Matrix reasoning tests in job assessments

Start in the right place

Which matrix test are you sitting?

The grid is the same shape everywhere and the instrument is not: the answer tray, the pacing and the rule mix all differ. If your invitation names one of these, practise that one. If it names none of them, the numbered series on this page is the general option and covers every rule family below.

The skill

What is a matrix reasoning test?

A matrix reasoning item is a grid of figures, usually three by three, with one cell left blank. Something changes as you move across each row, and something changes as you move down each column. Your task is to work out both, then pick the option that satisfies them together.

Pearson describes its Raven's Adaptive assessment in exactly those terms: three-by-three geometric designs, each with a piece missing, where you select the correct piece from a set of eight. Assessio describes Matrigma as consisting of matrices that require logical and spatial ability. Aon describes gapChallenge as understanding the pattern in the graphic provided and using logic to fill the empty cell.

Nothing in the grid is verbal or numerical, which is why employers reach for it: the same item works across languages and educational backgrounds. It is also why it feels unfair the first time. Preparation cannot change what the test is designed to measure, but it can make the item format and a consistent solving process familiar.

Try it before anything else

Free matrix reasoning practice questions

Six original questions covering the rule families above, each one a grid you have not seen. Commit to an answer before opening the solution — the solution walks the method and names the rule every wrong option breaks.

What employers are looking at

What do matrix reasoning tests measure?

Matrix items are the classic measure of fluid reasoning: solving an unfamiliar problem without drawing on learned knowledge. Pearson describes Raven's Adaptive as assessing lateral thinking, learning new concepts quickly, and solving new and complex problems without drawing on prior knowledge.

Assessio reports Matrigma as a general mental ability score on a nought-to-ten C-scale, grounded in Spearman's g-factor and Cattell's notion of fluid intelligence, and interpreted against a norm group. Assessio also notes that the non-verbal format helps reduce the disadvantage for candidates with dyslexia or dyscalculia.

In practice the test separates three things: whether you notice a change at all, whether you test it against every row and column instead of one, and whether you can hold two rules in mind at once.

Reading your result

How matrix reasoning tests are scored

A matrix score is rarely a count of grids solved. Providers convert responses into a standing against other people, and on the adaptive versions the difficulty of what you answered carries as much information as the amount.

• Results are norm-referenced, not marks out of ten. Assessio reports Matrigma as a General Mental Ability score on a nought-to-ten C-scale, interpreted against a norm group. The number the employer sees is a comparison with other people, not a tally of correct grids.
• No provider publishes a pass mark. Assessio's C-score is read against a norm group, and any threshold is set by the employer for the role. If a recruiter has disclosed a target, that is the only target worth preparing against.
• On an adaptive test, the question count is not the score. Matrigma and Raven's Adaptive change question difficulty with your responses, so the number of questions answered is not the score by itself. The provider's score report, rather than the count alone, is the right way to interpret the result.
• Difficulty is not feedback. Both adaptive assessments select the next grid from your last answer, and Aon states that gapChallenge may get more difficult as you progress. A harder grid often follows a correct answer, so treat rising difficulty as normal behaviour rather than as a verdict.
• The construct is general, and the weighting is the employer's. Assessio grounds the C-score in Spearman's g-factor and Cattell's notion of fluid intelligence, which is why a single non-verbal test is reported as General Mental Ability. How an employer weighs that number against the rest of your application is their decision rather than the test's.

Question formats

The rule families that fill the grid

The grid is constant. The rule families that fill it are a short list, and knowing the list is most of the preparation.

• Progression across a row or column. A count, size, angle or position increases or decreases by a fixed step in each successive cell.
• Distribution of three. Each row and each column contains each of three elements exactly once, in a different order. Very common, and easy to miss because nothing appears to move.
• Rotation and reflection. A figure turns by a constant angle at each step. Reflection distractors are supplied deliberately, because a mirrored figure matches the original feature for feature and no rotation can produce it.
• Addition, subtraction and overlay. The third cell combines the first two, with shared elements sometimes kept and sometimes cancelled. The cancelling version is often described as an exclusive-or rule.
• Movement of an element. A dot, line or marker travels a fixed number of positions at each step, sometimes wrapping around the figure.
• Constant along one axis. One attribute stays fixed along the row and changes down the column, or the reverse. Two independent rules stacked inside one grid.

