Number Memory Test: Find the Longest Number You Can Hold and Type Back
Digits appear one at a time, then you type the whole number back in order. Every number you get right adds another digit, up to twelve, and two misses at the same length end the run. You get the longest number you held, how many sequences you got right, and four editorial bands that describe this run. There are no norms for this task.

Number Memory Test is a free, confidential timed task: sequences that grow until you miss twice, about 4 minutes, entirely in your browser. You get your number length with a plain-language read across 4 ranges, plus a free PDF of your result. It is an original task for curiosity and self-reflection, not a clinical measure.
What a number length actually tells you
This is one of the oldest measurements in experimental psychology still in daily use. Joseph Jacobs ran it on schoolchildren in 1887 by reading out digits and asking for them back, and the method has barely changed. What it captures is narrow and real: how much unrelated material you can hold in order for a few seconds without writing anything down. It is not intelligence, it is not a clinical finding, and it is not a measure of how good your memory is in the rest of your life. This page is honest about one more thing most number memory tests are not: it runs the same mechanic as our working memory test, with a longer ceiling and a different read.
It starts short and grows
The first sequence is three digits. Get it right and the next one is four, then five, and so on. That staircase is the whole method: instead of guessing where your limit is, the task walks up to it and stops. Because it adapts, the number you end on is where your own run stopped, or the twelve-digit cap, rather than a pass or fail against somebody else's list. Digits are shown one at a time, each on screen for a moment, and no two neighbors repeat.
A twelve-digit ceiling, and we say so
The task stops after a correctly entered twelve-digit sequence or two misses at a length. Twelve is a design ceiling, not a population threshold. Published averages from other protocols provide background only and do not rank this implementation's scores.
Two misses at a length end it
After one incorrect submitted answer, you receive another sequence at the same length. A second miss there ends the run. Your result is the longest sequence entered correctly. Digits appear on a schedule, but response typing is untimed; a typing error still counts as a miss.
| Feature | Typical free quiz | Psychology.com |
|---|---|---|
| How the digits are shown | Often unstated, which changes the task completely | Stated plainly: one digit at a time, each briefly on screen, not the whole number at once |
| Where the bands come from | Percentiles or a memory age, with no sample behind them | Editorial bands, named as ours, with no percentile and no norm claimed |
| Honesty about overlap | Several near-identical tests sold as different things | We say this is the same task as our working memory test with a longer ceiling, and link to it |
| What the number is called | Sometimes framed as memory level, IQ or brain age | The longest number you held on this attempt, which is not intelligence and not a clinical finding |
| Answers stay on your device | Often collected | Scored in your browser; only summary scores if you request a PDF or report |
| Paid report at the end | Common | None on this page; the task result is free and complete on screen |
| Clinician-reviewed interpretation | Rarely | Yes, clinician reviewed |
How the result is read
Your result is read from your number length against 4 ranges. Higher is better. These ranges are educational reference points from the task design, not clinical norms.
| Result | Range | What it suggests |
|---|---|---|
| 0–4 digits | Four digits or fewer | The run ended with no correct sequence, or a longest correct sequence of three or four digits. |
| 5–6 digits | Five or six digits | Your longest correctly entered sequence contained five or six digits. |
| 7–8 digits | Seven or eight digits | Your longest correctly entered sequence contained seven or eight digits. |
| 9–12 digits | Nine digits or more | Your longest correctly entered sequence contained nine to twelve digits. |
How the number length is scored
Your result is the longest sequence you typed back in the correct order, and nothing else goes into it. Getting four digits right after missing a five does not help; getting nine right once is enough, even if you then miss ten twice. The two supporting numbers on the result screen, sequences correct and sequences attempted, are there so you can see whether your run was clean or whether you scraped past several lengths on a second try. A clean run and a scrappy one that end on the same number mean slightly different things, and the task is not precise enough to tell you which.
A result of zero is possible and it is not a finding. It means two submitted answers at three digits were wrong. The task does not record why, so it cannot tell you whether that was a typing slip, a moment of distraction or anything else; leaving the tab while digits are showing cancels that sequence instead of scoring it, so a backgrounded tab is not one of the explanations. Scores of one and two cannot happen at all, because the task never presents a sequence shorter than three. If you got a zero and the run felt wrong, start it again before reading anything into it.
