Abstract Reasoning Test: 5 Worked Questions with the Figures Explained
Five questions with the figures shown, not described. Each group opens with the method that cracks that question type, then gives you the diagram — work it out before opening the explanation, which also names the distractor you were most likely to pick.
An abstract reasoning test shows you figures and asks what rule connects them. There is no vocabulary and no arithmetic, which is exactly why employers like it: it is meant to measure how quickly you find a pattern rather than what you already know. It usually arrives as one paper in a battery alongside verbal and numerical reasoning.
The whole test rewards one habit: go feature by feature — shape, count, fill, rotation, position, size — instead of staring at the picture as a whole. Almost every question changes one or two features by a constant amount and freezes the rest, and almost every trap answer changes a feature that was supposed to stay frozen.
The questions below are original and drawn as inline diagrams. Every figure is also described in words, so the questions remain solvable if images do not load. When you want the same skills against a clock, the inductive reasoning test linked below runs 15 questions in 15 minutes with a score report.
Three question types, five worked examples
Sequences: find the one feature that moves
3 problems
Work feature by feature rather than looking at the whole picture: shape, number of elements, fill, rotation, position, size. Almost every sequence changes one or two of those by a constant amount and deliberately freezes the rest. Naming which features are frozen is as important as naming which move, because the most common distractor changes a frozen feature — it looks like 'something is different', which is exactly the impression a hurried solver is running on.
How it looks: "Which figure comes next?" — three or four frames, then a question mark
Q1
The figures show a triangle, then a square, then a pentagon. Which figure comes next?
AA hexagon (six sides)
BAnother pentagon
CA circle
DA filled pentagon
▶Show answer & explanation
Answer: A. A hexagon (six sides)
🐱 One feature changes and it changes by a constant amount: the side count goes 3, 4, 5, so the next is 6. The filled pentagon is the classic distractor — it changes a different feature (fill) while freezing the one that was actually moving, and it catches people who spot 'something changes' without pinning down what.
Q2
A line inside a circle turns by the same amount in the same direction at each step: it points right, then down, then left. Which figure comes next?
APointing up
BPointing right again
CPointing down-left (45° past left)
DPointing down
▶Show answer & explanation
Answer: A. Pointing up
🐱 Each step is a quarter-turn clockwise: right → down → left → up. The 45° option is the trap worth naming: it keeps the direction of travel but breaks the constant size of the step, and rotation questions are always testing both at once. Options that repeat a frame already shown can be eliminated instantly — a sequence that returns to a previous state would have to say so.
Q3
The three frames are: an empty box, then a box with one dot, then a box with two dots. Which figure comes next?
AA box with three dots
BA circle with three dots
CAn empty box again
DA box with four dots
▶Show answer & explanation
Answer: A. A box with three dots
🐱 Only one feature is moving here — the dot count, 0, 1, 2 — and the shape is deliberately held constant to test whether you check the frozen features instead of assuming they must change too. The four-dot option is the 'skipped a step' error and the circle option imports a change that never happened. When a sequence looks too simple, the question is usually testing exactly that suspicion.
Matrices: one rule per direction
1 problem
Matrix questions run two independent rules at right angles: one applies as you move across a row, another as you move down a column. The missing cell must satisfy both. Candidates who check only the row get an answer that is confidently, provably half right — and the test writers put exactly that option in the list every time.
How it looks: A grid with one cell missing — read across and down before choosing
Q4
In this 2×3 grid, read across the rows and down the columns before answering. The top row is a box with one dot, two dots, three dots. The bottom row starts with a circle containing one dot, then two dots. What belongs in the empty cell?
AA circle containing three dots
BA box containing three dots
CA circle containing four dots
DAn empty circle
▶Show answer & explanation
Answer: A. A circle containing three dots
🐱 Two rules run at right angles: moving right adds one dot, moving down changes the outer shape from box to circle. The missing cell is therefore a circle with three dots. This 'one rule per direction' structure is what matrix questions almost always use, and it is why you should read both a row and a column before choosing — the box-with-three-dots option is right on the row rule and wrong on the column rule, which is exactly how the trap is built.
