How it works

In 1935 John Ridley Stroop showed people cards of colour words and timed them. Reading the words was quick. Naming the ink the words were printed in was slower - and it got much slower when the word said one colour and the ink was another.

The usual explanation is that reading is automatic for a literate adult. You cannot choose not to read a word you are looking at. The word arrives whether you want it or not, and if it disagrees with the colour you have to actively hold it back before you can answer.

The interesting part is that this is one-way. A conflicting word badly slows you down when you are naming a colour. A conflicting colour barely touches you when you are only reading. This test measures both directions so the gap between them is yours, in milliseconds.

Nothing on screen says what to do. The panel behind the word does:

• Black panel — read the word and ignore the colour it is printed in.
• Grey panel — name the colour it is printed in and ignore what the word says.

A small reminder of the mapping stays on screen throughout, so nobody is being tested on their memory. What is being tested is whether you can read the panel, pick the rule, and apply it - before the word answers for you.

1. Black panel - read the word — The colour is trying to distract you. (14 trials)
   On a BLACK panel the rule is: read the word. The word is always printed in some other colour. Tap what the word SAYS and ignore the colour it is printed in.
2. Grey panel - name the colour — Now the word is trying to distract you. (14 trials)
   On a GREY panel the rule flips: name the colour the word is PRINTED IN, and ignore what the word says. Same stimuli as before, opposite job. Most people find this one markedly harder.
3. Panel changes - follow the rule — Both rules, and only the panel tells you which. (20 trials)
   The panel now changes from trial to trial, and it is the only thing telling you which rule to use. BLACK means read the word. GREY means name the colour it is printed in. Nothing else on screen will remind you which trial is which.

The word and the ink never agree, on any trial. That matters: it means the two rules always want different answers, so there is no trial you could answer correctly while following the wrong rule. Every single response tells us something.

• Cost of the colour rule - your median time on the grey panel minus your median time on the black panel. This is the headline number: what it costs you to go past the word and answer on the colour instead.
• Switch cost - in the last block, trials where the panel changed the rule against trials that kept it.
• Mixing cost - same-rule trials in the last block against the single-rule blocks. What it costs you merely to keep a second rule available.
• Rule slips - mistakes where you gave the other panel's answer. You read the stimulus correctly and applied the wrong instruction, which is a different kind of error from a random miss. Because the word and the ink never agree, every one of these is unambiguous.
• Inverse efficiency - mean time divided by accuracy, which catches people who look fast only because they are guessing.

The paid report draws each of your figures against the spread of results published for comparable keyboard Stroop and task-switching studies, and marks where you fall on it. The main anchor is a 2024 reliability study of a keyboard Stroop task, which reported incongruent colour naming at 796 ms (SD 124) with 93.1% accuracy, and a Stroop effect of 118 ms (SD 50.9). Switch costs in cued task-switching sit around 100 ms with wide individual spread.

Every curve in the report names its source, and says plainly when a figure is inferred from published numbers rather than lifted straight out of a paper. These are reference points, not clinical norms.

The score is out of 100, from four parts:

• Control under conflict, 40 points. Full marks for no interference, sliding to zero at 400 ms.
• Switching flexibility, 25 points. Full marks for no switch cost, sliding to zero at 500 ms.
• Accuracy, 25 points. Zero at 50% correct, full marks at 100%.
• Consistency, 10 points. From how spread out your reaction times are relative to their average.

Timing only counts on trials you got right. Answers faster than 150 ms are treated as anticipations and answers slower than 5 seconds as lapses; both are dropped from the timing figures but still count against accuracy.

Percentiles compare you against this site's own reference distribution for button-press tasks, seeded into the database. They are a rough guide, not clinical norms.

There is no account. A random id in a first-party cookie is what links this browser to its runs and its purchases - clear your cookies and the link is gone. An email address is only collected if you buy a report, and only to send it to you.

This is a for-interest test, not an assessment. Tiredness, screen, mouse, phone versus laptop and simple practice all move these numbers by more than most of what the test is trying to measure. Treat a single run as a conversation starter, not a finding.