Odd One Out Test
Four cells, three of them agreeing on one thing and a fourth that does not. A free online odd one out test — the wordless non-verbal reasoning kind: 12 puzzles, no account, everything generated and scored in your browser. Every answer is explained the moment you give it — right or wrong, you get the rule the three cells shared — and at the end the whole run is listed out: the four figures again, which one you tapped, which one broke the rule, and the decoy that was pulling your eye. The result itself is your count, your mean time and which of the three rules held you up.
Take the test
Three cells agree on one attribute. Tap the one that breaks it. Keys 1–4 work too.
0
of 12 figures found
A clean run. This format is built to be solvable — every item has exactly one defensible answer — so a full sweep is the expected outcome and the interesting number is the time beside it.
Nearly a clean sweep. What usually costs an item here is a decoy: the attribute that varies across all four cells and can never be the rule, but pulls the eye first anyway.
A little over half. Almost always the same cause — answering on the first difference you notice instead of the one shared by exactly three cells. The rules repeat, so a second run tends to look different.
Fewer than seven this time. One twelve-item run in a browser tab says nothing about you in general; it is a warm-up, and the same three rules come back on the next one.
- Odd shape
- — / 4
- Odd count
- — / 4
- Odd fill
- — / 4
Each explanation holds for up to 4.3 s — click the card, or press Space or Enter, to move on sooner.
12 items, easy ones first, each one explained as soon as you answer. There is no time limit — the clock exists only so the result can tell you how long you took. Nothing is uploaded; it all stays in this tab.
Your answers, with the rule for each
Every figure of the run you just took: what you were shown, what you tapped, what the answer was, and the rule behind it. The rules are read off the items themselves, so nothing here is a general description of the format — it is what these 12 were actually built from.
-
Your answer Correct answer
-
Your answer Correct answer
-
Your answer Correct answer
-
Your answer Correct answer
-
Your answer Correct answer
-
Your answer Correct answer
-
Your answer Correct answer
-
Your answer Correct answer
-
Your answer Correct answer
-
Your answer Correct answer
-
Your answer Correct answer
-
Your answer Correct answer
That was a single run. In the app it becomes a short daily session, with a Focus Index to show for it.
Point your camera at the code — the App Store opens.
What an odd-one-out question is
Four figures, one rule, and one cell that does not obey it. The work is to notice what three of them have in common before you can say which one is out.
Psychologists call this classification, and it is one of the standard ways of getting at inductive reasoning: forming a category from instances nobody named for you, then testing a candidate against it. Because not a word appears in the item, it is also the standard non-verbal reasoning question — the figures-only format that school entrance papers and job aptitude tests are built from. Nothing is recalled here and nothing is taught in advance: the rule exists only inside the four cells in front of you, which is why the format travels across languages and schooling so well.
The reason there is exactly one defensible answer is a piece of construction, not a judgment call. One attribute — the shape, how many figures a cell holds, or the fill — takes the same value in three cells and a different one in the fourth. A second attribute varies too, and that one is a decoy: it deliberately takes three different values across the four cells, so no value of it is shared by exactly three and it can never satisfy “three agree, one does not”. Every remaining attribute is identical in all four. The decoy is also what keeps the four cells from being the same picture twice.
The three rules you will meet
12 items to a run, drawn fresh each time from these three and served easy ones first — 4 of each, so no rule can be missed by luck of the draw. Each row below is a real item from the same generator the test uses; the odd cell is marked in color.
-
Odd shape
four outlined crosses.one outlined cross.The odd cell: three outlined circles.three outlined crosses.Three cells use the cross; the odd one uses the circle. How many figures a cell holds varies too — across three different values, so it never singles out one cell.
-
Odd count
four outlined circles.The odd cell: two outlined circles.four solid circles.four half-filled circles.Three cells hold four figures; the odd one holds two. The fill varies too — across three different values, so it never singles out one cell.
-
Odd fill
two outlined diamonds.four outlined diamonds.The odd cell: one solid diamond.one outlined diamond.Three cells are outlined; the odd one is solid. How many figures a cell holds varies too — across three different values, so it never singles out one cell.
The two rungs are not the same height. An odd odd shape is the one most people find without looking for it — a different outline pops out before you have decided to compare anything. The other two (odd count and odd fill) need an actual feature comparison across four cells, which is why they are served second.
Where the format comes from
Classification items are old, and they were built to be wordless on purpose.
