by the mattress maker)) |
Phone radio signals | No change | Phone signals haven't been shown to change REM | GSM-type exposure before or during sleep left sleep architecture unchanged (double-blind) (Regel 2007 (n=15); Fritzer 2007 (n=10)) |
Most environment levers cut REM rather than add to it. The two small positives, a partner in the room and heavy learning, come from studies of 24 and 10 people.
Substances and nutrition
Factor | REM effect | What it means | What the research found |
Alcohol | Negative (dose dependent) | Even a couple of drinks delay REM, and each extra drink cuts more | REM started 18.0 min later and ran 11.3 min shorter (REM share -2.8 points). Each 1 g/kg added 30.1 min to REM onset and cut 40.4 min of REM. REM delay began around 0.35 g/kg (about 2 drinks), REM loss around 0.5 g/kg (about 3) (Gardiner 2024 (meta, 27 studies, healthy adults)) |
Nicotine | Negative (dose dependent) | Nicotine cuts REM, and more nicotine cuts more | A 7 or 14 mg patch cut REM time and REM% dose dependently in nonsmokers; a 24 h patch lowered REM% in smokers (Gillin 1994 (n=12); Page 2006 (n=15)) |
Ramadan-style daytime fasting | Negative | Daytime fasting with a pre-dawn meal is linked to less REM, but the 3:30 am wake-up to eat may be the cause | REM share 25.6% at baseline vs 18.6% during Ramadan, with sleep times and calories fixed (p<0.05); a separate lab also found less REM (BaHammam 2013 (n=8); Roky 2001 (n=8)) |
Cannabis / THC | Mixed; rebound on quitting | Heavy THC can suppress dream sleep, and quitting brings vivid dreams back. Typical doses: unclear | High-dose THC reduced REM eye movements and, less so, REM time in small early studies; newer lower-dose studies show mixed or no change; REM rebound on withdrawal is consistent (Velzeboer 2025 (meta, 18 PSG studies); Feinberg 1975 (n=7)) |
Diet (carbs) | Positive (small) | More carbs nudges REM up a little. Low carb shifts sleep toward deep sleep | High-carb meals or diets raised REM time and share vs low-carb; low-carb raised deep sleep; REM latency unchanged (Vlahoyiannis 2021 (meta, 11 short-term trials, n=179)) |
Caffeine | No change | Caffeine doesn't cut REM. It cuts deep sleep | No significant change in REM% (k=19), REM minutes (k=17) or REM latency (k=14); deep sleep fell 11.4 min (Gardiner 2023 (meta)) |
Late dinner | No change | Eating dinner an hour before bed didn't reduce REM | REM 20.8% vs 20.6% (P=.87) with dinner at 22:00 vs 18:00 (the 18:00 arm had a 22:00 snack) (Gu 2020 (n=20, crossover)) |
Alcohol is the standout here, and the nightcap that trims your REM shows up in your recovery numbers too: here is why alcohol crushes your HRV and how long it takes to recover. For carbs, protein, fasting and every other food tested against REM, see foods that affect REM sleep.
Supplements
Factor | REM effect | What it means | What the research found |
Melatonin | Positive (low-REM patients only) | Helps REM only if yours is abnormally low. Otherwise it's a timing aid | 3 mg for 4 weeks raised REM% from about 14.7% to 17.8% (placebo fell to 12.0%) in sleep-disorder patients with low REM; in healthy men REM% did not change (Kunz 2004 (2 RCTs, n=14 total); Rajaratnam 2004 (n=8)) |
Ketone ester after hard exercise | Protects REM (athletes) | In trained cyclists, ketone drinks offset the REM loss from a very hard late training day. Untested in everyday life | A hard training day ending 1 h before bed cut REM from 125 to 93 min; 25 g ketone ester after each session and 30 min before bed (75 g total) kept REM at 117 min (p=0.011 vs placebo) (Robberechts 2023 (n=10, crossover); Poffé 2026 (n=24, 7 weeks)) |
Glycine (3 g) | No change | Glycine may help you fall asleep faster but doesn't change REM | Shortened time to fall asleep and to reach deep sleep; REM latency and stage proportions unchanged (Yamadera 2007 (n=11 poor sleepers, crossover; authors from Ajinomoto, a glycine maker)) |
Valerian | No change | Valerian doesn't shift REM either way | REM unchanged after 28 days of valerian-hops in mild insomnia, and in two small PSG studies (Morin 2005 (n=184, industry funded); Schulz 1994; Balderer 1985) |
CBD (300 mg) | No change | One dose of CBD didn't change REM | REM 17.6% placebo vs 16.6% CBD; REM latency also unchanged (Linares 2018 (n=26, crossover; authors hold CBD patents)) |
You can also browse these factors by category in the REM Sleep Factors explorer.
