Stroop Test: Measure Your Stroop Effect Interference in Milliseconds

How the score is calculated

The task opens with four practice trials that tell you whether each answer was right or wrong. Those four do not count toward anything; they exist so that learning which key belongs to which color happens before scoring starts. After the practice block, 40 scored trials run in random order: 20 congruent, where the word and the ink agree, and 20 incongruent, where they disagree. Each trial begins with a brief fixation cross so your eyes are in the right place when the word appears. You press R, G, B or Y, or tap one of four labelled buttons, for the color of the ink. Scored trials give no right or wrong feedback, which keeps you from adjusting your speed in response to how you think you are doing.

Three rules govern what counts. A response faster than 150 milliseconds is treated as an anticipation rather than a real response: it is not scored, and that trial is put back in the queue to be shown again. A response slower than 3,000 milliseconds still counts toward your accuracy figure but is left out of the medians, because one long pause would otherwise distort the comparison. Only correct trials feed the medians at all, since the time taken to give a wrong answer is not a measure of anything useful here. Leaving the browser tab cancels whichever trial is in flight.

Your headline number, the Stroop interference, is the median of your correct incongruent trials minus the median of your correct congruent trials. Medians are used rather than averages because a single distracted trial moves an average far more than it moves a middle value. The result page also shows both medians separately and your overall accuracy, so you can see whether a large or small difference came from fast matched trials, slow conflict trials, or a lot of errors in one condition.

What the Stroop effect actually is

In 1935, J. R. Stroop published a paper in the Journal of Experimental Psychology called Studies of interference in serial verbal reactions. He showed that people take noticeably longer to name the ink colors of color words when the word and the ink disagree than when they name the same colors printed as plain blocks or squares. The finding was neither obvious nor trivial at the time, and it has been reproduced so many times since that the effect carries his name.

The usual explanation is about automaticity. For a literate adult, reading is close to involuntary: the meaning of a familiar short word arrives whether or not you asked for it. Naming the color of ink is not automatic in the same way. It is a deliberate act that has to be assembled each time. When the automatic response and the deliberate one point at different answers, the conflict has to be settled before you can respond, and settling it costs time. That cost, measured in milliseconds, is the Stroop effect.

MacLeod's 1991 review in Psychological Bulletin pulled together roughly four hundred studies on the effect and remains the standard summary of what is and is not known about it. The size of the interference varies a great deal with how the task is built: whether words are presented one at a time or as a full page to be read aloud, whether the answer is spoken or given with a key press, how many colors are in play, and how many trials there are. This is one reason a number produced by one version of the task cannot be compared to a number produced by another.

Why this page is not a clinical Stroop test

Clinicians do use Stroop tasks, and those versions are not this one. The Stroop Color and Word Test published by Charles Golden in 1978 is administered on paper by a trained examiner, timed by hand, and scored against published norms that take age and education into account. The Color-Word Interference Test in the Delis-Kaplan Executive Function System is a four-condition task, also clinician-administered, with its own standardized materials and scaled scores. Both are copyrighted instruments. This page reproduces neither of them, and no score from this page maps onto either one.

The differences run deeper than the materials. A clinical administration happens in a controlled room with a person watching how you respond, not only how fast: whether you self-correct, whether you lose your place, whether you understood the instruction, whether you appear tired or unwell. That observation is part of the assessment. A result is then read alongside the rest of a battery, your history, and the reason you were referred. None of that exists here. A browser running on unknown hardware, with nobody present, produces a number and nothing else.

Scarpina and Tagini's 2017 review in Frontiers in Psychology is a good illustration of how unsettled even the scoring is among researchers. They compare several competing ways of calculating the Stroop effect and argue that the different methods are not interchangeable. If specialists still debate which arithmetic best captures interference under controlled conditions, a single unsupervised browser run should be read as an observation about one attempt, not as a measurement of you.

What can move your interference number

Your device is baked into every trial. Reimers and Stewart measured presentation and response timing in web experiments across nineteen computer systems and found real differences in how accurately a browser records when something appeared and when a key was pressed. Taking a difference between two medians helps here, because a delay that applies equally to congruent and incongruent trials largely cancels out, but it does not cancel perfectly, and a touch screen adds more variability than a keyboard does.

Strategy moves the number too, in both directions. Some people learn to defocus their eyes slightly or fix on one letter so the word registers less strongly, which shrinks interference without telling you anything about attentional control. Others read every word deliberately, which enlarges it. Fatigue, caffeine, distraction, and simply how many trials you have already done all shift response times, and the 40 trials here are few enough that random variation alone can move a difference score by tens of milliseconds between two honest attempts.

The task needs color vision. If you have any form of color vision deficiency, particularly red-green, the ink is harder to identify and the number this task produces will reflect that difficulty rather than interference. The same applies if the screen brightness is low or the room is dim. None of that is a fault in you, and none of it can be corrected for after the fact. If you want a broader picture of attention and speed, the Reaction Time Test, the Processing Speed Test, and the Attention Span Test each look at a different slice.

Methodology & sources

This is an original timed task built for Psychology.com, not a reproduction of any published clinical instrument. It uses the single-trial color and word conflict design that has been standard in experimental psychology since researchers moved the task from Stroop's original card-reading format onto computers: one color word at a time, printed in one of four inks, answered by pressing one of four keys. Four practice trials precede 40 scored trials, split evenly between congruent and incongruent and presented in random order. The headline figure is the difference between the median of correct incongruent trials and the median of correct congruent trials, one of the several scoring methods used in the literature.

The published research offers no cutoff we could honestly apply to a page like this. Computerized versions of the task that show one word at a time commonly report interference somewhere in the region of fifty to a hundred and fifty milliseconds, but that figure moves substantially with the number of colors, the response method, the proportion of conflict trials, and the participants, as MacLeod's review documents at length. Our five result bands are therefore educational reference ranges written to describe what a number of that size plausibly means on a browser task. They are not clinical cutoffs, they are not normed for age or education, and they should not be compared against a score from a clinician-administered Stroop.

Timing on the web is imperfect and we do not pretend otherwise. Reimers and Stewart showed that browsers and operating systems differ in how accurately they record stimulus onset and response times, and a touch screen adds further variability that a keyboard does not. Subtracting one median from another removes much of the constant portion of that delay, since it applies to both conditions, but it cannot remove the variable portion, and this task has no way to measure your particular setup. A result below zero is possible and is displayed as measured rather than adjusted. Treat any single run as one observation under unknown conditions, not as a stable characteristic.

Reviewed by Michael Callans, MSW
Reviewed to keep the separation clear between this browser interference task and the clinician-administered Stroop tests it is often confused with. Last reviewed September 23, 2026.

References

1. Stroop, J. R. (1935). Studies of interference in serial verbal reactions. Journal of Experimental Psychology, 18(6), 643 to 662. doi.org/10.1037/h0054651
2. MacLeod, C. M. (1991). Half a century of research on the Stroop effect: An integrative review. Psychological Bulletin, 109(2), 163 to 203. doi.org/10.1037/0033-2909.109.2.163
3. Scarpina, F., and Tagini, S. (2017). The Stroop Color and Word Test. Frontiers in Psychology, 8, 557. doi.org/10.3389/fpsyg.2017.00557
4. Reimers, S., and Stewart, N. (2015). Presentation and response timing accuracy in Adobe Flash and HTML5/JavaScript Web experiments. Behavior Research Methods, 47(2), 309 to 327. doi.org/10.3758/s13428-014-0471-1

Stroop Test FAQ