Number memory test at one digit a second

Around seven is the adult average for forward digit span, and the normal range is wide enough that anything from five to nine is unremarkable. Normative data for the Wechsler scales puts the mean forward span for healthy adults a little under seven, and it drifts down with age by roughly one digit between the twenties and the seventies. A single run is a noisy estimate of your own figure, though: attention, the time of day and how badly you want the round all move it by a digit in either direction.

By turning the digits into something that is not digits. The best-documented case is a college runner in a 1980 study by Ericsson, Chase and Faloon, known in the literature as SF, who reached 79 digits across about 230 hours of practice by mapping three-digit and four-digit groups onto running times — 3492 became a near-record time for the mile. The revealing detail is that when the same man was tested on letters his span dropped straight back to about six. He had not enlarged his memory; he had built a code for one kind of material.

Because the limit is on units held, not on digits, and grouping changes what counts as a unit. This is the point of George Miller's 1956 paper that everyone remembers by its title and misremembers by its content: the magical number seven was about chunks, and Miller was explicit that a chunk could be a digit, a word, a face or a chess position. Three digits recoded as one familiar year or area code cost you one slot instead of three, which is why a string that happens to contain 1999 or 007 is quietly much easier than a random one.

No, the first digit is always 1 to 9. A leading zero would make the length of the number ambiguous the moment you typed it back — 0472 and 472 are the same value and different answers — and rather than argue about that in the comparison, the generator never produces one. Zeros appear freely in every other position, and a run of them in the middle is one of the harder things this test throws at you, because repeated digits give rehearsal nothing to hold on to.

Because the exposure is one second per digit rather than a fixed window, which is the presentation rate the standard clinical subtest uses. Handing a twelve-digit number the same three seconds a three-digit number got would measure how fast you can read, and reading speed is not what the ladder is trying to find. Keeping the rate constant means every round asks the same question about a longer string, so the round you fail on is a span rather than an artifact of the timer.

Because span is defined as the longest string reproduced correctly, and repeated attempts at one length quietly measure something else. Given three tries at nine digits, most people eventually get one, and the figure then describes persistence. The clinical subtest does allow two trials per length and only discontinues when both fail, which means the number this page reports is the conservative one — if you want the comparable figure, treat a length you cleared on the second run as legitimately yours.

Yes, and by more than most people expect — span tracks how fast the digit names can be pronounced. Baddeley, Thomson and Buchanan showed in 1975 that what fits in verbal memory is roughly what you can say in about two seconds, and digit names vary a lot by language: the Chinese names are short and single-syllable, and Chinese speakers routinely reach nine or ten, while Welsh speakers score lower than English speakers on the same task for the same reason in reverse. If you are bilingual, rehearsing in whichever language has the shorter numbers is a genuine advantage rather than a trick.