Skip to main content

Raven's Progressive Matrices

The best-known measure of non-verbal reasoning, in which the examinee selects the figure that belongs in a blank cell of a matrix of shapes

Raven's Progressive Matrices is a visual reasoning task devised in the 1930s. Part of a matrix in which shapes are arranged according to a rule is left blank, and the examinee selects the figure that belongs there from a set of options. Its defining feature is a design that needs almost no instruction - looking at the item tells you what to do - and it has long been used as an index of non-verbal reasoning with comparatively little dependence on language or culture. Most of the shape-series items found on the web are simplified versions of this original.

Features of the Design

The reason this task has remained in use for so long is that it requires almost no knowledge. Items probing vocabulary or general knowledge turn differences in native language and schooling directly into differences in score, whereas changes in visual attributes such as form, count, orientation, and fill look the same to a speaker of any language. Items are arranged in graded order from easy to hard, and in the later stages several rules operate at once. The word progressive in the name refers to this construction of difficulty.

What It Captures

In the course of answering, the examinee repeatedly breaks a figure into its attributes to identify which element is changing, forms a hypothesis about the rule governing the change, and tests it against the next element. Because the hypotheses have to be held in mind throughout, working memory capacity bears on performance. What is being captured is the ability to grasp an unfamiliar rule on the spot - something close to fluid intelligence - which is a different property from the volume of knowledge accumulated. The correlation between performance on this task and on memory tasks is thought to come from the shared load of holding hypotheses.

Cautions When Taking a Simplified Version

Simplified versions on the web inherit the original's property of requiring a rule to be grasped on first sight without depending on knowledge, but they differ decisively in not being standardized. With no age-banded reference sample and no control over the testing environment or the number of retakes, the number produced cannot be interpreted as an intelligence quotient. Familiarity with a fixed format will raise the score, but that is fluency with the task. Using it to compare against your own earlier results taken under the same conditions, rather than against other people, fits what it can actually support.

Related Terms

Fluid IntelligenceIntelligence QuotientPercentileWorking Memory

Related Articles

What Pattern Reasoning Can Measure - Fluid Intelligence and Series Reasoning

Working out the rule behind a row of shapes taps into something close to fluid intelligence - handling an unfamiliar problem without leaning on stored knowledge. This article covers the distinction between fluid and crystallized intelligence, why Raven's Progressive Matrices came to be so widely used, and how to read a score that rises with practice.

What an IQ Test Actually Measures - How It Works and Where It Stops

An IQ score is not an absolute measure of how clever someone is. It is a relative position within a same-age population, rescaled so that the average sits at 100. This article explains the reasoning behind how the score is derived, the decisive difference between a standardized test and a quick test on the web, and what an IQ score can and cannot account for.