Odd one out: find the shared rule first
1 problem
The instinct is to hunt for the figure that looks different, and it fails because several figures usually look different in some way. Invert the task: find the property that exactly four of the five share, then the fifth falls out by itself. A rule that excludes two figures is the wrong rule, and a rule that excludes none is a description rather than a rule.
How it looks: "Which figure does not belong with the others?"
Q5
Four of these five figures share a property that the fifth does not. Which one is the odd one out? (The figures are: a triangle, a square, a pentagon, a hexagon, and a circle.)
AThe circle — it is the only figure with no straight sides or corners
BThe triangle — it has the fewest sides
CThe hexagon — it has the most sides
DThe square — it is the only one with four equal right angles
▶Show answer & explanation
Answer: A. The circle — it is the only figure with no straight sides or corners
🐱 The rule has to isolate exactly one figure. 'Has straight sides and corners' splits four against one; the circle is out. Every wrong option here is a true statement — the triangle really does have the fewest sides — which is what makes odd-one-out questions harder than they look: being distinctive is not the same as breaking the shared rule. Ask 'what do four of them have in common?' rather than 'which one looks different?'
What abstract reasoning actually measures
Abstract reasoning papers contain no words and no numbers, which is the point. Strip out language and arithmetic and what is left is the ability to find a rule in unfamiliar material and apply it — the thing employers are really screening for when they cannot assume a shared education or first language.
That also makes it the section candidates prepare for worst. There is no syllabus to revise. What there is instead is a short, finite list of dimensions along which figures can vary, and the whole test is combinations of them. Learn the list and the papers stop feeling arbitrary.
Shape — how many sides, which figure, whether it changes identity between frames.
Count — of shapes, sides, dots, lines, intersections.
Rotation — of the whole figure or of one marked element.
Fill or shading — solid, outline, hatched, and whether it alternates.
Position — which corner, which order, whether elements swap places.
Size — growing, shrinking, or alternating between two sizes.
Almost every question in every abstract paper is one to three of those six changing together. When a figure looks incomprehensible, you are not missing a seventh dimension — you are looking at three of these at once.
Scan in a fixed order, every single question
The single habit that separates high scorers is boring: they check the dimensions in the same order every time instead of looking at the whole figure and waiting for a pattern to announce itself.
Waiting does not work, because attention goes to whatever changed most dramatically, and test writers put the loud change in as a decoy while the rule lives in a quiet one. A fixed scan order defeats that by construction.
For each dimension ask one question: did it change from frame 1 to frame 2, and if so, did the same change happen from frame 2 to frame 3? A dimension that changes by the same amount twice is part of the rule. A dimension that changes by different amounts is not automatically an error: the step itself may be growing, as in a count going 1, 2, 4 or a rotation going 45°, 90°, 135°. Check for a pattern in the differences before assuming you misread the direction. A dimension that never changes is not the rule, however conspicuous it looks.
The scan takes about fifteen seconds once it is automatic, which leaves plenty inside a typical 45-second budget to apply the rule and match an option.
Never read the options first
This is the most expensive mistake available on an abstract paper, and it feels like efficiency while you are making it. If you look at five candidate figures before you have named the rule, one of them will look right, and you will then construct a justification for it. That is backwards reasoning and the distractors are designed to catch exactly it.
Do it the other way. Finish the scan, state the answer to yourself as a complete description — 'filled pentagon, top-left, pointer at ninety degrees, four dots' — and only then look for the option matching every part of that description.
When two options match on everything but one dimension, that dimension is what the question is testing. Go back to the frames and check that one property only, ignoring the rest. You will usually find you skipped it during the scan because it was the quiet change.
Matrices are two sequences sharing a corner
A three-by-three grid with the bottom-right cell missing looks like a harder question type. It is not; it is two independent sequences that happen to intersect.