The four bands at 4 and below, 5 to 6, 7 to 8, and 9 and above are editorial cutoffs written for this page. We chose them; they are not derived from any sample, and no standardization sample sits behind them. The published digit-span figures elsewhere on this page were measured with spoken digits and did not set these boundaries. They are not clinical thresholds either. One digit either side of a boundary moves the label without meaning much, because a single attempt at this task is noisy by nature.
Seven plus or minus two, and why your real number may be smaller
George Miller's 1956 paper is the most quoted number in psychology: the span of immediate memory is about seven items, plus or minus two. It is quoted more often than it is read. Miller's own point was that the limit is not a fixed quantity of information but a quantity of chunks, and that the chunk is the unit that matters, which is why a phone number you know is one thing and seven random digits are seven things.
Nelson Cowan revisited the question in 2001 and argued that once you strip out rehearsal, grouping and long-term knowledge, the core capacity is tighter: closer to three or four items than to seven. Both can be true. Seven is roughly what people manage on a digit task because digits are easy to rehearse and easy to group; three or four is what is left when you take those supports away. A number memory task sits at the generous end of that range, which is one reason it should not be read as a measure of your working memory in general.
The closest published work is Woods and colleagues (2011), who tested digit span by computer with lists that adjusted to performance. Their digits were spoken, not shown, and their second experiment started at five digits rather than three, so their numbers describe a different task from this one. For the record, their two-error maximum length, the metric closest to our stop rule, averaged 7.36 digits in a younger sample of 30 volunteers with more education and 6.35 in a broader community sample of 763 adults aged 18 to 65 who took a shorter version, and age and years of education both moved the number. Read those as background about digit span research, not as the middle of this scale.
Chunking is the strategy people reach for, and what it does and does not buy
Grouping digits is one strategy you can try: 4 9 2 7 3 8 can become three two-digit groups. The task does not record which strategy you used or establish how much it changes your score.
How far this can go is one of the best demonstrations in the literature. Ericsson, Chase and Faloon (1980) trained a single volunteer on digit span for more than 200 hours. He used his knowledge of running times to convert digit groups into race results, and his span rose from about seven digits to 79. His memory for other material did not improve. That is the honest moral of the story: practice buys you a bigger number on this specific task, not a better memory.
Which is also why repeating this test to track yourself does not work. The second run is easier because you have learned the format and probably found a grouping strategy, so a higher number cannot be told apart from practice. If you want a different reading rather than a repeat, Verbal Memory Test asks you to recognize a word list instead.
Word Memory Test uses four-choice recognition after intervening questions, while Visual Memory Test uses spatial patterns. Strategies may affect both; neither is a strategy-free check of your digit-span result.
Why this runs the same task as our working memory test
It is the same mechanic. Working Memory Test (Digit Span) uses the same digit sequences, the same one-digit-at-a-time display and the same two-miss stop rule. The differences are the ceiling, which is longer here, and the read: that page frames the number as a capacity estimate and sits inside a set of cognitive screens, while this page answers the plainer question people actually type into a search box, which is how long a number they can hold. We would rather tell you that than have you take both and feel sold the same test twice.
If you came for a capacity estimate and want it discussed alongside attention and processing, take the working memory version. If you have already taken it, your number here should land close to it, and this task cannot tell you whether a gap between the two comes from sleep, caffeine, noise, the device or something else. Our other two timed tasks measure different things: Reaction Time Test measures how fast you respond to a signal, and Processing Speed Test how quickly you can work through a simple matching rule.
None of these tasks, alone or together, can detect, rule out or monitor a condition. They are unproctored, unnormed browser tasks taken in whatever room you happen to be in. If memory or thinking has genuinely changed for you or someone close to you, Memory & Dementia Concern Screen covers the symptoms side properly, and a medical appointment with three concrete examples written down is worth more than any number on this page.
