Raven's Progressive Matrices: The Test That Was Never Meant to Be an IQ Test

What the test is

Every item is the same. A matrix of figures is displayed, arranged so that the figures change according to some rule across the rows and some rule down the columns. One cell is blank. Below sit six or eight candidate pieces. Exactly one of them satisfies both rules.

Nothing is written. No arithmetic is needed. No fact about the world helps. The rules must be inferred from the matrix itself, and they change from item to item — which is the point. Each problem is novel by construction, so what is measured is the capacity to find structure in something you have not seen before.

The word progressive describes the design: within each set, items get harder in a controlled way, so the test teaches you its own logic as you go and then tests how far you can carry it.

What Raven actually built, and why it matters

This is the part that almost every description of the test omits, and it changes how the results should be read.

Charles Spearman, whose general factor Raven set out to measure, distinguished two things. The first he called the eduction of relations and correlates — from the Latin educere, to draw out — meaning the ability to perceive an abstract relationship and apply it somewhere new. The second was the ability to store and reproduce a culture's accumulated store of verbal concepts.

Raven built a separate test for each. Progressive Matrices measures eductive ability. The Mill Hill Vocabulary Scale, published alongside it, measures reproductive ability. The two were designed to be administered together.

The modern equivalent of that split is the distinction between fluid and crystallized intelligence — eductive ability corresponds closely to Gf, reproductive ability to Gc.

Why it is called the best single measure of g

When a battery of diverse cognitive tests is factor-analysed, a general factor emerges, and each test loads on it to some degree. Raven's loads on it about as heavily as any test ever constructed. Give someone one test and nothing else, and matrices will tell you more about their general ability than any other single instrument.

That is a real and important property, and it is also the source of the confusion. High g-loading means the test correlates strongly with the general factor. It does not mean the test measures everything the general factor is extracted from. Those are different claims, and the second is false.

Is Raven's culture-fair?

It is culture-reduced. It is not culture-free, and the difference is not pedantic.

Removing words and numbers removes the most obvious cultural content. What remains is the abstract, two-dimensional, rule-governed format itself — and familiarity with that format is not evenly distributed. Sitting still with a pencil, understanding that a puzzle has exactly one defensible answer, treating an abstract figure as a problem rather than a picture: these are things schooling teaches. Performance on Raven's rises with education.

Something changed across generations about how readily people take to abstract, hypothetical, classificatory reasoning. Whatever that something is, Raven's is sensitive to it. That is precisely what a culture-free test would not be.

Where you can actually take Raven's

Not, in all likelihood, wherever you found it.

Raven's Progressive Matrices is a published, copyrighted psychological instrument, distributed by Pearson and restricted to qualified users. Its norms are the reason a score means anything, and the item content is protected because a matrices item that has circulated online is a matrices item that no longer works.

The free matrix tests that dominate search results are not Raven's. They are imitations built in the same format, usually with no published norms, no reliability data and no validation study. Some are perfectly reasonable puzzles. None of them is the instrument whose psychometric properties they borrow the reputation of.

Myths and facts about Raven's Progressive Matrices