How Matrice IQ arrives at a single number.
What we measure
Matrice IQ measures a single construct: fluid reasoning — your ability to recognize abstract patterns and apply rules to novel visual material with no advantage from prior knowledge, vocabulary, or schooling. In the Cattell–Horn–Carroll (CHC) model this is the Gf factor, and it is the cognitive ability that correlates most strongly with what people colloquially call “general intelligence.”
Pattern-matrix tests have been the gold standard for fluid-reasoning measurement since John Raven introduced the Progressive Matrices in 1938. They are language-independent, culture-light, and resistant to crammable knowledge.
Item structure
Every session is thirty-five pattern-matrix items — the entire authored bank, sequenced from easier to harder. Each item is a 3×3 grid with the bottom-right cell missing and six options from which to choose the cell that completes the pattern. The session is stratified across five designer-assigned difficulty levels (four introductory, six single-rule, eight compound, eight layered, nine bidirectional and meta-rule) so the curve is honest and not front-loaded.
The rule space draws on the formal grammar made explicit by Carpenter, Just & Shell’s 1990 analysis of the Advanced Progressive Matrices and the modern AI matrix datasets (PGM, RAVEN, ICAR), reframed for an authored bank:
- 01constant-in-axis (the simplest rule: one value per row or column)
- 02quantitative pairwise progression along an axis — count, size, polygon sides, ring fills, rotation, star points
- 03distribute-three (Latin square) on shape, fill, position, sides
- 04two- and three-attribute orthogonal rules running independently on the same cell
- 05row-wise figure operations — addition (OR), subtraction, XOR, AND — performed on line-sets, dot quadrants, and sub-grid masks
- 06bidirectional set operations — the same rule must hold simultaneously across rows and columns
- 07compound (shape-in-shape) Latin squares with two independently-Latin attributes
- 08whole-matrix arithmetic, where the cell at (r, c) is a function of both indices
Every item, including all five distractors, is original and authored in-house. Each distractor is rule-aware: it corresponds to a specific plausible misreading — wrong axis, operator swap, partial attribute match, input repeat, or foreign value. We do not reproduce items from Raven’s Progressive Matrices, Wechsler, ICAR, or other copyrighted instruments.
Scoring
Your score is a heuristic estimate. Each correctly answered item contributes its designer-assigned difficulty to a raw ability score. The raw score is normalized against the maximum possible difficulty sum and mapped to an integer estimate via a fixed lookup table.
Matrice IQ is not a norm-referenced score. We do not maintain a stratified norming sample of thousands of test-takers, and we make no claim of clinical validity. The mapping table is calibrated by hand against published distributions of matrix-reasoning scores and may be adjusted as we observe results across our population of users.
Limitations
- §1Not a clinical IQ test. Cannot be used for diagnostic purposes.
- §2Measures fluid reasoning only — not verbal ability, memory, or processing speed.
- §3Not norm-referenced. The number is an estimate, not a percentile rank.
- §4Single sitting, online, unproctored. Practice effects exist.
- §5Item bank is finite; repeat takers will eventually see all items.
What you get when you unlock
- ▪An integer estimate of your fluid-reasoning ability.
- ▪A categorical band: average, above average, top ten percent, etc.
- ▪A breakdown of your answers by item difficulty and rule family.
- ▪A copy of your result emailed to the address you provided.