See how much slower you get when you name the ink color instead of reading the word, in ms
The Stroop test asks you to look at the word "Red" printed in blue ink and answer Blue, the ink color, instead of reading the word. Reading is such a strong habit that it cuts in on its own, so when the meaning of the word and its ink color differ, answers come later and mistakes are more likely than when they match. The extra time is called the interference effect (the Stroop effect).
The default mode, "Pick the color", mixes 48 trials into congruent, neutral and incongruent trials, 16 of each, and computes the mean response time per condition from correct trials only. The result shows the interference effect, the incongruent mean minus the congruent mean (+100 ms in the example record on the first screen), its margin of error, and a chart with one dot for each trial. You can also choose the "Same / Different" mode from the older version, or a card-reading mode in which you read two cards aloud as in the original 1935 experiment.
Why try a Stroop test?
Once you can read, the meaning of a word comes to mind the moment you see it. The Stroop test lets you feel, in a few minutes, what it takes to hold that automatic reading back and answer what the task actually asks, the ink color. Since J. R. Stroop first published it in 1935, it has been studied so much that a review counted some 400 studies over more than 50 years (MacLeod 1991), and it is usually the first task mentioned when people talk about attention and inhibitory control.
It is worth trying in situations like these.
- Classes and talks: in a psychology or teaching setting you can feel the Stroop effect yourself and put your card-reading result next to the 1935 student results.
- Comparing your own condition: compare a day with little sleep to a well-rested day, or morning to late night, using the same mode and the same number of trials.
- Separating it from reaction speed: a simple reaction time test measures only how fast you press, while the interference effect shows the time it takes to filter out misleading information.
- Seeing practice effects: repeat the test under the same conditions and watch how the bars in this visit's records change. In the 1935 experiment, after 8 days of practice the time to read incongruent cards dropped by 33.9% compared with the first day.
How to take the Stroop test
- Choose a mode: "Pick the color" is the default, where you press the button for the ink color. In "Same / Different" you press Same if the word's meaning and its ink color match and Different if not. "Card reading (1935)" has you read two cards full of cells aloud while the time is measured.
- Trials and time limit: choose 24, 48 or 72 trials and a time limit per trial (None, 3 s, 2 s, 1.5 s). Running over the limit counts as a wrong trial. To compare records, use 48 trials or more and keep the conditions the same.
- Start: after a 3-2-1 countdown a + appears in the middle of the screen, and a word appears 0.4 to 0.7 seconds later. On a keyboard use D (Red), F (Blue), J (Green) and K (Yellow) or the numbers 1 to 4; on a phone use the buttons under the screen. A button press is timed at the moment your finger touches it.
- Stopping midway: pressing Esc or "Stop" ends the run without recording it. If you switch to another tab the test pauses automatically and the trial you were on is discarded; press "Resume" and it continues from that trial.
- Reading the result: the interference effect appears in large type, with bars for the mean of each condition and a chart of response times per trial below. Hover over a dot, or click the chart and use the ← → keys to step through the trials one by one.
- Full screen and records: "Full screen" shows only the task and buttons so that less distracts you. Finished runs pile up under "Records this visit", and "Copy record" lets you paste the result elsewhere.
How the result is calculated
Response time runs from the moment the word is drawn on the screen to the moment the button or key is pressed. Means use correct trials only. Responses faster than 100 ms are treated as presses made before seeing the word and are dropped, and so are values far from the median of the correct responses in the same condition. A study of the reliability of the Stroop task (Hedge et al. 2018) also dropped responses under 100 ms and values more than three times the median absolute deviation before computing its means. This tool multiplies the median absolute deviation by 1.4826, which puts it on the same scale as the standard deviation of a normal distribution, drops values beyond three times that, and takes the median and the median absolute deviation separately for each condition. Pooling the three conditions would also cut normal responses from the slower incongruent trials and make the interference effect look smaller. And so that ordinary responses are not dropped when your answers are very even, only values at least 200 ms from the median are dropped.
