Matrix Reasoning Rules: How to Read Grid-Based Logic

Matrix questions look wildly different on the surface, but the underlying logic is surprisingly repetitive. Learn a handful of rule families and a nine-cell grid becomes much less mysterious.

Matrix reasoning rules are the relationships that make the rows or columns of a visual grid behave consistently. One feature may increase, another may rotate, shapes may be distributed so each appears once, or two cells may combine to produce a third.

The important thing is to separate what the puzzle looks like from what operation it uses. A progression can use circles, arrows, colors, line lengths, or numbers. The pictures change; the relationship does not.

If you want a solving workflow rather than a rule catalog, see How to Solve Matrix Reasoning Questions. For worked problems, use Matrix Reasoning Examples.

Rule 1: Constant — One Feature Stays the Same

The easiest rule to overlook is often the one that does nothing.

A row might contain three different objects that all have the same color. Another row may use different colors but keep the same number of elements. The stable attribute is part of the logic even though your eye is naturally drawn to whatever changes.

In their classic cognitive analysis of Raven's Progressive Matrices, Carpenter, Just, and Shell identified constant in a row as one of the recurring relations used in the test. Their broader analysis showed that matrix items can often be described using a relatively small set of abstract relations rather than a different trick for every question.

Rule 2: Progression — A Feature Changes by a Regular Step

Progression means that some attribute changes consistently across cells. Common versions include:

• one dot → two dots → three dots
• small → medium → large
• 0° → 45° → 90°
• left → center → right

The step does not have to be numerical in the everyday sense. Orientation, position, size, and quantity can all progress.

Recent work on systematically generating Raven-like matrices still treats progression as a core rule family. A 2026 paper describing the matRiks rule-based matrix framework organizes matrix construction around clearly defined transformation rules spanning visuospatial features and logical operations. The continuity with older Raven analyses is useful: the visual clothing changes, but many of the underlying operations remain familiar.

Rule 3: Distribution — Every Value Appears Once

Distribution rules are easy to mistake for motion. Imagine three shapes: circle, triangle, square. The first row is circle–triangle–square. The second is triangle–square–circle. The third begins square–circle–?

You could describe several individual movements, but the cleaner rule is that every row and column contains all three shapes exactly once. The answer must therefore be a triangle.

Carpenter and colleagues called a common version distribution of three. The same logic can apply to color, shading, position, line direction, or number. Once you recognize distribution, stop tracking each object as though it has its own journey. Ask which value is missing from the set.

Rule 4: Addition — Two Cells Combine into the Third

In an addition rule, components from two cells are superimposed or otherwise combined in the third.

If the first cell contains a vertical line and the second a horizontal line, the third may contain a plus sign. If one cell contains a circle and another a triangle, the result may contain both.

Modern rule taxonomies often describe closely related operations using set logic. The matRiks framework, for example, includes operations analogous to union or superimposition, where features from earlier cells are carried into a resulting cell.

The practical question is simple: does the third cell look like the first two laid on top of each other?

Rule 5: Subtraction or Cancellation — Something Disappears

Subtraction is the mirror image of addition. Instead of asking what the first two cells contribute, ask what is removed.

One version removes the features in cell two from cell one. Another keeps only features that are different between the first two cells, causing shared components to cancel. In logical language, that second pattern resembles an exclusive-OR operation: keep what appears in one source or the other, but not in both.

This rule family is a common source of plausible distractors because an option may correctly preserve most of the shapes while failing to remove the overlapping component.

Rule 6: Transformation — Rotate, Reflect, or Move

A transformation changes the spatial state of an element while preserving its identity.

An arrow might rotate 90 degrees clockwise in every cell. A symbol may alternate between its original and mirror image. A dot may move around four corners in a predictable cycle.

Transformation rules are especially easy to confuse when two dimensions change at once. A shape can rotate while also gaining one dot per step. The safest approach is to track the transformations separately: orientation +90°; quantity +1.

If rotations are the part that slows you down, Cognitive Train's Mental Rotation Test isolates that neighboring spatial skill more directly.

Rule 7: Alternation — Two States Take Turns

Some matrix rules are cycles rather than progressions: filled–empty–filled, black–white–black, left–right–left.

Alternation becomes harder when it is staggered across rows or combined with another feature. For example, shading may alternate while the number of shapes increases. If you notice only one of those patterns, a distractor satisfying the other can still look convincing.

Hard Matrices Usually Layer Rules

The biggest jump in difficulty comes when the puzzle stops asking you to identify one relation and starts asking you to coordinate several.

Research on matrix reasoning often describes this in terms of relational complexity: more dimensions or relations must be considered jointly to reach the solution. In an fMRI experiment using problems adapted from Raven's Progressive Matrices, Christoff and colleagues compared 0-relational, 1-relational, and 2-relational problems. The 2-relational condition required participants to integrate multiple relations simultaneously, creating greater relational demands than the 1-relational condition.

That is why a difficult matrix may require something like:

• shape is distributed across each row,
• orientation rotates down each column,
• quantity progresses from left to right.

The missing cell is the intersection of all three rules.

If holding those simultaneous constraints is the main difficulty, Cognitive Train's N-Back Test gives you a separate look at working-memory updating rather than matrix logic itself.

How to Decide Which Rule to Test First

Start with the cheapest explanations. Check constant features, simple progression, alternation, and distribution before inventing a complicated operation. Then test addition/subtraction and spatial transformations.

A candidate rule should explain more than the one row containing the blank. Use complete rows and columns as evidence. If your explanation works beautifully once and fails everywhere else, it is probably a coincidence.

For strategy—when to construct an answer, when to eliminate choices, and how to track multiple constraints—continue to Matrix Reasoning Strategies.

Practice Reading the Rules

Use the embedded Matrix Reasoning Test below with a specific goal: before answering, name the rule family you think is operating. You do not need perfect terminology. “Rotation plus number increase” is enough if it accurately describes the grid.

That habit forces the pattern into an explicit form. It also makes wrong answers more informative: if your chosen option fails, you can ask whether the rule was wrong or whether you applied the right rule incorrectly.

Try the Full Matrix Reasoning Test

Cognitive Train's full Matrix Reasoning Test gives you a dedicated set of grid-based reasoning problems. For nearby formats, the Abstract Reasoning Test and Inductive Reasoning Test broaden rule detection beyond matrices.

For the larger family of visual-rule problems, browse the Pattern Recognition section or read about Progressive Matrices. Cognitive Train's brain tests collection covers related reasoning, memory, attention, and speed tasks, while the homepage brings together the broader set of brain training and cognitive training tools.