Number Sequence Test

3, 6, 12 is doubling, and the next term is 24. It is also a step that grows by three, and then the next term is 21. Both readings fit every number on the screen.

That is true of most short rows, and it is the reason this test works the way it does. Three terms are not a head start. They are a chance to commit early if you are willing to be wrong.

Simon and Kotovsky described the job in 1963 as two jobs rather than one. First you find how long the pattern is before it repeats. Then you find what it does inside that span.

A short row hides the first of those. Four terms of a woven series look like one series behaving oddly. Six terms look like two series behaving normally.

Answer wrongly and the row grows by one. The blank slides along, and you answer again on a longer series.

That is not generosity. The term you have just been handed is the term your score was trying to do without. A guess therefore costs you what asking would have cost.

The quiet button under the box does the same thing without the guess. It is there because somebody who can see a series is still open should not have to type a number they know is wrong.

• Three terms. The floor, and the best result available on any series.
• Four to six terms. Where the two-step and woven series tend to settle.
• Seven terms. Everything this test will ever put on the screen.
• Eight terms. What a series you never cracked counts as.

There are twelve series in the set, and they are not equally deep. Some fall to nothing more than a step that shrinks by one. 100, 90, 81 hands you that in three terms.

Others cannot settle before the fifth. 7, 50, 9 is two rows woven together, and three terms show you one of them twice and the other once.

The spread is deliberate. If every series settled at three terms the score would be a constant, and a measurement that returns the same figure for everybody measures nothing.

Holzman, Pellegrino and Glaser put this to the test on number series in 1983. What separated people was how quickly they settled on one rule out of the several that fitted the opening.

Every series names its rule once the item closes, in the words somebody would say out loud. That can be switched off in the settings above.

There is no percentile under your score, and one is not going to be invented for it.

Number series do appear on graduate aptitude papers, and those publishers hold norm tables. Each table belongs to its own item set, sat under its own time limit, and none of those items is ours.

The research does not fill the gap either. Kotovsky and Simon worked with letter series of a fixed length, and the length is the one thing this test moves.

The arithmetic is checkable instead. Three is the floor, eight is the ceiling, and a run of pure guessing lands near eight rather than near three.

For a figure that does have a measured spread behind it, what counts as a good digit span has one.

A low count means you read these twelve series on less evidence than a higher count needed. It means nothing past that.

No score here has been checked against any measure of intelligence, because nobody has run that study on this item set. A claim either way would be made up.

The count describes how you handled these twelve series today.

Practice will move the number, because twelve is a small set and you will start recognising them on sight. That is improvement on these twelve series. Whether puzzle practice carries any further is what the piece on brain training games goes through.