Device Bench

Device test · Input

Reaction Time Test

Wait for the panel to turn green, then click, tap or press Space as fast as you can. The test runs five trials and reports your best, your median and — the number that actually matters — the spread between them, because one fast trial is luck and five tell you something. It also measures the frame interval of the display you are using, so you can see how much of your score belongs to the monitor rather than to you. Everything runs in this tab, nothing is stored, and nothing is uploaded.

Last— Best— Median— Trials0 / 5 Your frame—

No trials yet.

Your trials

TrialReactionDisplay quantisation

The third column is not an error bar on your reflexes — it is the width of one refresh on this display. A green frame cannot appear between refreshes, so every trial carries up to that much delay that is nothing to do with you.

How this test works

The panel arms itself and waits a random interval between 1.5 and 4.5 seconds before turning green. The interval is randomised on every trial so that anticipation cannot be rewarded: if the wait were fixed you would learn its length and start timing the countdown rather than reacting to the colour. The moment the green frame is committed, the page records a timestamp from performance.now() — the browser's monotonic sub-millisecond clock — and the moment your pointerdown or key event arrives it records another. Your reaction is the difference. Activating before green is a false start: the trial is discarded and named rather than scored, because counting it would reward guessing and silently dropping it would look like a broken button.

Five trials are run because one is not a measurement. The page reports your best, your median and the spread between fastest and slowest, and the spread is the figure worth reading — a single quick trial is luck, while a narrow range across five says the rest were repeatable. The median is preferred to the mean for the headline because one distracted trial drags a mean of five a long way and barely moves a median. Nothing is stored: the trials live in memory for as long as the tab is open, there is no history, no account and no upload, and reloading the page starts a fresh set.

The limits are worth stating plainly, and the largest is that this number is not your nervous system. It is a chain: light leaves the panel, your eye and brain resolve it, your hand moves, a switch closes, USB polls, the operating system queues the event and the browser dispatches it. Two of those links are hardware and one of them is measurable here, which is why the page samples your display's frame interval directly and prints it (the refresh rate test reports the same quantity as a headline figure) — on a 60 Hz panel a green frame can only appear on a refresh boundary, so up to about 16.7 ms of any score is the monitor waiting its turn, and a high-refresh screen genuinely produces lower numbers for the same person. Wireless peripherals, browser choice and background load all add their own few milliseconds. And there is deliberately no comparison to anybody else on this page: without a server there is no population to compare against, so a percentile here would be an invented number. Compare a set from this page against another set from this page, and the comparison means something.

Frequently asked questions

What is a normal reaction time?
This page will not give you a figure to measure yourself against, because it does not have one. Reporting a population average would mean either inventing it or quietly borrowing someone else's sample collected on different hardware, and both are worse than saying nothing. What the test can tell you honestly is how you compare with yourself: run five trials, note the median and the spread, change one thing — a different mouse, a wired connection, a higher refresh rate, more sleep — and run five more.
Why does my score change so much between trials?
Because reaction time is a distribution rather than a value. Attention drifts, and a trial where your eye happened to be moving lands tens of milliseconds behind one where it was still. That is why the page runs five trials and shows the spread rather than crowning a single best: a 40 ms range across five trials describes you far better than the fastest of them does.
Does my monitor affect the result?
Yes, measurably, and this page is unusual in showing you by how much. A green frame cannot be displayed between refreshes, so on a 60 Hz panel up to about 16.7 ms of every score is the display waiting for its next refresh — before your eye is involved at all. The page samples the actual frame interval of the screen you are using and prints it beside each trial, so the hardware's share is visible rather than hidden inside the number.
Is a mouse click faster than the spacebar?
Usually slightly, though the difference is small next to the variation between your own trials. A mouse switch has a short travel and most mice poll at 1000 Hz, while a keyboard adds its own debounce and typically polls slower — the gap is a few milliseconds. It is worth testing both here rather than trusting the claim: run five trials each way and compare the medians, which is exactly the kind of comparison this page is built for.
Does clicking early help?
No, and the test is built so it cannot. Activating before the panel turns green is detected as a false start: the trial is discarded, the count does not advance, and the summary line tells you how many you discarded. Because the wait is re-randomised between 1.5 and 4.5 seconds on every trial, there is no rhythm to learn and no advantage in anticipating.

Related tools