The Coffee You Drank at Two O'Clock Is Still Awake at Ten

The claim arrives with great confidence, usually over the second cup: coffee doesn’t affect me, I can drink it at night and sleep like a stone. It is offered as a fact about a person. It is a fact about how badly people judge their own sleep, and it is one of the few folk beliefs that a wristwatch and a calculator can settle without anybody arguing.

Caffeine does not wear off at an hour. It leaves at a rate, proportionally, which is a different and much less convenient thing.

The arithmetic, which is not in dispute

Caffeine clears by first-order kinetics. Every half-life, half of what remains departs. For most adults that half-life is about five hours.

So the mug at two o’clock has not gone anywhere by seven. It has halved. By midnight it has halved twice. What is left at bedtime is

remaining = dose × 0.5 ^{(hours ÷ half-life)}

and for a 140 mg mug at 2pm against a 10pm bedtime, that is about 46 mg — call it a third of a mug, still circulating, at the moment you switch the light off.

The bedtime calculator

Bedtime Your ceiling at bedtime (mg)

What you drank

Drip coffee, mug · 12 oz Drip coffee, cup · 8 oz Espresso · 1 shot Instant coffee · 8 oz Cold brew · 12 oz Black tea · 8 oz Green tea · 8 oz Matcha · 8 oz Cola · 12 oz Energy drink · 16 oz Dark chocolate · 1 oz Decaf coffee · 8 oz Remove

+ another drink

One 12 oz mug at 2pm leaves roughly 46 mg on board at 10pm — about 33% of a mug, still circulating.

Half-life 4–6 h typical; 2–10 h across healthy adults. Drink figures are estimates with real spread. The ceiling is yours to choose, not a medical threshold.

Why “it doesn’t affect me” survives contact with reality

Because the thing caffeine damages is not the falling asleep. It is the sleep itself, and you are unconscious for that part.

The cleanest demonstration is Drake and colleagues (Journal of Clinical Sleep Medicine, 2013), who gave 400 mg — roughly three mugs — at zero, three, and six hours before bed. All three conditions measurably disrupted sleep against placebo. The six-hour dose still cost more than an hour of sleep on objective measurement. Participants frequently did not notice.

That is how it gets you. Self-report says fine. The polysomnograph disagrees. If your proof that coffee does not affect you is that you feel you slept well, you are using the instrument the drug interferes with.

Why the number on the box is nearly useless

The decay above is exact. The dose going into it is a guess, and often a poor one. “A cup of coffee” is not a unit of anything:

• The same nominal 8 oz drip coffee is measured anywhere from 70 to 140 mg, depending on bean, grind, dose and brew time.
• Instant is worse — surveys span 27 to 173 mg for one nominal cup.
• Matcha runs 38 to 176 mg, for the same green powder and the same water.

This is why the calculator above reports a band and not a single number. Anyone quoting you a precise milligram figure for a drink they did not weigh is performing arithmetic on an assumption.

The variation that is not measurement error

The half-life itself moves, and not slightly. Across healthy adults it runs from roughly two hours to ten, mostly on CYP1A2 activity. Smoking roughly halves it. Oral contraceptives roughly double it. Pregnancy pushes it past the slow end entirely.

Two people can drink identical coffee at an identical hour and go to bed with very different amounts aboard, and neither of them is doing anything wrong or imagining anything. It does mean that a friend’s tolerance tells you about that friend.

What The Method actually recommends

• Set a cut-off by the clock, not by feel. Working back from the arithmetic, eight hours before bed leaves a mug at roughly an eighth. Six hours — the interval that still cost an hour of sleep in the trial — leaves it at nearly half.
• Count everything. Tea, cola, matcha, the dark chocolate after dinner. The calculator adds them because your liver does.
• Treat 400 mg a day as a ceiling, not a target. That is the FDA’s figure for healthy adults, and it says nothing at all about timing, which is the part that costs you the sleep.
• Do not test the theory on yourself and grade your own paper. You are the one instrument in the room the drug is known to affect.

What should exist

A caffeine figure printed on the cup, the way alcohol is printed on the can. Not a range invented by marketing — a measured number for that drink, from that machine, at that dose. Every other psychoactive substance sold across a counter manages this. Coffee is exempt for no better reason than tradition.

---

Sources: half-life and CYP1A2 variation from standard pharmacokinetic literature; the six-hour sleep-disruption finding from Drake et al., J Clin Sleep Med 2013; the 400 mg/day figure from the FDA; drink content ranges from published surveys of measured servings. The decay equation is arithmetic. The doses are estimates and are shown as ranges for that reason.