The number memory test asks you to hold a sequence of digits long enough to type it back. Most people improve at it quickly with a simple change in strategy, then plateau. The interesting question is what that improvement means: a better score on this task is real, but it is not evidence of a general memory upgrade.
What you are actually training
Forward digit span is a working-memory task. You encode the digits, hold them briefly, and reproduce them in order. Performance depends on attention, encoding speed, grouping, and the short-term store. It is not a measure of long-term memory, general intelligence, or how well you remember a book. That narrowness is useful: because the task is constrained, a change in your score is easier to interpret.
Chunking is the main lever
Most people cannot hold nine or ten independent digits in mind. What they can do is group digits into chunks — two or three at a time — and hold those chunks. A sequence like 4-9-1-7-3-8-5-2 becomes easier as four short groups rather than eight separate items. Chunking does not increase raw capacity; it changes the units.
Practise grouping deliberately. Read the digits in pairs or triplets, give each group a small rhythm, and rehearse the groups rather than the individual digits. This is the strategy most likely to produce a real improvement on this task, and it is honest: you are learning a technique, not gaining a photographic memory.
Rehearsal and rhythm
Rehearsal is repeating the sequence silently to keep it active. A rhythm helps: group the digits into a pattern that sounds like a phone number or a short phrase. The rhythm is a memory aid, not a magic trick, and it works because it gives the sequence structure. If a group sounds like something familiar, the chunk becomes easier to hold.
Avoid trying to visualize a long string of digits as a single picture. Most people cannot maintain that image reliably, and the effort competes with the verbal rehearsal that the task rewards. Use the strategy that fits the task instead of the strategy that sounds most impressive.
Presentation speed matters
This test shows 3 digits for two seconds and adds 200 ms for each additional digit, capped at 3.2 seconds. The timing is part of the measurement: more time gives you more encoding and rehearsal. Two apps can both call their task "digit span" and produce different scores because they present digits at different rates or allow different rehearsal.
That is why your score here should be compared with your own score on this page, not with a number from a different test. Keep the test identical if you want to compare sessions.
What practice can and cannot improve
Practising digit span makes you better at digit span. A meta-analysis of working-memory training found reliable short-term gains on trained tasks, but little convincing evidence that those gains generalise to unrelated abilities such as reading, arithmetic, or general cognitive performance (Melby-Lervåg & Hulme, 2013). That is the realistic frame: task-specific improvement is plausible; a broad cognitive upgrade is not supported by the evidence.
This does not make practice pointless. It means the gain should be labelled for what it is: better performance on this particular memory task under these conditions. If that is your goal, practice works. If your goal is a general memory boost, look elsewhere — and be sceptical of products that promise one.
A short training routine
- Take a baseline without changing anything; note the sequence length and conditions.
- Practise chunking for five minutes with a separate digit list: read in pairs or triplets, say each group once, then recall the whole string.
- Take one more test and compare with the baseline; rest between attempts.
- Stop. A short, focused session is more useful than a long, tired one.
Keep the practice separate from the measurement. If you rehearse the exact sequences you will be tested on, you are training recognition, not recall. Use fresh digit strings for practice.
Sleep, attention, and state
Digit span is sensitive to attention and fatigue. A noisy room, a disrupted night, or a stressful morning can lower the score without any change in your underlying memory. Take the test under similar conditions if you want to compare sessions, and do not overinterpret a one-digit difference.
For the broader evidence on sleep and reaction tasks, see the sleep guide; for what training transfers in general, see the training-transfer guide.
Measuring progress fairly
Use the same device and browser, take several sessions on separate days, and compare medians rather than a single best score. A one-digit increase is within normal variation; a two-digit increase that persists across sessions is more likely to reflect a better strategy. Record the strategy you used, not only the number.
If you change your approach, keep everything else constant and see whether the change survives a second session. That is a small experiment, not a laboratory study, but it is enough to avoid fooling yourself. A stable, slightly lower score is often a better baseline than a lucky personal best. The goal of practice is not to memorise a particular sequence but to improve the strategy you apply to a new one; if a session only improves when you have seen the digits before, it has not trained the skill the test measures. Use fresh sequences, keep the strategy stable across a session, and let the median tell you whether it worked.
How the test scores a run
The test starts at three digits. A correct recall adds one digit and presents a new random sequence; the first wrong sequence ends the run. The score is the longest sequence you recalled correctly. Each sequence is shown for two seconds at three digits, plus 200 milliseconds for every extra digit, capped at 3.2 seconds. There is no time bonus for typing quickly; only the digits matter.
That scoring rule means one careless mistake ends the run, so attention and consistency matter as much as raw capacity. It also means a score is a threshold, not a full curve: a person who can reliably hold eight digits but not nine gets the same score as someone who barely held eight once. Compare medians across runs rather than treating a single score as a precise ability estimate.
Common myths about memory training
"This will give me a photographic memory." No. Digit span is a narrow verbal working-memory task; the strategies that help are chunking and rehearsal, not a photographic buffer.
"A higher digit span makes me smarter." Digit span correlates with some cognitive measures, but it is not an intelligence test and should not be used to rank people.
"Brain-training games transfer to school or work." The meta-analytic evidence on working-memory training finds near transfer and limited far transfer; claims of broad cognitive enhancement are not well supported. Practise the task you care about, and be sceptical of general promises.
"More practice is always better." Short, focused sessions with rest are usually better than long, tired ones, especially for a task that depends on attention.
A four-week practice plan
If you want to improve at this specific task, keep the plan simple and repeatable. In week one, take a baseline and practise chunking for five minutes, three times. In week two, keep the same routine but experiment with pair versus triplet grouping. In week three, add a second short session on two days and keep the rest of the week unchanged. In week four, repeat the baseline conditions and compare medians.
The plan is deliberately modest: it changes one variable at a time and gives the strategy time to settle. If the score improves, note whether the improvement survived a second session; if it did not, the change was probably noise or a temporary practice effect.
What to record
Record the score, the strategy you used (pairing, triplets, rhythm), and the conditions: quiet or noisy, rested or tired, same device. A score with no context is a number; a score with a strategy and a state note is a data point you can learn from.
If you change the strategy, keep the rest of the session identical and give the new approach a second session before deciding. A one-digit difference is usually noise.
When to stop
Stop when you notice you are rehearsing the test rather than practising the skill, or when the session has become a grind. A short session that ends while you are still accurate is more useful than a long one that ends in frustration. The goal is a slightly better strategy and a more stable score, not exhaustion.
A one-page practice card
Five minutes: read a fresh digit string in pairs, say it once, recall it, then check. Do not rehearse the exact test sequences. Track the longest sequence recalled correctly and the strategy used. Stop when accuracy falls. After four sessions, compare medians rather than a single high score.
For the measurement side of memory scores, see the ReflexTimer guides.
Sources
- Miller, G. A. (1956). The magical number seven, plus or minus two. Psychological Review, 63(2), 81–97. PubMed record.
- Cowan, N. (2001). The magical number 4 in short-term memory. Behavioral and Brain Sciences, 24(1), 87–114. PubMed record.
- Melby-Lervåg, M., & Hulme, C. (2013). Is working memory training effective? A meta-analytic review. Developmental Psychology, 49(2), 270–291. PubMed record.
- Number Memory Test — the test itself and scoring rules.
- Reaction Time Training: What Transfers — why practice is task-specific.