Reaction Speed Test

A reaction speed test measures the delay between a visual signal and your response. In this page, you wait for the test area to change state, then click or press a key as quickly as possible; the result is shown in milliseconds. People use it to compare recent attempts, warm up before games, demonstrate attention and latency, or simply learn how fatigue and focus change performance. The number is not a pure measure of your nervous system: monitor refresh rate, mouse or keyboard latency, browser scheduling, display brightness, and early-click penalties can all affect it. For reliable comparisons, use the same device, same input method, and several attempts rather than judging one isolated result.

The clock that times the reaction is performance.now(), the high-resolution timer from the W3C High Resolution Time spec. In a top-level browsing context it returns a DOMHighResTimeStamp with sub-millisecond precision (Chromium reports 100 μs steps after Spectre mitigations, Firefox 1 ms by default in private windows). The legacy Date.now() is only ~1 ms resolution and is therefore unsuitable for sub-frame measurements. Both the random pre-stimulus delay (1-5 s) and the elapsed click delta are computed in the performance.now() timebase. The stimulus state change is driven by requestAnimationFrame, which the browser pegs to the display's vsync signal. At 60 Hz the frame interval is 16.67 ms, at 120 Hz it is 8.33 ms, at 144 Hz it is 6.94 ms, and at 240 Hz it is 4.17 ms; the result therefore quantises to whatever the panel can show. Below ~150 ms most of the measurement is no longer the user — it is the input pipeline: a 1000 Hz USB-polling wired mouse adds about 1 ms, a Bluetooth mouse adds 7-30 ms, and OS event coalescing can stretch this further on a battery-saving laptop. Real human simple-reaction-time data sits in a roughly normal distribution with a mean around 250 ms and a standard deviation around 30 ms, walking the same chain every time: photon hits retina, signal traverses the LGN to V1, motor planning fires in the premotor and motor cortex, the corticospinal tract drives the finger flexor, and the click registers. A premature click is therefore filtered as a 'too early' state to keep that distribution from being polluted with anticipation guesses that would otherwise read as fake 50-100 ms records.

• Clock: performance.now() (DOMHighResTimeStamp), 100 μs steps in Chromium top contexts, 1 ms in some private windows; Date.now() is only ms-precision and is not used.
• Frame quantisation: requestAnimationFrame is locked to vsync, so the smallest measurable delta is one frame: 16.67 ms at 60 Hz, 6.94 ms at 144 Hz, 4.17 ms at 240 Hz.
• Input latency floor: ~1 ms for a 1000 Hz USB-polling wired mouse, 7-30 ms for a typical Bluetooth mouse, +OS event-coalescing jitter in power-saving mode.
• Human baseline: simple visual reaction time is roughly normal with mean ≈ 250 ms and SD ≈ 30 ms; sub-150 ms records are almost always device-floor, not nervous-system gains.
• Neural chain: retina → LGN → V1 → premotor/motor cortex → corticospinal tract → finger flexor, summing to the irreducible biological latency band.
• Anti-cheat: clicks before the stimulus are flagged as 'too early' so anticipation guesses do not pollute the distribution as artificially fast records.