Methodology & sources
This is an original timed task on our own runner, not an administration of a published instrument. Digit span itself is a public method with no owner: Jacobs (1887) introduced it, Miller (1956) gave it its famous interpretation, and it has appeared in countless batteries since. Nothing from a copyrighted test is reproduced here, there is no validation study for this implementation, and there are no norms, age corrections or reliability estimates behind it. The digits are generated at random in the browser at the moment they are shown, with no two neighbors repeating, so new runs use randomly generated numbers, which can repeat. Each digit is displayed briefly and on its own; the whole number is never shown at once, which makes this a serial presentation task and not the whole-number flash some other sites use. That difference matters, so we state it rather than leaving you to guess.
The run starts at three digits, adds one after each correct answer and ends after two misses at the same length or successful completion of twelve digits. The score is the longest correct sequence. Bands of 0 to 4, 5 to 6, 7 to 8 and 9 to 12 are editorial groupings. Published digit-span studies used other samples and procedures; they do not establish an average, percentile or clinical cutoff for this implementation.
This page uses the same digit-span mechanic as our working memory test, with a longer ceiling. One browser attempt cannot establish capacity or change in cognitive health. Conditions and familiarity may affect a run, but this implementation has not quantified their effects. Hiding the tab cancels an in-flight sequence without scoring it; typing the response is untimed.
References
- Jacobs J. Experiments on "prehension". Mind. 1887;os-12(45):75–79. doi.org/10.1093/mind/os-12.45.75
- Miller GA. The magical number seven, plus or minus two: Some limits on our capacity for processing information. Psychological Review. 1956;63(2):81–97. doi.org/10.1037/h0043158
- Cowan N. The magical number 4 in short-term memory: A reconsideration of mental storage capacity. Behavioral and Brain Sciences. 2001;24(1):87–114. doi.org/10.1017/S0140525X01003922
- Woods DL, Kishiyama MM, Yund EW, Herron TJ, Edwards B, Poliva O, Hink RF, Reed B. Improving digit span assessment of short-term verbal memory. Journal of Clinical and Experimental Neuropsychology. 2011;33(1):101–111. doi.org/10.1080/13803395.2010.493149
- Ericsson KA, Chase WG, Faloon S. Acquisition of a memory skill. Science. 1980;208(4448):1181–1182. doi.org/10.1126/science.7375930
Cite this source
Psychology.com. (2026, December 1). Number Memory Test. Psychology.com. https://psychology.com/tests/number-memory-test
Number Memory Test FAQ
What is an average number memory score?
No average has been established for this browser implementation. Woods and colleagues reported averages from different samples and protocols, which provide background but are not norms for your run. Your score is the longest sequence typed correctly, from zero or three through twelve digits.
Why does it show one digit at a time instead of the whole number?
Because that is what our runner does, and the two formats are different tasks. Showing a whole number for about a second lets you read it as a visual block, which some people find easier and others much harder. Showing digits one at a time forces you to hold each one as it arrives, which is the classic digit span method. Neither is more correct, but a score from one format should not be compared with a score from the other, and sites that do not say which they use make that comparison impossible.
How is this different from your working memory test?
It is the same mechanic with a longer ceiling and a different read. Working memory test frames the number as a capacity estimate and sits alongside our other cognitive screens; this page answers the plainer question of how long a number you can hold. Take one, not both. If your numbers differ by a digit or two across the two pages, that is the ordinary wobble of a single attempt, and this task cannot tell you where it came from.
Can I improve my number memory?
You can improve this number, which is not quite the same thing. Chunking is the usual method: break a long string into groups of two or three and hold the groups instead of the digits. This task does not record which strategy you used, so it cannot tell you how much yours changed your score. Ericsson, Chase and Faloon (1980) trained one volunteer from about seven digits to 79 over more than 200 hours of practice, and his memory for other material did not improve. So practice buys a bigger span on this task. For everyday forgetting, writing the thing down at the moment you learn it beats any span exercise.
Does a low result mean something is wrong with my memory?
A low result cannot identify a condition or explain why the run ended. Zero means two incorrect submitted sequences at length three. It does not establish that instructions were missed. If thinking or memory has changed in daily life, bring examples to a clinician. Sudden confusion requires immediate emergency medical help.
Is the test really confidential?
Yes. It runs entirely in your browser. Your individual answers are scored in your browser and are never sent to us or linked to you. They stay in this tab so you can return to your result, and closing or clearing the tab clears them. If you ask us to email the PDF, we receive your summary scores with your email address, never your individual answers. No account is needed.