Each condition mean is rounded to 1 ms before the means are subtracted, so the numbers on screen always add up. The table below shows how to read the example record on the first screen (48 trials, 2 wrong).
| Result row | Example value | How to read it |
|---|---|---|
| Congruent mean (14 correct) | 633 ms | Trials such as "Red" shown in red ink |
| Neutral mean (16 correct) | 694 ms | A word unrelated to color, such as "Ship" |
| Incongruent mean (14 correct) | 733 ms | Trials such as "Red" shown in blue ink |
| Interference effect | +100 ms (+16%) | The time lost because meaning and color clash |
| Incongruent − neutral | +39 ms | The delay from a word naming a different color |
| Neutral − congruent | +61 ms | The speed-up from meaning and color matching |
| Margin of error (95%) | ±54 ms | Smaller than the interference effect, so the difference is unlikely to be chance |
In card reading, the two card times are rounded to 0.1 s and subtracted to give the interference time, and both are scaled to 100 cells so that different cell counts can be compared with the 1935 experiment. For example, if two 50-cell cards took 30.8 s and 52.4 s, the interference time is 21.6 s (+70%), and scaled to 100 cells the times are 61.6 s and 104.8 s.
What the original 1935 experiment found
Stroop ran three experiments in his 1935 paper. The key result was that the ink color barely disturbed reading words, but the meaning of the word disturbed naming the ink color a great deal.
| Experiment | What was done | Result |
|---|---|---|
| Experiment 1 (70 students) | Reading color names printed in black ink and color names printed in other colors aloud | 100 cells: +2.3 s (5.6%), not a reliable difference |
| Experiment 2 (100 students) | Naming the colors of squares, then naming the ink color of color words | 100 cells: 63.3 s → 110.3 s (+47.0 s, 74%) |
| Experiment 3 (32 students) | Practicing ink-color naming for 8 days, 200 trials a day | Half sheet (50 cells): 16.8 s (33.9%) less than the first-day average |
The original experiment used five colors (red, blue, green, brown and purple) and arranged them so that the same color never came twice in a row. Card reading here uses four colors that are easy to tell apart on a screen (red, blue, green and yellow) and scatters them at random so that the same ink color does not appear in two neighboring cells. Because the number of colors and the medium (print versus screen) differ, treat the 1935 numbers as something to compare against, not as a standard.
Comparing the three modes
| Mode | How you answer | What is measured | Good for |
|---|---|---|---|
| Pick the color | Ink-color buttons (4) or D, F, J, K | Condition means, interference effect (ms), margin of error | Comparing your own records, seeing the interference effect as a number |
| Same / Different | Same or Different buttons, F and J | Mean response time, gap between same and different trials, accuracy | A light run of the format from the older version |
| Card reading (1935) | Read aloud and press "Done reading" | Two card times, interference time (s), scaled to 100 cells | Class demonstrations, setting your result next to the 1935 results |
The number of trials per condition and the time needed depend on the trial count, roughly as below. The estimate assumes 0.4 to 0.7 seconds of fixation for each trial, the time you take to answer, and 0.26 seconds of feedback.
| Trials | Pick the color (per condition) | Same / Different (per condition) | Estimated time |
|---|---|---|---|
| 24 trials | 8 trials | 12 trials | about 40 s |
| 48 trials | 16 trials | 24 trials | about 1 min 15 s |
| 72 trials | 24 trials | 36 trials | about 1 min 50 s |
Common mistakes
- Reading the word silently: if you try to take in the whole word, its meaning comes in first. Resting your eyes on one spot of a letter's stroke makes you less likely to be swept along.
- Rushing and making more errors: the interference effect is computed from correct trials only, so with many wrong trials fewer trials enter the mean and the value becomes unstable. If accuracy is below 90%, try again a little more slowly.
- Treating the first run as the baseline: in the first few trials you are still finding the buttons, so you tend to be slow. Treat the first run as practice and start comparing from the second record.
- Comparing records made under different conditions: a different device, screen size, keyboard versus touch, trial count or time limit changes response times. The bars in this visit's records are also scaled only among records of the same mode.
- Reading the Same / Different result as an interference effect: "Different − same trials" in that mode is the difference in a task that compares two pieces of information, so it is not the same value as the interference effect of "Pick the color".
Limits and cautions
- This tool is for hands-on experience and comparing your own records. It is not a medical or psychological test or a diagnosis. If you are worried about attention or cognitive function, see a professional.
- Response times include the delay before the screen is drawn and the input delay of the keyboard or touch screen. These can differ by tens of ms between devices, so it is better not to compare records measured on different devices.
