How to Solve Classification Problems: Spot the Odd One Out in 3 Lenses

Classification problems are solved by decomposing figures one attribute at a time and finding where the shared rule breaks. Staring at them holistically gets you nowhere, but fixing the order in which you check attributes makes your accuracy stable. This article covers three lenses and the concrete procedure.

What classification problems actually measure

Classification is the ability to find the rule shared by a set of figures and identify the one that does not follow it. IQ tests commonly use two formats: 'pick the odd one out,' and the Latin-square type where three attribute values each appear exactly once per row and column.

Carpenter, Just, and Shell (1990), analyzing the rules behind Raven's Progressive Matrices, classified these tasks under the rule type 'distribution of three / distribution of two.' In other words, item writers tune difficulty by deliberately combining multiple attributes — which means that decomposing by attribute will always get you there. You can see worked examples on our classification sample problems page.

Lens 1: Break the figure into four attributes

Figure-based problems draw on essentially four attributes. Check them one at a time, in this order:

• Shape: circle, triangle, square, and so on
• Count: the number of figures or dots
• Rotation: angle of rotation, axis of symmetry
• Fill: solid, outline, hatched

The key is to never look at two attributes at once. Compare all options on shape alone; if they share it, move on to count. Human working memory has a hard limit on how much it can hold simultaneously, and accuracy drops the moment you mix attributes (see working memory).

Lens 2: Pin down the majority rule first

In an odd-one-out item, four or five of the five or six options follow the same rule. What you are hunting for is the majority rule, not the exception.

The procedure has three steps. (1) Compare the first three options and find one shared attribute. (2) Check whether that attribute holds for the remaining options. (3) If exactly one option fails, that is your answer. If two or more fail, the attribute you picked is wrong, so switch to another one. Searching for the exception first makes candidates proliferate; establishing the majority rule narrows things down in a single pass.

Lens 3: For Latin squares, read both rows and columns

When three attribute values are arranged across a 3x3 grid, the rule is that each value appears exactly once per row and once per column. If checking rows leaves you with two candidates, always check the columns too. These items are designed so that overlaying the column constraint resolves the tie.

This 'overlay two constraints' habit is shared with matrix reasoning, and the 3-step method in how to solve matrix reasoning problems transfers directly.

Narrowing down when stuck, and managing the clock

If all four attributes fail to settle it, a compound rule is in play — shape combined with fill, for example. Here it helps to restrict yourself to a subset first: pull out only the triangles, then compare fill within that group.

If it still looks like 30 seconds or more, the right move is to answer provisionally and move on. Most IQ tests get harder toward the end, so fixating on one item costs you later ones directly (see pre-test preparation).

Try it on real problems

Reading a method does not make it stick. BrainRank's free IQ test scores spatial reasoning, pattern recognition, logical reasoning, and classification separately, so your classification score appears on its own in the results. Twenty questions, about 10 minutes — enough to see how well these three lenses worked for you.