Raymond Cattell published the Culture Fair Intelligence Test in the 1940s around a claim that has aged better than most: if a reasoning task carries words, it also carries the vocabulary and the schooling of whoever wrote them. His answer was four subtests made of figures alone — series, matrices, conditions, and classification, which is the one on this page. It was an intelligence test, and it is fair to say so as history; it is not what this page is, and nothing here converts your count into a score of that kind.
The wordless part is why the format survives in modern batteries and why the app can ship it in every locale without translating a single item. It is also honest about its own limits: “culture fair” was a design goal, and the decades since have shown that no test is entirely free of the world it was written in.
How to read your result
Three numbers, and they are worth more together than apart.
-
The count
How many of the 12 you found. With 4 cells per item, tapping at random is worth about 3 of them, so the count only starts saying something above that — and this format is built to be solvable, which is why the bands below sit high.
-
The per-rule tally
Which of the three rules actually cost you something. 4 items turn on the shape, 4 on how many figures a cell holds and 4 on the fill — missing one kind and sweeping the others is a different result from a scatter, even at the same count.
-
The mean time
Seconds per item you got right. Wrong answers are left out on purpose: a fast wrong answer is a fast wrong answer, and folding it in would make guessing quickly look like fluency.
The verdict word under your count is a rough banding and nothing more — the cuts are at 12, 10 and 7 found. They sit higher than on the harder tests in this catalog because this format is the easy end of figural reasoning: a full sweep is the ordinary outcome, so the band exists to give the number a sentence, not to congratulate you.
What this version cannot tell you
-
12 items is short
Long enough to meet all three rules 4 times each and see which one slows you down, not long enough to be a measurement. A run where one lucky tap lands reads a point higher than it should.
-
It is the easy end of the format
One attribute separates the odd cell, and the same three attributes come back every run. Classification items on published batteries stack several attributes at once and get much harder than anything here.
-
Your conditions are in the result
A tab in the background, a phone that buzzed, the end of a long day — all of it lands in the time and some of it in the count. Compare your own runs on the same evening rather than your run to anyone else’s.
-
It is not a diagnosis or a score
Classification items appear inside professionally administered batteries, always beside other item types and read by someone qualified. One format on one page is not that, and nothing here turns your count into a score of any kind.
Questions about the odd one out test
Does it show the answers and explain them?
Yes. Every one of the 12 items is explained the moment you answer it — right or wrong — by naming what the three cells shared and what the fourth did instead. When the run ends the whole thing is listed under the card: the four figures again with the odd one and your own tap marked, plus the decoy — the attribute that also varies, across three values rather than one, which is exactly why it can never be the rule and exactly why it pulls the eye first.
Do I need an account?
No. The odd-one-out puzzles are generated in your browser and scored there too — nothing is uploaded, nothing is stored, and the result is gone when you close the tab. There is no email box anywhere on this page.
Is this an IQ test?
No. It is one item format — the classification format that has appeared on culture-fair intelligence batteries since the 1940s — and it reports only what it actually measured: how many of the 12 odd figures you found, how long you took, and which of the three rules gave you trouble. It does not estimate an IQ, and no single-format test in a browser tab honestly could.
What counts as a good score?
We publish no norms for our version of the odd-one-out test, and anyone quoting a number for a browser test is guessing. Two things are worth knowing. Four cells mean about 3 of the 12 come from chance alone. And only 4 items in a run turn on the shape, the rule most people find without looking for it, so anything above 4 means at least one count or fill item came apart for you. Beyond that, the comparison worth making is with your own earlier runs.
I can see more than one difference. How can there be one answer?
Because only one difference splits the four cells three against one. A second attribute does vary — that is the decoy, and it deliberately takes three different values across the four cells, so no value of it is shared by exactly three. Everything else is identical in all four. The rule is always "three agree, one does not", and exactly one attribute can satisfy it.
How do you solve an odd-one-out puzzle?
Go through the attributes one at a time instead of staring at the four cells as pictures. Shape first: are three outlines the same and one different? Then how many figures each cell holds. Then the fill. Exactly one of those will split the cells three against one, and the moment you find it you are done. On the harder non-verbal reasoning papers several attributes vary at once, and the same routine still works — it just takes longer.
Do the puzzles repeat if I take it again?
The figures do not — every run is generated from the clock, so you get 12 fresh items. What stays the same is the recipe: the same three rules, 4 items of each, and roughly the same easy-to-hard order (neighboring items may swap), which is what makes two of your runs worth comparing.
Twelve figures is a snapshot. The trend is the point.
In NEURA, odd-one-out is one task type inside a daily session of 20, picked for you each morning and pushed toward whatever you keep missing — and after three sessions the Focus Index opens, one score with a 14-day trend behind it. That is the part a single page cannot do.