- The interference effect varies quite a lot even within one person. In a study with 240 trials per condition measured twice, three weeks apart, the standard error of measurement of the interference effect was still 21 to 24 ms (Hedge et al. 2018). The margin of error in this tool shows only the variation within the current run.
- If you have trouble telling colors apart, the result is not accurate. In the 1935 experiment too, participants took the Ishihara color vision test, and the results of participants with red-green difficulty were left out. If you are worried about your color vision, take a color blindness test first.
- Records stay on this screen only and are not saved to a server or to browser storage. The address contains only settings such as mode, trial count, time limit and cell count.
Reference date The research results in this article were checked on 2026-10-16 against the full text of Stroop (1935), the full text of Hedge et al. (2018) and the abstract of MacLeod (1991).
Sources
- 01Stroop JR. Studies of interference in serial verbal reactions. Journal of Experimental Psychology 1935;18(6):643–662Experiment 1 (70 students, +2.3 s, 5.6%), experiment 2 (100 students, 63.3 → 110.3 s, 74%), experiment 3 (32 students, 8 days of practice −16.8 s, 33.9%), five colors, Ishihara color vision test
- 02Hedge C, Powell G, Sumner P. The reliability paradox. Behavior Research Methods 2018;50(3):1166–1186Four-color key press, 240 congruent, neutral and incongruent trials each, two sessions three weeks apart, exclusion under 100 ms and beyond 3 median absolute deviations, interference effect ICC .60 and .66, standard error of measurement 21 and 24 ms
- 03MacLeod CM. Half a century of research on the Stroop effect: An integrative review. Psychological Bulletin 1991;109(2):163–203Reviews some 400 studies over more than 50 years and identifies 18 reliable empirical findings (abstract)
Frequently asked questions
What does the Stroop test measure?
It measures how much longer you take to name the ink color when the meaning of the word differs from it. This tool computes the mean response time for trials where word and color match (congruent) and trials where they differ (incongruent) separately, and shows the incongruent mean minus the congruent mean as the interference effect. The larger the value, the more your automatic habit of reading cut in on the color judgment.
How many ms of interference effect is normal?
This tool uses no cutoff between normal and abnormal. The interference effect depends on the device, the screen, the button layout, the number of trials and how much you have practiced, and even in a study with 240 trials per condition measured twice three weeks apart, the standard error of measurement of the interference effect was 21 to 24 ms (Hedge et al. 2018). Repeat the test on the same device in the same mode and compare your own records with each other.
Why are neutral trials mixed in?
They are trials where a word unrelated to color, such as Ship or Advice, is shown in an ink color. Incongruent minus neutral shows the delay caused by a word naming another color, and neutral minus congruent shows the speed-up from meaning and color matching. These are the rows "Incongruent − neutral" and "Neutral − congruent" in the result.
How many trials should I use?
The first time, use 24 trials to get used to the mode; to compare records, 48 trials or more are recommended. With few trials, a couple of late presses can swing the mean a lot. If the margin of error in the result is larger than the interference effect, a note says the difference may be chance, and then you should try again with 72 trials.
How are wrong, too fast and too slow responses handled?
Means use correct trials only. Responses faster than 100 ms are treated as presses made before seeing the word and are dropped, and so are values far from the median of the correct responses in the same condition (more than 3 times the median absolute deviation × 1.4826, and at least 200 ms away). The number excluded is shown in the row "Responses excluded from the mean" and as faded dots in the chart.
How is card reading different from Pick the color?
Card reading follows Stroop's 1935 experiment: you read the cells aloud in order. On card 1 you name the colors of colored squares, on card 2 the ink color of color words, and you press the button yourself when you finish to time it. In the 1935 experiment, 100 cells took 63.3 s on average and then 110.3 s, 74% longer, and this tool scales your record to 100 cells and shows it next to that.
Can I compare Same / Different results with Pick the color results?
It is better not to compare them directly. In Same / Different you check both the meaning and the ink color and then judge whether they match, which is a different task from picking only the ink color. When comparing records, use the same mode, the same number of trials and the same time limit.
Are my records saved?
No. Records from this visit stay on this screen and disappear when you reload, and nothing is kept on a server or in browser storage. To keep a result, use "Copy record" and paste it somewhere. The "Same-settings link" contains only settings such as mode and trial count, so another person can start under the same conditions.