Matrix Reasoning Task

It is a visual reasoning task built from matrices of abstract figures. One cell of the matrix is missing, and the task is to work out which figure belongs there. Solving an item means detecting the relations that run through the matrix — a shape that changes systematically across columns, a colour that is fixed within each row, a size that increases from left to right — and combining them into a single answer. No language, arithmetic or prior knowledge is required, which is why the format is used across cultures and age groups.

No. This page is an experiment, and it produces no IQ score. An IQ or percentile number requires a norm sample: a large, defined group whose performance defines what "average" means for each age. This experiment has no such sample, so your accuracy here can be described but not translated into a standard score. The item bank whose structure we follow makes the same point about itself, which is why we describe the task rather than ranking you on it.

Each form contains 24 items and takes roughly eight minutes, though there is no time limit and you can pause between items for as long as you like. Three interchangeable forms are available, with the same structure but different items, so running the experiment again is not simply a repeat of the same puzzles.

It counts how many independent relations have to be combined to determine the missing cell. If only the size varies across the matrix, that is one dimension. If size, colour and rotation all change according to their own rules, that is three. The label is a property of the item's construction, so a high number means the item is built to demand more integration, not that you did badly.

Because difficulty here is structural. Adding a second rule does more than add a second thing to notice: the rules interact, and a distractor that satisfies one relation may violate another. Items near the end of a form combine several relations at once, and they are designed to be failed by most people rather than solved by everyone — the form is meant to keep discriminating as far up the ability range as it can.

Not during a run. Feedback between items would turn the task into a learning exercise and change what the later items measure, so the correct option is not revealed as you go. The results screen shows accuracy overall, accuracy split by the number of relational dimensions, and accuracy split by rule family, which is enough to see which kinds of relation you handled well without turning the page into a tutoring tool.

The two share a format — a matrix of abstract figures with one cell missing — but they are different instruments and this experiment uses neither the items nor the norms of Raven's tests. If you are looking for the published Raven's versions, their structure, scoring and interpretation are covered separately, and any real percentile there also depends on a norm sample rather than on an online practice page.

No. The task runs entirely in your browser: the items are fetched once as a data file, your choices and response times are computed on your device, and they disappear when you leave or reload the page. Nothing is submitted, and there is no account, ranking or leaderboard attached to the results.

No. The items are identical and only the three colour values are substituted. Both sets use a palette chosen so that the colours stay distinguishable for viewers with the common forms of colour vision deficiency, which matters because colour is one of the relations an item can be built on. If two colours are hard to tell apart on your screen, switching sets is usually the quickest fix.

Familiarity with the rule families does help, and that is the honest version of the claim. Practising matrices is mostly practising recognition: learning that a set of figures can differ by count, by rotation, by fill, by size, or by a permutation across rows and columns, and that these can be combined. It is a trainable skill in the same way that any pattern-recognition skill is, but gains are largest for people who have not met the format before, and they should not be read as a change in general reasoning ability.