Memory
Number Memory Test
Digits appear one per second; you type them back in order. The string grows by one digit each time you get it right. Your score is the longest string you recalled correctly — a count, not a percentile, and not a claim about your memory in general.
Number memory
Digits appear one at a time, one per second. When they stop, type the whole string back in order. Each time you get it right the string gets one digit longer. Two misses at the same length ends the test.
What your equipment contributed
- Your display
- not measuredassuming a typical 60 Hz panel
- Each digit shown for
- 700 ±16.7 msquantised to whole frames
- Effect on your score
- nonethe result is a count, not a duration
A digit can only appear and disappear on a frame boundary, so a 700 ms exposure is really 700 ms give or take one frame. Unlike our reaction test, that error does not meaningfully reach your score here — a few milliseconds either way on a 700 ms exposure does not change how many digits you can hold, and the result is a count of digits rather than a measured duration. We state it because the same equipment limit applies to every test on this site, and because telling you when a limit doesn't matter is the only thing that makes it credible when we say one does.
How this is measured
How the digits are presented
One digit at a time, on a one-second cadence: each digit is shown for 700 ms and followed by a 300 ms gap. The gap is not decoration — without it a repeated digit like 77 would look like one long 7, and the test would quietly under-present exactly the strings that are hardest to hold. Digits are drawn uniformly at random from 0–9, independently each time, so repeats occur as often as chance dictates.
Why one per second, rather than the whole string
Because the reference range below was collected at one digit per second, and a string shown all at once is a measurably easier and different task — you can re-read it, chunk it into pairs, and rehearse the chunks. Showing the whole string and then comparing the result to digit-span research, which is what most versions of this test do, compares two things that were never measured the same way. Matching the rate is what earns us the right to cite the figure at all.
How the test ends, and what the score is
The string starts at three digits and grows by one every time you are right. Two misses at the same length end the test — a single slip is not enough to establish that a length is beyond you, which is the same reasoning behind the reference study's two-error rule. Your score is the longest string you recalled correctly, and every round you played is listed on the results panel, including the strings you missed.
What we cannot tell
We cannot distinguish a typo from a forgotten digit. A transposed pair and a string you never encoded at all are both scored as a miss, and we do not guess between them — showing you the string next to what you typed is the honest alternative to inventing a correction. We also cannot tell whether you wrote the digits down, looked away, or were interrupted mid-string. The score is only worth what the conditions you took it in were worth.
What your equipment contributes
Very little, unusually for this site. Your display quantises each digit's 700 ms exposure to whole frames, so it is really 700 ms give or take one frame — but the result is a count of digits rather than a measured duration, and a few milliseconds of exposure either way does not change how many digits you can hold. The panel above states the figure anyway, because a site that only mentions equipment error when it is convenient is not being careful, it is being selective.
How your result compares
Forward digit span is one of the most heavily studied measures in psychology, so a defensible reference range exists. In the larger of the two samples we cite, 763 adults aged 18–65 averaged a forward span of 6.35 digits (SD 1.15). A smaller, younger sample of 30 adults aged 18–46 averaged 7.36 digits (SD 0.88), with individual spans ranging from 5.7 to 9.0. Forward span did not correlate significantly with age in either sample.
What we matched, and what we could not
Matched: the presentation rate of one digit per second, and the principle of the two-error stopping rule.
Not matched: their digits were spoken through headphones and recalled aloud; yours are read on a screen and typed. Heard digits reach phonological memory directly, while digits you read have to be recoded into sound first — so the two tasks load short-term memory differently, and not by a constant that we could subtract. Their digits were drawn from 1–9; ours from 0–9. Their stopping rule was a calibrated maximal-length procedure rather than our plain two-misses rule.
Read 6.35 digits as a bearing on what typical looks like for a closely related task — not as a line you have passed or failed.
No percentile is shown. The source gives a mean and a standard deviation but not the shape of the distribution, and span scores are bounded integers rather than a smooth normal curve, so a percentile computed from those two numbers would be arithmetic dressed up as a measurement. The modality difference above compounds it: even a perfectly characterised distribution for the spoken task would not tell us where you sit on the typed one. And a percentile drawn from BlinkBench's own visitors would be worse than either — self-selected, unvetted, and impossible to verify.
Source: Woods, D. L., Kishiyama, M. M., Yund, E. W., Herron, T. J., Edwards, B., Poliva, O., Hink, R. F., & Reed, B. (2011). Improving digit span assessment of short-term verbal memory. Journal of Clinical and Experimental Neuropsychology, 33(1), 101–111.
Last reviewed: July 2026
Frequently asked questions
Why are the digits shown one at a time instead of all at once?
Because the research this page compares you against presented digits one per second, and a string shown all at once is a different task — you can re-scan it, group it, and rehearse the groups, which sequential presentation does not allow. Most number memory tests online show the whole string and then compare you to digit-span figures anyway. Matching the presentation rate is the thing that makes our comparison mean anything, so we matched it.
Why do I get two attempts at each length?
Because stopping at the first mistake would measure one unlucky round as much as it measures memory. The reference study used a two-error rule for the same reason: it takes more than a single slip to establish that a length is genuinely beyond you. Two misses at the same length ends the test, and your score is the longest string you got right.
Can you tell whether I forgot a digit or just mistyped it?
No, and we don't pretend to. A transposed pair and a completely lost string are both scored as a miss. Every string you were shown is listed alongside what you typed, so you can see for yourself which kind of miss it was — but that judgement is yours, not something the score encodes. Nothing stops you writing the digits down either; the number only means something if you don't.
Is a digit span of 7 good?
It is close to typical for the research task, which is not quite this task. In the largest sample we cite, 763 adults aged 18–65 averaged 6.35 digits (SD 1.15) — but those digits were spoken through headphones and recalled aloud, while yours were read on a screen and typed. Reading and typing engage the phonological loop differently from hearing and speaking, so the two numbers are not interchangeable. Treat the published figure as a rough bearing.
Related tools
Five trials measuring how fast you respond to a colour change, reported as a median with the share your own display could account for.
Open tool →Hearing Range TestThe highest frequency you can still hear, measured on a stepped ladder with silent catch trials — and bounded hard by your headphones, your volume and your device's sample rate.
Open tool →Sequence Memory TestThe longest sequence of block positions you can reproduce in order, presented at the rate the board-based research used — and reported as a screen score, not a clinical span.
Open tool →