Human reaction time, explained
A reaction test looks like the most objective thing on the internet. A box turns green, you click, and a number appears. No opinion, no judgment, just milliseconds. In practice that number is one of the noisiest measurements a web page can produce, because it is the sum of a biological process and a stack of hardware and software that all take their own time. This guide walks through where the milliseconds actually go, so that you know what your score is made of and how much weight it can carry.
The chain between the light and the click
When the pad on your screen changes color, a surprising number of separate things have to happen before the click is registered, and each one costs time.
- Light from the changed pixels travels to your eye and lands on the retina.
- The retina transduces that light into nerve signals — a chemical and electrical process, not an instant switch.
- Those signals travel along the optic pathway to the visual areas at the back of the brain.
- Your brain has to actually recognize that a meaningful change occurred, as opposed to noise or an irrelevant flicker.
- A decision to act is made and a motor command is issued.
- The command travels down the spinal cord and along the nerves to your arm and hand.
- The muscle contracts, your finger moves, and the switch under it finally closes.
Two features of that list matter more than the details. First, the transmission steps are limited by how fast nerve signals propagate, which is fast but nowhere near instant, and the distance from your brain to your fingertip is not trivial. Second, the recognition and decision step is the one that varies most, both between people and from trial to trial in the same person. It is also the step that attention affects.
As a very rough order of magnitude, the whole chain for a simple visual signal lands somewhere in the low hundreds of milliseconds for most adults. We are deliberately not assigning numbers to the individual stages. Those values differ depending on the signal, the measurement method and the person, and quoting a precise figure per stage would imply a precision that a browser cannot deliver and that we cannot verify here.
Simple versus choice reaction time
There are two very different things people call reaction time, and mixing them up makes comparisons meaningless.
Simple reaction time has one signal and one response. Green means click. There is nothing to work out, so you are measuring detection plus movement.
Choice reaction time asks you to pick. Green means click left, red means click right; or click only when the shape is a circle. Now your brain has to identify which signal appeared and select the matching action before anything moves.
Adding that decision reliably adds time, and the effect is large enough that it is never in doubt. Broadly, the more alternatives there are to choose between, the longer the response takes. This is why a driving reaction, a fighting-game reaction and a green-box reaction are not comparable quantities even in principle. Our test is deliberately a simple reaction test: one signal, one response, nothing to decide.
The hardware tax nobody mentions
This is the part that most reaction-time articles skip, and it is often the difference between two scores that people treat as a difference between two people.
The display only updates on a cycle
Your screen does not change the instant the browser decides something should change. It redraws on a fixed refresh cycle. On a 60 Hz display those cycles are roughly 16 milliseconds apart, so a change that becomes ready just after a refresh has to wait for the next one. That waiting is pure quantization, it has nothing to do with you, and on its own it can add up to about 16 milliseconds on a 60 Hz panel. On a 120 Hz or 144 Hz screen the same effect exists but the gaps are smaller. Panels also take time to physically change their pixels, which adds a little more.
Your input device samples at its own rate
A mouse reports its state at a polling rate rather than continuously, and a touchscreen scans for contact on its own schedule. Your click exists in the physical world the moment the switch closes, but it only exists to the computer at the next sample. A wireless mouse or a Bluetooth controller adds transmission latency on top of that.
The browser has its own scheduling
Between the operating system receiving the input event and your JavaScript handler running, the browser is doing work: dispatching events, running its own frame loop, painting. Timestamps are taken at particular points in that pipeline, and browsers deliberately limit timer precision for privacy reasons. None of this is a bug. It is just latency that lands inside your measured number.
Add it together and you get the practical conclusion: the same person will score differently on a phone and on a desktop, and comparing your number to a stranger's number is close to meaningless. You are not comparing two nervous systems, you are comparing two entire setups. Comparing your own score across sessions on the same device is a much more honest exercise, and even that has limits.
All figures here are rough orders of magnitude, not published statistics. As an estimate from our own testing across a handful of devices, moving between a phone and a desktop can shift the same person's average by a few tens of milliseconds, and a 60 Hz display contributes up to roughly 16 milliseconds of refresh quantization by itself. Treat both as ballpark reasoning about the equipment, not as a measurement of you.
Why the delay before the signal has to be random
A reaction test is only a reaction test if you cannot predict when the signal will arrive. If the pad always turned green two seconds after you started, you would quickly learn to click at two seconds and your score would measure your sense of timing, not your reaction. That is anticipation, and it can produce numbers that look impressive and mean nothing.
So the wait before each signal is randomized across a range wide enough that guessing does not pay. A consequence of that design is the early click: if you click while the pad is still in its waiting state, you did not react to anything, because there was nothing there yet. There is no valid measurement to record. Our test therefore discards that attempt and repeats the trial rather than logging a suspiciously fast score. It is not a punishment, it is the only correct way to handle a trial with no stimulus in it.
If you find yourself getting several early clicks in a row, that is useful information in itself: it usually means you have started predicting instead of waiting.
Things that plausibly move your number
Reaction time is not a fixed personal constant. It moves around, and several ordinary factors are widely understood to nudge it. We are describing directions here, not magnitudes.
- Fatigue. Tired and sleep-deprived attempts tend to be slower and, more noticeably, more variable.
- Caffeine and general alertness. Where on your day's arousal curve you happen to be affects how ready you are to respond.
- Warm-up. The first trial or two of any session are often slower than the rest, simply because you are still settling into the task.
- Anticipation. Leaning into an expected signal shortens the number without improving anything real, and it produces early clicks when you guess wrong.
- Whether your finger is already down. Holding a button pressed and releasing it is a different physical action from pressing from a resting position, and the two are not interchangeable.
- Where you are looking. A signal in your peripheral vision is not the same task as one you were staring at.
What the number does not mean
Reaction time is not a measure of intelligence. It is not a measure of skill at any particular game. It is not a diagnostic of anything, and it cannot tell you whether something is wrong with you. It is a rough timing of one narrow task on one particular set of equipment.
It is also statistically thin. A single session of five trials carries enough variance that small gaps between sessions are noise. As a rule of thumb, only differences on the order of tens of milliseconds are worth talking about at all, and even those deserve more than one session before you believe them.
One practical recommendation: prefer the median of your trials over the mean. One distracted attempt where you looked away or your hand drifted can be far slower than the rest, and because the mean adds that whole excess to the total, a single bad trial drags it noticeably. The median just steps past it and reports your typical attempt, which is what you actually wanted to know.
Our reaction test is not a clinical or diagnostic tool, and nothing on this page is medical advice. Boss Games makes no health claims. If you are worried about your reflexes, your vision, your alertness or your coordination, please speak to a qualified professional rather than drawing conclusions from a browser tab.
How to run a session that tells you something
If you want the number to be at least internally consistent, hold the setup steady. Use the same device, the same browser, the same hand and the same posture each time. Do two or three throwaway trials before the ones you count. Sit at a similar time of day. Then record the median of a full run rather than your single best attempt, because the best attempt is the one most likely to have been a lucky anticipation.
Compared against itself, under those conditions, your score becomes a mildly interesting personal metric. Compared against a leaderboard full of strangers on unknown hardware, it is entertainment. It is worth knowing which one you are looking at.