Pattern recognition: what it is, and what a test of it can honestly tell you
Noticing what fits, what breaks, and what is missing. It is one of the seven domains this site measures, and it is the one most often sold to you as something else.
There is a test on this site. Eighteen grids, one tile in each breaks the rule, no timer.
Nothing is stored or sent. Your answers stay in your browser.
What pattern recognition means here
In our domain model it is the ability to notice that something does not belong — that one item breaks a rule the others follow, that a sequence has a structure, that a piece is missing from an arrangement.
It is deliberately a narrow definition. It is not "seeing the big picture", it is not intuition, and it is not a general capacity for insight. Our methodology page maps each of the seven domains to what the tasks in it actually require of you, and this one requires you to compare things and find the one that is different.
The thing you should know before you take any pattern recognition test online
Most of the pattern recognition tests you will find are IQ tests, or are sold by companies that sell IQ tests.
That is not a slur, it is a description of a real overlap: the classic item type in this space is matrix reasoning — a grid with one cell missing, and you pick the piece that completes it — and matrix reasoning is a core component of several intelligence tests. A site offering you "pattern recognition" is very often offering you an IQ estimate with the label filed off.
We deliberately do not use that task. Our test asks you to find the tile that breaks a rule, not to complete a matrix, and it says on its own page that it is not an intelligence test and produces nothing IQ-like.
We can say something here that almost nobody else answering this question can: we have no commercial interest in the answer being yes. Is a pattern recognition test an IQ test? →
What moves a pattern recognition score that is not pattern recognition
- The screen. Fine visual differences on a small or dim display are harder to see. That is optics, not cognition, and any honest test of this domain has to admit it.
- Colour vision. Tasks that vary fill or hue are not neutral. Roughly speaking, if a task can be solved by seeing a colour difference, then it partly measures whether you see that colour difference.
- Familiarity with the item type. Anyone who has taken aptitude tests before has seen these shapes. That is practice, and practice inflates scores on this kind of task more than most.
- What "odd" was defined to mean. If a task never tells you the rule, then part of what it measures is how quickly you guess which kind of rule is in play. Change the mix of rules and you change the score.
Why two pattern recognition tests give you different numbers
Because they are different tasks with different rules, different item counts and different scoring — and none of that is disclosed on most of them.
Our test scores one point per grid where your first choice was the odd tile, out of eighteen, with no weighting and no adjustment. A test that weights harder items, or allows retries, or uses a timer, is measuring something else and its number is not comparable to ours. Neither number is more correct. They are answers to different questions.
Why your odd one out score will not match another site's →
What we can tell you about your own score
Very little, and we would rather be clear about that than dress it up. Our reference sample for this task is 0 people. We store no results — not yours, not anyone's — so there is no distribution to place you in and no percentile we could compute even if we wanted to.
What a score here supports is narrow: what you did on this task, on this screen, today, and a starting point for comparing yourself with yourself later under conditions you can name. The instrument register states the rung, the sample size and the version for every test on this site, and what it would take to change any of them.
What we do not know
- How people score on our task. Nobody's result has ever reached us. This is stated as 0 rather than left vague.
- Whether performance here relates to anything outside the task. We have not measured that and we are not going to assert it.
- Whether our band boundaries are meaningful. They come from the structure of the grids, not from data about people, and the test page says so where the bands are published.