ReflexTimer

Reaction Time Test Accuracy: What the Milliseconds Really Include

By oogwaaay (ReflexTimer operator) · Last updated: November 7, 2026

A reaction time score from a browser is not a pure measure of nerve speed. This guide explains every stage that contributes to the number, what you can control, and how to compare scores honestly.

When a page reports "your average was 240 ms", the obvious interpretation is that your nervous system took 240 ms to respond. The more accurate interpretation is that the entire measurement chain took 240 ms: the browser changed the screen, the display showed a new frame, your eyes detected it, your brain selected a response, your hand moved, the input device reported the action, and the browser timestamped the event. The difference between those two interpretations is the subject of this page.

Where the milliseconds come from

A browser reaction test measures signal-to-input latency, not isolated neural conduction. Two people with identical nervous systems can legitimately produce scores tens of milliseconds apart on different setups. That does not make the test useless; it makes it a personal benchmark whose conditions matter. If you keep the setup constant, changes in your score are informative. If you change devices between sessions, you are partly measuring the devices.

Where the milliseconds go

A single attempt passes through roughly these stages:

  1. Signal generation. The test decides the signal should appear and updates the page state.
  2. Rendering and compositing. The browser paints the new state and the operating system composites it with other windows.
  3. Display. The monitor receives a frame, scans it out, and the panel's pixels respond.
  4. Perception and decision. Light reaches your eye, your visual system registers the change, and you choose to respond.
  5. Motor response. Your hand or finger moves and activates an input.
  6. Input reporting. The device sends the event, the operating system processes it, and the browser dispatches a click or tap.
  7. Timestamping. The test reads a high-resolution clock when the input event is handled.

The measured value is the difference between the clock readings at the start and end of that chain. Every stage can add latency, and only some of them are under your control.

Display: refresh rate is a floor, not the whole delay

A display does not show changes continuously. It refreshes at a fixed rate: a 60 Hz panel refreshes about every 16.7 ms, a 144 Hz panel about every 6.9 ms, and a 240 Hz panel about every 4.2 ms. A change requested between two refreshes waits for the next one, so the refresh interval is a lower bound on how long the signal can be invisible after the browser has decided to paint it.

The visible delay is usually longer than that bound. The frame must be scanned out, the panel must complete its pixel response, and the operating system may add a composition step. Some monitors also apply processing such as overdrive, scaling, or motion smoothing, which can add latency even when the refresh rate is high. This is why a higher refresh rate tends to reduce both the average and the variance of browser scores without changing your nervous system: you are removing a source of waiting, not becoming faster.

Input: polling, debounce, and the path back

On a mouse, button presses are sampled rather than delivered as continuous data. A mouse polling at 125 Hz reports every 8 ms; at 500 Hz every 2 ms; at 1,000 Hz every 1 ms. That interval is a small but real part of the return path. Switch debounce — the short window during which the mouse ignores further contact because the switch is settling — can also delay or reject a very fast press, though for a single deliberate click it is usually not the dominant term.

Keyboards have their own scan rates and debounce settings. Touchscreens introduce a further step: the browser may need to distinguish a tap from a scroll or a multi-touch gesture before dispatching a click. This is one reason phone scores often differ from desktop scores even when the person is equally alert. The test can only timestamp the event it receives; it cannot see the physical contact that happened earlier.

Browser timing is precise, but it is not the hardware's clock

Modern browsers expose a high-resolution timer such as performance.now(), with sub-millisecond resolution in typical desktop configurations. That precision is useful for measuring the interval between two events inside the page. It does not make the browser aware of what the display or the device did before the event arrived. The timestamp marks when JavaScript handled the event, not when the photon left the screen or when your finger first moved.

This is also why the page's own implementation details matter. A test that starts its clock before the browser paints the green frame will include some rendering delay in the score; a test that starts the clock on the next animation frame includes a different amount. The number is meaningful, but it is a property of the whole system — test code, browser, operating system, and hardware — not a universal physical constant.

Why five attempts help

Any single attempt is a sample from a noisy process. Attention drifts, a frame lands at an unlucky moment, a finger moves a little differently. Averaging five attempts reduces the influence of one outlier. The median shown alongside the average is even less sensitive to a single very slow attempt, which is why the result card shows both. Neither number eliminates the device latency; they only reduce the noise in your own estimate.

This is also why comparing a single fast attempt to a friend's single fast attempt is close to meaningless. Compare medians from several rounds, on the same setup, under similar conditions — that is a comparison of people rather than a comparison of lucky frames.

What you can control

What you cannot control

You cannot remove the panel's internal processing, the operating system's compositor, the physical distance light travels from the screen to your eye, or the time your visual system needs to register a change. You also cannot make a browser event timestamp include hardware events that happened before the browser saw them. Those limits are why a browser score should never be presented as a laboratory-grade measurement of neural conduction time.

How to compare scores honestly

Compare like with like. A phone score and a desktop score are different measurements. A 60 Hz laptop and a 240 Hz monitor are different measurements. Even two browsers on the same machine can differ in how they schedule rendering and input.

The practical rule is simple: keep one setup, record it, and compare your own results over time. If you want to compare with other people, compare within the same device class and say so. Published research numbers usually come from controlled laboratory equipment, so a browser score is expected to read higher; that is a difference in method, not evidence that the population is slower than the literature.

A five-point accuracy checklist

  1. Use the same device, browser, and display for every session you intend to compare.
  2. Note the refresh rate and whether the mouse is wired; these belong in your record.
  3. Run more than one round; use the median of several rounds as your baseline.
  4. Do not compare a single fast attempt with a single slow attempt; compare distributions.
  5. When you change hardware, change one thing at a time and re-baseline before drawing conclusions.

For a practical guide to checking your own display and input settings, see display and input latency calibration. For the exact scoring rules used on this site, see Methodology. For published ranges by age and stimulus, see Average Reaction Time. If you want to improve the controllable part of the chain, the Improve Reaction Time guide separates practice effects from device effects.

Common ways people misread the number

"I got faster after changing my mouse." You may have measured less input latency, not faster nerves. That is a legitimate improvement in your setup, but it is not a change in biological reaction time. If you want to separate the two, test the old and new device under identical conditions before concluding anything about yourself.

"My phone is slower than my laptop, so I am slower on mobile." The phone is a different measurement chain: touch sampling, gesture disambiguation, display timing, and browser scheduling differ from a desktop mouse. A slower mobile score is expected and does not describe a different nervous system.

"One round gave me 190 ms, so that is my reaction time." That is your best sample from one round, not your typical value. Single attempts vary for reasons that have nothing to do with skill; use the median of several rounds and report the conditions.

"The research says 200 ms, so my 240 ms is slow." Laboratory simple-reaction studies and browser tests measure different chains. A browser score includes device latency and is expected to read higher than a controlled lab number. The comparison is only meaningful when the method is comparable — which is why this page explains the chain rather than presenting a single universal threshold.

The same measurement chain is covered from other angles in the full guides index.

References