Read the top row left to right as a three-frame sequence and name its rule. Read the left column top to bottom and name that one. They are independent — the row rule knows nothing about the column rule, and trying to find a single rule that explains the whole grid is the standard way to lose two minutes.
Then use the structure: the missing cell can be reached two ways, by applying the row rule to the bottom-middle cell or the column rule to the middle-right cell. Do both. Agreement confirms the answer in the same motion that produces it. Disagreement tells you one of your two rules is wrong, and the suspect is whichever was harder to put into words. This is the only question type on the paper where verification is free.
Odd one out: the rule is usually a relation
Beginners look for the figure that appears different. That fails on well-written papers, because the four conforming figures are deliberately varied in shape, fill and position — everything looks different from everything else.
The rule is normally a relation between two dimensions rather than a property of one: the number of dots equals the number of sides; the figure is filled exactly when its pointer is vertical; the shape sits in the corner matching its count. Pick two dimensions, test the relation across all five figures, and count how many satisfy it. Four out of five means you have found the rule and the fifth is your answer. Five out of five means the relation holds but is not what the question is about. Fewer than four means try another pair.
There are fifteen pairs from six dimensions, but in practice the first two you try — count against sides, fill against orientation — cover most of what these papers set. If forty seconds pass without a relation surfacing, pick the figure that stood out on the dimension you checked last, flag it, and move on.
How abstract reasoning is used and scored
You will meet this paper under several names, which is a genuine source of wasted preparation. Abstract reasoning, diagrammatic reasoning, inductive reasoning, non-verbal reasoning and logical reasoning are used by different publishers for overlapping formats. If you have been told to prepare for one of them, practising any of the others is usually the right move.
Graduate schemes, consultancies and technology employers use it as an early screen, often as the first sitting before anything human sees your application. Scores are converted to percentiles against other applicants for the same role rather than absolute pass marks, and there is typically no penalty for a wrong answer.
Two practical consequences follow, and both favour speed. Never leave a blank, because an informed guess beats an omission at zero risk. And treat finishing as part of the score: if most applicants complete the paper and you leave four questions untouched, your percentile drops even though your accuracy looks fine.
Pacing, and what the wrong answers tell you
Budget 40 to 60 seconds a question depending on the paper, and impose a hard ceiling around 75 seconds. Abandoning a question is a skill worth practising deliberately, because the cost of refusing is not one mark — it is the two easy questions at the end you never reach.
When you review, sort your wrong answers by the mistake rather than by the question type. Four slips cover nearly all of them, and each one has a different fix.
Rule run backwards — you read the frames right to left, or rotated the page in your head. Fix: read the frames aloud in order.
Rule applied once too often — you counted the question mark as a frame. Fix: point at the last given frame before applying.
Wrong dimension changed — you found one rule and assumed the rest stayed constant. Fix: finish the whole scan even after a hit.
Previous frame repeated — the time-out pick. Fix: the 75-second ceiling.
A cluster of one slip type is a habit, not bad luck, and habits are faster to fix than skills. That is why abstract reasoning scores move more with a few hours of the right review than with weeks of extra papers.
A non-verbal aptitude test: you are shown figures and must identify the rule linking them, then apply it. Employers use it because it needs no subject knowledge, so it is used to compare candidates from different backgrounds. It is also called diagrammatic, inductive or non-verbal reasoning depending on the publisher.
How do I get better at it quickly?
Adopt the feature checklist — shape, number, fill, rotation, position, size — and run it in the same order every time. Most people improve fast simply by stopping the habit of looking for a general impression, because the impression is what the distractors are designed to exploit.
Is abstract reasoning the same as an IQ test?
They overlap in format but not in purpose. Employer abstract reasoning papers are scored against a comparison group of applicants for the same kind of role, not converted to an IQ figure.
How long do I get per question?
Typically somewhere between 45 and 75 seconds, depending on the publisher and the level of the role. That is enough time for the feature checklist and not enough time to stare — which is the point.
Are these real test questions?
No. They are original questions drawn for this page, built on the same rule types commercial papers use. That means practising here cannot spoil a real assessment for you.