Stroop test: method

What it measures

Interference control / selective attention: the slowing caused by a conflicting colour word (Stroop interference effect), with accuracy.

Procedure

Four colours: red, green, blue, yellow. Each trial shows one colour word printed in one of the four inks; the player answers the ink colour with one of four buttons, the keys 1 to 4, or (where the language allows) the first letter of the colour word, for example R, G, B, Y in English. 4 practice trials with feedback (700 ms); new players get practice by default, everyone can skip it. Then 40 counted trials, exactly half congruent (word = ink) and half incongruent, in a shuffled order, without feedback. Response window 3,000 ms; no answer counts as an error. 500 ms between trials. Leaving the browser tab ends the run. About 2 minutes.

Parameters

Scoring

Accuracy = correct answers / all counted trials (timeouts are errors). Mean reaction time of correct congruent and of correct incongruent trials (arithmetic means in ms, no trimming of slow or fast answers). Primary: interference ('mismatch cost') = mean correct incongruent RT minus mean correct congruent RT, in ms; lower = less interference. It is reported only when accuracy is at least 70%; otherwise accuracy becomes the main number. A negative value is shown as 0 on the result screen (the stored value keeps its sign). No player comparison is shown.

Valid-run rules

The page sends every counted trial (word, ink, answer, reaction time). Colours outside the set are rejected; at most 200 trials per log. The scorer itself has no lower reaction-time bound and no outlier trimming; very fast answers are only caught by the automation checks. The server rebuilds every run from the raw log (and the run's seed where there is one) and computes the score itself; anything the page says about right or wrong is ignored. A log that breaks the game's rules is rejected, and nothing is stored. Stored runs can carry quality flags that are never shown and never block a result: 10 or more runs from one IP address, a missing visitor id, tab switches, and automation signals (automated browser, impossible speed, no input events, machine-regular timing).

Differences from the classic task

Stroop (1935) timed people reading whole printed cards aloud. This version shows one word at a time and uses button or key presses, like most computer versions, with only 40 trials. There is no neutral condition (colour patches or non-colour words), so the score is incongruent minus congruent, which includes any speed-up on matching words, not incongruent minus neutral.

Limitations

Colour vision deficiency, screen colour settings, night mode and dark-mode filters change the task. Input latency differs between devices. With 20 trials per condition the interference score is noisy, and difference scores like this have low test–retest reliability even in the lab (Hedge et al. 2018). Practice lowers the effect. Reading fluency in the page language matters. Not a clinical test.

Languages and review status

Colour words in the page language for English and 99 other languages (native colour words; scoring uses stable colour codes, so it is the same in every language). Letter keys where a set of letters works, otherwise the keys 1 to 4 only.