Launch coverage

Which providers use matrix reasoning tests?

Three of Cognivy's launch families deliver matrix items, under three different names and three different delivery models.

• Matrigma. Assessio's help material describes Matrigma as consisting of matrices requiring logical and spatial ability, delivered adaptively, with a one-minute limit per question inside a twelve-minute total. A less common classic version runs to 40 minutes. Results are reported as a nought-to-ten C-score for General Mental Ability.
• Raven's Adaptive matrix reasoning. Pearson lists Raven's Adaptive as three-by-three geometric designs with a piece missing, chosen from a set of eight, delivered by computer-adaptive testing at 15 items in approximately 12 minutes.
• gapChallenge. Aon files gapChallenge under deductive-logical thinking and describes the task as understanding the pattern in the graphic provided and using logic to fill the empty cell. Its published practice material is more specific: a four-by-four or five-by-five grid with some squares filled and one marked with a question mark, where no shape is repeated in any row or column. Aon states there is no set number of questions and that the test may get more difficult as you progress.

Provider names identify the assessment format. Cognivy is independent and is not affiliated with or endorsed by these providers.

What varies, and what does not

Timing, adaptive delivery and scoring

Major matrix assessments are often short, and some use adaptive delivery. Both details change how you should pace yourself.

• The clock is often per item, not per test. Assessio gives Adaptive Matrigma a one-minute limit for each question inside a twelve-minute total. A grid you have not started within the first fifteen seconds is a grid to leave.
• Some major matrix assessments are adaptive. Both Matrigma and Raven's Adaptive select the next item from your last answer. An adaptive test may follow a correct answer with a harder item, but you cannot read your score from how hard the grids feel, so treat rising difficulty as normal behaviour rather than as feedback.
• The classic long form still exists. Assessio's less frequently used classic Matrigma runs 40 minutes with progressively harder items. Your invitation should tell you which version you have been sent.
• Scores are norm-referenced, not marks out of ten. Assessio reports a nought-to-ten C-score for General Mental Ability against a norm group. No pass mark is published, and any threshold belongs to the employer.

A repeatable approach

A procedure that catches the second rule

Matrix items reward a fixed procedure, because what goes wrong is a check that was skipped rather than a rule that was too hard.

1. 1

   Scan attributes in the same order every time. Shape, count, size, position, rotation, shading, line style. A fixed order is what stops you missing the one attribute that moved.

2. 2

   Solve the rows first, then the columns. State each direction as a sentence before looking at the options. If only one direction has a rule, you have not finished the grid.

3. 3

   Predict the answer before reading the options. Describe the missing cell in words. Options are written to be attractive, and reading them first anchors you to one of them.

4. 4

   Reject on a broken rule, not on appearance. Take each option and name the rule it violates. An option that fits the row and breaks the column is the classic trap on this format.

5. 5

   Check for the second rule before committing. A grid often carries two rules at once, and finding one and stopping leaves an option that satisfies half the pattern looking correct.

What slows progress

Common mistakes, and the fix for each

• Solving only across the rows. Fix: force yourself to state the column rule before opening the options. If you cannot state it, you have not solved the grid.
• Stopping at the first rule. Fix: assume two rules until you have proved there is one. If your single rule eliminates fewer than all but one option, another rule is running.
• Confusing rotation with reflection. Fix: pick one asymmetric feature and track where it lands. A rotation carries it around the figure; a reflection swaps its side.
• Choosing the option that looks right. Fix: name the violated rule for every option you reject. If you cannot name one, you are recognising rather than reasoning.
• Repeating grids you have already seen. Fix: practise on unseen items. Remembering an answer feels like improvement and measures nothing, and an adaptive test will not serve you the grid you rehearsed.

In practice

How do you decompose a three-by-three matrix under time pressure?

Run the attribute scan in a fixed order on every grid: shape, count, size, position, rotation, shading, line style. The order itself is not the insight — any order works if it never varies — but fixing it is what stops one moving attribute from hiding behind another. Sample question 5 runs a moving dot and a darkening fill at once, and a candidate who scans ad hoc finds one of them, commits, and walks into the option built for exactly that.

