Reaction time test

Measures by sight, hearing or choice, reports the median instead of a mean one lapse ruins, and tells you how many milliseconds your kit adds.

Mode Attempts

Press as soon as the box turns green.

Results

Nothing measured yet.

What your kit adds to every measurement

This number is not your reaction time: it is your reaction time plus the trip the signal takes through your equipment. The page can only paint the green on a refresh boundary, so the signal shows up half a frame late on average; the panel takes its own time to change colour, and the keyboard or mouse only reports when its polling turn comes round.

Measured refresh

measuring…

Estimated addition

13.3–44.7 ms

Device polling

Office keyboard 125 Hz, gaming mouse 1,000 Hz. The browser cannot read it — pick it yourself.

It is a range rather than one figure because two of the three delays cannot be seen from inside a browser. Subtracting them properly needs external hardware: a 1,000 fps camera filming screen and finger, or a latency measurement board. Nothing inside a tab can do it.

The ruler method, with no computer at all

A ruler in free fall is a clock. Someone holds it by the top end, you put open fingers at the zero mark without touching it, and they let go without warning. Where you catch it tells you how long you took.

Caught at cm works out as 202 ms

A 30 cm ruler runs out at 247 ms, so the method cannot measure a slow reaction.

Warning: Sight mode changes colour abruptly. If you have photosensitive epilepsy, use hearing mode — nothing flashes there.

Generated right here, in your browser. What you type never leaves your device.

Keep going with Click speed test Online ruler Refresh rate test Audio test

A reaction time is the gap between a signal arriving and your body doing something about it. For a simple visual signal most adults land somewhere between 200 and 250 milliseconds, and almost none of that is the nerve impulse travelling — the signal covers the distance from eye to finger in about twenty milliseconds. The rest is your brain deciding.

This page measures that gap three ways, because they are not the same measurement. Sight is the slowest, because the retina itself takes time to turn photons into a nerve impulse. Hearing is faster. Choosing between two answers is slower again, and the difference between the two is a number psychology has been reading since 1868.

It also tells you something the pages ranking above it do not: the figure you get here is not purely yours. Your screen, your keyboard and your browser each add a few milliseconds, and the panel below the test estimates how many.

Manual

How to use it

1
Pick a mode

Sight turns a box green. Hearing plays a tone, so you can close your eyes and take the screen out of the measurement entirely. Choice shows an arrow and asks for the matching key, which measures a decision as well as a reflex.
2
Wait for the signal, do not guess it

The delay is random between 1.2 and 4.5 seconds on purpose. A fixed delay stops measuring reaction and starts measuring rhythm: after three attempts you would be counting, and a counted press can beat the nerve.
3
Press as soon as it changes

Space, Enter, a click or a tap all work. In choice mode use the left and right arrows, or tap the side the arrow points at.
4
Do all five attempts

One attempt is noise. Five gives a median worth quoting, and the list underneath shows each one so you can see which was the outlier.
5
Read the median, then the spread

The median is your headline number. The spread says how steady you were: two people can share a median of 250 ms and have completely different runs behind it.
6
Check what your kit is adding

The panel measures your actual refresh rate and estimates the milliseconds your display and input device contribute. On a 60 Hz screen with an office keyboard that is a meaningful slice of the total.

In depth

Things worth keeping in mind

The three reactions, and why they differ

Donders named them a, b and c, and the labels have stuck in the literature even though nobody uses them conversationally. A simple reaction has one signal and one response: the fastest thing a person can be asked to do. A choice reaction has several signals each with its own response, and costs the time of deciding which. A go/no-go reaction has several signals but only one response, so it costs the time of deciding whether.

The interesting quantity was never any single one of these — it was the difference. Subtract the simple reaction from the choice reaction and you have, at least in Donders' argument, an estimate of how long the act of choosing took. The assumption that mental processes stack in neat separable stages has been argued about ever since, but the measurement remains useful.

Simple · one signal, one response · fastest
Choice · several signals, matching responses · adds a decision
Go/no-go · several signals, one response · adds an inhibition
Auditory · consistently 30 to 50 ms faster than visual

What moves the number, and by how much

Ranked by effect size, the things that change a reaction time are not the ones people expect. Sleep loss and alcohol dominate everything else. Age matters over decades rather than months, with a slow decline from the mid-twenties. Caffeine gives a small real improvement. Training gives a modest one, most of which is task-specific rather than general.

Equipment sits awkwardly in the middle of that list, which is why this page measures it rather than ignoring it. Moving from a 60 Hz panel with a 125 Hz keyboard to a 240 Hz panel with a 1000 Hz mouse can shift the reported figure by more than a decent night's sleep would.

