How Reaction Time Tests Actually Work
Three engineering decisions decide whether your score is meaningful.
TL;DR — A trustworthy reaction test randomises stimulus delay, syncs rendering to monitor refresh and timestamps the input at the JavaScript event itself. Tests that skip any of those three produce systematically biased numbers.
Most online reaction tests are honest about scores but quiet about their accuracy. A well-built test should explain its timing source and refresh-rate assumptions.
Decision 1 — Randomised stimulus delay
The cue must appear after a random wait. Best practice is uniform distribution between 1 and 5 seconds. Fixed delays let the brain predict the cue and produce sub-floor times (sometimes under 100 ms) that are anticipation, not reaction. SERO uses random 1.2–4.5 second delays.
Decision 2 — Render synced to monitor refresh
The cue must appear exactly when the browser repaints. Drawing it inside a setTimeout produces 0–16 ms of phantom jitter on a 60 Hz display. The correct method is requestAnimationFrame, which fires immediately before the next repaint.
Decision 3 — Timing at the input event
The stop timestamp must be recorded inside the input event handler itself, not later in a state update. JavaScript event loops can delay state updates by 1–5 ms — small but non-zero. The cleanest approach is performance.now() captured inside the keydown or pointerdown handler.
Sources of unavoidable error
| Source | Typical magnitude |
|---|---|
| Display refresh (60 Hz) | ±8 ms |
| Display refresh (144 Hz) | ±3 ms |
| Display refresh (240 Hz) | ±2 ms |
| Mouse polling (1000 Hz) | ±0.5 ms |
| USB processing | ±1 ms |
| Browser scheduling | ±1–2 ms |
What makes a result trustworthy
- Average of at least 5 attempts.
- Reasonable variance (under 60 ms between best and worst).
- No suspiciously fast outliers (sub-140 ms on vision).
- Consistent stimulus delay range disclosed.
Frequently Asked Questions
Why does my reaction test give different scores on different sites?
Mostly because of unrandomised delays, different timing sources and different display refresh rates.
Can browser tests match clinical tests?
Within 10–20 ms on modern hardware. Clinical EMG is still the gold standard for sub-millisecond accuracy.
Are mobile tests less accurate?
Slightly. Touch latency adds 10–25 ms compared to a wired mouse on the same person.
Test Your Reaction Time Now
Free. 90 seconds. Global leaderboard. No download.
Start Test