DeviceBenchFree online reaction time test
Reaction time test that refuses a false start
Click once to arm the field, wait while it is red, and click again the instant it turns green — five times over, for a median instead of one lucky attempt. The clock starts on the frame the green was actually painted rather than on the code that asked for it, and a press that lands while the field is still red is thrown away rather than scored. What comes back is the whole loop and not only you: a tab can time from paint to press but cannot see the milliseconds your panel and your mouse add at either end, so the page measures the refresh rate it is being drawn at and prints that share of the number separately.
- 100% free
- No signup
- 5 attempts
- Median, not best
- False starts refused
Click to begin
5 attempts. Click, wait for green, click again.
Attempt 1 of 5. Space or Enter works instead of a click.
Nothing scored yet. Each attempt is listed here as it lands, because a run of 250, 255, 251 is a different hand from 190, 340, 220.
Measuring your display now — the figure appears once enough frames have been timed. If it stays empty, this tab is in the background and the browser has stopped sending frames.
Every figure here is the whole loop: eye, brain, hand, mouse, browser and panel, in that order, and none of it can be unpicked from inside a tab. On a phone the touchscreen adds its own scan on top, which is why a tablet score and a desktop score are not the same measurement. Nothing about the number is medical — it moves with caffeine, sleep and the hand you used.
How to take the test without fooling yourself
The measurement is easy; not gaming it is the part worth the three steps.
Click once to arm it, then leave your hand still
The field goes red and stays red for somewhere between 1.2 and 5 seconds, redrawn at random every attempt so there is no rhythm to learn. Resting a finger on the button is fine; pressing it is not, and the interval is deliberately long enough that holding your breath for it does not help. If you use a keyboard, focus the field once and Space or Enter does the same job as a click.
Click on green, and accept a refusal when it comes
Two presses get thrown away rather than scored: one that lands while the field is still red, and one that lands under 100 ms after green, which is quicker than sight and muscle can manage together. Both replay the same attempt rather than costing you one, so a run of five scored figures is always five genuine responses. The count of refusals is kept and shown, because a run with four false starts in it says something the timings do not.
Read the median, then the gap between fastest and slowest
The median is the headline because one attempt where you blinked drags a five-attempt mean by tens of milliseconds and the median by nothing. The spread underneath it — slowest minus fastest — is the figure worth watching over several runs: a tight 20 ms spread means you have found a rhythm, and a 120 ms spread usually means you were anticipating on some attempts and genuinely responding on others.
Technical specifications
| Scored attempts | 5 per run; a refused press replays the attempt rather than using one up |
|---|---|
| Wait before green | Random between 1,200 and 5,000 ms, drawn again for every attempt |
| Where the clock starts | The task after the frame the green was painted on — a requestAnimationFrame callback followed by a MessageChannel message, so a frame the eye never saw is not charged to you |
| Refusal floor | 100 ms; retinal and motor delay together rule out anything faster |
| Headline figure | Median of the five, with mean, fastest, slowest and the spread beside it |
| Display allowance | Refresh rate estimated from about 200 frame intervals, then shown as the per-frame wait in ms |
| Accepted input | Primary mouse button, a finger, or Space and Enter on the focused field |
| Data kept | None — no leaderboard, no saved best, nothing survives the tab |
Frequently asked questions
What counts as a normal result?
Most adults clicking a mouse in a browser land between 200 and 300 ms, and the middle of that band is where the majority of runs sit. Under 200 ms is quick for this setup and under 180 ms is unusual. Laboratory figures are faster — a dedicated response button, a chin rest and a CRT put simple visual reaction near 190 ms — and the difference is not the person, it is everything a web page has to borrow: an operating system, a mouse and a panel that were all designed for something other than being timed.
Why was my 90 ms result rejected?
Because it was already on its way before the green appeared. The retina alone spends roughly 20 to 40 ms converting light into a signal, the visual pathway and the decision add more, and the motor half — spinal cord to finger — costs tens of milliseconds again, so a genuine response cannot arrive in under about a tenth of a second. A test that scores those numbers rewards guessing at the right moment, and the fastest score on the page stops meaning anything.
