What is a normal reaction time?

There is no single number this page can hand you that would mean the same thing on your hardware and on someone else's. What it measures is the whole loop: your screen drawing the change, your eye and hand, and your browser reporting the press, and the screen and the input path are a real part of that figure on any machine. Nothing is uploaded from this page, so it also has no population of visitors to compare you against and does not invent one. The comparison that does hold is your own: run the five trials on the same machine on different days, or before and after a change to your setup, and compare the medians.

Why does it report a median instead of my best time?

Because the best of five is the luckiest trial, not the typical one, and it is the one figure a guess can buy: press a moment before the signal and, if it lands just after, you get a very low number that says nothing about how fast you react. The median is the middle of the five sorted times, so one lucky trial cannot move it and neither can one distracted one. Both are shown, the median first. If you stop before the fifth trial, the median of an even number of trials is the mean of the two middle ones.

How can you test reaction time?

Three of the tests here time you directly. Colour change is the classic one. Start lights turns on five red lights one a second and you respond when they all go out. Sound plays a beep with nothing to see. The fourth, Ruler drop, is the one commonly used in sport and PE. It needs a partner and a ruler: they let it go, you catch it, and the distance it fell becomes a time. Choose one with the buttons at the top of the tool.

Does it work on a phone or with a keyboard?

Yes. It runs in any current browser on a phone, tablet or computer, with nothing to install. Tap the stage on a touchscreen, or press the space bar or Enter on a keyboard; a key held down does not count twice. Use the same input when you compare runs, because a touchscreen, a mouse and a keyboard each add their own delay to the figure.

Does my screen or mouse affect the result?

Yes, and nothing is deducted for either. A trial is timed from the frame the colour was actually drawn in to the moment the browser reports your press, so the refresh rate of the display, the response time of the panel and the polling rate of the mouse or the touchscreen are all inside the figure. That is why the accuracy line calls it comparable between your own runs on one machine rather than against a laboratory instrument. Some tests subtract a fixed number of milliseconds for display lag; that is a guess about hardware a web page cannot see, so this one does not.

What happens if I click too early?

It is a false start. The trial is not recorded, the stage says what happened, and you press again to run that same trial. A response that lands at the exact moment of the signal counts as early too, because there was no drawn frame before it to react to. False starts are counted on screen for the run, so a run with three of them is visibly a run with three of them rather than a suspiciously fast median.

Why was one of my trials thrown away?

Almost always because the tab stopped being visible during it, which is the one case the page voids on purpose: a browser stops painting a tab that is not on screen and holds its timers back, so the signal you would have reacted to was never drawn and the elapsed figure would be measuring your return to the page. Three other trials are refused rather than recorded: a response that lands more than ten seconds after the signal, which is somebody who looked away rather than somebody reacting, a sound trial whose beep the browser would not play, and the rare case where the clock does not give a usable pair of readings. In every case nothing is added to the list and the message says which it was.

Where are my times kept?

In this browser, on this device, and nowhere else. The trial times and the wait range you chose sit in local storage, one key for each timed test, and the ruler drop converter keeps nothing; there is no account, nothing is sent anywhere, and no result of yours reaches any analytics. The Start over button at the top of the page clears them, and so does clearing site data. If your browser blocks storage, the test still runs exactly the same way and simply forgets between visits.

How do reaction-time challenges work?

In the colour change test, choose Start five-trial challenge to use five seeded waits in the 1–5 second range. A refused trial repeats the same wait; Reset repeats the whole sequence. After five accepted responses, copy or share your median and exact challenge link, or download a PNG result card. Recipients receive the same waits. Practice results and restored times do not create challenge links. Challenge runs are not saved over your practice data. Scores are self-reported and include display and input latency; repeating a sequence can also make it predictable.

Reaction Time Test

Click the moment the stage changes colour, five times, and read the median of the five rather than a lucky best, with every trial kept on screen beside it. The wait before the signal is random, one to five seconds or two to eight, so the press cannot be timed rather than reacted to. A response before the signal is a false start, and a trial during which the tab was hidden is voided rather than scored. The same tool can also run five start lights that go out together, a beep with nothing to see, or turn a ruler drop catch into a time. Nothing is uploaded and there is no account: your times stay in this browser.

Median so far

not yet

Best

not yet

Wait before signal

The stage takes a click, a tap, the space bar or Enter, and the press is what is timed, not the release. A second press in the moment after a trial ends is ignored, so a double press cannot start the next trial while you are still reading the last one. Changing the wait range cancels a trial that is waiting; the trials already recorded stay. Your times and the range you chose are kept in this browser and nowhere else. If you are using a screen reader, the signal is announced after it is painted, so the reading will be higher than what your hand did.

