Matrices are the signature format of aptitude testing and the one that defeats people fastest. You look at nine figures, see a tangle of shapes, and pick something at random.
Yet almost every matrix is built on the same principle. Understanding it changes the experience completely.
The principle: independent attributes, stacked
A figure in a matrix is not one thing. It is a stack of independent characteristics. Typically three to five of:
- shape — circle, square, triangle, star;
- fill — empty, half, solid, hatched;
- rotation — in steps of 45° or 90°;
- number of elements;
- size or position;
- a small detail — a dot, a bar.
Each of these follows its own simple rule. What makes the whole thing hard is never the complexity of any one rule — it is trying to read all of them at once.
The method
Step 1: inventory what varies. Scan the grid and note what changes from cell to cell. Anything that never changes is not a rule, it is scenery. Discard it immediately.
Step 2: take one attribute, only one. Follow it along the first row, left to right. Then the second row. If both rows tell the same story, you have that attribute’s rule.
Step 3: if rows give nothing, read columns. Many matrices put one attribute on the rows and another on the columns. It is in fact the most common construction: the row fixes the shape, the column fixes the fill.
Step 4: repeat for each attribute. You end up with a list of constraints — “shape: triangle”, “fill: solid”, “rotation: 90°”.
Step 5: eliminate. Do not look for the right answer; remove the wrong ones. Take your first constraint and strike out every option that breaks it. Then the second. Usually only one survives after two or three constraints.
That inversion — eliminating rather than searching — is what separates competent performers from strong ones on this format.
Four constructions that cover nearly everything
Progression. An attribute increases steadily in one direction: the count goes 1, 2, 3; the figure turns 45° each step.
Distribution. Three values of an attribute each appear exactly once per row and per column — the sudoku principle. If you see three different shapes per row with no apparent order, think of this straight away.
Combination. The third cell results from the first two: superposition, subtraction, or a logical rule. The most frequent case is an element appearing in the third cell only when it is in exactly one of the first two — an exclusive-or in disguise.
Cross-borrowing. The third cell takes its outline from the first and its fill from the second. Obvious once named, invisible until you know the pattern.
Two real traps
The continuous series. Some matrices are not read row by row but as a single run of nine cells, left to right and then down — the way you read text. If rows and columns both fail, try this before giving up.
The near-miss distractor. Options are built to satisfy every constraint but one. This is exactly why elimination beats searching: hunting for the right answer, you stop at the first option that “looks right”; eliminating, you are forced to check each constraint.
Practising usefully
Three pieces of advice worth more than hours of repetitive drilling.
Say the rule out loud before answering. If you cannot state it in one sentence you have not found it — you have a hunch, and hunches are what the distractors are designed to catch.
When you get one wrong, do not just look at the correct answer: identify the attribute you missed. Errors cluster on one or two specific attributes, usually rotation or a small detail.
Finally, work on original matrices. The classic published sets circulate so widely that redoing them measures recall, not reasoning. Every figure in our test was drawn for this site, and each is explained afterwards with the rule spelled out.