Number Memory Test

The famous number, and why it is wrong

Almost everyone has heard that short-term memory holds seven items, plus or minus two. That comes from a 1956 paper by George Miller, and it has become perhaps the most quoted figure in psychology.

Miller's own argument was subtler than the number it produced. He noticed that the limit was not on the amount of information but on the number of chunks — and that people expand what a chunk contains through practice. Later work, particularly Nelson Cowan's, put the capacity for genuinely unrelated items closer to four.

So a test like this is not measuring a fixed slot count. It is measuring how well you can package digits into fewer, larger pieces in the second or two before they fade.

What the good scores are actually doing

Chunking, and it is not cheating — it is the entire skill. Somebody who reads 4 9 1 2 6 3 8 5 as eight separate digits is holding eight things. Somebody who reads it as 49 12 63 85 is holding four, and someone who spots that 1963 is a year is holding fewer still.

Three strategies do most of the work:

• Group in twos or threes. The single highest-return change, and the reason the number is displayed in groups here.
• Attach meaning. Years, ages, house numbers, familiar sequences.
• Say it aloud, or subvocalise. The phonological loop rehearses sound more reliably than it rehearses shapes.

This is also why practice raises the score so quickly and then plateaus: you are learning a technique, and once you have it there is not much more to gain.

How the timing works

The number is shown for roughly one second per digit — three digits for three seconds, ten for ten. A fixed duration would make the early rounds trivial and the later ones impossible, and would end up measuring reading speed as much as memory.

Your score is the last length you got right. Failing at nine digits scores eight, because eight is what you demonstrably held.

Why the score is the last length you got right

Failing at nine digits scores eight, because eight is what you demonstrably held. Some versions score the length you failed at, which flatters everyone by one and makes results incomparable between sites — so it is worth knowing which convention a number came from before comparing it to anything.

There is also no partial credit. Getting eleven of twelve digits right is scored as a failure at twelve, not as eleven, because the task is holding the whole string. A test that awarded partial credit would be measuring something closer to recognition than to span.

Digit span in the wild

The task is old and unusually well studied. Versions of it appear in the Wechsler intelligence scales as a subtest, in dementia screening batteries, and in a large research literature on working memory — usually in two forms: forwards, as here, and backwards, where you repeat the digits in reverse.

Backwards span is consistently shorter, typically by about two digits, and it is the more demanding version because you have to hold the string and manipulate it at the same time. That gap between forwards and backwards is often more informative than either number on its own, which is a good illustration of why a single figure from a browser test should not be over-read.

What it does not tell you

This is not a diagnostic test. Digit span correlates only weakly with working memory measures generally, and barely at all with anything most people mean by intelligence. It is a narrow, trainable skill. Treating a score of six as a verdict on yourself is not something the research supports, and neither is treating a score of twelve as evidence of anything much beyond a good chunking habit.

It is also affected by the ordinary things: tiredness, noise, whether you were rehearsing something else a moment earlier. Compare your own results across a fortnight rather than against a single attempt or against a table.

And because there are no accounts here, everything above is stored only in this browser. Nothing is uploaded, and there is nothing to sign up for.