What Raven's Progressive Matrices Is
Each item shows a grid of abstract figures — usually three by three — with the bottom-right cell missing. Below it sit six or eight candidate pieces. One completes the pattern.
There are no words, no numbers, and no recognisable objects anywhere in the test. The instructions can be conveyed by pointing.
John C. Raven published it in 1938, and it has been in continuous use ever since — one of the very few psychological instruments of that age still considered current rather than historical.
Why "progressive"
Items are ordered so that each teaches the logic needed for the next. Early items establish that patterns continue across a row; later ones assume it and add a second rule running down the columns.
The test is, in a limited sense, self-instructing. That property is what makes it administrable without shared language.
Where It Came From
Raven worked within Charles Spearman's tradition. Spearman's 1904 observation that performance across unrelated cognitive tasks is positively correlated implied a general factor — g — and implied that some tasks would measure it more purely than others.
Spearman held that g was largely about perceiving relationships between things. A matrix of abstract figures presents relationships with nothing else attached.
The instrument was therefore designed from theory first and validated afterwards, which is the reverse of how most tests are built and part of why it is theoretically respected.
Why It Is So Heavily Used
- It is the most g-loaded brief instrument available. No other test of comparable length correlates as strongly with full-scale battery scores.
- It crosses languages. One form works across populations that share no language, which is impossible for anything verbal.
- It is cheap to administer. Group-administrable, objectively scored, no trained interviewer required.
- The normative base is enormous. Eight decades of data across dozens of countries and populations.
The consequence of that combination
Raven's became the default instrument in cross-cultural research, in large-scale epidemiology, and in any study needing a quick, defensible index of general ability.
Much of what is confidently claimed about intelligence across populations rests, in practice, on this one test — which is a reason to understand its limits rather than only its strengths.
The Versions
Three forms cover different ability ranges, and confusing them produces nonsense.
- Standard Progressive Matrices (SPM). The original: 60 items in five sets of twelve, spanning the general adult range.
- Coloured Progressive Matrices (CPM). Easier, printed on coloured backgrounds, for young children, older adults, and people with cognitive impairment.
- Advanced Progressive Matrices (APM). Harder, built to discriminate at the top of the range where the Standard form ceilings out.
Later editions added parallel forms and revised norms. Employer tests described as "Raven's-style" are usually not Raven's at all — they are commercially produced matrix tests using the same item form.
What a Matrix Item Actually Requires
The task decomposes into four steps, and difficulty is manufactured at a specific one.
- Encoding. Registering what varies across the grid — shape, count, shading, rotation, position.
- Rule induction. Inferring what governs the change along each dimension.
- Rule combination. Holding two or three rules simultaneously and applying them together.
- Application and check. Generating the missing cell and matching it against the options.
Hard items are hard mainly at step three. A single rule is rarely difficult; three rules operating at once exceed what most people can hold in working memory while still searching the options.
Why the distractors matter
The wrong options are constructed, not filler. Each typically satisfies some of the rules and violates one — so an answer that "looks right" usually means one rule was never noticed.
This is why checking every option against every identified rule is worth the time it costs.
Is It Culture-Fair?
The claim was made early and has been qualified ever since.
The strongest counter-evidence is the Flynn effect. Scores rose substantially across the twentieth century, and the gains were largest on abstract instruments including Raven's itself.
A measure of a stable, culture-independent trait should not drift that far within a few generations. Something environmental — schooling, exposure to diagrams and abstract symbols, familiarity with the very idea of a timed test — is being measured alongside reasoning.
The defensible description
Culture-reduced. Removing language and content removes several real sources of bias, which matters. It does not remove the advantage conferred by an education that trains abstract visual reasoning.
Can You Practise Raven's?
Yes, and the practice effect is well documented — which is precisely why the materials are restricted.
Repeated exposure raises scores. The gain comes from recognising the recurring rule types, not from any change in reasoning ability, so a practised score overstates the thing it is supposed to estimate.
What this means in each setting
- Clinically. A serious assessment records when the person last took the test, and repeat administration within a short window is treated as compromised.
- In hiring. Employers using matrix tests expect candidates to have practised, and set norms accordingly — which shifts the advantage to whoever practised most, not whoever reasons best.
- For yourself. Familiarity with the format is worth having before a real assessment. Just do not read the improvement as growth in ability.
What a Score Does and Does Not Tell You
A Raven's score is a percentile against a norm group, and the norm group is doing most of the interpretive work. The same raw score maps to different percentiles across countries, age bands and publication dates.
What it indexes is fluid reasoning — the capacity to work out novel relationships without relying on stored knowledge. That is genuinely predictive of academic and occupational outcomes, and it is one ability among several.
It is silent on knowledge, verbal skill, motivation, conscientiousness and everything interpersonal. A profile of several measures says far more than any single number from any single instrument, this one included.
The practical caution
A properly administered Raven's involves standard conditions, correct version selection, current norms, and a qualified interpreter. An unproctored online matrix test, however well built, is a rehearsal and a rough indication — not a clinical result, and not something to build a self-assessment on.