How Sleep, Caffeine and Alcohol Change Your Reaction Time

Your score is a range, not a number

Test yourself five times across one week and you will not get five similar numbers. The spread within one person across a week is often wider than the gap between two different people, which is why a single session tells you very little and a tracked median tells you a great deal.

Sleep researchers settled this argument decades ago with the psychomotor vigilance task, a simple visual reaction test that has served as the field's standard measure since 1985. Its appeal is precisely that it has no learning curve worth speaking of: nobody gets good at it, so whatever changes in the score is the state of the person, not their skill. Every click test on this site is a cousin of that task.

That makes the practical instruction simple. Compare like with like: same device, same time of day, same posture, warm-up rounds thrown away. Your stats page keeps every median, so the only thing you have to control is the condition you are testing under.

Sleep is the biggest lever you own

Nothing else on this page comes close. A 2024 meta-analysis of one night of restricted sleep found reaction and attention tasks slowing by around 29 ms on average, which on a typical 250 ms baseline is more than a tenth of your score gone to a single bad night.

The comparison that lands hardest, though, is the alcohol one. In a controlled study of sustained wakefulness, 17 to 19 hours awake produced impairment equivalent to a blood alcohol concentration of 0.05%, the legal driving limit in much of the world. Push on to 24 hours and performance matched roughly 0.10%, which is over the limit everywhere. Wake at 7am and by midnight you are, in performance terms, driving drunk on sleep alone.

The debt you cannot feel

One heavy night is at least obvious. The genuinely expensive pattern is the ordinary one: six hours a night, indefinitely, while believing you have adapted.

The landmark experiment held people to 4, 6 or 8 hours in bed for two weeks. The six-hour group's vigilance lapses climbed steadily the whole time and finished level with people who had gone two full nights without any sleep at all. Their own ratings of sleepiness, meanwhile, had barely moved after the first few days. Subjective alertness stabilises while objective performance keeps sliding, so the one instrument you would naturally trust, your sense of how sharp you feel, is the first one to stop working.

Recovery is slower than the folklore suggests too. In a separate restriction study, groups held to 3 and 5 hours a night had not returned to baseline reaction speed after three consecutive nights of 8 hours of recovery sleep. One long weekend does not clear a working week. If you want to know where you actually stand, run five rounds of the Classic Reaction Time Test on a Wednesday, not after a holiday.

What caffeine buys, and when

Caffeine does not add energy. It blocks adenosine, the molecule that accumulates in the brain across a waking day and produces the pressure to sleep. Occupying those receptors hides the signal without clearing the backlog, which is why the tiredness is still waiting when the dose wears off.

The effect on reaction speed is real and reasonably well characterised. Reviews of caffeine's cognitive effects put the useful range at 1 to 6 mg per kilogram of body weight, with vigilance and reaction time responding at the lower end of that band: for a 70 kg adult, somewhere between one strong coffee and two. A 2024 dose-response study in first-person-shooter players measured both faster reactions and better shooting, worth roughly 10 to 20 ms on simple reaction tasks. Doubling the dose does not double the benefit, and past a point the tremor costs you more than the alertness returns.

Timing matters more than dose for most people. Caffeine peaks in the blood 30 to 60 minutes after you drink it and has a half-life of about 5 hours in healthy adults, so half of a 4pm coffee is still circulating at 9pm.

The morning coffee that only breaks even

Here is the finding that reframes the whole subject. When researchers compared habitual caffeine consumers, tested after an overnight abstinence, against people who rarely touch the stuff, the consumers given caffeine ended up no more alert than the non-consumers given nothing.

Overnight abstinence is mild withdrawal. The morning cup does not lift a habitual drinker above baseline so much as return them to it, and the lift they feel is relief from a deficit they created the previous day. Genuine net gains belong mostly to people with little tolerance, or to anyone using caffeine deliberately and occasionally rather than by the clock.

None of that makes caffeine useless. It makes it a tool with a repayment schedule. Used sparingly before something that matters, it is one of the few interventions on this page with a clear measurable upside. Used continuously, it mostly keeps you level while quietly taxing the sleep that would have done the job properly.

Alcohol charges you twice

The first bill is the obvious one. Reviews of driver performance research find measurable impairment on some tasks from around 0.04% blood alcohol, and reliable, broad impairment by 0.05%, covering exactly the skills a reaction test isolates: divided attention, tracking, and reaction time itself.

The second bill arrives the next morning, and it is the one people forget to account for. A systematic review of next-day effects found sustained attention and psychomotor speed still impaired during hangover with blood alcohol back at zero. Sober, legal, and slower. The Driving Reaction Time Test is the blunt way to see what that costs, since it converts your milliseconds straight into metres.

Combine the two effects with the sleep table above and the arithmetic gets ugly quickly. Alcohol also fragments sleep, so a heavy night hands you the hangover deficit and a short-sleep deficit at once, from a single decision. The stopping distance figures are a useful reminder of what these milliseconds turn into outside a browser tab.

The clock decides before you do

Even perfectly rested, sober and uncaffeinated, your reaction speed moves through the day on a fixed schedule. Alertness and performance track core body temperature, which bottoms out in the small hours and peaks in the late afternoon and early evening. That is not a habit you can train away, it is the clock your physiology runs on.

Two dips are worth planning around. The early-morning trough sits somewhere between 3am and 6am for most people and is the slowest you will be all day, sleep debt aside. The post-lunch dip in the mid-afternoon is smaller, real, and happens whether or not you actually ate lunch, which rules out the obvious explanation.

Add sleep inertia at the front of the day and the practical advice writes itself: do not benchmark yourself in the first half hour after waking, and do not set a 6am score against a 6pm one and conclude anything about your training.

Water, movement and a warm-up

Three modest effects are worth collecting once the big ones are handled.

Dehydration past roughly 2% of body mass measurably impairs attention, executive function and motor coordination, and 2% is easier to reach than most people assume across a hot afternoon or a long training session. Light exercise helps afterwards rather than during: a meta-regression of exercise and cognition found performance dips briefly while you are actually exerting yourself, then improves modestly once you stop, which makes a short walk a better pre-test ritual than a hard set.

The warm-up is the cheapest of the three and the one people skip. The first round or two of any reaction test run slower than everything that follows, so a session you count from the very first click is a session biased against yourself. Throw away two rounds, then start recording. The same goes for switching tests: the Choice Reaction Time Test and the Go/No-Go Test each need their own settling-in before the number is worth keeping.

How to take a score you can trust

You cannot control your genetics or your age, and you should not try to change your sleep, your caffeine and your testing hour all in the same week. Move one variable at a time and let the median do the arguing.