A Fifth of a Second: The Real Limits of Human Reaction Time

The square turns green, you click, and the number says 250. A quarter of a second. It does not feel like a quarter of a second — it feels instantaneous, like the click and the colour happened together. They did not. By the time you were aware the square had changed, it had already been green for roughly a tenth of a second, and your hand had not started moving yet.

That gap is the interesting thing about reaction time. It is not a measure of how fast you think. It is mostly a measure of how long it takes a signal to physically get from the front of your eye to the muscles in your finger, and the honest summary is that there is very little you can do about it.

The typical numbers are worth knowing before you judge your own. For a simple visual reaction — one signal, one response, no decision to make — healthy adults land somewhere around 200 to 250 milliseconds. Sound is consistently faster than sight, at roughly 140 to 160 milliseconds, which surprises people until you consider the machinery. A sound is a pressure wave that mechanically wiggles hair cells in your inner ear, and mechanical things are fast. Light has to be converted into a nerve signal by a chemical cascade in your photoreceptors, and chemistry takes time.

So where does a fifth of a second actually go? Something like 20 to 40 milliseconds disappears in the retina before the signal has left your eye at all. The trip to the visual cortex and the recognition that something changed takes a few tens of milliseconds more. Then a motor command has to travel back down through the spinal cord and along the nerves to your arm, which happens at somewhere between 50 and 120 metres per second — fast, but not instant over the length of a human being. Finally the muscle itself has to convert an electrical signal into actual movement, and that electromechanical delay alone is another 30 to 50 milliseconds. Add it up and the deliberate, conscious part of the process is a small minority of the total. The rest is plumbing.

This is why the rules of sprinting look so strange from the outside. World Athletics treats any response registered less than 0.100 seconds after the starting gun as a false start, even if the runner was still in the blocks. The reasoning is not that the athlete cheated in the ordinary sense, but that a hundred milliseconds is below what a human can plausibly do as a genuine response to a sound — so a faster reading means the runner was already going, and happened to guess well. Some researchers argue the threshold is set slightly too high and that a rare athlete can beat it honestly, which tells you how tight the physiological ceiling is even when it is disputed.

Anticipation is the loophole in every reaction test, and it is the reason this one waits an unpredictable amount of time. The delay before the square turns green is randomised between two and five seconds. If it were a fixed three seconds, you would stop reacting after the first round and start counting, and by the fourth round you would be producing scores that no longer measure anything about you. The randomness is not there to be annoying. It is what makes the number mean something.

Which brings up the honest caveat about the version on CerdikCerdik. If you click during the red phase it tells you off and lets you retry, and that round simply is not recorded — there is no time penalty and no score for it. A lab would log that as a false start and hold it against you. Here you can jump the gun as many times as you like at no cost, so the score is only as honest as you are. If you want it to mean something, treat an early click as a failed round rather than a free rewind.

There is a second kind of reaction time that behaves quite differently, and confusing the two is why people misread their own results. A simple reaction has one stimulus and one response. A choice reaction makes you pick. The moment there is more than one thing that might happen and more than one thing you might do, an extra step appears, and it is not a small one. Reaction time grows with roughly the logarithm of the number of options — the relationship is known as Hick’s law — which means going from one choice to two costs you far more than going from eight to nine. This is why Whack-a-Mole feels slower than the green square even though the movement is the same, and why Color Match is slower still: there you are not just choosing, you are choosing while suppressing something. We went through that particular cost at cerdikcerdik.com/blog/stroop-effect-colour-words.

The other thing sitting inside your score is your hardware, and it is not a rounding error. A touchscreen samples your finger at a fixed rate rather than continuously. The display only redraws itself so often, and on a 60Hz screen the green can be ready to show for up to about 17 milliseconds before you are given the chance to see it. The browser then has to schedule the event. None of this is anybody’s fault, but it means a web-based reaction test typically reads slower than laboratory equipment by a noticeable margin. Your number is you plus your phone. Compare it to your own score on the same device, and treat cross-device leaderboards with a healthy amount of suspicion.

What genuinely moves the number is less flattering than most people hope. Reaction time is quickest in the early-to-mid twenties and declines slowly and steadily after that, in everyone, forever. Sleep matters a lot, fatigue matters a lot, and caffeine helps modestly. Practice helps a little and then stops helping almost immediately, because unlike mental arithmetic or typing there is no technique to acquire — you cannot learn to make your optic nerve shorter. If you improve substantially over a week, what improved was your understanding of the task, not your nervous system.

This is why the score here is the average of five rounds rather than your best of five. Your fastest round is largely luck: it is the one where your attention happened to be exactly where it needed to be at the moment the colour changed. The average is harder to fake, and the spread between your rounds is arguably the most informative thing on the results screen. Consistency is a real property of an attentive brain. Someone posting 230, 240, 235, 250, 240 is in a genuinely different state from someone posting 190, 400, 210, 380, 215, even though the two averages are close. Those long rounds are lapses of attention, and they are what actually gets people into trouble in the situations this skill is supposed to model.

So the useful way to play is not to chase a personal best. Take a baseline when you are rested and unhurried, and remember roughly what it was. Then run it again at eleven at night after a long day, or the morning after a bad night’s sleep, and look at both the average and the worst round. The absolute numbers were never the point — the difference between rested you and tired you is the finding, and it is usually larger than people expect.

It takes under a minute. Try it at cerdikcerdik.com/games/reaction-time, and be strict with yourself about the early clicks. When you want the version with a decision bolted on top, cerdikcerdik.com/games/whack-a-mole is the same reflex plus a target that moves, and the gap between your scores on the two is roughly what choosing costs you.