The Stroop effect, ninety years on

Show someone the word RED printed in blue ink and ask them to say “blue”. It takes measurably longer than if the word had been BLUE, and the delay is large enough to feel as a moment of resistance.

That is the Stroop effect, and it has survived ninety years of attempts to explain it away.

Where it comes from

John Ridley Stroop described it in 1935, in a paper that remains among the most cited in experimental psychology. His design was straightforward: compare the time to name colours on a card of coloured patches against the time to name the ink colours of conflicting colour words.

The result has since been reproduced in essentially every population tested, in every language with colour terms, and with variations substituting numbers, directions and emotional words for colours. It is about as robust as findings in this field get.

Why it happens

The standard explanation is automaticity. Reading, for a literate adult, is not optional. You cannot look at a familiar word in a familiar script and decline to process its meaning — the meaning arrives whether you want it or not.

So on a conflicting trial two responses are activated at once: the automatic one from reading, and the deliberate one from naming the ink. The automatic one is faster. It has to be suppressed before you can answer, and suppression takes time.

Two pieces of supporting evidence make this convincing:

• The effect is asymmetric. Naming the ink is slowed by the word, but reading the word is barely slowed by the ink. The interference runs one way, from the automatic process to the deliberate one.
• It develops with reading. Young children who are learning to read show a much smaller effect. It grows as reading becomes fluent, which is the opposite of what you would expect if it were a general limitation of attention.

What the interference cost actually measures

Your raw response time on this task is mostly about how fast you press buttons, what your mouse is like, and how awake you are. None of that is the Stroop effect.

So the number worth reporting is the difference: your average on conflicting trials minus your average on matching ones. Both halves carry the same button-pressing overhead, so subtracting cancels it and leaves the interference itself.

Our Stroop test splits its twenty trials exactly half and half rather than randomly, so both averages are computed over the same number of trials and the difference is comparable between runs.

Can you beat it?

You can reduce it, and how people do so is revealing.

The techniques that work all involve not reading: defocusing your eyes slightly so the word is not legible, fixating on a single letter rather than the whole word, or looking slightly to one side of it. Some people report imagining the word as a shape.

What does not work is trying harder. The interference happens before deliberate control gets involved, which is exactly why it remains interesting — effort is applied at the wrong stage.

A very low score therefore usually means you have found an avoidance strategy rather than that you are immune. That is not cheating so much as a different task.

Where it is genuinely used

Stroop tasks appear as one measure among many in research and clinical work on executive function, attention and inhibitory control. They show up in studies of ageing, ADHD, schizophrenia and traumatic brain injury, and in “emotional Stroop” variants substituting threatening words for colour words.

Two things separate that use from a web page. Those settings use standardised administration — same instructions, same lighting, same equipment — and normative data collected on comparable populations. A twenty-trial browser test has neither.

So a high interference cost on a Tuesday afternoon means you were tired or rushing. It is not evidence of a condition, and it should not be treated as a screening result. If your attention genuinely worries you, that is a conversation for a doctor.

The variants, and what they were for

The basic finding turned out to generalise, and the variations are half of why it is still studied.

Numerical Stroop. Show 3 3 3 and 5 5, and ask how many digits there are. People are slower when the digit’s value conflicts with the count, which suggests number magnitude is processed as automatically as reading.

Directional Stroop. The word UP printed at the bottom of the screen, and you must name the position. Same interference, no reading of colour words involved.

Emotional Stroop. Colour-naming of threatening or personally significant words rather than colour words. Slower responses to those words have been used to study anxiety and addiction, though the interpretation is more contested than the classic version.

What the variants share is the structure: an automatic process producing an answer you did not ask for, competing with a deliberate one you did. That structure, rather than colour specifically, is the finding.

Reading the word is not optional, and that is the point

It is worth sitting with how strange the finding is. You are looking at a shape. Nobody asked you to extract meaning from it. And you cannot stop yourself.

That automaticity is not innate — it is built by years of practice, which is why it grows as children learn to read and why it is absent in a script you cannot read. Show the same task in an alphabet you have never learned and the interference disappears entirely.

So the effect is a measure of expertise rather than of weakness. The reason you are slow at naming the ink is that you are very good at reading, and the skill runs whether or not it is wanted.

Why it survives practice

Most psychological effects shrink substantially when people repeat the task. This one barely does, which is unusual enough to be part of why it stayed interesting.

Practice on the task helps — you get faster at pressing the right button, and your raw times fall. But the difference between congruent and incongruent trials, which is the effect itself, is remarkably persistent. You get faster at both halves rather than closing the gap.

The exception is people who develop an explicit avoidance strategy, and it proves the rule: they reduce the effect by not reading, rather than by reading and resisting.

What the effect is not evidence of

It is not evidence that you are easily distracted, and it is not a measure of willpower. Everyone shows it, including people with exceptional attention, and the size of your effect on a given day is dominated by how tired and how careful you are.

It is also not stable enough from a twenty-trial run to compare against anyone else. The useful comparison, as with every test here, is you against your own history on the same equipment — which is why the number is stored and charted rather than presented as a verdict.

The colour-blindness problem

Any colour-naming task excludes people who cannot distinguish the colours, and that is a substantial group — red-green deficiency affects roughly one in twelve men.

Our version deliberately avoids pairing red with green, uses red, blue, yellow and purple, and names every colour in text on the answer buttons so the controls are readable regardless. That helps with the interface. It does not solve the underlying problem: the task itself requires separating ink colours, and if you cannot do that reliably the score will not mean anything.

Saying so plainly is better than handing someone a number that does not measure what they think it does — which is the same reason the reaction time test reports how much of your result is your hardware.