Why this number cannot be perfectly honest

Everything a browser can time, it times from inside the browser. The clock starts when the page asks for the box to be green and stops when the key event arrives, and neither of those moments is the physical event you care about. Between them sit the compositor, the display's own processing, the panel's pixel response, the key's debounce and the USB polling interval.

The tool estimates the part it can reason about — half a frame of refresh, plus half a polling interval — and states the rest as unknown rather than guessing. If you need the real figure, the instrument is a high-speed camera filming the screen and your hand in the same frame, or a purpose-built latency board. Both exist, both cost money, and neither fits in a tab.

History

The assistant who was fired for being slow

In 1796 the Astronomer Royal, Nevil Maskelyne, dismissed his assistant David Kinnebrook from the Royal Greenwich Observatory. The offence was recording the moment stars crossed the meridian about eight tenths of a second later than Maskelyne did. Maskelyne assumed carelessness. Kinnebrook could not correct it, and left.

Twenty years later the German astronomer Friedrich Bessel read the case and suspected something else entirely. He compared his own transit timings against other astronomers and found the same kind of stable, personal offsets between every pair of observers. The difference was not carelessness — it was that different people take different amounts of time to perceive and record the same event. Bessel called it the personal equation, and observatories began measuring and subtracting it as routine.

That is the accidental birth of reaction time as a measurable human quantity: not a psychology experiment, but astronomers trying to work out why their clocks disagreed. In 1868 the Dutch ophthalmologist Franciscus Donders turned it into method. He measured a simple reaction, then a reaction requiring a choice, and subtracted the first from the second — arguing that the difference was the duration of the mental act of choosing. The subtractive method is still what the choice mode on this page is doing.

The public got their turn in 1884, when Francis Galton set up an Anthropometric Laboratory at the International Health Exhibition in South Kensington and charged visitors threepence to have themselves measured, reaction time included. Over 9,000 people paid. The data sat largely unexamined for a century until researchers went back to it in the 1980s and noticed something awkward: the Victorians appeared to be slightly faster than modern samples measured with electronic equipment.

1796: Kinnebrook dismissed over 0.8 s
1868: Donders subtracts simple from choice
9,000+: Visitors measured by Galton in 1884
0.1 s: The FIA's assumed human floor

How it works

What happens in those 250 milliseconds

The chain has more links than it feels like it has. Light hits the retina, where photoreceptors convert it into a nerve signal — and this first step is the slowest single one in the visual path, taking roughly 20 to 40 milliseconds on its own. The signal then travels the optic nerve to the visual cortex, gets recognised as "the thing I was waiting for", crosses to the motor areas that plan a press, runs down the spinal cord, and finally reaches the muscle, which itself takes a few milliseconds to develop force.

Raw travel time is the cheap part. Myelinated nerve fibres carry impulses at 50 to 120 metres per second, so the roughly one and a half metres from eye to fingertip costs something like 15 to 30 milliseconds. Everything else — the great majority of your 250 ms — is processing.

Sound skips the slowest step. Hair cells in the cochlea convert vibration to nerve signal much faster than photoreceptors convert light, which is why auditory reaction times come out around 140 to 160 ms against 180 to 200 for vision. It is not that you care more about sounds; it is that the transducer is quicker.

Choice mode adds a decision. When there is more than one possible answer, the time grows with the logarithm of the number of options — that is the Hick–Hyman law, published by W. E. Hick in 1952. Two options do not double anything; they add a fixed slice for the choosing.

By hand

The ruler drop

A ruler in free fall is a clock, and a very good one, because gravity is the same in every classroom. The tool has a converter for it, but the measurement itself needs no computer at all — which also means no screen and no keyboard adding milliseconds to your result.

1

One person holds a 30 cm ruler by its high end, hanging straight down, with the zero mark at the bottom.
2

The other person puts thumb and forefinger either side of the zero mark, ready to pinch, without touching the ruler.
3

No countdown, no "ready". A countdown turns this into the same rhythm problem the random delay exists to avoid.
4

The number at the top of your pinch is the distance the ruler fell.
5

The ruler falls d = ½gt², so t = √(2d/g). Twenty centimetres is 202 milliseconds. Ten is 143. Five is 101. The scale is quadratic, which is why the marks crowd together at the fast end and why a 30 cm ruler runs out at 247 ms.

Diagnosis

If something is not working

Symptom:My times are much slower here than on other reaction tests

Likely cause:: Different pages measure with different amounts of equipment in the way, and some start their clock at a different moment. A laptop on battery saver also throttles the browser, which delays the frame the signal is painted on.
What to do:: Plug the laptop in, close other tabs, and try hearing mode — it takes the screen out of the loop entirely. Comparing your own sight and hearing numbers on the same machine is more informative than comparing either against a stranger's.