Then state each direction's rule as a sentence before you open the options: the tile turns a quarter-turn clockwise, shared marks cancel, the dot moves one corner. A rule you cannot say in words is a rule you have not found, and a spoken prediction turns the answer list from a menu into a check — you are no longer browsing for what looks right, you are confirming what must be there.

The clock settles the rest. Adaptive Matrigma allows one minute per question, and a grid you have not started within the first fifteen seconds is a grid to leave. A fixed routine — scan, state, predict — is worth having precisely because a per-question clock punishes staring: each step either produces something or tells you to move.

In practice

Row rules or column rules — which do you check first?

Rows first, as a habit. Reading order makes a left-to-right change the cheapest to spot, and most grids state their clearest rule across the rows. But first is the operative word: a grid is not solved until the columns confirm what the rows claim, and the standard distractor on this format — right in one direction, silent in the other — exists because stopping at half the evidence feels like finishing. Sample question 1's plain triangle is exactly that option.

Sample question 6 shows the stronger case: the row leaves two shapes standing, the column leaves a different two, and only the shape on both lists survives. Neither direction is sufficient on its own, which makes the order you check them in matter far less than the refusal to commit after one.

The family that punishes a rows-only habit hardest is the distribution of three, because nothing appears to move: each row and each column simply holds each of three elements once, in a different order. There is no motion for the eye to catch, so the rule only surfaces when you audit both directions deliberately. If a single rule leaves more than one option alive, a second rule is running — and it usually lives in the direction you have not checked.

In practice

How do you tell rotation from reflection in a matrix grid?

Track one asymmetric feature and note where it lands. A rotation carries a feature around the figure and preserves the way its parts wind; a reflection swaps its side and reverses that winding. The reliable check is handedness: stand at a fixed point of the figure, swing from one named part to another, and record whether the swing runs clockwise or anticlockwise. No turn within the plane can reverse that swing, and every mirror does.

That check is what unpicks sample question 3. The mirror card matches the correct tile feature for feature — long arm up on both — and only the handedness betrays it: swinging from the long arm to the short arm runs anticlockwise on the mirror where every tile in the grid runs clockwise. Reflection distractors are supplied deliberately, and they are built to survive a glance and fail a check.

So on any grid with a turning figure, test the options by handedness first. It rejects every mirror in one step, with nothing rotated mentally, and it is immune to the corner trap: two different orientations can put the corner in the same place — sample question 3's Cards 3 and 4 both sit in the bottom right — and only following the arms tells them apart.

A clear route

How to prepare in the days you have

Matrix preparation is short and specific. Learning the rule families is most of it.

1. 1

   First, learn the rule families one at a time. Progression, distribution, rotation, overlay, movement. Work untimed on grids using only one, until you can name it within a few seconds.

2. 2

   Next, move to two-rule grids. Almost every real item stacks two. Practise stating both before choosing, and treat a single-rule answer as an incomplete answer.

3. 3

   Fix the attribute scan. Run the same checklist on every grid until it is automatic. Consistency is what prevents the missed-change error.

4. 4

   Add the per-item clock. If your assessment is Adaptive Matrigma, practise with a one-minute ceiling per grid so that abandoning becomes a decision rather than a panic.

5. 5

   Rehearse on unseen items. Repeating known grids inflates your sense of readiness and does nothing for an adaptive item bank.

Cognivy uses your assessment date to choose the route rather than asking you to predict a study schedule. When you sit down to practise, you choose the session length that fits that day.

Pace

Getting faster without losing accuracy

On a per-item clock, speed is about deciding when to stop rather than about scanning faster.

• Predict before you read the options. It is faster overall, and it removes the pull of an attractive wrong figure that satisfies one of the two rules.
• Use the options to test, not to browse. Once you have a prediction, one pass through the options is enough. Needing a second pass usually means the prediction was not specific.
• Abandon at your ceiling. With a one-minute limit per item, a grid you have not started in fifteen seconds is unlikely to resolve in the remaining forty-five. Choose the best fit and move.
• Practise the checklist, not the puzzles. The transferable skill is the attribute scan. Grids you have memorised transfer nothing to an adaptive item bank.

Keep exploring

Related test types and how they differ

Matrix items sit inside a wider non-verbal family. These are the neighbouring formats and how they differ.

Direct answers

Matrix reasoning test FAQs