How matrix reasoning puzzles actually work
The grid is not a picture. It is a rule, drawn out.
A matrix reasoning puzzle shows you a grid — usually three by three — in which every cell but one has been filled in. Your job is to work out what belongs in the empty cell and pick it from a set of options. That is the entire format, and it has barely changed since Raven designed the first version in 1936.
It looks like a picture puzzle. It is not. Every one of these grids was produced by applying a rule, and the shapes you can see are just what that rule happens to look like when you draw it. Once you understand that, the puzzles stop feeling like guesswork and start feeling like reading.
Why this format is used so widely
Matrix puzzles turn up in school entrance tests, cognitive assessments and job screening because they measure something unusual: reasoning that does not depend on what you already know. There is no vocabulary to have learned, no formula to have memorised, no cultural knowledge that gives one student an advantage over another.
That makes them fairer than most tests. It also makes them frustrating in a particular way — when you cannot see the rule, there is nothing to fall back on. You cannot half-remember your way to the answer. Either the rule clicks or it does not.
The good news, and the thing most people are never told, is that the rules come from a surprisingly small family. Once you can name them, you start recognising them instead of rediscovering them.
The rule families
Nearly every matrix puzzle you will meet is built from one or more of these:
- Progression. Something increases or decreases steadily across a row or down a column — a count, a size, an angle, a thickness.
- Distribution. Three values appear once each in every row and every column, in some order. Think of a tiny sudoku.
- Addition or subtraction. The third cell is the first combined with the second — shapes overlaid, or shared parts cancelled out.
- Constancy. One attribute never changes at all. Easy to overlook precisely because it is doing nothing.
- Transformation. A shape is rotated, reflected or moved by a fixed amount at each step.
That is close to the whole vocabulary. A hard puzzle is rarely a sixth exotic rule — it is two or three of these running at the same time on different attributes, which is covered in its own article.
Two rules are running. Across each row the count goes 1, 2, 3 — that is a progression. Down each column the shape stays whatever its row started with — that is a constancy. The missing cell needs three squares. Notice that neither rule alone gets you there: the count tells you 'three of something', the constancy tells you 'squares'.
Read rows, then columns, then diagonals
Work in that order, every time, and do it deliberately rather than letting your eye wander over the grid.
- Rows first. Most puzzles put their main rule across the rows, because we read left to right and the designers know it.
- Columns next. If the rows look like nothing, the rule may be vertical. Some puzzles use both at once.
- Diagonals last. Rarer, and usually a distribution rule hiding in plain sight.
The reason the order matters is that it stops you staring. Staring at a grid you do not understand produces nothing except the growing feeling that you are bad at this. Having a fixed procedure means you are always doing something, and one of the three passes usually turns the rule up.
Say the rule out loud before you look at the options
This is the single habit that separates people who are quick at these from people who are not, and it feels backwards at first.
The instinct is to glance at the answer choices to get a sense of what kind of thing is expected. Resist it. The options in a well-designed matrix puzzle are not neutral — they are distractors, built specifically to look correct to someone who found part of the rule. There is usually one that is right for the row rule but wrong for the column rule, one that is a mirror image of the right answer, and one that is correct in every way except a single detail like shading.
Look at those before you have committed to a rule and they will pull you toward whichever near-miss matches the fragment you noticed. Decide first, in words — three squares, unshaded, rotated ninety degrees — and then go find that. If it is not among the options, you have learned something specific and can go back and check which part of your rule was wrong.
If you cannot state the rule in a sentence, you have not found it yet. A vague feeling that an answer 'looks right' is exactly what the distractors are engineered to produce.
When you are genuinely stuck
Everyone gets stuck. What matters is having something better to do than stare harder.
- Pick one attribute and follow only that. Ignore shape entirely and track only how many things are in each cell. Then only shading. Then only orientation. Complexity usually comes from several simple rules stacked, and looking at all of them at once is what makes it look impossible.
- Check what is NOT changing. A constancy is a rule too, and it is the one people skip past.
- Count everything. Sides, dots, lines, intersections. Numeric rules are common and are invisible if you are looking at the pictures as pictures.
- Eliminate rather than select. You often cannot see the right answer but can prove four options wrong, and that is just as good.
How to practise so it actually transfers
Doing hundreds of puzzles quickly is much less useful than doing thirty slowly and understanding every one. The skill you are building is rule-recognition, and you only build it by naming rules.
After each puzzle — including the ones you got right — say what the rule was. If you got it right but cannot name the rule, you guessed well, and guessing well does not improve. If you got it wrong, find the specific step where your rule broke rather than just reading off the correct answer.
Speed arrives on its own after that. It is a side effect of recognising a rule family instantly, not something you can train directly by rushing.