What the supplement aisle gets wrong about REM
Five supplements in this list have REM measured in a sleep lab, and three of them did nothing to it: glycine, valerian and a single 300 mg dose of CBD. Glycine may help you fall asleep faster, but REM latency and stage proportions did not move.
The two that did move REM come with big conditions. Melatonin raised REM from about 14.7% to 17.8% of the night, but only in sleep-disorder patients whose REM was abnormally low; in healthy men it did not change. A ketone ester protected REM in trained cyclists after a very hard late training day, which says little about an ordinary night.
So if a product page claims it "boosts REM", the honest question is whether it cites REM from a sleep study in people like you. For the same evidence lens across the whole shelf, see supplements ranked by the metric they actually move.
How to protect your REM tonight
Roughly in order of how much they moved REM in the studies above:
Protect the back end of your night. REM is back-loaded, so a short night with a fixed alarm cuts it first.
Skip the nightcap. REM delay starts around 2 drinks and REM loss around 3, and each extra drink cuts more.
Keep the bedroom from getting hot. A 32°C room cut REM compared with 22°C. Staying cool protects REM, but going colder than normal did not add more.
Finish hard workouts earlier. A hard session ending 0.5 to 4 hours before bed trimmed REM that night. Moderate sessions and morning workouts did not.
Keep a steady clock. REM is timed by your circadian rhythm, so jet lag and shift work misalign it.
Watch nicotine and constant noise. Both cut REM in controlled studies.
And you can stop worrying about the afternoon coffee, at least for REM. Caffeine's cost lands on deep sleep instead.
See it in your own data, not just the studies
Research gives you the average. Your body gives you the exception. The studies say alcohol cuts REM by 11 minutes on average, but your personal number could be double that, or you might be unusually resilient. The only way to know is to line up your own inputs against your own REM data.
That is what Kygo does. It connects your food, drinks, and supplement logs to the REM data from your Oura, Apple Health, Garmin, Fitbit, or WHOOP, then runs correlation analysis across 12 to 36 hour windows to surface patterns like "your REM drops an average of 14 minutes on nights after two or more drinks." It does not guess from averages. It checks your data.
REM sleep factors FAQ
What influences REM sleep the most?
How long you sleep and when. REM is back-loaded into the last hours of the night, so a short night with a fixed alarm cuts it before any other stage. After that, the most reliable REM suppressors are alcohol, nicotine, a hot bedroom, altitude and hard late workouts.
Does alcohol reduce REM sleep?
Yes. A 27-study meta-analysis found alcohol delayed REM by 18 minutes and cut 11.3 minutes of REM on average. REM delay began around 0.35 g/kg (about 2 drinks) and REM loss around 0.5 g/kg (about 3), and each extra drink cut more.
Why do I get more REM in the early morning?
REM propensity peaks just after your body temperature low point, on the rising part of the curve, which puts most REM in the last hours of the night. That is also why cutting a night short removes REM first.
Does caffeine reduce REM sleep?
Not in the pooled data. A meta-analysis found no significant change in REM percentage, REM minutes or REM latency. Caffeine's cost lands on deep sleep instead, which fell by 11.4 minutes.
