How Long Does Alcohol Affect Your Sleep? The Dose, the Timing and the Rebound

Short answer: the alcohol itself is gone in a few hours, but the sleep disruption outlasts it. A 2025 meta-analysis of 27 studies found REM sleep damage beginning at two standard drinks, while the sedative effect that actually shortens sleep onset requires about five. Blood alcohol falls at roughly 0.015 g/dL per hour, so a moderate evening clears somewhere between about two and a half and six hours later, well before the second half of the night, which is when the worst of it happens.

The awkward part is that the effect people want and the effect they get are on opposite ends of the dose curve.

The dose response, which almost nobody states correctly

Gardiner and colleagues pooled 27 studies and separated the effects by dose. The result inverts the folk model of a nightcap.

Their conclusion, in their words: "a low dose of alcohol will negatively impact (i.e., reduce) REM sleep. It appears that high doses of alcohol may shorten sleep onset latency, however this likely exacerbates subsequent REM sleep disruption."

So the two-glass evening that feels harmless is doing the REM damage without buying you the sedation. And the amount that genuinely knocks you out is the amount that ruins the rest of the night.

The same analysis found "large uncertainty" in the effect on total sleep time, sleep efficiency and wake after sleep onset. That is honest and worth repeating: the architecture changes are established, the total-hours changes are not.

The two halves of the night

This is the mechanism that explains why you wake at 4am after drinking and cannot get back down.

Alcohol front-loads your sleep. In the first half of the night it increases slow-wave sleep, delays the first REM period, and consolidates sleep. Then, as blood alcohol falls toward zero, the second half becomes fragmented.

Ebrahim and colleagues summarised it as a consistent finding across doses: alcohol causes "a more consolidated first half sleep and an increase in sleep disruption in the second half of sleep." They also note that "the onset of the first REM sleep period is significantly delayed at all doses" and that this is the most reliably observed effect of alcohol on REM.

A controlled study makes the magnitude concrete. Chan and colleagues gave 24 healthy young adults a mean dose of 0.828 g/kg, with 22 providing analysable data, reaching a breath alcohol concentration of 0.084% at lights out. In the first half of the night, REM sleep fell from 13.3% of the period on placebo to 6.56% on alcohol, roughly halved.

The finding that matters most came next. In the second half, wake after sleep onset increased, sleep efficiency fell, slow-wave sleep fell, and there was no REM rebound. The REM you lose in the first half is not repaid later that night.

That study also found that alcohol did not reduce sleep onset latency. Consistent with the dose data above: 0.828 g/kg sits just below the threshold where sedation shows up.

How long until it has cleared

Blood alcohol falls at about 0.015 g/dL per hour on average, with most people between 0.01 and 0.025. That range is the useful part: the fastest metabolisers clear roughly two and a half times faster than the slowest, which is why the same three drinks land completely differently on two people at the same table.

Working from that rate alone, without assuming anything about your body size: if an evening puts you at 0.06 g/dL, you are looking at about four hours to clear at the average rate, and closer to six at the slow end. Peak concentration depends on your body size, your sex and whether you ate, but the descent rate is comparatively stable.

Which sets up the important point. Even if the alcohol is gone by 2am, the second half of the night is still disrupted. Clearance is not recovery. The fragmentation happens as blood alcohol approaches zero, not while it is high.

A low dose of alcohol will negatively impact (i.e., reduce) REM sleep. It appears that high doses of alcohol may shorten sleep onset latency, however this likely exacerbates subsequent REM sleep disruption.

Gardiner et al., Sleep Medicine Reviews, 2025

The nightcap trap, measured

If you drink to sleep, this is the study to know about. Roehrs and Roth gave people with insomnia alcohol before bed and recorded polysomnography.

A 0.6 g/kg dose increased total sleep time and stage 3 to 4 sleep on night 2. By night 6, those effects were gone.

Then the second experiment. After six nights of alcohol pretreatment, participants were given a free choice between alcohol and placebo before bed across seven nights. The alcohol-pretreated group took significantly more refills than the placebo-pretreated group.

The benefit disappears within a week and the pull to drink more does not. The authors call these "the first data to explicitly show the risks associated with the use of alcohol as a 'sleep aid'."

One nuance keeps this from being too neat. A 2024 study running three consecutive alcohol nights found the REM suppression also partly attenuated across nights, from about an 11-minute REM deficit on night 1 to about 4 minutes on subsequent nights, while the distortion of when slow-wave sleep accumulated persisted throughout. So it is not simply that the sedation tolerates and the damage does not. Some of the damage tolerates too. What persists is the misshaping of the night.

Breathing

If you snore or have sleep apnoea, this section matters more than the rest of the article.

A meta-analysis of 14 studies found that alcohol significantly increased the apnoea-hypopnoea index and reduced mean blood oxygen saturation. The increase was much larger in the groups who could least afford it:

Seven additional breathing events per hour, every hour, in people already diagnosed. The reviewers describe alcohol as "a modifiable risk factor that can result in the development or worsening of OSA."

A separate meta-analysis of 21 studies found higher alcohol consumption raised sleep apnoea risk by 25% (RR 1.25, 95% CI 1.13 to 1.38), though it also detected significant publication bias, so treat that specific figure as softer than the experimental data above.

What about "stop drinking three hours before bed"?

