Skip to main content
Back to Blogs
Tips8 min read

Abstract Reasoning: The Eight Pattern Types That Cover Almost Every Question

Abstract and inductive reasoning questions reuse a small set of underlying rules. Learn the eight families and a fixed inspection order, and most items become recognition rather than discovery.

1
Scan the rows
Left to right across each row, looking for something that changes consistently. Most items resolve here and it costs five seconds.
2
Scan the columns
Top to bottom. Matrix items frequently carry one rule along rows and a second down columns, and finding both is what separates the two remaining options.
3
Then count, rotate and compare
Only if the first two passes fail. Count elements, check for rotation and reflection, then look at the answer options for what they have in common. Options are information: a rule none of them tests is not the rule.

Abstract reasoning looks like it could contain anything, which is why candidates stare at items rather than working them. It cannot. Item writers build from a small set of underlying rules, and once you can recognise them, most questions become a matter of identifying which family you are in rather than discovering a pattern from nothing.

A teaching device, not an official taxonomy
The eight families below are a way of organising what turns up in practice, not a scheme any publisher endorses or publishes. Different tests emphasise different families, real items frequently combine two, and the boundaries are ours rather than anyone else's. Use it as a checklist to run through, not as a claim about how tests are built.

The eight families

Family 1
Rotation
An element turns by a consistent amount per step: 45, 90 or 180 degrees. The classic trap is a shape whose rotation is invisible because it is symmetrical, so look for an asymmetric marker such as a dot or a notch and track that instead of the whole shape.
Family 2
Reflection
A mirror image rather than a rotation, and the distinction is the whole item. A mirrored asymmetric shape can never be rotated back into the original, no matter how many turns you allow. Spotting handedness early ends the question.
Family 3
Counting
The number of elements, sides, intersections or enclosed areas changes in sequence. Count something specific rather than eyeballing quantity. Enclosed areas are the most commonly used and the least commonly noticed.
Family 4
Positional movement
An element travels through the grid: one cell per step, around the perimeter, along a diagonal, or wrapping from one edge to the other. When the path seems to break, check for a wrap rather than abandoning the rule.
Family 5
Alternation
Two states flipping back and forth: filled and unfilled, present and absent, large and small. Easy to spot in a sequence, easy to miss in a matrix because it can alternate along rows and down columns simultaneously.
Family 6
Addition, subtraction and overlay
Two cells combine to produce a third. Common variants: everything from both appears, only what appears in exactly one appears, or only what appears in both appears. On a three-by-three matrix the third column is very often the combination of the first two.
Family 7
Attribute association
A property is tied to another property rather than to a position: every triangle is shaded, every shape with four sides is black, arrow direction follows element count. Common in the odd-one-out and rule-detection formats used by Aon and Test Partnership.
Family 8
Progressive change
Something grows, shrinks or accumulates steadily. Size increasing, an element gaining a side each step, shading progressing from empty to full. Distinguished from counting by being continuous rather than discrete.

Why a fixed inspection order beats intuition

When people find abstract items hard, the usual reason is not that the rule is difficult. It is that they are searching an unbounded space with no order, so they notice the wrong feature first, follow it, and have to start again.

The order in the intro block, rows then columns then everything else, is not magic. What makes it work is that it is fixed, so you are never deciding what to look at, only executing a routine. That single change typically halves the time on items you were going to get right anyway, and the time you save is what lets you attempt the ones you would otherwise never reach.

Reading the answer options as data

This is the most underused technique in abstract reasoning, and it is nearly free.

The options tell you which rules are actually being tested. If every option contains exactly three elements, counting is not the discriminating rule and you can stop looking at it. If four options differ only in the orientation of one shape, orientation is the rule and you can ignore everything else in the stimulus.

Ten seconds spent looking at what varies between the options frequently identifies the rule without your ever having to derive it from the sequence. On a test where the average item budget is under a minute, that is a substantial saving.

The three formats you will meet

Complete the sequence

A row of figures with the next one missing. Usually one or two rules operating along the row. The most forgiving format because there is a single direction to check.

Complete the matrix

A three-by-three grid with the bottom-right cell missing. Almost always two rules, one along rows and one down columns. This is the format used by Assessio's Matrigma, by classic matrix tests, and by a great many graduate reasoning batteries. Checking only the rows is the commonest way to land on a plausible wrong answer.

Odd one out and rule detection

Several figures, all but one obeying a hidden rule. Aon's scales ix is the best-known example: nine objects, eight of which follow the rule, five minutes for twenty items. The approach is different here, because you are looking for what the majority share rather than for a progression. Find the commonality first, then find the violation.

Adaptive delivery, and why the last items feel impossible

Many abstract tests are adaptive: get one right and the next is harder, get one wrong and the next is easier. Over a sitting this converges on the difficulty where you are correct about half the time.

Which means finding the last few items brutal is the normal experience of doing well, and finding the whole test comfortable often means you were converged downwards early. Knowing this in advance is worth a surprising amount, because the moment where a candidate panics is usually the moment the test has found their level. More on how that scoring works in what a good psychometric test score is.

The guessing rule, again

It differs by vendor and it matters more on abstract items than anywhere else, because abstract items are the ones you are most likely to be genuinely stuck on.

  • SHL inductive, Test Partnership inductive, Cubiks abstract: no penalty for a wrong answer, so a blank and a wrong answer score the same. Never leave one empty, and eliminate what you can before guessing.
  • Aon scales ix and cls: marks deducted for wrong answers. If you cannot rule out at least two options, skipping beats guessing, and the arithmetic on that is not close.
  • Adaptive matrix tests such as Adaptive Matrigma: a per-item time limit alongside the overall one, so the worst outcome is not a wrong answer but a single item consuming the budget for three.

How to practise this specifically

  1. Untimed first, with the families in front of you. Work twenty items with the checklist visible, naming the family for each one. The goal is recognition, not speed.
  2. Then timed, checklist away. The families should now surface without being consulted. If they do not, go back to step one for another twenty items.
  3. Review only the ones you got wrong, and name the family you missed. Almost everyone has one blind family, most commonly reflection or attribute association. Finding yours is worth more than another hundred mixed items.

Our SHL inductive, Aon scales ix and Adaptive Matrigma drills cover all three formats at their real pace and with their real marking rules, which is the part that decides how you should be playing them.

Drill the families until they are recognition

The gain here is real and it comes from repetition. Once a rule family is recognised rather than derived, an item takes ten seconds instead of forty.

Practise Abstract Reasoning