Symptom:Every attempt comes back as a false start

Likely cause:: You are pressing on expectation rather than on the signal. The delay is deliberately unpredictable, so a press timed by counting will usually land early.
What to do:: Rest your finger on the key without pressure and watch the box rather than the clock. If it keeps happening, switch to hearing mode: a tone is harder to anticipate than a colour you can see approaching in your peripheral vision.

Symptom:Hearing mode does not play anything

Likely cause:: Browsers refuse to start audio until the page has been interacted with, and some devices route the tab to a different output than you expect.
What to do:: Click the box once to start a run, which counts as the interaction. If it is still silent, check the volume mixer and the output device, then try the audio test on this site to confirm the tab can make sound at all.

Symptom:The results jump around between attempts

Likely cause:: That is normal, and it is why the spread is shown. Attention drifts, and a single drift shows up as one slow attempt.
What to do:: Look at the median and the list of attempts rather than any single figure. If the spread is above about 50 milliseconds, run it again when you are less tired — the variation usually shrinks before the median does.

Symptom:My phone gives worse numbers than my computer

Likely cause:: A touchscreen samples your finger at its own rate and then has to decide the touch was a tap rather than the start of a scroll, which adds time before the browser ever hears about it.
What to do:: Compare phone against phone and computer against computer. If you want the fairest possible measurement on either, use hearing mode and the ruler drop, which removes the digital path completely.

Trivia

Things worth knowing

Astronomy invented it by accident

Reaction time entered science because observatories could not get their clocks to agree. The personal equation — the stable difference between two observers timing the same star — was a correction table long before it was a psychology topic.

Galton charged threepence

At the 1884 International Health Exhibition, visitors paid threepence to be measured at Galton's Anthropometric Laboratory: height, weight, grip, hearing, and reaction time. More than nine thousand did it, producing the first large dataset of human reaction times.

More options cost a logarithm, not a multiple

Hick's law, published in 1952, says choice reaction time grows with the logarithm of the number of alternatives. Going from two choices to four does not double the decision time; it adds one more fixed step.

A tenth of a second is a legal threshold

Formula 1 does not judge a jump start by whether the car moved before the lights went out, but by whether the driver reacted faster than 0.1 seconds afterwards. The rule encodes an assumption about the human floor into the regulations.

The nerve is not the bottleneck

Impulses travel myelinated fibres at up to 120 metres per second, so the eye-to-finger journey costs perhaps 20 milliseconds of a 250 millisecond total. Nine tenths of your reaction time is deciding, not transmitting.

Questions

Frequently asked questions

What is a good reaction time?

For a simple visual signal, 200 to 250 milliseconds is the ordinary adult range, and under 200 is genuinely quick. But compare like with like: a number measured on a 240 Hz monitor with a 1000 Hz mouse is not comparable with one from a 60 Hz laptop, and the difference can be 15 milliseconds or more before your nerves are involved at all.

Why is my hearing time faster than my sight time?

Because the ear is a faster transducer than the eye. Converting light into a nerve signal in the retina takes tens of milliseconds; converting vibration into one in the cochlea is much quicker. The rest of the chain is roughly the same, so the head start survives all the way to your finger.

Why does this page lead with the median instead of the average?

Because reaction times are skewed in one direction only. Nothing can make an attempt faster than your nervous system allows, but a moment of distraction can make one arbitrarily slower. One lapse in five attempts can pull the mean up by 80 milliseconds while leaving the median untouched. Both are shown, and the gap between them tells you whether you blinked.

Does my monitor really change the result?

Yes, and measurably. The page can only turn the box green when the screen next refreshes, so on a 60 Hz display the signal is on average 8.3 milliseconds late before you have seen anything. The panel adds the polling interval of your keyboard or mouse and shows the total as a range.

Can I train my reaction time?

A little. Practice on a specific task reliably improves the number for that task, mostly by removing hesitation rather than by speeding up the nerve. Sleep, alcohol and fatigue move it far more than training does, and it slows gradually with age from the mid-twenties onwards. Anyone promising large permanent gains is selling something.

Is this the same test Formula 1 drivers do?

The idea is the same and the conditions are not. An F1 start has five red lights that go out after a random hold, and the FIA treats anything under 0.1 seconds as a jump start on the assumption that no human can legitimately react faster. Drivers typically register around 0.2 s — with their hands already on the wheel, their eyes already on the lights, and no browser in between.

Why do I keep getting false starts?

Because you are anticipating instead of reacting, which is exactly what the random delay is designed to catch. The wait is drawn fresh between 1.2 and 4.5 seconds every time, so there is no rhythm to lock onto. False starts are counted separately and kept out of the statistics rather than silently dropped.