You have read this rule everywhere. It has not been tested.

A 2026 review of the literature from 2015 to 2025 addressed it directly: "The often-cited recommendation to avoid alcohol within approximately 3 to 4 h of bedtime should be interpreted as harm-reduction guidance rather than a proven physiological threshold."

There is no experiment comparing, say, drinking at 6pm against drinking at 10pm on measured sleep. The reasoning behind the rule is sound, since more time means more clearance before the biological night, but the specific number is convention rather than evidence.

The closest real-world support comes from a very large wearable dataset: 5,109,185 person-days from 20,968 people. Drinking one drink more than personal average, compared with one less, was associated with nocturnal resting heart rate rising 2.8 bpm in women and 2.4 bpm in men, and heart rate variability falling 3.8 and 3.3 ms respectively. Effects were larger in women and in younger adults. Drinking earlier in the day reduced them, but only slightly, and earlier drinking was also associated with slightly shorter sleep.

That study was funded by the wearable manufacturer, and seven of its eleven authors are company employees. The dataset is extraordinary and the conflict is real; hold both.

What does not help

Drinking water. Mackus and colleagues revisited the dehydration theory and concluded that water "was not effective to alleviate the alcohol hangover", and that hangover and dehydration are "two co-occurring but independent consequences of alcohol consumption." A separate crossover trial adding water during drinking found no difference in ethanol kinetics or in cognitive test scores. Both measured hangover rather than sleep, and nobody has tested whether water before bed changes sleep architecture after alcohol. So the accurate statement is: it does not fix the hangover, and its effect on your sleep has never been measured.

Eating. Food genuinely speeds ethanol clearance. It has never been shown to protect sleep, which is a different claim.

Believing you slept well. Strüven and colleagues tracked 40 adults through three alcohol-free days, three evenings of moderate drinking and three days after. Nocturnal resting heart rate rose from 63.6 to 66.6 bpm and subjective sleep quality fell, while the smartwatch's own sleep architecture readings showed no significant change. Different instruments disagree about this, which is a reason to be careful about concluding from your tracker that alcohol does not affect you.

The guidance changed, and most articles have not caught up

Worth knowing if you are checking your intake against official advice. The Dietary Guidelines for Americans, 2025 to 2030, published in January 2026, removed the numeric daily limits that had stood for years. Where the previous edition specified two drinks or fewer per day for men and one or fewer for women, the current edition says only: "Consume less alcohol for better overall health," followed by a list of people who should avoid it entirely.

There is now no US numeric threshold. If you see "the official limit is two drinks" in an article, it is quoting a superseded edition.

The World Health Organization's position, unchanged since its 2024 fact sheet, is that "there is no form of alcohol consumption that is risk-free" and that "alcohol is an established carcinogen." The Group 1 classification people often attach to that sentence comes from the International Agency for Research on Cancer, not from the fact sheet itself.

What to actually do with this

Separate the two questions. How much you drink is a health question with its own literature. This article answers a narrower one: what tonight's drinks do to tonight's sleep. Even a person entirely comfortable with their drinking benefits from knowing the answer is "more than you would guess, at doses lower than you would guess."

If you drink to fall asleep, that is the specific thing to stop. The evidence on tolerance and escalation is unusually clean, and it comes from people with insomnia rather than from people with a drinking problem.

Earlier is better than later, on general principle. Just do not treat three hours as a safety line that has been established. It has not.

If you snore, the effect on you is several times the average. That is the one group for whom the numbers here are genuinely large.

Test it if you track. Two weeks with your normal pattern, two weeks with nothing after 6pm, comparing nocturnal heart rate rather than the sleep score. Heart rate is the measure that moved most consistently in the real-world data.

Questions people ask

How long does alcohol affect your sleep? Blood alcohol clears at about 0.015 g/dL per hour, with most people between 0.01 and 0.025, so a moderate evening is gone in roughly two and a half to six hours. The disruption outlasts it, because the fragmentation happens in the second half of the night as levels fall toward zero.

How many drinks affect sleep? Roughly two. The meta-analysis found REM disruption beginning at doses up to 0.50 g/kg, about two standard drinks, and worsening from there.

Does alcohol help you sleep? It shortens sleep onset only at high doses, around five drinks, and at that point the rest of the night is substantially worse. At the doses most people drink, it does not shorten sleep onset at all.

Why do I wake up at 3 or 4am after drinking? Alcohol consolidates the first half of the night and fragments the second. Wake after sleep onset rises and sleep efficiency falls in the second half, with no REM rebound to compensate.

Does the effect go away if I drink every night? Partly, and not in a good way. The sedative benefit was gone by night 6 in a controlled trial, while participants pretreated with alcohol went on to choose more of it. Some REM suppression attenuated across nights, but the distortion of sleep architecture persisted.

How long before bed should I stop drinking? The three to four hour rule is sensible convention rather than a tested threshold. No study has compared drinking times on measured sleep, and a 2026 review says so explicitly.

Does drinking water before bed help? Not for hangover, where it has been tested and failed. For sleep specifically, it has never been tested.

This article summarises published sleep research for general information and is not medical advice. If you are using alcohol to fall asleep, drinking more than you intend to, or have symptoms of sleep apnoea such as loud snoring and daytime sleepiness, speak to a doctor.