Skip to the test
HumanBenchTests

Reflex · five trials, median reported

Reaction Time Test

Click the panel, wait for it to change, then hit it as fast as you can. This reaction time test runs five trials and reports your median in milliseconds — not your best, and not your average. Most adults land between 230 and 320 ms. Timing uses performance.now() sampled after the frame that actually presents the stimulus, so the reading is not flattered by a frame of display latency.

Trial 1 of 5
False starts 0

How is reaction time measured here?

Three details separate a reaction time test that means something from one that does not.

  1. A monotonic clock. performance.now() has sub-millisecond resolution and, unlike Date.now(), cannot jump backwards when the system clock is adjusted mid-measurement.
  2. Timing from the paint, not from the state change. When the code decides to show the stimulus, nothing is on screen yet. The start timestamp is taken inside a double requestAnimationFrame, which lands after the frame containing the stimulus has actually been presented. Timing from the decision instead would silently subtract up to a full frame from every score.
  3. A random wait. The delay before the stimulus is drawn uniformly between 1.4 and 4.6 seconds. A fixed delay would let you learn the rhythm and time your click against the clock rather than against the signal.

How much of your reaction time is the hardware?

Every browser reaction reading contains display latency and input latency. The display component is calculable: a frame is held for 1 ÷ refresh rate, and the stimulus may arrive anywhere within one. Add roughly 5 to 20 ms for mouse sampling and event dispatch and you get the ranges below.

Overhead added by the display and input chain, by refresh rate
Refresh rateFrame timeTypical overhead added
60 Hz16.7 ms13.3–45 ms
75 Hz13.3 ms11.7–40 ms
120 Hz8.3 ms9.2–32.5 ms
144 Hz6.9 ms8.5–30.4 ms
240 Hz4.2 ms7.1–26.3 ms

This is why a score measured on a 240 Hz gaming monitor with a wired mouse is not comparable to one measured on a 60 Hz laptop with a trackpad, and why the difference can be 30 ms or more before the person is considered at all. Compare your own scores across sessions on the same hardware; treat cross-device comparisons with suspicion.

What your number means

Bands cut on the reference distribution used by this test
MedianBandApprox. percentileInterpretation
120–200 msExceptional95thTop few per cent of the modeled distribution. At this level display latency is a large share of what you are measuring.
200–240 msFast74thClearly quicker than the middle of the distribution and hard to sustain when tired.
240–300 msTypical28thThe band that contains most adults. The modeled median is 270 ms.
300–380 msRelaxed3rdCommon when tired, distracted, or on a laptop trackpad rather than a mouse or a touchscreen.
380+ msSlow for this test< 3rdUsually a divided-attention or hardware effect rather than a reflex one. Repeat it rested, on a wired input.

The reference distribution is a log-normal model with a 270 ms median, matching the medians published by large browser-based click tests. It is stated as a model wherever it appears here rather than dressed up as measured data.

Reaction time versus aim

The number on this page is a simplereaction time: one stimulus, one response, no decision. Hitting a target that appears somewhere on screen is a different task — it adds a ballistic movement plus one or more corrective sub-movements, and it obeys Fitts’s law rather than being a fixed latency.

That is why the aim trainer reports 400 to 800 ms per target while this page reports 200 to 300 ms, and why the two numbers should never be compared directly. If you want to see the split for yourself, run both: the difference between them is roughly your movement time.

For the other side of latency — what your display and input hardware actually contribute — scrolltest.io measures the device instead of the person.

Reaction time questions

What is a good reaction time?

On a browser click test, most adults land between 230 and 320 milliseconds, with the middle of the distribution around 270 ms. Under 200 ms is genuinely fast — and at that level a meaningful share of what you are measuring is your monitor and mouse rather than your nervous system.

What is the average human reaction time?

Laboratory studies of simple visual reaction time report means in the region of 200 to 250 milliseconds for young adults. Jain et al. (2015), in the Int J Appl Basic Med Res 5(2):124–127, measured mean visual reaction times of roughly a quarter of a second and found auditory reaction times consistently faster — sound reaches the cortex through a shorter path than vision.

Why does this test use the median instead of the average?

Because reaction time distributions are right-skewed. One lapse of attention adds 200 ms to a single trial and drags a five-trial mean well away from the value that describes you. The median ignores that outlier. Both numbers are shown so you can see the gap — a large one means attention, not reflexes.

How much of my score is the computer?

More than most people expect. A 60 Hz display holds each frame for 16.7 ms, so the stimulus can wait up to a full frame before your eye receives it, and the click has to be sampled by the mouse, delivered to the browser and dispatched to the page. Expect 20 to 40 ms of overhead on typical hardware — the table below breaks it down by refresh rate.

What counts as a false start?

Anything faster than 100 ms after the panel changes. That is not a strict reaction: the retina alone takes tens of milliseconds to transduce light, and the motor path adds far more on top. A response under 100 ms was already in flight before the stimulus existed, so the trial is discarded rather than counted.

Can I improve my reaction time?

The floor set by nerve conduction and motor delay is fixed, but most people are nowhere near it. Sleep, caffeine timing, screen refresh rate, using a mouse instead of a trackpad and simply expecting the stimulus all move a browser reading by tens of milliseconds. Consistency — a small spread between trials — improves with practice more reliably than the raw number does.

Other instruments on this bench

Testing the hardware rather than the person? device tests — scroll, touch, refresh rate and input latency lives on a sister site. And if you want to see through someone else’s eyes rather than screen your own, use the simulate how an image looks to a protan, deutan or tritan viewer.