Raven's Progressive Matrices: What They Measure and How They Work

In 1938, a British psychologist named John C. Raven published a test unlike most intelligence measures of the time. No vocabulary questions. No arithmetic. No reading required. Just a series of visual puzzles — grids with a missing piece — that got progressively harder as you went through them. Nearly ninety years later, Raven's Progressive Matrices remains one of the most used and most trusted measures of human cognitive ability in the world.

Understanding what this test actually measures, and why it works so well, tells you something important about the nature of intelligence itself. If you want to get a feel for the format before reading further, the Matrix Reasoning Test on this site uses the same 3×3 grid structure.

What the Test Looks Like

Each item in Raven's Progressive Matrices presents a 3×3 grid of visual patterns with one cell missing — always the bottom-right. Your job is to study the patterns across the rows and columns, identify the underlying rules, and select the correct missing piece from a set of options (typically six or eight choices).

The patterns themselves involve shapes, sizes, colors, quantities, orientations, and relationships between elements. The simplest items have one rule to find. The hardest require tracking several simultaneous rules — for example, shape changes along one axis while quantity changes along another and a shading pattern rotates along a third. The test consists of increasingly difficult pattern matching tasks and has little dependency on language abilities, which is a large part of what makes it so useful across cultures and educational backgrounds.

The items are organized into sets that increase in difficulty — hence "progressive." The Standard Progressive Matrices has 60 items. There's also a Coloured version (easier, used with young children or clinical populations) and an Advanced version for higher-ability testing.

What It Actually Measures

Raven designed the test to measure what psychologist Raymond Cattell later called fluid intelligence — the ability to reason through new problems that can't be solved using stored knowledge or learned routines. This is distinct from crystallized intelligence, which reflects what you've accumulated through education and experience.

Fluid intelligence is what you use when you encounter something genuinely unfamiliar and have to work it out from scratch. The matrices capture this well because each puzzle is self-contained — there's nothing to remember from outside the test, no vocabulary to know, no prior knowledge that helps. You either see the rules or you don't.

Research has consistently found that performance on the matrices correlates strongly with performance on other reasoning tasks, which is why it's often used as a proxy for general cognitive ability. Working memory capacity shares around 50% of the variance with general fluid intelligence, and the matrices tap directly into this — harder items require holding multiple rules and sub-goals in mind simultaneously while searching for the solution.

Why Working Memory Is Central

The reason harder matrix items are harder comes down to working memory. Researchers Carpenter, Just, and Shell analyzed how people actually solve Raven's problems and found that the key cognitive demand is managing a hierarchy of goals and sub-goals — identifying one rule, holding it in memory, moving to the next attribute, checking whether the candidate rule works across all rows and columns, and so on. Performance on Raven matrices reflects the ability to induce abstract relations and manage problem-solving goals in working memory.

On easy items, you might need to track a single rule — every row has the same shape. On the hardest items, you might be tracking four or five rules across multiple attributes at once, while also eliminating distractor answer choices that satisfy some rules but not others. This is why people with higher working memory capacity tend to score better on the matrices — not because they're faster, but because they can hold more of the problem in mind at once without losing track.

This also explains something counterintuitive: the test doesn't feel like a memory test. But memory is doing a lot of the work behind the scenes. You can explore how working memory interacts with pattern detection through the Pattern Memory Test, which isolates the visual encoding side of this process.

Three Versions, Three Populations

The Raven's family of tests comes in three forms, each targeting a different population:

Coloured Progressive Matrices (CPM) — designed for children aged roughly 5–11 and for adults with cognitive impairments or low baseline ability. The grids use color to make patterns more perceptually distinct. It's widely used in developmental research as a measure of fluid intelligence in young children.

Standard Progressive Matrices (SPM) — the general population version with 60 black-and-white items across five sets. This is the form most commonly encountered in research and occupational assessment. It's well-suited for ages 6 through adulthood across a broad ability range.

Advanced Progressive Matrices (APM) — a harder version designed to differentiate among people at the upper end of the ability distribution. It's used in academic selection, high-ability research, and contexts where ceiling effects would make the standard version too easy.

Why It's Considered Culture-Fair

One reason Raven's test has been used so extensively across countries and cultures is its minimal reliance on language and culturally specific knowledge. A child in rural Kenya and a university student in Norway face the same visual puzzles, with no translation needed and no disadvantage from differing educational backgrounds — at least in theory.

In practice, the picture is more nuanced. Familiarity with multiple-choice test formats, exposure to geometric shapes and visual puzzles, and general test-taking experience do appear to affect scores somewhat. But compared to vocabulary or general knowledge tests, the matrices are substantially less dependent on what you've been taught, which makes them valuable in cross-cultural cognitive research and international assessments.

What a Score Means (and Doesn't)

Raven's scores are typically expressed as percentiles relative to an age-matched reference group. A score at the 75th percentile means you performed better than 75% of people your age in the normative sample — it says nothing about an absolute level of ability.

It's worth being clear about what the test doesn't measure. It captures fluid reasoning and abstract pattern detection well, but it says little about emotional intelligence, creativity, practical problem-solving, or domain-specific knowledge. Many things that matter in real-world performance fall outside what any single test can capture. A high score suggests strong abstract reasoning; it doesn't predict everything about how someone thinks or performs.

The matrices are also sensitive to practice effects — doing similar puzzles repeatedly tends to improve scores somewhat, not because fluid intelligence has increased, but because you become more familiar with the format and the types of rules that typically appear. This is one reason the Visual Pattern Test and Matrix Reasoning Test can serve as useful training tools as well as assessments.

How It's Used Today

Raven's Progressive Matrices is used in an unusually wide range of contexts. In research, it appears constantly as a measure of fluid intelligence or general cognitive ability. In occupational settings, it's used in screening and selection — particularly for roles requiring strong analytical reasoning. In clinical neuropsychology, it helps assess cognitive function in patients who may have language or motor impairments that would make verbal tests invalid. In education, it's used in gifted identification and learning disability assessment.

It has also become a benchmark in artificial intelligence research. Developing AI systems that can reliably solve Raven's items at human levels has proven more difficult than expected, and the test is widely used to evaluate machine reasoning capabilities — partly because it demands the kind of abstract relational thinking that has historically been harder for AI to replicate than pattern matching or memorization.

For anyone interested in understanding how their own abstract reasoning compares, the Matrix Reasoning Test offers the same core format. For a broader look at what pattern recognition ability means and how it develops, the articles on what pattern recognition is and why some people see patterns others miss cover the underlying cognitive science in more depth.

The Bottom Line

Raven's Progressive Matrices has lasted nearly ninety years because it measures something real and does so efficiently — the ability to find rules in unfamiliar visual material without relying on language, memory for facts, or prior knowledge. It's not a perfect measure of intelligence (no single test is), but it's probably the cleanest measure of fluid reasoning currently available.

Understanding the test also illuminates something about cognition more broadly: that a large part of what we call "being smart" is the ability to detect structure in new information and hold multiple relationships in mind simultaneously. That's a trainable skill, not a fixed trait — and it's what the tools on this site are built to develop.