N-Back test online: 1, 2 and 3 places back

Letters appear one at a time, and you respond when one matches the letter 1, 2 or 3 places earlier. Your hits and false alarms show how well you keep a moving list up to date in your head.

3 levels · about 6 min · space bar or tap

Start the N-Back test

Your result appears after the 3-back block, and a free account keeps it.

B, F, B is a match2-back · the letter 2 places back is outlined

How it works

3 levels, from 1-back to 3-back, each with its own walkthrough and practice before the scored block. About 6 minutes in all.

1 · Level 1
1-back
Respond when a letter repeats the one just before it. The scored block has 21 letters, 6 of them matches.
2 · Level 2
2-back
Compare each letter with the one 2 places earlier, so K, B, K is a match. 22 letters, 7 matches.
3 · Level 3
3-back
Now look 3 places back. 23 letters, 7 matches. This is the hardest level, and some misses are expected.
4 · Before each level
Walkthrough, practice, countdown
An animated example you can replay, a practice round of 7 to 9 letters with feedback, then a 3-second countdown into the scored block.

Each letter stays on screen for half a second; a cross then holds its place, and the next letter arrives 2.5 seconds after the previous one appeared. You can respond at any point in that window, and only your first press counts. On a computer you press the space bar; on a phone you tap the Match! button at the bottom of the screen.

Practice rounds show a strip of the last few letters with the one n places back outlined, and tell you after each press, and after each match you let pass, whether you were right. The scored blocks drop both the strip and the feedback, so every comparison is made from memory. The letters come from a set of 12 consonants: B, C, D, F, G, H, J, K, L, M, R and T.

Wayne Kirchner described the task in 1958 to study age differences in memory for rapidly changing information. It has since become one of the most used working memory tasks in brain imaging, where a meta-analysis linked it to a network of frontal and parietal regions (Owen et al., 2005).

What it measures

At every letter you compare, decide and then update: the oldest letter drops out of the set you are holding and the new one joins it.

Working memory updating

Keeping the last 1, 2 or 3 letters in mind, in the order they came.
Swapping the oldest letter for the new one every 2.5 seconds.
Checking the new letter against the one at the right distance back.

Control and attention

Holding back on near-repeats: a letter can also recur at the wrong distance, such as 1 back during the 2-back block, and that is not a match.
Staying with the stream for about a minute per block, with no pause.
Responding within the 2.5-second window, before the next letter replaces the current one.

The load rises with each level, which is why most people find 1-back easy and 3-back hard. Researchers use the n-back mainly as a measure of updating, and performance on it relates to fluid reasoning, especially at higher loads (Jaeggi et al., 2010).

It is not interchangeable with span tasks such as Digit Span. A meta-analysis found that n-back and complex span scores correlate only weakly (Redick and Lindsey, 2013), so the two are best read as related but different views of working memory.

A test, not a diagnosis.

This is one run of one task. It cannot confirm or rule out any condition, and struggling at 3-back is common rather than a sign of anything.

What you get

The result screen opens as soon as the 3-back block ends.

On the result screen

A summary line and a marker on a strong-to-weak scale, both based on your 2-back d-prime.
For each level: your d-prime, the share of matches you caught and the share of other letters you responded to.
A full data table per level: hits out of matches, false alarms, accuracy, d-prime and mean response time on hits.
A downloadable CSV with one row per letter, practice included.

After you save it

Stored in your results once you create a free account, and there on any device you sign in from.
Placed next to the other tests you take, so working memory can be read beside other areas.
Followed by a suggested next test, so your set of results keeps growing.

How this version is scored

Every scored letter ends in one of four outcomes. The result screen counts them per level and builds each score from those counts.

Hits and misses: A hit is a press on a match; a miss is a match that passes without one. Only the first press on a letter counts.
False alarms and correct rejections: A false alarm is a press on a letter that is not a match. Letting a non-match pass is a correct rejection.
Accuracy: Hits plus correct rejections, divided by all scored letters in the level, as a percentage. Because about 70% of the letters are non-matches, a block with no presses at all still scores close to 70%.
d-prime (d′): How well you separate matches from non-matches, independent of how readily you press: d′ = z(hit rate) − z(false alarm rate), where z turns a proportion into a standard score. Higher is better, and 0 means your presses did not track the matches at all.
Extreme rates: A perfect hit rate or a false alarm rate of zero would make d′ infinite, so the runner applies the log-linear rule (Hautus, 1995): hit rate = (hits + 0.5) ÷ (matches + 1), and the same for false alarms over non-matches.
Mean response time: The average time from the moment a letter appears to your press, over hits only, in milliseconds.
Main score: The 2-back d′ is saved as this test's main result, with 2-back accuracy beside it. The 1-back and 3-back levels are reported alongside.

Why d-prime leads. Most letters in each block are non-matches, so accuracy rewards pressing rarely, while pressing often catches every match at the cost of false alarms. d′ weighs the two against each other. In a clinical sample, Haatveit and colleagues (2010) found the 2-back d′ a better working memory index for desktop testing than two Wechsler span subtests.

One summary, three levels. The summary line uses fixed cut-offs on the 2-back d′ and nothing else. Your 1-back and 3-back results do not move it; they appear in the level-by-level rows just below.

Before you start

Each scored block runs for about a minute without a pause, so settle in before the countdown.

For a steadier run

On a computer, rest a finger on the space bar before the block starts.
On a phone, hold it steady and keep your thumb near the Match! button.
Silence notifications: one lost letter can break the list you are holding.
Replay the walkthrough if a level's rule is not clear yet.

Interpret cautiously when

Something interrupted a block, or you were tired, unwell or in a hurry.
Fast responses are hard for you, for example with slower assistive input.
You have done n-back tasks before; practice raises scores on this task.
Your device lags or the screen is small, which can shift the timing or make letters harder to read.

Questions

What is a good score on the N-Back test?

It depends on the level and on the version. Scores fall as the level rises, and published n-back studies differ in stimuli, timing, the share of matches, the number of trials and whether they report accuracy or d-prime, so their averages do not transfer from one version to another. This page shows no reference table for that reason. Your result screen gives your own d-prime and rates for each level; compare repeat attempts made under similar conditions rather than holding one result against a single cut-off.

What is the difference between 1-back, 2-back and 3-back?

The number is how far back you compare. In 1-back you only check whether the letter repeats the previous one. In 2-back you hold 2 letters and update them with every new one, and in 3-back you hold 3. Each step adds to what you must keep in mind and reorder, which is why accuracy usually drops from level to level.

Is this the same as dual n-back training?

No. Dual n-back shows two streams at once, usually a square's position and a spoken letter, and training versions raise or lower the level as you go. This test uses one stream of letters at 3 fixed levels and measures a single attempt. It is a test, not a training program.

Can I take the N-Back test on my phone?

Yes. On a phone a Match! button appears at the bottom of the screen in place of the space bar. Timing and screen size can affect quick-response tasks, so if you compare attempts, use the same kind of device each time.

Can I repeat the N-Back test?

Yes, whenever you want, from the tests hub. N-back scores improve with practice as the task becomes familiar, so a better second run does not always reflect a real change. Leave some time between attempts and keep the same setup.

Sources

By Cristian Dominguez Rein-Loring, Creative Technology Lead · Updated October 10, 2026