← All guides

What a reaction time is made of

A quarter of a second is not hesitation. It is the speed limit of the wetware, and every millisecond of it is spoken for.

Updated August 31, 2026

Tap the screen the instant it turns green, and about a quarter of a second passes before anything happens. That gap feels like slowness. It is not; it is the itinerary of a signal that has real distance to cover and real work to do on the way, and none of it is optional.

The itinerary

Light from the green field hits your retina, and the first cost is paid immediately: photoreceptors and retinal circuits take roughly 20 to 40 milliseconds to turn light into a neural signal worth sending. The signal travels up the optic nerve and reaches the visual cortex at the back of your head at around 50 to 80 milliseconds. Only now does your brain contain the fact that something changed.

Recognizing that the change is the one you were waiting for, and releasing the response you had prepared, takes the biggest and most variable slice: something like 60 to 120 milliseconds of cortical work. This is the part that attention, fatigue and caffeine push around, and the part that varies from run to run even when you feel identical.

Then the motor command travels from your motor cortex down the spinal cord and out to the muscles of your hand, at conduction speeds that sound fast and are not free: another 10 to 20 milliseconds. The muscle itself contracts, your finger moves the last few millimeters, and the switch under it closes. In total: about 180 to 250 milliseconds for a practiced adult on a good day, before the hardware between you and the site adds its own share.

This is why the reaction time test refuses any press under 100 ms. Nothing human crosses eye to muscle that fast; a sub-100 press is a guess that got lucky, and counting it would make the score partly a lottery.

One signal, no decisions

What the simple test measures is deliberately stripped: one signal, one prepared response, nothing to decide. You are not choosing to click; you decided to click minutes ago, and the click is sitting in your motor system like a drawn bow. The green light only releases it. That is also why the wait is random: if the flip could be counted out, you would not be reacting to the light at all but predicting it, which is a different skill with its own test.

The Dutch physiologist Franciscus Donders worked this out in 1868 with nothing but a rotating drum and patience: measure the simple case, then add ingredients one at a time and subtract. His method, a century and a half old, is still how reaction tests are built, including the ones on this site.

Add a decision, pay for it

Give the signal four possible identities, each demanding its own key, and the response can no longer be pre-loaded. Now the brain must identify what it saw, select the matching response, and only then move. The choice reaction test measures exactly this, and shows you the price as a number: the decision lag, typically one to two hundred milliseconds on top of your baseline.

The price is lawful. Hick’s law, from 1952, says choice reaction time grows roughly linearly with the logarithm of the number of options: every doubling of the alternatives costs about the same extra slice. Four colors is two doublings past one; eight would cost one more slice, sixteen another. Your keyboard’s home row runs out before the law does.

Add a “don’t”, pay differently

There is a third ingredient: sometimes the right response is nothing at all. The go/no-go test makes most signals demand a fast tap and a minority forbid it, and measures whether you can cancel an action your body has already prepared. This is inhibition, and it fails in a characteristic way: the slip, the tap that escapes on a red light. Slips are not slowness; fast, confident responders slip more, not less, because their tap is further out of the barn when the red arrives.

The escape hatch

Athletes returning 200 km/h serves do not have supernatural reaction times; measured in the lab, elite athletes sit only modestly below average. What they have is prediction: they read the server’s toss and shoulders, and their response is planned before the ball exists as a stimulus. That skill, acting so your move lands exactly when the world needs it, is anticipation, and it has its own test here, scored in milliseconds of forecast error rather than milliseconds of transmission.

The practical order of the three numbers surprises people: a good anticipation miss (tens of milliseconds) is far smaller than a good reaction time (about two hundred), which is smaller than a good choice reaction (four to six hundred). They are not competing measurements of one speed. They are three different jobs, and the milliseconds mean different things in each.

What the number is not

A browser reaction test reads your nervous system through a screen and an input device, both of which add real, measurable milliseconds that belong to the machine and not to you. It also reads you today: sleep, caffeine, time of day and simple practice all move the number by more than most people expect, which is why the test averages five rounds and why comparing yourself with yourself, same device, same hour, is the only comparison this site will defend. Where the percentile curve comes from, and how much of it is still assumption, is declared on the Method page.