Two drinks is already enough to cut REM#
A drink before bed really can get you to sleep faster — at a dose most people would describe as a heavy night. At ordinary doses it does something else entirely: it suppresses rapid-eye-movement sleep. In the largest quantitative synthesis of the question, pooling 27 studies, REM sleep was both delayed in onset and reduced in duration starting from a low dose of roughly 0.50 g of ethanol per kilogram of body weight — about two standard drinks — and worsened progressively as the dose climbed1.
The sedation people believe they are buying showed up only at the top of the ladder. Shorter time to fall asleep, and a shorter run-up to deep sleep, appeared in that analysis only at doses of 0.85 g/kg or above — around five standard drinks. Which places the common arrangement, one or two glasses to take the edge off the evening, squarely in the window where you pay the REM cost without collecting the sedative benefit that supposedly justifies it. This article is about that trade, and about the second half of alcohol's reputation — the calories and the next-day appetite — which turns out to be smaller and stranger than the internet suggests. The broader machinery it plugs into, how a bad night reaches your eating at all, is the pillar's subject.
| Dose | Roughly | What the pooled evidence shows |
|---|---|---|
| ≤0.50 g/kg | ~2 standard drinks | REM onset delayed, REM duration reduced; no measurable sedative benefit |
| 0.50–0.85 g/kg | ~2–5 drinks | REM disruption worsens progressively with dose |
| ≥0.85 g/kg | ~5 drinks | Sleep onset and time-to-deep-sleep shorten — alongside the worst REM disruption |
| Any dose | — | Total sleep time, sleep efficiency and wake-after-sleep-onset: too uncertain to call |
Alcohol does not steal your deep sleep. It takes the other kind.#
The folk version of this has the stage wrong. "A nightcap robs you of deep sleep" is close to the reverse of what polysomnography shows in healthy drinkers: across studies, dose ranges, ages and both sexes, the majority found slow-wave sleep increased in the first half of the night relative to baseline, with total-night slow-wave sleep increased at high doses2. That review judged the slow-wave finding more robust than the REM finding, and specifically not an artifact of REM being displaced.
That detail matters more here than it looks, because slow-wave sleep is the stage this cluster keeps pointing at — it is where most of the measurable metabolic action lives. Alcohol largely leaves that stage alone, or adds to it, and takes REM instead. REM's best-documented roles sit in emotional processing and memory consolidation rather than glucose handling, which is a decent prior for expecting alcohol's next-day damage to land on mood, judgement and drive rather than on your resting burn.
The fragmentation everyone quotes is the shakiest part#
Here two syntheses of substantially the same literature reach different conclusions, and what separates them is method rather than data.
Ebrahim's review states that at all dosages alcohol shortens sleep onset, consolidates the first half of the night and increases disruption in the second half. Gardiner's meta-analysis found the sleep-onset effect only at high doses — and, for the disruption claim, reported that the effect on total sleep time, sleep efficiency and wake-after-sleep-onset simply could not be determined, with large uncertainty on all three.
Both teams were reading much the same set of small crossover experiments. The 2013 paper was a qualitative review, which counts which direction studies point; the 2025 paper pooled them with confidence intervals, so disagreement between underpowered trials surfaces as a wide interval instead of as a trend. Vote-counting converts heterogeneity into a conclusion. Pooling converts it into an error bar. The earlier review also drew a formal published critique in the same journal — titled, bluntly, "Alcohol and sleep review: flawed design, methods, and statistics cannot support conclusions"3, which the original authors answered in print. Worth knowing that exchange exists before treating the 2013 conclusions as settled.
The claim everyone repeats — that alcohol shatters the back half of your night — is the one a 27-study pooling could not pin down. The REM suppression, which almost nobody mentions, is the part that survived.
None of this makes your 4am wake-up imaginary. It means the measured support for second-half fragmentation is thinner than the confident version circulating, while the REM result is firm and starts low. If you are going to carry one sentence away, carry that one.
The awakenings that are well evidenced come from your airway#
There is one route by which alcohol converts into night waking with a clean mechanism behind it, and it is a breathing problem rather than a sleep-architecture problem — which is exactly why it is invisible on a hypnogram. Alcohol reduces genioglossal muscle tone, the tone that holds the upper airway open, and raises upper airway resistance. In a systematic review of 31 studies, 21 of which could be pooled, higher alcohol consumption was associated with a 25% higher risk of sleep apnea (RR 1.25, 95% CI 1.13–1.38)4.
Spend that carefully. Heterogeneity was high (I² = 82%), the underlying studies are observational, and the outcome is a sleep apnea diagnosis rather than a count of awakenings in an average drinker. What travels well is the mechanism: relax the muscle keeping the airway open, then lie down, and obstruction gets more likely on precisely the nights you drank. If you already snore, this is the lever with your name on it.
