Skip to content

Odd One Out Across Words, Numbers and Shapes

An odd one out puzzle asks for a dimension on which one item differs. Why several answers are often defensible, and how a setter removes the ambiguity.

By Published Updated 6 min read

Reign Creative LLC makes IQ Test & Brain Training Games, which this article links to, so we have a commercial interest in it. How we write and check articles.

Key takeaways

  • An odd one out item does not ask which item is different — everything is different from everything. It asks for a dimension on which all but one of the items agree.
  • Because any small set of items shares many dimensions, several answers are usually defensible, and the intended answer is the one whose dimension the setter fixed in advance.
  • For numbers, run a fixed checklist: parity, primality, perfect squares and cubes, digit count, digit sum, and membership of a familiar family such as the triangular numbers.
  • For words, the useful dimensions are category, part of speech, morphology and spelling or sound — and a single set can split differently under each.
  • For shapes, definitions do the work, and definitions vary by country, so a quadrilateral puzzle can have a different answer in British and North American usage.
  • Our own app, IQ Test & Brain Training Games, lists vocabulary games on word meaning and verbal relationships and logic games built on patterns, but its listing does not describe an odd-one-out mode as such.

An odd one out puzzle is not really asking which item is different. Every item in any set differs from every other in countless ways. It is asking you to find a dimension — a single property — on which all but one of the items agree. Once you put it that way, the notorious feature of the format stops being a flaw and becomes a structural fact: because a small set of items shares many dimensions, several answers are usually defensible, and the "right" one is whichever dimension the setter fixed before writing the item.

That makes the format unusual among puzzle types. A sequence puzzle at least pretends to a unique continuation; an odd one out item is openly a search through unstated criteria. Knowing which criteria to search is most of the skill.

The dimensions worth testing, by medium

Different media reward different checklists. The value of a checklist is that it finds the split that gives four against one, rather than the first difference you happen to notice.

Numbers. Parity, primality, perfect squares and cubes, digit count, digit sum, divisibility by small factors, and membership of a well-known family such as the triangular or Fibonacci numbers. Divisibility is the most productive of these and the least often checked; the rules for testing it quickly are in our guide to mental division.

Words. Semantic category, part of speech, morphology (is one a compound, a plural, a past participle?), and spelling or sound (letter count, a silent letter, a sound no other member has). A set that splits cleanly by meaning will often split differently by spelling, and the setter has to say which is intended.

Shapes. Symmetry, closure, number of sides, number of right angles, and formal definitions. Definitions are where shape puzzles get genuinely slippery, for reasons below.

Three worked sets, each with more than one defensible answer

Numbers: 4, 9, 16, 25, 36.

All five are perfect squares, so that dimension does not separate anything. Work the checklist and three separate answers appear.

  • 36 is the only triangular number. The triangular numbers run 1, 3, 6, 10, 15, 21, 28, 36; none of the other four appears in that list.
  • 25 is the only one ending in 5.
  • 16 is the only fourth power: 16 = 2 x 2 x 2 x 2. Note that 4 is a power of two but not a fourth power, so it does not compete here.

Parity gives nothing, because 4, 16 and 36 are even while 9 and 25 are odd — two groups, not four against one. Digit count gives nothing either, because 4 and 9 are both single digits. Three of the five numbers can be justified as the answer, which is the point. (If squares as a family are the part you want more of, the shortcuts for squaring in your head are in multiplication shortcuts explained.)

Words: copper, iron, gold, bronze, silver.

  • Bronze is the only one that is not a chemical element. It is an alloy, principally of copper and tin.
  • Iron is the only one outside the copper–silver–gold group of the periodic table, the three metals traditionally used for coinage.

Both answers are correct, and they use different kinds of knowledge — one chemical, one taxonomic. A solver who reaches "bronze" and stops has answered the question; a solver who reaches "iron" has also answered the question. Only the setter's intention separates them.

Shapes: square, rectangle, rhombus, parallelogram, trapezium.

  • Under British usage, a trapezium has exactly one pair of parallel sides, so it is the only one of the five that is not a parallelogram.
  • Square is the only one that is both equilateral and equiangular — the only regular figure in the list.

There is an extra wrinkle here that is honest to flag: in North American usage the shape with one pair of parallel sides is called a trapezoid, and "trapezium" means something else. A shape puzzle can therefore have a different answer depending on which convention the setter learned, which is a good argument for stating definitions in the item.

Why the ambiguity is structural, not sloppy

Two bodies of work explain why this keeps happening.

