← All guides

What choosing costs: Hick's law and the choice reaction test

A simple reaction is a reflex with a clock on it. Add a decision and the time stops being constant, and starts obeying a rule.

Updated October 30, 2026

Press a key when the screen changes color: that is one stimulus and one response, and there is nothing to decide. Now let the screen turn one of four colors, each with its own key. The eyes and the finger are the same, but something new sits between them: the brain has to work out which color it saw and which key that color owns. The time that step takes is not a rounding error. It is the subject of one of the oldest tidy results in experimental psychology.

What Hick found

In 1952 William Hick measured how long people took to press the key that matched a signal, as he changed how many signals were possible. The time did not climb in proportion to the number of options, which would make ten options ten times as slow as one. It climbed with the logarithm of the number: each doubling of the alternatives added about the same extra slice. Hick framed this in the language of information theory, the same move Fitts made two years later for aiming (the Fitts guide tells that story). A choice among two equal options is one bit, among four it is two bits, among eight it is three, and reaction time rose roughly in a straight line with the bits.

The formula people quote has a small twist. It takes the log of the number of options plus one, because the person is also uncertain about whether a signal will come at all. For the four-key test that detail does not change what you see on screen, only that the law is about uncertainty, not about counting keys.

It is not really about the number of keys

Ray Hyman tested that reading in 1953. Instead of only changing how many signals there were, he changed how predictable they were: some more likely than others, or following each other in patterns. Reaction time tracked the information in the signal, in bits, and not the raw count. A signal you can see coming is cheaper to answer than one you cannot, even if the set of options is the same size.

That is why the law is better stated as “time grows with how much you were unsure of”. The choice reaction test draws each color at random, so there is no pattern to lean on and the uncertainty is the plain four-way kind. In a real task, a pattern you can anticipate would make you faster, and a test that let you learn one would be measuring something else.

The mapping matters too

In 1953 Fitts and Seeger turned to what links the two. They paired three layouts of signals with three layouts of responses, every one with every other. Each layout of signals was answered fastest, and with the fewest errors, by the response layout that matched its own, and more slowly by the ones that did not. They called it stimulus-response compatibility. Later work kept the signals and the responses fixed and changed only which goes with which, and came to the same lesson. A light on the left that asks for the left key is cheap. A color that asks for the third key from the left is not, until you have learned it.

This is the part of the test that practice reaches. The page labels each answer with its color’s name so that hue is never the only cue, but you still have to learn which key is which. A first run carries that learning; a tenth run mostly does not.

What the lag on this site is made of

The choice reaction test plays two rounds. The first is eight flashes answered with any key: your plain reaction on this screen with this hand. The second is sixteen flashes in four colors on D, F, J and K, or four buttons. The score is the four-key average, and the decision lag under it is round two minus round one. Whatever your screen, browser and input device add rides both rounds equally, so it cancels. What is left is the cost of identifying the color, picking the response and starting it: the choosing.

Three honest limits on reading it. The lag is a difference of two averages from a handful of taps, so a single run’s lag is noisy and can even come out near zero; the card says so rather than hiding it. Four options is a single point on Hick’s line, so the test cannot show you a slope, only the height of one point. And the page’s own typical figure for the choosing, about 200 ms, comes from the provisional curve, which is a derivation until enough runs have been measured. The Method page declares it.

Using this

Sources

  • Hick, W. E. (1952). "On the rate of gain of information". Quarterly Journal of Experimental Psychology, 4(1), 11–26.
  • Hyman, R. (1953). "Stimulus information as a determinant of reaction time". Journal of Experimental Psychology, 45(3), 188–196.
  • Fitts, P. M., & Seeger, C. M. (1953). "S-R compatibility: spatial characteristics of stimulus and response codes". Journal of Experimental Psychology, 46(3), 199–210.

Keep reading

  • What a reaction time is made ofA quarter of a second is not hesitation. It is the speed limit of the wetware, and every millisecond of it is spoken for.
  • Why stopping is harder than goingReading cannot be switched off and a loaded finger cannot be recalled for free. The attention tests measure the price of overruling yourself.