Why colour-choice personality tests do not work

What a colour test claims to do

The format is familiar. You are shown a set of coloured patches and asked to pick them in order of preference, without thinking about it, ideally quickly. The order you produce is then read back to you as a description of your character, your current emotional state, and often your unmet needs and your sources of stress.

The best-known version has a clear origin. A scoping review of colour preference research describes it in one sentence: the Swiss psychotherapist Max Lüscher developed a colour test to identify personality traits and the current emotional state of an individual through the unconscious selection of colour preferences from a set of coloured plates.

Two claims are packed into that sentence, and they are worth separating because only one of them is even slightly plausible. The first is that colour preference is stable and meaningful enough to carry information about a person. The second is that the ordering happens unconsciously, so that it reveals something you would not report if asked directly. The second claim is what makes the test feel like a magic trick, and it is also what makes it unfalsifiable by anyone taking it: if the result does not match, you are told your conscious self-image is what is wrong.

The free versions on the internet add a third claim that the original did not: that this works through a screen.

What the published record actually found

The research literature on this is not the tidy demolition people expect, and pretending otherwise would be the same sin as overselling it. The same scoping review reports that while some studies supported the use of the test for the diagnosis of psychological disorders, others argued that it is an unreliable diagnostic tool. A divided literature is a real finding and it is not the finding that decides this.

The one that decides it is methodological, and the review states it in a single sentence: the test is not standardised, and neither its CIE coordinates nor its Munsell values have been documented.

That sentence is worth unpacking because it is the whole argument. CIE coordinates and Munsell values are the two standard ways of saying exactly which colour a colour is, in numbers, so that a laboratory in one country can reproduce the stimulus used in another. Without them, nobody knows precisely which orange was shown to which participants in which study. Every result in the literature is therefore about one printing of one set of plates under one set of lights, and there is no way to check whether the next study used the same thing.

The review finds the same weakness across the wider field, noting that many studies did not account for the standards recommended in colour science, and that colours were often described only with names such as red, green and blue, without specification of hue, lightness or chroma. Its overall verdict on the evidence base is that scientific evidence remains limited and the methodological quality was moderate, and that the patterns it found should be interpreted with caution.

Now apply that to a browser. A screen shows colour differently depending on the panel, the brightness, the colour profile, the ambient light in the room and whether a night-shift filter is running. If the original stimulus was never specified in numbers, a version rendered in a tab is not a slightly degraded copy of it. It is an unrelated set of colours with the same names.

Why it feels accurate anyway

None of the above explains the experience, which is that these results often feel uncannily right. There is a well documented reason for that, and it has nothing to do with colour.

It is usually called the Barnum effect, after the showman, or the Forer effect, after the psychologist who demonstrated it. A study of personality assessment use describes it as the tendency of people to accept general personality descriptions as their own characteristics, and adds that people easily accept general and ambiguous personality interpretations and believe they accurately express their unique characteristics.

The same paper names the two mechanisms behind it. The Barnum effect works due to the subjective validation effect and the flattery effect: the first is the phenomenon in which a person feels a description is appropriate because it is personally meaningful or crucial to them, and the second is the tendency that most people are willing to believe things that make them appear more positive.

Read a typical colour-test result with those two mechanisms in hand and the machinery is visible. The statements are compound, so at least half of each one lands. They describe a tension rather than a trait, and everybody contains tensions. They are flattering in the specific way that makes flattery credible, which is to say they flatter you for a difficulty rather than for a virtue. And they are framed as having been uncovered rather than offered, which is exactly the condition under which people rate a description as more accurate.

This is not a fault of the reader. The effect is robust, it works on people who know about it, and it is the engine under most free personality content on the internet, colour or otherwise. The same paper notes in passing that the retest reliability of one of the most popular commercial type instruments did not exceed 0.6, which is a reminder that feeling accurate and being stable are not the same property.

How to test a personality claim in about a minute

The useful thing to take from this is not scepticism about colour specifically. It is a short procedure you can run on any result any site hands you.

• Reverse it. Take the sentence and write its opposite. If the opposite would also have felt true, the sentence carried no information. Most Barnum statements fail here immediately.
• Ask who it excludes. A real claim about you is a claim that some people are not like that. If you cannot name the kind of person the sentence would be wrong about, it is not describing you, it is describing people.
• Check whether anything could disconfirm it. Ask what you would have to do this week for the result to be shown wrong. If there is no such thing, the result is not a finding, whatever its confidence.
• Look for the flattered difficulty. Statements of the form you are harder on yourself than others realise are the highest-yield Barnum shape there is, because they are almost universally true and they feel like a secret.
• Take it twice, a fortnight apart, without looking at the first result. Instability you can observe yourself is more convincing than any reliability figure a page could print.

Run those five on a colour result and on a questionnaire result, and the difference shows up quickly. A questionnaire that asks about your behaviour and reports back what you said about your behaviour can survive the reversal test, because you could have answered the other way. A result derived from the order you liked some squares cannot.

Why this site does not publish a colour test

There are two reasons, and the second is the real one.

The first is rights. The original test is not free to copy. Its rights holder states that the original colours, shapes and texts are legally protected by international copyright since 1947, under a copyright notice covering 1949 to 2023, and that its own site is the only one authorised by the foundation that holds the mark. That covers both halves of what a clone would need: the stimulus and the interpretive text.

The second reason is the one that would still apply if the rights were free tomorrow. We could commission our own colour set, write our own interpretations, and ship it in a week. It would be popular. It would also be a machine for producing the effect described three sections above, on a site whose entire proposition is that it does not do that. A page cannot argue that percentiles without norms are dishonest and then hand out character readings from coloured squares.

So the honest form of this search is the page you are reading. It is on the entertainment shelf of this site because that is where colour choice belongs, and because saying so is the part the competitors leave out.

What a colour preference does tell you

Something, but not what the test claims. You prefer some colours to others, that preference is real, and it is worth roughly what a preference is worth: it tells you what to paint a room and it predicts almost nothing about how you behave under pressure.

Colour preference is also a live research area rather than a closed one, and the scoping review is careful about where it might go. It concludes that colour has the potential to be used as a complementary marker in the diagnosis and monitoring of mental health disorders, but that a necessary step in this direction is using stimuli that are colorimetrically characterised. In other words, the first thing the field has to do is the thing the famous test never did.

That is a reasonable position to hold about the subject, and it is a long way from a page that reads your character out of four squares. If you want a personality result with a published evidence base behind it, the trait questionnaires in this catalogue are the honest version, and they come with their own limits printed on the page rather than in a guide three clicks away.