Number Memory Statistics

How this compares with the classic span figures

Laboratory digit span tasks usually report averages a little below what a test like this one produces, and the reason is procedural rather than mysterious. A clinical span task reads digits aloud at one per second and stops after two failures at a length; this test shows them written, all at once, and lets you group them however you like.

Seeing the digits rather than hearing them is the larger of the two differences. A written number can be chunked visually into pairs and threes at a glance, which is a strategy that a spoken sequence delivered at a fixed rate largely denies you.

So the distribution here is best read as the span of a self-paced, visually presented number in a browser. That is a real thing to measure and it is what the test measures, but it is not interchangeable with a figure from a clinical instrument.

The shape, and the bump that strategy puts in it

The bulk of attempts fall in a narrow band, which is what a capacity limit looks like when it is real. Unlike reaction time, the distribution is close to symmetric around its peak, because there is no equivalent of a slow round dragging a tail out to the right.

What the curve does have is a secondary rise further up, where deliberate chunkers sit. The gap between remembering digits and remembering groups of digits is large enough to show as a change in slope rather than a smooth continuation.

Below the peak the distribution falls away quickly. Very short spans are rare and, as everywhere on this site, a very low result is more often an abandoned attempt or a mistyped entry than a measurement of anybody.

Why an average span is a weak thing to quote

A mean of eight point six digits sounds precise and tells you almost nothing, because the interesting question is nearly always positional: what share of attempts reach eleven, or what span puts somebody in the top quarter. Those are the numbers written out below the chart.

The mean is also sensitive to the top tail in a way that misleads. A handful of very long spans from people using a strong grouping strategy lifts it above the level most attempts actually reach.

The median and the quartiles are given first for that reason, and the sample size is stated alongside them. A percentile without a sample size behind it is a claim rather than a statistic.

The reverse span is the second figure on this page

Reverse writes a length too, and it is a different achievement from a forward span of the same digits. Those attempts sit on the second chart so the cost of walking the number backwards stays visible.

The axis is still whole digits. A miss on the first reverse is stored as one, not zero, so the left edge is a finished attempt rather than an empty.

Two hundred attempts before that second curve as well. Until then the honest display is the count.

This is a measurement toy, not a medical or psychological test. No score here says anything about your health, attention or cognition, and it is not a screening tool of any kind.

Questions

What is the average number of digits people remember?

On this test most attempts land in the high single figures to low tens, which is somewhat above classic spoken digit-span averages because the digits here are shown in writing and can be grouped visually.

Why do browser results differ from clinical digit span scores?

Different procedure. Clinical tasks present digits aloud at a fixed rate; this one shows them written and self-paced, which allows chunking strategies that lift the score.

Is a high digit span a sign of intelligence?

It is one narrow capacity measured under one particular procedure, and much of the variation between people on this test comes down to whether they group the digits. It is not an assessment of anything broader.

Is a seven here the same as a seven on reverse?

No. Same unit, different task, different chart. Reverse is the second figure on this page.