Responded before the signala false start
the trial does not count, and the page says why
Tab hidden mid-trialthe browser stopped drawing it
the trial is voided rather than timed
0 of 5 trials donemedian: the middle time
no trials recorded yet

What the number includes. This measures the whole loop: the screen drawing the change, your eye and hand, and the browser reporting the press. Display and input lag sit inside that figure and nothing is deducted for them, so it is comparable between your own runs on one machine rather than against a laboratory instrument. A response that cannot be a reaction is refused rather than scored: one that arrives before the signal, which had nothing to react to, and one from a trial during which the tab stopped being visible, because a browser stops painting a hidden tab.

Common questions

What is a normal reaction time?: There is no single number this page can hand you that would mean the same thing on your hardware and on someone else's. What it measures is the whole loop: your screen drawing the change, your eye and hand, and your browser reporting the press, and the screen and the input path are a real part of that figure on any machine. Nothing is uploaded from this page, so it also has no population of visitors to compare you against and does not invent one. The comparison that does hold is your own: run the five trials on the same machine on different days, or before and after a change to your setup, and compare the medians.
Why does it report a median instead of my best time?: Because the best of five is the luckiest trial, not the typical one, and it is the one figure a guess can buy: press a moment before the signal and, if it lands just after, you get a very low number that says nothing about how fast you react. The median is the middle of the five sorted times, so one lucky trial cannot move it and neither can one distracted one. Both are shown, the median first. If you stop before the fifth trial, the median of an even number of trials is the mean of the two middle ones.
How can you test reaction time?: Three of the tests here time you directly. Colour change is the classic one. Start lights turns on five red lights one a second and you respond when they all go out. Sound plays a beep with nothing to see. The fourth, Ruler drop, is the one commonly used in sport and PE. It needs a partner and a ruler: they let it go, you catch it, and the distance it fell becomes a time. Choose one with the buttons at the top of the tool.
Does it work on a phone or with a keyboard?: Yes. It runs in any current browser on a phone, tablet or computer, with nothing to install. Tap the stage on a touchscreen, or press the space bar or Enter on a keyboard; a key held down does not count twice. Use the same input when you compare runs, because a touchscreen, a mouse and a keyboard each add their own delay to the figure.
Does my screen or mouse affect the result?: Yes, and nothing is deducted for either. A trial is timed from the frame the colour was actually drawn in to the moment the browser reports your press, so the refresh rate of the display, the response time of the panel and the polling rate of the mouse or the touchscreen are all inside the figure. That is why the accuracy line calls it comparable between your own runs on one machine rather than against a laboratory instrument. Some tests subtract a fixed number of milliseconds for display lag; that is a guess about hardware a web page cannot see, so this one does not.
What happens if I click too early?: It is a false start. The trial is not recorded, the stage says what happened, and you press again to run that same trial. A response that lands at the exact moment of the signal counts as early too, because there was no drawn frame before it to react to. False starts are counted on screen for the run, so a run with three of them is visibly a run with three of them rather than a suspiciously fast median.
Why was one of my trials thrown away?: Almost always because the tab stopped being visible during it, which is the one case the page voids on purpose: a browser stops painting a tab that is not on screen and holds its timers back, so the signal you would have reacted to was never drawn and the elapsed figure would be measuring your return to the page. Three other trials are refused rather than recorded: a response that lands more than ten seconds after the signal, which is somebody who looked away rather than somebody reacting, a sound trial whose beep the browser would not play, and the rare case where the clock does not give a usable pair of readings. In every case nothing is added to the list and the message says which it was.
Where are my times kept?: In this browser, on this device, and nowhere else. The trial times and the wait range you chose sit in local storage, one key for each timed test, and the ruler drop converter keeps nothing; there is no account, nothing is sent anywhere, and no result of yours reaches any analytics. The Start over button at the top of the page clears them, and so does clearing site data. If your browser blocks storage, the test still runs exactly the same way and simply forgets between visits.
How do reaction-time challenges work?: In the colour change test, choose Start five-trial challenge to use five seeded waits in the 1–5 second range. A refused trial repeats the same wait; Reset repeats the whole sequence. After five accepted responses, copy or share your median and exact challenge link, or download a PNG result card. Recipients receive the same waits. Practice results and restored times do not create challenge links. Challenge runs are not saved over your practice data. Scores are self-reported and include display and input latency; repeating a sequence can also make it predictable.

Timed with the browser's high-resolution clock, to the millisecond it reports. The figure includes your screen or speaker presenting the signal and your device registering the press, so it is comparable between your own runs on one machine rather than against a laboratory instrument. A ruler drop time is free-fall arithmetic, only as precise as the mark you read at the catch.