Raven’s Progressive Matrices is a non-verbal reasoning test: each item shows a pattern of shapes with one piece missing, and you choose the piece that completes it from six to eight options. Known simply as Raven’s matrices or the Raven test, it was first published in 1938 by the British psychologist John C. Raven. It measures abstract reasoning — often called fluid intelligence — with almost no reliance on language, and it is one of the most widely used intelligence tests in research. It is a strong measure of reasoning, but not a complete IQ test on its own.
What the test looks like
Every item is a design with a gap. In the simplest items the design is a single continuous pattern and the task is close to visual matching. In harder items the design is a grid — often three rows by three columns — in which shapes change across the rows and down the columns according to one or more rules: a shape rotates, a line is added, a figure moves from one position to the next. You have to find every rule and combine them to pick the missing cell.
The items get steadily harder, which is where the word progressive comes from. In the classic Standard version there are 60 items in five sets, labelled A to E, of 12 items each, and each set is harder than the one before. The original test was designed as an untimed “power” test: what matters is how difficult a problem you can solve, not how fast you work. Matrix items of this kind are now a standard part of many tests; our guide to IQ test questions walks through a worked example.
What Raven’s Progressive Matrices measures
Raven built the test to measure one of the two components of general ability described by Charles Spearman. His son John Raven, who led the test’s later research, calls it eductive ability — from the Latin educere, “to draw out” — and describes it as “the ability to make meaning out of confusion”. The companion Mill Hill Vocabulary Scale was built to measure the other component, reproductive ability: absorbing and recalling information. Within an age group the two correlate at about .5 — related, but far from the same thing.
A 1990 study by Patricia Carpenter, Marcel Just and Peter Shell looked at how people actually solve the items, using verbal protocols, eye-fixation patterns and errors. Everyone used the same step-by-step strategy of spotting one regularity at a time. What separated higher from lower scorers was mainly “the ability to induce abstract relations and the ability to dynamically manage a large set of problem-solving goals in working memory”. That is why Raven’s scores are closely tied to working memory.
Is Raven’s a pure measure of intelligence?
Raven’s is often described as the best single measure of the g factor. A 2015 study by Gilles Gignac tested that claim across several large samples and found a more modest picture:
| Source of variance in Raven’s scores | Share (Gignac 2015) |
|---|---|
| General intelligence (g) | about 50% |
| A fluid-reasoning factor separate from g | about 10% |
| Reliable variance specific to Raven’s itself | about 25% |
Gignac concluded that Raven’s “is not a particularly remarkable test with respect to g” and advised researchers not to use it alone when they want a valid estimate of general intelligence; a short combination of four Wechsler subtests did better. In plain terms: Raven’s is an excellent reasoning test, but a single test type always carries some skill that belongs to that test rather than to intelligence in general. Our guide to how IQ tests work explains why full batteries combine many subtests.
The versions of the test
| Version | Who it is for | Notes |
|---|---|---|
| Standard Progressive Matrices (SPM) | From about age 6 into old age | 60 items in five sets (A–E) of 12; published 1938 |
| Coloured Progressive Matrices (CPM) | Younger children and people who find the standard items too hard | Easier items, printed in colour |
| Advanced Progressive Matrices (APM) | More able adolescents and adults | Two parts, Set I and Set II; harder items to avoid a ceiling |
| Raven’s 2 (current edition) | Ages 4 to 90 | Paper or digital; reports a standard score and a percentile rank |
John Raven explains why the family exists: the Standard version could be used at almost any age and was short enough to give in homes, schools and workplaces, but it “had limited discrimination at the upper and lower levels”. The Coloured and Advanced versions were developed to fix that for the less and more able. Researchers have also built shortened forms; a 2022 study by Dutch researchers, for example, developed and validated a much shorter version of the Standard test for children and adolescents.
How Raven’s is scored
Each correct answer scores one point, so the raw score on the Standard test runs from 0 to 60. A raw score means nothing until it is compared with norms — the scores of a representative sample of people of the same age. The comparison turns the raw score into a percentile (the share of the norm group scoring lower) and, in modern editions, a standard score on the familiar IQ-style scale where 100 is average. Our IQ scale guide shows how standard scores and percentiles line up.
The age of the norms matters more than most people expect. When the test was re-standardized in Britain in 1979, the scores had risen so far since the 1938 norms that, at the 5th percentile, 10½-year-olds in 1979 scored about as well as 14-year-olds had in 1938. James Flynn’s 1987 survey of 14 nations found gains of 5 to 25 points in a single generation, with some of the largest on culturally reduced tests such as Raven’s — the Dutch data alone showed a gain of 20 points. This is the Flynn effect, and it means a score read against an old table can overstate a person’s standing by a wide margin.
Is Raven’s culture-fair?
It is culture-reduced, not culture-free. Because it uses shapes rather than words, it depends far less on language and schooling than a vocabulary or general-knowledge test, which is why it has been used in so many countries. In the 1979 British standardization the item difficulties behaved in much the same way for children from different socioeconomic backgrounds, and analyses in several US school districts found that the test worked in the same way within different ethnic groups.
But the rising scores are the strongest evidence that environment matters: nobody’s genes changed between 1938 and 1979. Flynn’s own conclusion was that the tests measure something with only a weak causal link to intelligence itself. National norms also differ, so a score is only meaningful against norms for the population the person belongs to.
Can you take the real Raven’s test online?
Not as a member of the public. The current edition, Raven’s 2, is sold by its publisher only to buyers who meet its professional qualification requirements, and is given on paper or through the publisher’s own digital platform. The many free “Raven tests” online are unofficial: they use their own items, or items of unknown origin, and their conversion to an IQ number is not based on the published norms. They can be good practice with the item type, but their result is not a Raven’s score.
The IQ Revealed test is also a matrix-reasoning test — 40 original items — but it is not Raven’s Progressive Matrices, uses none of its items and is not normed against it, so its result should never be read as a Raven’s score. Our guide to how accurate IQ tests are covers what an online reasoning test can and cannot tell you.
Trust and scope notes
This page explains a psychological test in general terms. Raven’s Progressive Matrices is also used by psychologists within clinical and educational assessments; only a qualified professional can administer and interpret it for an individual, and nothing on this page, or our own test, can stand in for that. Our IQ test is for adults aged 18 and over.
IQ Revealed is not affiliated with, endorsed by or connected to Pearson, the publisher of Raven’s Progressive Matrices and Raven’s 2, or to the publishers of the Wechsler scales, the researchers and the universities named on this page. “Raven’s” and the test names are used only to identify the tests being described.