Caffeine Against Sleep Pressure: What It Blocks, and What It Cannot

Two processes, and why the afternoon dip is not tiredness

The standard account of sleep regulation has two moving parts. Process S is homeostatic: a pressure to sleep that rises steadily for every hour you are awake and falls only while you sleep. Process C is circadian: an oscillation driven by the internal clock that raises and lowers alertness on an approximately 24-hour cycle, and that runs whether you slept last night or not. Borbély set this out in 1982, and the 2016 reappraisal he wrote with Daan, Wirz-Justice and Deboer is the version most people cite now. Sleepiness at any moment is the gap between the two, which is why the model explains things a single-variable account cannot.

The clearest of those things is the early-afternoon dip. If sleepiness were simply hours-awake, it would rise monotonically all day and be worst just before bed. It is not: most people are sleepier at around three in the afternoon than at eight in the evening, despite having been awake five hours longer by the evening. That is process C dipping and then rising again into the evening — the so-called circadian wake-maintenance zone, a couple of hours before habitual bedtime when the clock is actively holding alertness up against a large and growing pressure. It also explains the second wind of a very late night: process C starts rising again in the small hours even as process S is far higher than it has ever been.

Caffeine blocks the report, not the pressure

Adenosine accumulates in the extracellular space during waking, as a by-product of neuronal energy metabolism, and declines during recovery sleep. It is not the whole of process S — nobody claims a single molecule is — but it is the best-characterised molecular signal of it, and its sleep-promoting effects run mainly through two receptor subtypes, A1 and A2A. Caffeine is a competitive antagonist at both. Reichert and colleagues, reviewing the field in the Journal of Sleep Research in 2022, put the timing at roughly thirty minutes from oral intake to central-nervous-system effect. The consequence is precise: the receptors are occupied, so the signal is not transmitted, and you do not feel the pressure you are under.

The distinction between blocking the signal and discharging the pressure is the point of this whole article, and there is direct evidence for it in the electroencephalogram. Slow-wave activity in non-REM sleep is the standard physiological marker of process S — it is higher after a long day and higher still after sleep deprivation. Landolt and colleagues showed in Neuropsychopharmacology in 2004 that caffeine attenuates those waking and sleep EEG markers of sleep homeostasis: the readout goes down while the underlying need does not go away. An earlier study from the same group, in Brain Research in 1995, gave nine healthy men 200 mg at ten past seven in the morning and still found altered sleep and reduced power in the lowest delta band that night, by which time saliva caffeine had fallen to 3 µmol/l. The drug was almost gone and the night was still different.

Why the crash arrives, and why it cannot be a repayment

If caffeine occupies the receptors while adenosine continues to accumulate, then the moment the caffeine falls away the accumulated signal reaches receptors that are suddenly free. That is the standard mechanistic account of the crash, and it fits the timing people report: the drop is sharper than the natural rise in sleepiness would be, because it is a step change in signalling rather than a gradual build. It also explains why a second dose taken into the crash works, and why the crash after that one is worse — process S has kept climbing throughout.

The same mechanism is why caffeine cannot repay a sleep debt. Process S discharges during sleep and only during sleep; nothing in the pharmacology of an adenosine antagonist removes the accumulated pressure, and the EEG evidence says caffeine dims the marker rather than clearing the state. Caffeine restores alertness and measurably restores performance on many tasks, which is real and useful, but it does so while the debt continues to run. The practical consequence belongs in a road-safety sentence rather than a physiology one: a driver who is seriously sleep-deprived and feels fine after a coffee is a driver whose warning light has been disconnected. If you are fighting to stay awake at the wheel, the answer is to stop, not to drink something.

What is still circulating at bedtime

Caffeine clears by first-order kinetics — a constant fraction per unit time, so the amount left is the dose multiplied by one half raised to the number of half-lives elapsed. The half-life article in this section carries the pharmacology; the number that matters here is that the mean half-life in healthy adults is around five hours and the published range runs from roughly one and a half to nine and a half, with pregnancy and oral contraceptives lengthening it and smoking shortening it. That spread is not a detail. It is the difference between two people drinking the same coffee at the same hour and going to bed in completely different states.

The table above runs the arithmetic for a bedtime of eleven at night. Two hundred milligrams — about two moderate cups — taken at four in the afternoon leaves 40 mg at bedtime in a fast metaboliser, 76 mg in an average one and 120 mg in a slow one. Three hundred milligrams at five in the afternoon leaves 131 mg on the average half-life, which is more than a strong single espresso taken at the moment of switching the light off. And Drake's experimental condition — 400 mg six hours before bed — leaves 174 mg circulating at lights out on the average half-life, and 258 mg in a slow metaboliser. Seen that way, the finding that it cost 41 minutes of sleep stops being surprising.

