How Choice Reaction Scoring Works, and Why Accuracy Is Part of It
A fast wrong answer is not a good result. Choice reaction time is the clearest example on this site of a score that is meaningless on its own.
The mode presents one of four colours and you respond with the matching key. It runs 20 trials and reports accuracy alongside timing — because timing a set of guesses would produce an excellent and completely worthless score.
Why accuracy belongs in the score
There is an obvious exploit if you ignore accuracy: press all four keys as fast as possible as each colour appears. You will register many very fast presses and a high average, and be completely wrong.
This mode prevents that structurally rather than by asking nicely. Only correct trials contribute to the reported average — a wrong press does not record a fast time, it records an error. Guessing therefore produces an accurate reflection of your indecision, which is exactly what a choice reaction test is meant to show.
The scoring rule
accuracy = round(correct / 20 × 100)
reported = mean(correct trial times only)If every trial is wrong, accuracy is 0% and no time is recorded at all — the result screen shows the miss rather than a flattering average. That is the intended behaviour: a run of guesses did not measure choice reaction time.
What a gap between correct and incorrect looks like
Consider two groups of correct trials — deliberately quick ones and deliberately hesitant ones. The values below are illustrative rather than recorded, but the arithmetic is the mode's own.
| Set | Times | Mean |
|---|---|---|
| Decisive | 412, 398, 425, 401, 409 ms | 409 ms |
| Hesitant | 521, 498, 534, 512, 527 ms | 518 ms |
| Difference | — | 109 ms |
109 ms of separation from decisiveness alone — comparable to the entire effect people are usually looking for. It is also why the averaging is restricted to correct trials: mixing in the two seconds a hesitant person spends on wrong answers would flatten that into a single uninformative number.
Why choice is slower than simple reaction
Responding to one signal is a perception-and-response task. Identifying which of four colours appeared adds a selection step, and that step grows with the number of options.
Hick's 1952 finding — cited on our sources page — describes that growth as proportional to the logarithm of the number of choices, which means each additional option adds progressively less:
| Options | Added decision time | Shape |
|---|---|---|
| 1 | +0 ms | |
| 2 | +90 ms | |
| 3 | +139 ms | |
| 4 | +180 ms | |
| 5 | +214 ms |
Illustrative values showing the logarithmic shape, not measurements from this site.
This is precisely why the mode fixes four colours rather than escalating. Going to six options would add roughly 58 ms of decision time and increase errors — making a longer, fussier test without making a more informative one.
Reading your own result
- Accuracy first. Below about 85%, the timing is not trustworthy — the average is only over correct trials, so a low accuracy means a small sample.
- Then the time, against its own band. Choice reaction scores are not on the simple-reaction scale. This mode has its own reference band.
- Then consistency. Unlike simple reaction, choice timing varies a lot with momentary attention, so a wide spread is common and expected.
- Not comparable across modes. Do not read a choice reaction score against a visual reaction scale, or against an audio one.
Try it with accuracy in mind
Twenty trials, four colours, accuracy reported alongside your average. Rest all four fingers on the keys to cut travel time.
Trial count, accuracy formula and the correct-trials-only averaging are this site's own, frompublic/test-scripts/choice-reaction-test.js. The Hick’s-law table is an illustrative log₂ fit, not data from this site. Hick (1952) is listed on the sources page.