Short answer: in a controlled trial, 400 mg of caffeine taken six hours before bed still cut sleep by about an hour — and participants didn't notice. Alcohol shortens time to fall asleep, then fragments the second half of the night and suppresses REM sleep. Usable rules: last coffee eight to ten hours before bed, last drink at least three hours before, and alcohol-free nights during the week.
Caffeine: the problem isn't falling asleep
Everyone knows a late coffee can delay sleep onset. That is not the main problem, and it is why intuition fails here.
A controlled trial published in the Journal of Clinical Sleep Medicine in 2013 gave 400 mg of caffeine — two to four cups depending on the brew — at three different times: at bedtime, three hours before, and six hours before. All three conditions degraded measured sleep. At six hours out, the reduction in total sleep time came to roughly an hour.
The decisive finding sits elsewhere, and it is uncomfortable: participants did not perceive the degradation. Their subjective estimate did not match what the measurement showed. That is why "coffee doesn't affect me" is not usable information. It isn't bad faith; it is a limit of perception.
The realistic rule
Eight to ten hours before bed. For an 11pm bedtime, that puts the last coffee between 1pm and 3pm. Caffeine half-life varies widely between people — genetics, smoking, some medications, pregnancy — but the safety margin is more useful than a personalised calculation, precisely because self-assessment does not work.
And if the only way to get through the day is a 5pm coffee, the coffee isn't the thing to look at: it's the sleep debt that made the coffee necessary. The loop closes on itself.
Alcohol: the sedative that steals the second half
Alcohol gets you to sleep faster. That is true, measurable, and exactly what misleads.
A review published in Alcoholism: Clinical and Experimental Research in 2013 synthesised the effects on normal sleep. The picture is consistent: shorter time to fall asleep, more deep sleep in the first part of the night, then, as the alcohol is metabolised, fragmentation of the second half and suppression of REM sleep.
REM is not a decorative extra: it is the portion associated with memory consolidation and emotional regulation. Hence the ordinary experience of waking at 4am after a night out, on a night that seemed to start well. You fall asleep better and you recover less.
Rule: last drink at least three hours before bed, and alcohol-free nights during the week. If you wear a tracker, this is one of the few effects large enough to see within a week of data.
The other two evening substances
Nicotine Grade B
A polysomnographic analysis published in Sleep Medicine in 2012 compared the sleep of smokers and non-smokers: smokers' sleep is lighter, with less deep sleep. Nicotine is a stimulant, and overnight withdrawal fragments the end of the night. Vaping included — it is the molecule that counts, not the delivery.
Cannabis Grade B
A review published in Current Psychiatry Reports in 2017 summarises what the literature shows: short-term sedation, reduced REM sleep, tolerance developing with repeated use, and rebound insomnia on stopping. Cannabis sedates; it does not restore. Many people using it to sleep are in fact treating an insomnia that would respond to the behavioural protocol.
The four, side by side
| Substance | What it does to sleep | Rule |
|---|---|---|
| Caffeine | Cuts total sleep time, without being noticed | Last coffee 8–10 h before bed |
| Alcohol | Fast onset, fragmented second half, less REM | Last drink 3 h before; alcohol-free weeknights |
| Nicotine | Lighter sleep, less deep sleep | None in the evening, vaping included |
| Cannabis | Sedates, cuts REM, tolerance and rebound | Not a sleep treatment |
The only honest test
Since perception fails on all four, the only reliable method is a comparison over time: two weeks with the rule, two weeks without, and a diary kept for one minute a day. A standardised diary exists for exactly this kind of measurement. It isn't elegant, but it is the only thing that settles the question — and it is also the first step of any behavioural sleep protocol.
Worth knowing. Stopping caffeine abruptly causes headaches for a few days: tapering is more sustainable. And drinking used to manage mood or anxiety, rather than for pleasure, is a reason to talk to someone in its own right — the motive is what counts, not the amount. Do not change any prescribed sleep medication on your own.
How Sillon handles it
All four practices exist in the library with their grade, dose and source: "Last coffee" (grade A), "The evening drink and your night" (grade B), "Evening nicotine", and "Cannabis sedates, it doesn't restore". The app does not hand you all four at once: it puts one in front of you, with the reason it is that one today.
Sources
- Drake C et al., "Caffeine effects on sleep taken 0, 3, or 6 hours before going to bed", Journal of Clinical Sleep Medicine (2013)
- Ebrahim IO et al., "Alcohol and sleep I: effects on normal sleep", Alcoholism: Clinical and Experimental Research (2013)
- Jaehne A et al., "How smoking affects sleep: a polysomnographical analysis", Sleep Medicine (2012)
- Babson KA et al., "Cannabis, Cannabinoids, and Sleep: a Review of the Literature", Current Psychiatry Reports (2017)
- Carney CE et al., "The consensus sleep diary: standardizing prospective sleep self-monitoring", Sleep (2012)
One card a day, not a list of ten
Sillon picks the practice that matters for you today, and says why. Free, offline, no account.