The reference figure: a mean caffeine half-life of 5.7 hours in healthy adults.
Coffee doesn't leave the body the moment the mug is empty. It has a half-life, the time it takes for half of it to clear, and that number is longer, and more variable, than most people assume.
The Number, and Where It Comes From
A frequently cited reference figure comes from a small 1980 clinical study: three healthy adults had a mean serum caffeine half-life of 5.7 hours. That's a genuinely small sample, a case study rather than a large trial, but the figure has held up as a rough baseline in the decades since.
A much larger 2022 systematic review, pooling data from 141 publications and 4,714 individuals, didn't converge on one single universal number (the underlying studies were too varied in method for that), but it confirmed the same broader pattern: half-life genuinely differs from person to person, and several specific factors reliably shift it. Smoking speeds caffeine clearance up. Oral contraceptive use slows it down, an effect the review describes as consistent across more than 50 years of data. Liver disease, particularly cirrhosis, slows it dramatically. In the original 1980 study, two patients with alcoholic liver disease had half-lives of 60 and 168 hours, ten to thirty times longer than the healthy baseline. Obesity, notably, showed no measurable effect on clearance.
What a Half-Life Actually Means
Half-life doesn't mean caffeine disappears at that point. It means half of it is still there. Take a coffee at 2pm with roughly 5.7 hours as a rough baseline: by around 7:45pm, roughly half the caffeine is still circulating. By around 1:30am, a quarter of it still is. The exact numbers move a lot by person, but the shape of the curve, a slow, long tail rather than a sharp cutoff, holds regardless of exactly where someone's own half-life sits.
What Actual Sleep Studies Show, Not Just Blood Level
A half-life curve describes how much caffeine is in the blood. It doesn't, on its own, say what that does to sleep. A randomised, double-blind, placebo-controlled crossover trial published in 2024/2025 looked at that question directly: 23 healthy men with moderate habitual caffeine intake took a placebo, 100mg, or 400mg of caffeine at three points before bed, 12 hours, 8 hours, and 4 hours, with their sleep measured overnight.
100mg, roughly a single normal coffee, produced no measurable difference at any of the three timings. 400mg did. Taken 4 hours before bed, it reduced sleep efficiency by 9.5%, added just over 14 minutes to the time it took to fall asleep, and added roughly 26 minutes of wakefulness during the night. Taken 8 hours before bed, the effect on falling asleep eased, but night-time wakefulness was still measurably higher and deep sleep was still reduced. Even at 12 hours before bed, the larger dose still measurably reduced deep sleep.
This is a small study (23 people), all male, and conducted outside the UK, so it shouldn't be read as a universal number any more than the half-life figures above should be. What it adds is a different kind of evidence: not how much caffeine is still in the blood, but what a specific measured dose, at a specific time, actually did to a night of recorded sleep.
What It Doesn't Tell Us
Neither of the two half-life sources above says anything about how caffeine's presence in the blood relates to how alert or tired someone feels at a given moment; that's a separate question they don't address. The sleep study above looks at measured sleep itself, not felt alertness, and only at the specific doses and timings it tested, not a general rule for every amount at every hour. Nor does any of it say what any individual reader's own half-life actually is. Genuinely knowing that would need direct testing, not a population average. This is a description of how caffeine clears the body and, separately, what one study found it did to sleep, not advice about when anyone specifically should or shouldn't have a coffee.
Why This Matters for the Second Coffee
Most people don't think about the first coffee still being active when the second, or third, gets poured. The working day defaults people into another coffee because it's there, not necessarily because the moment calls for it. Understanding that the first one hasn't actually left yet is a small piece of information that makes the decision a genuine decision, rather than an automatic one.
We're not telling anyone to change what they drink or when. We're interested in the moment right before that choice gets made on autopilot; that's the territory we're building 0800Labs around. If that's useful, the waitlist is below.
Questions
Does afternoon coffee affect sleep?
It can, and the effect shows up further before bedtime than most people assume. A 2013 randomized trial gave healthy adults 400mg of caffeine (roughly the equivalent of several cups of coffee) at 0, 3, or 6 hours before their normal bedtime, compared with a placebo. All three timings measurably disrupted sleep relative to placebo; taken 6 hours before bed, it reduced objectively measured total sleep time by more than an hour (from 7.68 to 6.50 hours on average). That 400mg dose is higher than a single coffee, so this isn't a claim that any one afternoon coffee, at any dose, wrecks sleep for everyone. It's evidence that caffeine's reach extends well past when it's easy to notice, consistent with the half-life data above. For more on sleep itself, see The Short Nap.
What are the limitations of using caffeine to address tiredness?
Caffeine doesn't address the underlying rhythm this article and The 3pm Slump both describe; it doesn't reset the roughly twice-daily dip in alertness, it works around it temporarily. It also doesn't clear the body quickly: at a mean half-life of 5.7 hours, a coffee taken to fix afternoon tiredness is often still active well into the evening, which is a separate trade-off against sleep (see above), not a reason the afternoon dose "doesn't work." The sources behind this article didn't measure how caffeine affects felt alertness or tiredness directly; that's a distinct question the pharmacokinetic data alone doesn't answer.