Matrix Reasoning Items: How They Work
Matrix reasoning items — the 3×3 grids of figures with one cell missing that you select an answer for — are the workhorse of fluid intelligence assessment. They were popularized by John Raven's Progressive Matrices in 1938 and remain the single most g-loaded item type in modern test design.
Matrix reasoning items — the 3×3 grids of figures with one cell missing that you select an answer for — are the workhorse of fluid intelligence assessment. They were popularized by John Raven's Progressive Matrices in 1938 and remain the single most g-loaded item type in modern test design. Almost every adult IQ battery includes a matrix-reasoning subtest, often weighted heavily in the fluid-reasoning composite.
The format works because it minimizes confounds with prior knowledge while maximizing demand on novel rule detection. The figures are abstract geometric shapes that no test-taker has memorized; the rules are general (counting, rotation, addition of features, alternation, intersection); and the answer is constrained to one of a small set of options, so guessing is bounded. A test-taker who has never seen a matrix item in their life can still solve the easy ones from the format alone.
Difficulty in matrix items is a function of how many simultaneous rules the test-taker must detect and integrate. Easy items use a single rule along a single axis (rows OR columns). Medium items use two rules, often one per axis. Hard items combine three or more rules, often with one rule applying to the whole matrix and another applying only to specific subsets of cells. Distractors at hard difficulty levels typically satisfy one of the rules but break another, ensuring that test-takers who only detect part of the structure are pulled toward wrong answers.
Performance on matrix items correlates strongly with academic and occupational success — Hunter and Schmidt's meta-analyses estimate operational validities of 0.4 to 0.5 for predicting job performance in cognitively demanding roles, comparable to or slightly higher than general intelligence measures based on more diverse subtests. The format is also remarkably resistant to coaching: targeted practice can produce 0.3 to 0.5 standard deviation improvements on matrix scores, but the gains plateau quickly and transfer to other reasoning tasks is modest.
The Open-Source Psychometrics Project distributes a 25-item matrix-reasoning test based on the format pioneered by Raven; the ICAR catalog includes matrix-reasoning items in its public-domain pool. Both resources are widely used in academic research and provide normative data that test designers can draw on for offline scoring conventions. The matrix items in this site's quiz are original constructions that follow the same item-design conventions.