Skip to main content

N-back test: a free working memory game

Letters appear one at a time. Press Match when the letter is the same as the one N steps back. The level adapts to you after every block, and your result is reported as d′ — the measure researchers use — rather than a percentage that rewards doing nothing.

Start levels
1, 2, 3, 4, 5-back
One sitting
5 blocks of 20 scored letters
Pace
one letter every 2.5 s
Score
d′ (signal detection)
Loading the game…

How to play

Pick a starting level and press New game. A letter appears every 2.5 seconds. If it is the same letter you saw N letters ago, press Match — or the space bar. If it is not, do nothing: staying silent is an answer, and the right one most of the time.

At 2-back, for example, the sequence K · T · K is a match on the third letter, because the K two steps back is the same letter. K · T · T is not a 2-back match, even though two Ts appear in a row — that would be a 1-back match.

A sitting is 5 blocks. Each block has 20 scored letters after the first N, and 6 of them are matches (30%, the standard rate).

Why the level changes as you play

The level you pick is where you start, not where you stay. After each block the game counts your errors — missed matches plus false alarms. Two or fewer and the next block is one level harder; five or more and it is one level easier. That is the standard adaptive staircase, and it keeps the task difficult enough to measure something as you improve.

The level is re-derived on our server from your responses, so the level you reached is a level you earned. Your result reports both where you started and how far you got.

How the score works: d′, not percent correct

Percent correct is misleading on an n-back task. Because only 30% of letters are matches, a player who never presses anything is "right" about 70% of the time. A score that rewards doing nothing is not measuring memory.

So the result is d′ (d-prime), from signal detection theory: the distance between your hit rate (matches you caught) and your false-alarm rate (non-matches you pressed on), in standard-deviation units. Pressing nothing scores close to zero, pressing everything scores close to zero, and only telling matches from non-matches scores well. Perfect blocks would make d′ infinite, so the standard log-linear correction is applied (Hautus, 1995).

Sequences built so the score means something

Random letters are not enough. The game uses 8 letters chosen to be hard to confuse by shape or sound, and every block is checked before it is played: no accidental matches on letters that are not meant to be targets (which would mark you wrong for being right), no runs of more than three identical letters, and exactly 3 "lures" per block — letters that match one step too early or too late, which is where n-back is genuinely hard.

What the n-back task measures

The n-back task was introduced by Wayne Kirchner in 1958 to study short-term memory, and it is now one of the most widely used working-memory tasks in cognitive neuroscience: a meta-analysis of brain-imaging studies found it reliably engages frontal and parietal regions associated with working memory (Owen et al., 2005).

It is not a pure memory test. Performance also depends on attention control and on resisting interference from the lures, and its correlation with other working-memory measures is weaker than its popularity suggests (Kane et al., 2007). Treat your d′ as a measure of how well you do this task.

Does n-back training raise IQ?

Probably not, and we would rather say so. A 2008 study reported that training on a dual n-back task improved fluid intelligence (Jaeggi et al., 2008), which made the task famous. Later meta-analyses found the evidence much weaker: improvements on the trained task are reliable, but transfer to intelligence tests is small at best and tends to disappear in studies with active control groups (Melby-Lervåg & Hulme, 2013; Melby-Lervåg, Redick & Hulme, 2016). The most favorable meta-analysis, from the original study’s group, still put the gain at a small effect, about a quarter of a standard deviation (Au et al., 2015).

Play n-back because it is a good, hard attention-and-memory workout, or to track your own performance over time — not because it will raise your IQ.

Frequently asked questions

What is a good n-back score?

There is no universal cut-off, because d′ depends on the level you played. As a rough guide, a d′ above 2 at a given level means you are reliably telling matches from non-matches; around 1 means you are above chance but making many errors; close to 0 means guessing. Compare your d′ with your own earlier sittings at the same level.

Is this dual n-back?

No — it is single n-back with letters. Dual n-back adds a second stream (usually positions on a grid) that you track at the same time. Single n-back is simpler to learn and is the version most often used in research on working memory.

Which level should I start at?

Start at 1-back if the task is new to you; the staircase will move you up within a block or two if it is too easy. Experienced players usually start at 2- or 3-back.

Do I need an account?

No. The 1-back start level is free without an account. Signing in is free too and opens every level and keeps your results; anything you played as a guest is moved to your account when you sign in.

Will n-back make me smarter?

It will make you better at n-back. Research has not shown that it reliably improves intelligence or other skills; see the section above for the studies.

Sources

  1. Kirchner, W. K. (1958). Age differences in short-term retention of rapidly changing information. Journal of Experimental Psychology, 55(4), 352–358.
  2. Owen, A. M., McMillan, K. M., Laird, A. R., & Bullmore, E. (2005). N-back working memory paradigm: A meta-analysis of normative functional neuroimaging studies. Human Brain Mapping, 25(1), 46–59.
  3. Kane, M. J., Conway, A. R. A., Miura, T. K., & Colflesh, G. J. H. (2007). Working memory, attention control, and the n-back task: A question of construct validity. Journal of Experimental Psychology: Learning, Memory, and Cognition, 33(3), 615–622.
  4. Jaeggi, S. M., Buschkuehl, M., Jonides, J., & Perrig, W. J. (2008). Improving fluid intelligence with training on working memory. Proceedings of the National Academy of Sciences, 105(19), 6829–6833.
  5. Melby-Lervåg, M., & Hulme, C. (2013). Is working memory training effective? A meta-analytic review. Developmental Psychology, 49(2), 270–291.
  6. Au, J., Sheehan, E., Tsai, N., Duncan, G. J., Buschkuehl, M., & Jaeggi, S. M. (2015). Improving fluid intelligence with training on working memory: A meta-analysis. Psychonomic Bulletin & Review, 22(2), 366–377.
  7. Melby-Lervåg, M., Redick, T. S., & Hulme, C. (2016). Working memory training does not improve performance on measures of intelligence or other measures of "far transfer". Perspectives on Psychological Science, 11(4), 512–534.
  8. Hautus, M. J. (1995). Corrections for extreme proportions and their biasing effects on estimated values of d′. Behavior Research Methods, Instruments, & Computers, 27(1), 46–51.