One arithmetic aside that will matter in a moment. Because the decay is first-order, doubling the dose does not double the time to clear — it adds exactly one half-life. Starting from 100 mg, it takes 5.0 hours to fall below 50 mg; from 200 mg, 10.0 hours; from 400 mg, 15.0; from 800 mg, 20.0. Each doubling buys five hours, not a doubling. Alcohol behaves in the opposite way, and the sobriety article in this section works through what that changes.

What a late dose actually costs

Drake and colleagues ran the experiment properly in the Journal of Clinical Sleep Medicine in 2013. Twelve healthy normal sleepers took three pills a day for four days — one at six hours before bed, one at three hours, one at bedtime — of which exactly one was 400 mg of caffeine and the other two were placebo. Sleep was recorded with an in-home monitor as well as a diary. Objectively measured total sleep time fell by 67 minutes when the caffeine was taken at bedtime, 65 minutes at three hours before, and 41 minutes at six hours before, all significant against placebo.

The detail that makes the study memorable is what the diaries said. At bedtime and at three hours before, both the self-report and the objective measure caught the disturbance. At six hours before, only the objective measure did: the participants lost the better part of an hour of sleep and did not notice. That is the honest form of the advice usually given as "no caffeine after two in the afternoon". The reason for the rule is not that the drink keeps you awake in a way you can feel. It is that it does not.

Tolerance, and what it does not restore

Regular caffeine intake upregulates adenosine receptors: the brain compensates for the chronic blockade by making more of them, so the same dose occupies a smaller share and the subjective effect shrinks. That is the mechanism behind the familiar experience of a habitual drinker who reports that coffee "does nothing" for them, and behind the headache that follows a missed morning cup — with more receptors and no antagonist, the signal is briefly louder than baseline.

Two things about tolerance are worth keeping straight. The first is that tolerance to the subjective alerting effect and tolerance to the sleep-disrupting effect are not the same thing and do not necessarily develop together: a habitual drinker may genuinely not feel a late coffee and still lose sleep to it, which is the Drake finding in another form. The second is that tolerance changes the size of the effect and not its direction. It does not turn caffeine into something that discharges process S, because nothing in this class of drug does that. The only thing that discharges sleep pressure is sleep.

Frequently asked questions

How many hours before bed should I stop drinking coffee?

Six hours is the figure that comes directly out of the Drake study, which found significant disruption at that interval with 400 mg. But the honest answer depends on your dose and your half-life: 400 mg six hours out leaves 174 mg circulating on an average half-life, while 80 mg three hours out leaves 53 mg. If your half-life is long — pregnancy and oral contraceptives lengthen it — six hours may not be enough, and if it is short, a smaller dose earlier in that window may be fine. The reliable test is not how you feel at bedtime, since the study's participants felt nothing.

I fall asleep fine after an evening coffee. Does that mean it does not affect me?

No, and this is the finding that most surprises people. In the Drake study the six-hours-before condition produced a 41-minute loss of objectively measured sleep that the participants' own diaries did not register at all. Falling asleep quickly is only one part of the picture; total sleep time, wake after sleep onset and sleep depth all move too. Sleep-onset latency is also the part of the night most influenced by how tired you were to begin with, which is exactly what caffeine is masking.

Can caffeine make up for a bad night's sleep?

It can restore alertness and improve performance on many tasks, which is genuinely useful and is why it is used. It cannot discharge the sleep pressure that built up. Process S falls only during sleep, and caffeine acts on the receptors that report the pressure rather than on the pressure itself — the EEG work shows the marker being attenuated while the underlying state persists. So a coffee buys you working hours; it does not settle the account, and the debt is still there when the caffeine goes.

Why am I sleepy at three in the afternoon but wide awake at nine at night?

Because the two processes are pulling in opposite directions at those two moments. Sleep pressure rises all day, so it is higher at nine in the evening than at three in the afternoon. But the circadian process dips in the early afternoon and then rises into a wake-maintenance zone in the couple of hours before habitual bedtime, actively holding alertness up. The net feeling is the gap between the two, not the pressure alone — which is precisely why a model with one variable never worked.

Does tolerance mean a late coffee stops affecting my sleep?

It shrinks what you feel, which is not the same as shrinking what happens. Chronic intake upregulates adenosine receptors, so a given dose blocks a smaller share of them and the subjective lift fades. Tolerance to the alerting effect and tolerance to the sleep-disrupting effect are separate, though, and the second is the one that matters at bedtime. A habitual drinker who no longer notices an evening coffee is in exactly the position of the Drake participants who lost 41 minutes of sleep and reported nothing.

Why does doubling my coffee not double how long it keeps me up?

Because caffeine clears by first-order kinetics: a constant fraction of what is left disappears per unit time, not a constant amount. Each doubling of the dose therefore adds exactly one half-life to the time needed to reach any given level. Starting from 100 mg it takes 5 hours to drop below 50 mg; from 200 mg, 10 hours; from 400 mg, 15; from 800 mg, 20. Alcohol is the opposite case — it is eliminated at a roughly constant mass per hour, so doubling the dose does double the time — and the sobriety article in this section works that through.