The next-day cost that actually shows up is movement, not appetite#
The biggest real-world look at this used within-person comparisons — each person's drinking nights against their own non-drinking nights — across 5,109,185 person-days from 20,968 wearable users. One drink above someone's personal average, versus one below, raised resting heart rate during sleep by 2.8 bpm in women and 2.4 bpm in men, and lowered heart-rate variability by 3.8 ms and 3.3 ms respectively. Sleep was shorter, the effects were dose-dependent, and they ran larger in women and in adults under 305.
Audit that before you spend it. The study was funded by WHOOP, several authors are WHOOP employees, and six of them hold stock options — all disclosed, and all pointing the same commercial direction, since a finding that the company's recovery metric detects a real physiological insult is a finding that sells the metric. The sleep numbers are device estimates rather than lab-scored polysomnography, and consumer wearables are genuinely unreliable at scoring sleep stages. Take the direction and the sign; leave the decimals.
The interesting result is which next-day variable moved. The received story is that a hangover makes you eat more. What this dataset measured was physical activity, and activity fell after drinking, more so at higher doses. For anyone running a deficit, a day of subtracted incidental movement is a quieter leak than an extra meal — and a more predictable one, because it requires no decision to happen.
The appetite leg is thinner than its reputation. Alcohol does suppress leptin: in 14 healthy non-obese adults given alcohol or water in a randomized crossover, the overnight incremental leptin area fell to 53 ± 18 after alcohol versus 113 ± 15 after water (P < 0.01), and the authors speculated this might explain alcohol's appetite-stimulating effect6. Fourteen people, one night, and leptin is the hormone in this cluster least willing to predict what someone actually eats. The better-supported route needs no hormone at all: alcohol raises food intake in the same sitting, which is where its calorie arithmetic is worked out.
What this is worth against a deficit#
Three costs, of decreasing certainty. Ethanol's own calories are the one you can compute rather than estimate. The REM suppression is well established, begins around two drinks, and is unlikely to change your metabolic rate — its plausible route to your eating runs through the following day's judgement, the same territory as tired food decisions. The next-day intake effect, the one the headline promises, is the least nailed down of the three.
One timing detail is worth having. In that same wearable dataset, drinking 60 minutes earlier than usual was associated with a smaller overnight heart-rate rise and higher variability. It is observational, within-person, and small in magnitude — but it costs nothing to move the last drink earlier, and it is the only modifier in this literature with any supporting numbers attached. That puts alcohol in the same practical family as caffeine, where the timing matters more than most people assume: the substance is not the whole problem, the hour you take it is part of it.
FAQ#
Does a nightcap actually help you sleep?#
Only at doses that hurt you elsewhere. Pooled across 27 studies, shorter time to fall asleep appeared only at roughly 0.85 g/kg or more — about five standard drinks — while REM suppression began around two1. One or two drinks sits in the gap: enough to cut REM, not enough to sedate. The feeling of dropping off faster after a small amount is not something the measured evidence supports.
How many drinks does it take to affect sleep quality?#
About two. That is the low-dose threshold at which REM onset was delayed and REM duration reduced in the meta-analysis, with the effect getting worse as the dose rose1. Notably, deep sleep is not what goes missing — reviews of polysomnography find slow-wave sleep increased in the first half of the night after drinking2.
Does drinking earlier in the evening reduce the damage?#
The available evidence says probably, modestly. In a within-person analysis of nearly 21,000 wearable users, drinking 60 minutes earlier than usual was associated with a lower overnight resting heart rate and higher heart-rate variability5. It is observational and industry-funded, so treat it as a cheap adjustment worth making rather than a proven fix.
Sources#
- Gardiner C, Weakley J, Burke LM, et al. The effect of alcohol on subsequent sleep in healthy adults: A systematic review and meta-analysis. Sleep Med Rev. 2025;80:102030.
- Ebrahim IO, Shapiro CM, Williams AJ, Fenwick PB. Alcohol and sleep I: effects on normal sleep. Alcohol Clin Exp Res. 2013;37(4):539-549.
- Pressman MR, Grunstein RR, Mahowald MW, et al. Alcohol and sleep review: flawed design, methods, and statistics cannot support conclusions. Alcohol Clin Exp Res. 2015;39(5):941-943.
- Simou E, Britton J, Leonardi-Bee J. Alcohol and the risk of sleep apnoea: a systematic review and meta-analysis. Sleep Med. 2018;42:38-46.
- Grosicki GJ, Robinson AT, Joyner MJ, et al. Real-world effects of alcohol on heart rate, sleep, and physical activity by age and sex. PLOS Digit Health. 2026;5(3):e0001284.
- Röjdmark S, Calissendorff J, Brismar K. Alcohol ingestion decreases both diurnal and nocturnal secretion of leptin in healthy individuals. Clin Endocrinol (Oxf). 2001;55(5):639-647.