Does melatonin increase REM sleep?
Only if your REM is abnormally low. In sleep-disorder patients with low REM, 3 mg for 4 weeks raised REM from about 14.7% to 17.8% of the night. In healthy men REM% did not change, so for most people melatonin is a timing aid.
Does room temperature affect REM sleep?
Heat does. A 32°C bedroom lowered REM compared with 22°C, and a cooling mattress topper restored the lost REM. At normal room temperature, extra cooling added deep sleep, not REM.
Does exercise help or hurt REM sleep?
Only hard late sessions hurt it, and only that night. One session at 77% or more of max heart rate ending 0.5 to 4 hours before bed cut REM by about 2 points of total sleep and delayed REM onset by 7.7 minutes. Moderate, light and morning sessions showed no REM change.
Do naps reduce nighttime REM?
No. Naps taken between 09:00 and 18:00 did not reduce the next night's REM in young or older adults. An early-evening nap made REM start sooner but did not change the amount.
Why do I get vivid dreams after quitting cannabis?
REM rebounds when you stop. High-dose THC reduced REM eye movements and, less so, REM time in early studies, and a REM rebound on withdrawal is the most consistent finding. Effects at typical lower doses are mixed.
Does REM sleep decline with age?
Very little. An older meta-analysis found REM% fell about 0.6 points per decade through most of adulthood, and a 2019 meta-analysis with modern scoring found no significant change. Total sleep and sleep efficiency carry most of the age decline.
The bottom line
REM is the most back-loaded stage of your night, which makes it the easiest to lose and fairly easy to protect. The research is clearest on a handful of levers: get the full night, skip the nightcap, keep the room from overheating and move hard workouts earlier. Much of what gets blamed or sold for REM, from caffeine and naps to glycine and CBD, did not move it in controlled studies. To see how the same inputs show up in your recovery, compare them with the factors that move your HRV.
Knowing which factors move REM on average is only half the job. The other half is finding out which ones move yours.
Sources
Age: Floyd 2007
Sex / hormones: Boulos 2019
Menstrual cycle (luteal): Shechter 2010
Late pregnancy: Wilson 2011
Obesity (BMI 30+): Pessoa 2025
Acute illness: Pollmächer 1993
Genetics: Leocadio-Miguel 2025
Evening intense exercise: Yue 2022
Cold water immersion after exercise: Robey 2013
Short sleep / early alarm: Shechter 2012
Bedroom heat: Pasquier 2025
Unfamiliar room (first night): Ding 2021
Continuous background noise: Basner 2026
Evening screens: Chang 2015
Presleep stress: Feng 2023
Altitude: Nussbaumer-Ochsner 2011
Sleeping sitting up: Roach 2018
Hot bath right before bed: Horne & Shackell 1987
Sleeping next to a partner: Drews 2020
Presleep learning: De Koninck 1989
REM deprivation: Endo 1998
Circadian timing: Czeisler 1980
Daytime napping: Campbell 2005
Mattress type: Chen 2014
Phone radio signals: Regel 2007
Alcohol: Gardiner 2024
Nicotine: Gillin 1994
Ramadan-style daytime fasting: BaHammam 2013
Cannabis / THC: Velzeboer 2025
Diet (carbs): Vlahoyiannis 2021
Caffeine: Gardiner 2023
Late dinner: Gu 2020
Melatonin: Kunz 2004
Ketone ester after hard exercise: Robberechts 2023
Glycine (3 g): Yamadera 2007
Valerian: Morin 2005
CBD (300 mg): Linares 2018
Disclaimer: Kygo is a personal data aggregation and insights platform designed for informational purposes only. The information provided by Kygo, including correlations, patterns, and trends identified in your data, does not constitute medical advice, diagnosis, or treatment. Always consult a licensed healthcare provider with any questions regarding medical conditions.
Which factor surprised you most, and have you seen it show up in your own REM data? Let us know.