The first is the psychology of categorisation. Edward Smith and Murray Grossman, reviewing the evidence in Neuroscience and Biobehavioral Reviews, argue that people categorise through more than one system: a rule-based route that attends selectively to a defining attribute, and a similarity-based route that compares an item against remembered examples. The two can reach different verdicts on the same set. Ask for a rule and people give a categorical answer; prompt similarity and membership becomes graded. An odd one out item does not say which mode it wants, so two solvers can apply different machinery to the same five items and both be reasoning properly.

The second is the philosophy of similarity. The Stanford Encyclopedia of Philosophy's entry on analogy and analogical reasoning makes the point that evaluating any similarity claim depends on identifying which similarities are relevant to the conclusion being drawn. Relevance is not in the items; it is supplied by the question. Remove the question and you remove the only thing that could have made one answer better than another.

This is also why the format is used in teaching the way it is. NRICH, the mathematics education project at the University of Cambridge, publishes a sorting activity for young children in which nine houses can be grouped by door colour, roof colour, window count, floor count and more, and its teachers' notes say plainly that any way of sorting is valid so long as the categories are explained. The explanation, not the selection, is the assessed part.

How a setter removes the ambiguity

If you write these items rather than solve them, three techniques do most of the work.

  1. State the domain. "Which is the odd one out by meaning?" cuts the search space in half and costs nothing.
  2. Make one dimension dominate. Choose items that agree strongly on everything except the intended property. The five squares above fail this test on purpose; a set of four triangular numbers and one square would pass it.
  3. Mark the reason, not the letter. Accepting any answer accompanied by a stated criterion turns an ambiguous item into a well-posed one, and it is what the classroom versions of this format do.

Solvers can use the same list in reverse. If you find two answers, you have not failed — you have found an under-specified item, and naming both criteria is a better response than guessing which one the setter meant.

What our app has, and what it does not

Being direct about it: the Google Play listing for IQ Test & Brain Training Games, our own app, does not describe an odd-one-out mode. What it does list nearby are vocabulary games on synonyms, antonyms, word meaning and verbal relationships, logic games built on analogies, patterns and rule discovery, and an IQ-style test whose areas include logic, patterns, vocabulary and spatial reasoning; the listing calls the test's results estimates based on app activities, not clinical or professional assessments. The app's games, tests and progress tracking are described on the IQ Test & Brain Training Games app page.

The ambiguity described here has a close cousin in figural puzzles, where the question is which transformation is changing rather than which property is shared; that is covered in visual pattern puzzles. The related task of naming a relationship before testing options is the subject of analogy questions, and the general procedure for rule-finding starts with number sequence puzzles. Everything in this subject is collected in the logic and pattern puzzles hub, and our wider writing on practice and reasoning is in the education and brain training section.

Good to know

Frequently asked questions

Why do odd one out puzzles often have more than one answer?
Because the question leaves the dimension of comparison unstated. Any five items agree on some properties and differ on others, so picking a different property picks a different outlier. A set of perfect squares can single out one number because it is the only triangular one, another because it is the only one ending in five, and a third because it is the only fourth power. All three are correct under their own criterion.
What should I check first in a number odd one out?
Work a fixed checklist rather than staring. Parity first, then primality, then whether the numbers are perfect squares or cubes, then digit count, then digit sum and divisibility, then membership of a familiar family such as the triangular or Fibonacci numbers. A checklist beats inspiration because it finds the dimension that splits four against one rather than the first difference you happen to notice.
How do good setters make an odd one out unambiguous?
Three ways. They state the domain, so you know whether to think about spelling or meaning. They choose items that agree strongly on one dimension and vary freely on the rest, so only one split gives four against one. And they accept any answer with a stated reason, which is how the format is usually used in classrooms, where the explanation is the point rather than the letter.
Does the app have odd one out puzzles?
The Google Play listing for IQ Test & Brain Training Games does not describe an odd-one-out mode. The nearest things it lists are vocabulary games on synonyms, antonyms, word meaning and verbal relationships, logic games built on analogies, patterns and rule discovery, and an IQ-style test covering logic, patterns, vocabulary and spatial reasoning. The listing calls the test's results estimates based on app activities, not clinical or professional assessments.

Sources

  1. Multiple Systems of Category Learning (opens in a new tab)Neuroscience and Biobehavioral Reviews, via PubMed Central — accessed
  2. Sort the Street (opens in a new tab)NRICH, University of Cambridge — accessed
  3. Analogy and Analogical Reasoning (opens in a new tab)Stanford Encyclopedia of Philosophy — accessed

About the author

— The independent app studio that builds the apps covered on this site and publishes every article here. On Google Play it publishes as Reign Collective Apps.

We use AI tools to help research, draft and edit our articles. Each article lists the sources its factual claims rely on, and details about our apps come from their live Google Play listings. No article on this site has been reviewed by an independent expert unless the reviewer is named on it.

Spotted an error? Tell us — how corrections are handled is set out in our editorial policy.

Keep reading