Pattern Recognition Test

A tile here carries four properties and no others. That closed set is deliberate, and it is what turns a guessing game into a puzzle you can actually solve.

You are never hunting for what to look at. You are working out which of the four is moving, and how many tiles it takes to come back round.

• Shape. A circle, a square, a triangle or a diamond.
• Fill. Solid or hollow, with nothing in between.
• How many. One, two or three copies of the shape in the tile.
• How big. Small, medium or large.

There are six levels, and each runs more rules at the same time. A rule gives one property a cycle, such as circle, square, diamond and back to circle.

1. One property alternating between two values.
2. One property cycling through three.
3. Two properties, on cycles of three and two.
4. Two properties, on cycles of four and three.
5. Three properties at once, on cycles of four, three and two.
6. Four properties at once, on cycles of four, three, three and two.

By level six no two tiles in the strip are alike. Matching a whole tile against one you have already seen stops working, and each property has to be read on its own.

Simon and Kotovsky put numbers on that in 1963. What predicted how often people got a sequence wrong was the number of rules running at once. The length of the cycle mattered as well. How clever the shapes looked did not.

Holzman and colleagues found the same thing again on series completion twenty years later. So the ladder is built out of the two things the research says actually make this harder.

Anything no rule touches holds still for the whole strip. A property that changes is therefore always part of the answer, and one that does not is never a red herring.

The ladder is ours. We chose the six levels and we chose the cycles, and nobody has measured a thousand people on this exact arrangement.

Published norms for sequence puzzles do exist, and none of them fits. Raven's tables describe raw scores on the Progressive Matrices, a different test with a different answer set. Hanging our levels on that curve would be an invented number wearing a citation.

What you can check instead is the arithmetic. Every question offers four tiles, so a guess is right one time in four. On two tries a level falls to a guesser about 44 per cent of the time. All six falling that way happens about seven times in a thousand.

For a number that does have a real distribution behind it, what counts as a good digit span has one.

This page gets confused with a matrix test, and the two ask for different work.

A matrix puzzle lays out a three by three grid with the bottom right cell missing. Rules run across the rows and down the columns at once, and the answer comes from eight options rather than four.

Carpenter, Just and Shell took that test apart in 1990. What separated people was not spotting a rule. It was holding several part-finished goals in mind while checking the next one.

A strip runs in one direction only, so there is far less to hold. This is the simpler machine, on purpose, and the level you reach here says nothing about the grid version.

Sequence puzzles turn up on intelligence batteries, and that is the whole reason this section exists. Appearing on one does not make this page one.

No score here has been checked against any measure of intelligence. Nobody has run that study on this item set, so a claim in that direction would be made up.

It is not a screening test either. Nothing here detects or rules out any condition, and no level is a diagnosis. If something has genuinely changed in how you think or remember, that belongs with a doctor.

What the number is: the hardest arrangement of rules you worked out today, on our ladder, in a browser.

A run stops at the first level you cannot clear. One lapse can therefore end it well below what you would have managed on a better minute.

The settings exist for that. Two tries at each level is the default and forgives a single slip. One try makes the ladder strict, and scores from the two are not comparable.

The time limit is the other lever. Leave it off and you can stare at a strip for as long as you like. The median time on the result card still records how long you took.

Any measurement taken once has this problem. The piece on what a good reaction time is makes the case with a number that has been measured properly.