Does a 144 Hz monitor actually improve the number?
Yes, by around 5 ms on average, which is real but smaller than people expect. A frame that is ready has to wait for the display's next refresh: at 60 Hz that wait averages 8.3 ms and can reach 16.7 ms, and at 144 Hz it averages 3.5 ms and tops out at 6.9 ms. The panel's own pixel response and the monitor's internal processing sit on top of both figures and vary far more between models than the refresh rate does. This page measures the refresh rate it is being drawn at and prints the frame allowance beside your result, so the part that is your display is at least visible.
Why does the wait before green vary instead of being fixed?
Because a fixed wait stops measuring reaction and starts measuring your internal clock. Given the same interval three times, most people begin anticipating it and press on the count rather than on the cue, which produces impossibly fast figures followed by false starts. A window from 1.2 to 5 seconds, redrawn each attempt, is wide enough that anticipation costs more than it gains.
Why five attempts rather than one?
Because a single reaction figure has a spread of 50 to 100 ms around it, and whichever one you happen to get first is as likely to be your worst as your best. Five gives a median that stops moving much between runs while still being short enough that nobody stops halfway. Anything past about ten attempts starts measuring boredom instead: the interval creeps upward as attention drifts, which is a real effect but not the one you came for.
Is this a health check?
No, and no browser test can be. The figure moves with sleep, caffeine, time of day, which hand you used and whether you had just been staring at the field or had glanced away — swings of 30 to 50 ms from those alone are ordinary in the same person on the same machine. It is a way to compare your own runs, or one machine against another, and nothing here is diagnostic of anything.
Does clicking a phone screen give the same figure as a mouse?
No, a touchscreen usually reads slower, and the difference is the panel rather than your thumb. A touch digitizer samples at 60 to 120 Hz and the browser has to wait long enough to rule out the start of a scroll or a pinch before it delivers the press, which stacks tens of milliseconds onto a figure a mouse would have reported immediately. Compare phone runs with phone runs.
About what a reaction figure is made of
What this measures has a name in the literature: simple visual reaction time, one stimulus and one response with no choice in between. It is the fastest of the reaction families, and it is not even the fastest sense — a sound reaches a decision in roughly 140 to 160 ms against 180 to 200 ms for light, because the auditory pathway is shorter and the cochlea transduces faster than the retina does. Add a choice, as in press left for red and right for blue, and another 100 ms or so appears from nowhere. So a figure only means something next to another figure gathered the same way, which is the argument for running this several times on one machine rather than comparing a screenshot with a stranger’s.
The rest of the number is your equipment, and it is a bigger share than most people assume. A press has to reach the browser: a mouse polling at 125 Hz can sit on it for up to 8 ms before reporting, which drops to 1 ms at 1000 Hz. The green frame has to reach your eye: the browser composites it and the display shows it at its next refresh, then the panel needs a few more milliseconds to change the pixels — none of which any web API exposes. That machine half is worth measuring on its own rather than guessing at, and two pages here do it with the human taken out of the loop: the input lag test times a press to the pixel that answers it, and the FPS test shows how often your browser gets to draw anything at all.
The last thing worth knowing is how much a single person varies. Run this five times in an evening and the medians will move by 20 to 40 ms without anything about you changing except attention — reaction figures drift upward through a long session, drop after a break, and are measurably worse on the non-dominant hand. That variability is exactly why the headline here is a median of five rather than a best of one: a best-of figure rewards the attempt where you happened to be anticipating, which is the same instinct the false-start rule exists to catch. If you want the version of this where the cue is a position rather than a color, the aim trainer adds a movement to the response, and the chimp test replaces the reflex with something to remember.
What happens to your five timings
Every number on this page is worked out by JavaScript running in the tab you are reading it in. Nothing you type, paste or open is uploaded, logged or kept, which is also why the tools carry on working after you disconnect from the network.
There is no leaderboard here, so there is nothing for a result to be submitted to. The five timings and the display measurement live in this tab’s memory for as long as the tab does, and closing it is the entire deletion process. Copy puts the summary on your clipboard; nothing else moves it anywhere.